Abstract:Currently, fresh market apples are picked manually around the world, which requires a huge seasonal labor force to complete the job within the narrow harvest window. In addition, the shortage of skilled labor and high labor cost are increasing every year, which have threatened the sustainability of apple production. To address these issues, it is essential to develop mechanical harvest solutions for fresh market apples and other tree fruits. Shake-and-catch harvesting system offers an alternative solution to potentially obtain high harvest efficiency. Research on shake-and-catch harvesting for fresh market apples has been conducted for several decades. However, commercial machine is still unavailable. The crucial problem is that the damage rate and degree induced during harvesting is too high. Apple orchards in USA are being planted in modern trellis trained architecture, including formally trained horizontal limbs. The ready access to the limbs and fruit in formally trained orchard creates the potential for controlled localized fruit removal at the limb level. To verify the hypothesis that the newest apple tree canopies provide opportunities for targeted shaking and catching with minimal fruit damage, a new targeted shakeandcatch method was proposed. To realize this method, we developed a set of limb shaking device and corresponding capturing platform. A set of data acquisition system based on computer was also integrated to the shaking device to monitor and control the forced vibration frequency of a targeted limb. According to the capturing platform with/without separation buffer strips and horizontal/tilted capturing, four capturing patterns were tested in a commercial-grade ‘Jazz’ apple orchard (Prosser, Washington State, USA), in which trees were trained to a vertical fruiting wall architecture with seven layers. The harvesting test was conducted by shaking a tree limb and capturing the removal fruit right underneath the targeted limb. Each capturing pattern was involved in 20 samples, each of which including two adjacent limbs growing in a same layer of two trees. During testing, the shaking location was approximately at the middle of the targeted limb. Shaking frequency selected 20Hz with 5 seconds duration time for all test; shaking amplitude of the device was 3.2cm. According to USDA (United States Department of Agriculture) standards for fresh market apples, the fresh-market percentage was adapted to evaluate fruit quality. The percentage defined was that the number of apples satisfying fresh market grade divided by the capturing number in one sample shaking. Obtained results showed that fruit removal efficiency was around (81±14.5)% under the 20Hz shaking frequency and 3.2cm shaking amplitude; the fresh-market percentage in the four capturing patterns ranged from 89.5% to 96.3%; there was no significant difference in terms of the fresh-market percentage among the four capturing patterns. These results indicated that the targeted shake-and-catch harvesting method was feasible and showed a promise for mechanical solution for mass harvesting of fresh market apples; it could be permitted to adjust tilted angle within a certain range that couldn’t significantly affect fruit quality for the special trellis trained ‘Jazz’ apple trees.

Abstract:In order to obtain the deposition value of the droplets on the leaves of the fruit trees after spraying, a detecting system of droplet deposition on fruit leaves used in low altitude pesticide spraying was developed. The system consisted of dielectric leaf wetness sensor, data transmission module and host computer detection software. Through the calibration experiment of dielectric leaf wetness sensor, the regression equations of three solutions, which were tap water with 553μS/cm of EC, thiophonate-methyl with 860μS/cm of EC, and KH2PO4 foliar fertilizer with 1525μS/cm of EC, were established. The accuracies of the regression equations were verified by spectra photometry. Then, a wireless data transmission sensor system based on ZigBee was developed. Moreover, the Qt platform was used to write the host computer program including data analysis and display functions. Finally, the contrast test between water sensitive paper and this system was carried out by using WSZ-4X type plant protection UAV in a cherry orchard. The results showed that the coefficient of correlation between the density values of the droplet deposition obtained by two methods was 0.9626. For a single measurement point, the average error of the droplet deposition density was 22.8%. In the orchard experiments, under the influence of the wind speed and the unmanned aerial vehicle air flow and other environmental factors, the droplet distribution on the sensor and the water sensitive paper appeared certain differences. The consistency of the droplet deposition density obtained by two methods was preferable if the influence of environmental factors was ignored. However, the droplet deposition on leaves was acquired quickly, conveniently, timely and simply by using the developed system.

Abstract:The traditional axial-flow air-assisted orchard sprayer is lack of guidance for air-assisted spraying, which would lead to the waste of pesticide. The 3WZF-400A orchard air-assisted sprayer uses two linear guide plates to guide its air-velocity field, which could partly solve the problem, but it still could not make the distribution of air-velocity match the canopy shape at the target. To this matter, the air-velocity field distribution model was established by CFD simulation, and the optimized angle was selected with combination of the two linear guide plates as the basis for the further improvement. Short guide plates with different angles and quantities were set up to subdivide the air-velocity field above. Proper combination of the two linear guide plates could make the airflow concentrate to the canopy range, the optimized angle are 45° for the upper and 5°for the lower. Short guide plates could effectively subdivide the air-velocity field, and could concentrate the airflow to the target. But too many short guide plates would lead to the interference of airflow. The optimized quantities and angels for short guide plates are 2, 15° and -15°. The experimental verification and spray performance test were carried out. The maximum relative error between analog value and measured value was 22.87%, which caused by the attenuation of the air-velocity and the instability of the airflow source. The droplet coverage rate was used to evaluate the spray performance of the improved air-assisted sprayer, the comparison experiment between previous sprayer and the improved sprayer showed that the droplets could be reasonably distributed according to the canopy shape in the vertical plane.

Abstract:According to modern orchard planting pattern, a new self-propelled electric sprayer with multi-angle variable speed swing nozzle function was developed to accommodate the small space and the poor traffic-ability of conventional orchards which are densely planted with dwarf plants. A 4 factors and 3 levels of test was performed according to the Box-Behnken center-united experimental design principles. To find influence factors of droplets deposition distribution, the parameters were analyzed by using response surface methodology including spraying flow rate,spraying distance, angular swing speed and sprayer forward velocity. Subsequently, a mathematical model was established by using Design-Expert software,then the effects of various parameters and their interactions were discussed. The results showed that the effects order of the four parameters on distribution of droplets deposition ranked from large to small as: sprinkler swing speed,spraying distance,spraying flow rate, and sprayer forward velocity. The optimal spraying parameters should be spraying flow rate at 375.20mL/min, spraying distance of 1.72m, sprayer forward velocity of 0.14m/s and angular swing speed of 16.19(°)/s, then the coefficient of variation of droplets distribution of 11.471% was obtained.

Abstract:In order to realize agricultural machinery automatic navigation control, and taking into account the system cost and operation efficiency, the automatic navigation control decision method was studied, and a navigation software system was designed and developed. Firstly, the system conducted path planning according to the field boundary, the field shape and working requirements. Secondly,the simplified two-wheel vehicle kinematic model was used and the fuzzy control was adopted for navigation control decision. And the input parameters of the fuzzy controller were the lateral deviation and the heading deviation of agricultural machinery, and the output parameter was the steering angle data. Finally, using the steering angle data, the machine was controlled by the steering wheel through PLC controller. The modular design ideas were adopted in software development. The software mainly consisted of four modules: serial data communication, data analysis and processing, data and graphic display and data storage. It was developed based on C++/MFC programming language. The software can analyze and process the received data, such as GNSS positioning data, angle sensor data, attitude sensor data and PLC controller data, then send the corresponding control decision information to the PLC controller. In addition, the system can store the deviation data for error analysis after the navigation. The experimental results demonstrated that the automatic navigation control decision method can achieve preferable control precision. The software has user-friendly interface, stability communication and relatively complete function, so that the proposed automatic navigation control decision and software system can meet the field operation requirements for agricultural machinery.

Abstract:The spatial variability of crop yield reflects the effects of environment and management factors on yield, and obtaining accurate yield spatial distribution information is the prerequisite for optimal input of resources on demand. In order to obtain spatial information on grain yield, a real-time monitoring platform was designed and developed for the spatial distribution of farmland grain yield based on the mobile terminal .The platform can realize the remote monitoring of the real-time position, operation status and yield data of combine harvester, and then analyze the spatial distribution of yield data. The platform consists of four modules, including data reception and storage, data transmission, data display and data analysis. Among them, the data reception and storage module is used to receive the operation status information packet such as the position of the harvester, the grain flow rate, the ascending speed, the barn temperature and humidity, and the cutting width. The data is analyzed and stored in the database. The data transmission module provides Web service for the mobile terminal, and extracts the corresponding data from the database for the front end call. The data display module can display the position and operation status of combine harvester in real time on mobile terminal. The data analysis module uses the ArcGIS Server GP service to create spatial distribution of grain yield. The testing results indicate that the monitoring platform is stable, and can display the yield spatial distribution information in real time.

Abstract:During the work of GNSS (Global navigation satellite system) controlled land leveling system, the GNSS location data is not only the basis of topographic survey and land leveling datum design, but also affects the accuracy of land leveling in real time. In the land leveling work, if the GNSS location data is not accurate enough, the results of the terrain measurement can not truly reflect the actual terrain data and affect the entire farmland land leveling process. In order to analyze the GNSS location data error and improve the location accuracy, we proposed a method to analyze the error source of GNSS location data in the process of land leveling work. Combining the multi-path effect and random noise in the process of GNSS location data, we proposed an error correction method of vibration error caused by terrain fluctuation in land leveling work. We use the Kalman filter algorithm and Wavelet transform to process the location date and correct the data error, we can improve the positioning accuracy. We do farmland comparison experiments, and the analysis results showed that, in this method, the height location accuracy increased, during the land leveling work the GNSS positioning actual height fluctuation error range reduced by 20%, so we can use the method to guide land leveling work better, and can give the follow-up research scientific and rational positioning data, provide reliable data support.

Abstract:In order to realize the autonomous navigation collaboration of multiple agricultural machinery in farmland environment, the multi-machine cooperative navigation communication system was designed and developed based on TD-LTE network in this paper. The system consists of four parts: navigation positioning sensors, wireless communication module, vehicle control terminal and remote communication software. The navigation positioning sensors include the global navigation satellite system (GNSS) receiver, the inertial measurement unit (IMU), and the angle sensor, which are used to obtain the geographical position of each agricultural machinery, and their attitude and vehicle steering angle information. The wireless communication module uses 4G Data transfer unit (DTU) as the system communication equipment, which connects to the serial port of vehicle terminal, and realizes RS232 serial port conversion to TD-LTE network. After the serial port parameters and other information of 4G DTU was configured by configuration software, it connected the destination server IP address and port number. Then the sensor data collected from vehicle were transferred to the remote server communication software through the TD-LTE network, using the designed communication protocol. The vehicle control terminal is an industrial personal computer (IPC), which is used to achieve automatic control of agricultural machinery and human-computer interaction. Through applying the socket network programming, the data receiving display and data transmission function module was developed. The system is able to upload the status information of each agricultural machine in real time, and also receive the remote server control commands. The communication can be selective and the online agricultural machine have higher priority. Four Levuo tractors were used as the experimental platform, and the frequency of communication was 5Hz. The testing results showed that the packet loss rate is 0.1%, and there is no delay. Therefore, the system has a high reliability and real-time performance.

Abstract:In order to realize the real-time remote monitoring of multi-machine cooperative navigation operation, a Web-GIS based remote monitoring platform was designed and developed for cooperative operation of multi-machine. The platform mainly includes data transceiver, data storage, data query, data display and data analysis module. The data transceiver module adopts Socket technology, is used to receive real-time multi-machine position and attitude information and other operations, and also can send remote control commands to the vehicle terminal. The data storage module is responsible for storing the received job information into the corresponding SQL Server data table. The data query module is used to query the history information of multi-machine operation, and to present the query result in the web. The data display module combines Web-GIS technology, and can realize real-time visualization of multi-machine operation track through real-time interaction with Baidu map server. The data analysis module can analyze and process the location and attitude information of multi-machine in real time, and make decision analysis and task scheduling for each machine. The experimental results show that the platform has good stability, it can display multi-machine operation track and job information in real-time, and can realize multi machine task scheduling. Therefore, it can meet the demand of multi machine cooperative operation.

Abstract:The laboratory simulation of the sowing depth consistency control system is mostly testing parallel four-bar linkage or up-and-down movement due to surface fluctuation, while the actual farmland surface morphology varies irregularly in both elevation and slope. In view of this, a farmland-terrain simulation system suitable for rugged topography is designed to study the irregularities of farmland surface morphology led by both topographic fluctuation and terrain tilt. The system consists of terrain simulation mechanism, hydraulic system, control system, etc. The terrain simulation mechanism is connected to the rack through hydraulic cylinder to simulate farmland surface fluctuation. The hydraulic system controls the hydraulic cylinder to drive the terrain simulation mechanism through the electro-hydraulic proportional directional valve. The control system controls hydraulic system and drives the terrain simulation mechanism according to the topographic data. The physical parameters of the terrain simulation mechanism are achieved by mathematic modeling of the profiling mechanism and building the geometrical relationship between the telescopic gradient angle and the expansion or contraction of the hydraulic cylinders. Based on the force analysis of hydraulic cylinder, the parameters of the hydraulic system are determined by theoretical calculation. In the simulation at 2.0m/s operating speed, the average elevation error is 1.61mm and the average slope error is 0.56°. The experimental results indicate that the system showed rapid and accurate performance on terrain elevation and slope simulation, and it can meet the requirements of farmland terrain simulation.

Abstract:In order to solve the fertilizer layered problem caused by the different density of mixed pellet fertilizer in Heilongjiang province, a four-element variable rate fertilization control system was developed based on 2BJM fertilizer machinery. According to the four different speeds of fertilizer feed shafts，the control system could work on either setting value or fertilizer maps condition by electro-hydraulic proportional control technology. It also integrated a submeter-scale differential GNSS device. The system could calculate the target speed of the hydraulic motor in real time according to the target value, and send the speed instruction to the fertilizer controller synchronously. Once the system received motor speed signal, the opening of proportional valve could be adjusted automatically. And then the four-element synchronization variable fertilization (nitrogen, phosphorus, potassium, and micronutrient fertilizer) could be carried out. The results of field experiment showed that the errors of the fertilizer tubes were less than 3.0%, and the variation coefficient was less than 0.05. Compared with the traditional machinery area, though the maize growth indexes of height, weight, aboveground biomass and SPAD did not increase significantly, all of these variable coefficients reduced obviously. The nitrogen and phosphorus fertilization contents reduced from 217kg/hm2, 232kg/hm2 to 150kg/hm2, 200kg/hm2 respectively. The potassium fertilization content increased from 79kg/hm2 to 108kg/hm2. The final yield data was 12200kg/hm2, increasing by 1.81%. All the experiments showed that the four-element variable rate fertilization control system could solve the fertilizer layered problem and satisfy the need in practice.

Abstract:The engine is primary power source of harvester, and its punch combination position has a direct impact on machine assembly quality and further affections on operating efficiency and reliability of the whole machine. At present, the engine punch combination position measurement is mainly performed manually with quite low precision and bad consistency. Focused on the automatic measurement of engine punch combination position, a machine vision based online measurement method for engine punch combination position of large-scale harvesters was proposed in this paper through establishing a punch combination position error model. A certain number of cameras were used to capture the two-dimensional images of each energy mounting hole, and the punch combination position error was measured and calculated real-timely through online calibration, image enhancement, feature extraction, coordinate transformation and so on. On this basis, a measurement and control software was developed based on Labwindows/CVI, and realized the rapid measurement of punch combination position. In this paper, engine punch combination position of a certain type of harvesting chassis of corn harvester was studied. The experiment results showed that the punch combination position relationships could be achieved effectively, and the error analysis and assessment could be fulfilled by the established punch combination position error model. Furthermore, its measuring results were superior to traditional calipers measuring tools readings in resolution and accuracy with a highly efficiency. This proposed online measurement method could meet the demand of automated production-line measurements.

Abstract:Laser pulse launching by LiDAR sensor has strong penetrability and sun shine as well as extreme weather has little influence on it, because of which, it can genuine acquire the three-dimensional information on the ground. It is an ideal data source for crop canopy structure information extraction. In this paper, based on the airborne laser radar data, as the goal was to extract the corresponding feature ground point. Using the TerraSolid software to classify the whole points, the points were divided into different classification, such as ground, vegetation, wire power and wire line. Meanwhile, RANdom SAmple Consensus (RANSAC) was applied to fit the plane segmentation model based on the Point cloud library (PCL), which optimized the obstacles extraction results. The TerraSolid software classification results, PCL plane segmentation fitting results with initial classification of point cloud for confusion matrix were obtained, respectively. Confusion matrix for precision evaluation was concluded. Correlation analysis was carried out on two kinds of precision evaluation. Research results show that it is better for TerraSolid to deal with block rather than the whole point cloud data. The results of TerraSolid and PCL are similar for the same point cloud. Its operation is fast and efficient but poor for the visibility. We can combine both advantages in extracting obstacles. This study basically achieved the anticipated goal of flying obstacles extraction, to provide security for the unmanned aerial vehicle (UAV) flight and help with the flight path planning.

Abstract:In the automatic fruit picking system, it is one of the most important aspects to recognize apples, especially green apples. Quick and accurate identification directly affects realtime operability and reliability of picking robot. In order to realize recognition of green apple in natural illumination condition, images of apple trees in natural growth period were taken, and random forest algorithm was used to classify and identify green apples. To solve complexity and fuzziness of green apples and fruit trees and complex background’s color and texture features, especially similarity of green apples and leaves on many characteristics, the Otsu threshold segmentation method was applied to remove the background noise and tree trunk and branches in images in RGB space so that images contained only green apples and leaves were obtained. After filtering processing on images, the grey level information and texture features of apples and leaves were extracted respectively, and they were used to train and build the green apple identification model based on random forest algorithm. Then green apple prediction experiments were carried out for the sample images by using template pixel scanning, and the predicting accuracy reached 90%. Finally, ten green apple tree images were chosen to execute green apple recognition by using the model, and with Hough transform method to mark the identified apples. It illustrated that the green apple recognition rate reached 88%. The results showed that the method had a good robustness, stability and accuracy, and it could be used to recognize green fruits under natural illumination conditions.

Abstract:The chlorophyll content of crop could indicate the growth status of crop and the spatial model of the chlorophyll content in the field could help to make the decision on the field management. In order to develop a spatial distribution mapping system of winter wheat’s growth parameter, the research was conducted to find a method to establish the model with high accuracy and sparse sample data points. The chlorophyll content and the GPS data were measured in the field. The 67 samples were involved and divided into the calibration and validation samples randomly. Firstly, according to the calibration data, the distribution maps of chlorophyll content were drawn by using inverse distance weighting (IDW) method and ordinary Kriging(OK) method respectively. The fitting data could be extracted at the same GPS position where the validation data located. And then, the accuracy analysis was discussed between the fitting data and the measured validation data. The results show that there is a positive correlation between them based on both IDW and OK map. However, compared with the OK method, the correlation coefficient between IDW fitting and validation data is higher with r is 0.722 and covariance is 1.361.The possible reason caused error in the research was also discussed. It could help to improve the mapping accuracy when the sampling points are unevenly distributed and closed to each other.

Abstract:The pruning of apple trees plays an important role in fruit set and tree shape. A five year old apple tree was selected as the research object, and the simulation model was established and compared with the actual results. First of all, the 3D laser scanner was used to scan the fruit trees after winter defoliation to obtain the point cloud data. After denoising, resampling and software processing, the skeleton nodes of apple trees were obtained to establish the apple tree structure model.Secondly, four different types of pruning treatments were carried out on apple trees. After one growth period, the growth data of the pruned branches were recorded, and the qualitative and quantitative analysis was carried out.The results showed that with the increase of fruit tree pruning, the latent bud rate and short branch rate decreased gradually, but the middle branch rate did not change significantly, but the long branch rate increased gradually.According to a parent type and each type after the corresponding branch pruning branches of the node changes, generating polynomial function corresponding to the mathematical rules for provide a virtual tree pruning simulation.Finally, the apple tree pruning simulation software for early fruit tree pruning simulation was established, a consistent change rule after pruning a simulation with actual results generated by pruning, pruning can provide for early fruit tree fruit or shaping reference for next year.

Abstract:The canopy leaf is an important part for the fruit tree development, and the spatial distribution and morphological structure of canopy plays a vital role in fruit yield and quality. The study on three-dimensional (3D) reconstruction of apple tree leaf blade can not only analyze the morphological characteristics, but also provide the theoretical basis for the calculation of canopy light distribution and pruning of fruit trees. With the rapid development of laser scanner technology, three-dimensional laser scanner is widely used in 3D reconstruction of different fields due to the advantages of non contact, high precision and high efficiency. A method was proposed for 3D reconstruction of apple tree leaf blade based on point cloud. And the accomplishment of the 3D reconstruction method was on the basic of presented point cloud simplification method. First of all, 3D laser scanner was used to obtain point cloud of apple tree leaf blade. Secondly, search the K-neighborhood points based on the bounding box method and calculate the average distance between the midpoint and its neighbors. The noises were identified according to distance threshold and then they were removed . Thirdly, extract the boundary feature points and retain them as a mandatory feature area. For the non-boundary points, according to the relationship between the midpoint normal vector and its neighbor normal vector, the points were reduced to different degrees. Finally, the 3D reconstruction of apple tree leaf blade is realized. The results showed that the constructed leaf model can preserve the 3D morphological features of the leaf, which can provide a basis for further canopy reconstruction and light distribution calculation.

Abstract:Leaf temperature is the key factor that reflects the growth conditions of winter wheat, but it is difficult to obtain the dynamic variation and profile distribution of wheat field leaf temperature. The simultaneous heat and water (SHAW) model describes heat and water movement in a plant-snow-residue-soil system in detail.Taking Shangqiu as the study area, based on the localization,the SHAW model was used to simulate the leaf temperature time series curve and profile from 0cm to 60cm separated by 10cm from jointing period to heading period of winter wheat.The simulation results were analyzed by combining with field observed data at the same time. The results showed that the SHAW model can be used effectively to simulate leaf temperature time series curve and profile, and the coefficient of determination reached to 0.8476. The night simulation results were significantly better than the daytime,and the coefficient of determination were 0.8622 and 0.7602 respectively. The analysis of the average, minimum and maximum values of leaf temperature simulation showed that RMSE ranged from 1.36℃ to 4.09℃. The minimum temperature simulation effect was the best, and the average value was second, and the maximum temperature error was the greatest. The analysis of the leaf temperature profile showed that height decision coefficient all achieved above 0.82, and increased with the height. RMSE was ranged from 2.41℃ to 3.35℃, and MBE were all less than 0℃. The leaf temperature generally showed a tendency to decrease with the height in the night, while increase with the height during the day.

Abstract:Wheat plant water content is one of factors that affects the feeding rate, separation loss and selection cleanliness of combine harvester. A rapid nondestructive detection method for wheat plant water content measurement was proposed based on near-infrared photoelectric sensor. In this system, the near-infrared detectors of different wavelength were designed and the system was built to sense the reflection intensity of wheat plant by the method of median filter and reference real-time correction. Multiple sets of samples were tested and analyzed, then multivariate linear regression, multivariate stepwise regression, partial least squares, and least squares support vector machines were used to build the model respectively. The results showed that the model based on least squares support vector machines performed best, whose correlation coefficient of the correction set came up to 0.9742. Other sets of samples were tested by the established mode, and found out that the correlation coefficient between the true and predicted value of wheat plant water content was 0.9337 and standard deviation of predictive sets were less than or equal to 3.00%. This paper provided a rapid and nondestructive detection method and equipment for wheat plant water content, which was beneficial to improve the work performance of combine harvester by real-timely adapting the operating parameters according to the wheat plant water content.

Abstract:Leaf area index (LAI) is an important parameter for evaluating crop growth. It is also the main index to determine the spraying dosage of UAV. In order to establish a real-time measurement method of crop leaf area index (LAI), a rapid measurement system of winter wheat leaf area index was developed based on Android mobile platform. Under the condition of field, 10 evenly growing experimental areas were selected, and the canopy images of wheat were obtained by using the Android mobile platform and the ADC multi-spectral camera at different growth stages. Meanwhile, the actual leaves’ areas of wheat were artificially measured. According to the different measurement results, three kinds of leaf area index were calculated: (1) Converting Android mobile phone images from RGB to HSV, then calculating the green leaf area (IArea) after image segmentation on H-V dual channel combination image. (2) The leaf area index ALAI retrieved by the normalized vegetation index (NDVI) and the adjusted soil vegetation index (SAVI) data obtained by the ADC multi-spectral camera software. (3)The leaf area index (LAI) of actual manual measurement. With the different growth stages of wheat, the correlation analysis and modeling analysis of the above three leaf area indices showed that the R2 between the IArea obtained by the Android mobile platform and the actual measured leaf area index LAI was R2>0.84(P<0.01). And the R2 between the leaf area index ALAI obtained by the ADC and the actual measured leaf area index LAI was more than 0.83.

Abstract:The Diptera fruit fly adults of B.dorsalis Hendel, the B.tau Walker and the B.cucurbitae are the dominant species in the south of China. Because of its wide host range and high risk, it has been the most serious pest in the citrus growing areas in South China. Under the premise of accuracy, how to reduce the human and material resources for monitoring insect pests is an urgent problem to be solved. From the view of image recognition, this paper studied the morphological characteristics of the harmful flies, and proposed a classification algorithm. In the algorithm, Hough transform was used to detect the lines of fly wings to correct the direction of fly and define the effective area of the stripe by lines. Filtering in HSV space was used to detect the scutellum of fly waist and abdomen. A combination of the two ways separate the mesonotum from the whole fly. According to definition formula of characteristic factor of the central stripe, four shape feature parameters are extracted to form the feature vector after digital processing. Feature data sets were built by collecting feature vectors in 90 sample images, and the BP neural network was trained to get the neural network model parameters for the classification of the flies. Experimental results showed that the recognition effect of this method on Diptera fruit fly adults had a good accuracy and real-time, under the condition that the distribution of the wings of the flies and the distribution of the pectoral fin stripes were clear. It greatly reduces the requirement of image clarity, and is more suitable for dynamic identification of video streaming devices. The recognition accuracy of B.dorsalis was 95.45%, the B.tau Walker was 93.33%, the B.cucurbitae was 97.83%. The overall accuracy rate was 95.56%.The average time of single recognition was about 500ms, which can meet the needs of practical applications. The identification model proposed in this study has good expansibility for Diptera adults.

Abstract:The identification of individual plant maize core at seedling stage is the key to complete the operation according to the plant. It can improve the accuracy of variable fertilization for individual corn, and can further improve the utilization rate of fertilizer. A fast recognition method of maize core based on top view image was purposed. Firstly, by using the super green factor to enhance maize plant of seedling stage, the maize plants were separated from soil and shadow. And based on the enhanced images, Otsu method was used to automatically determine the optimal threshold for image segmentation, in effect of avoiding shadow influence and separating the maize plants correctly. Then by using the image brightness of the plants at seedling stage as one-dimensional coordinate, the elevation map of maize was drawn, the central region of each maize plant showed as a shape of water basin. The level set was used to determine and locate the central area of each maize plant. And a method of dividing and conquering was used to reduce the level set scale and search the minimum value area of each maize plant. Through data validation, the results showed that the recognition rate reached 96%. It indicated that the algorithm was feasible in real-time. In addition, since the method and level set method were combined to determine the center area of each maize plant, the algorithm was adaptive and not affected by weather factors, which improved the robustness of the algorithm in the field operation. The time complexity of the algorithm was O(lgn), which could meet the real-time field operation.

Abstract:Remote sensing images are generally large images. For the subsequent analysis of this kind of image, there is a common method of dividing image into blocks, while the boundary effect is easy to occur in block processing. Therefore, the elimination of boundary effects is a problem that needs to be studied in block processing. The most common way to eliminate the boundary effects is to extend the image. Symmetry extension, zero extension and periodic extension are the common extension methods. The conventional extension method is not applicable because texture in farm remote sensing images carries important information. Thus, according to the line characteristics shown on remote sensing images, a new extension method based on texture orientation was proposed in this paper. Here, we used the method of multi-scale interpolation wavelet to solve the partial differential equation, according to the change of gray level of the image. In this method, external collocation points were chosen adaptively. Thus the computational efficiency could be greatly improved. Then, the texture direction of farm remote sensing images was identified by using bounding boxes, and the texture is further extended along the texture direction. Experimental results show that the image extension method proposed effectively overcomes the shortcomings of the conventional extension method, greatly improve the efficiency of calculation and the boundary effect is effectively eliminated.

Abstract:Because of the influence of nature meteorological condition and background environment during the acquisition of the plant leaf image, the image degradation is always unavoidable with the salt and pepper noise. The image of plant leaf is generally characterized by rich textures and welldefined edges. It is unfavorable to the subsequent processing of color image with noise pollution. Although there are several filtering methods such as average filtering, wiener filtering, gauss filtering and median filtering, they do not satisfy the requirment on effective repairation and texture reservation of image. Consequently, to repair the image successfully with the textural details preserved and the edges clear, a new model for color image inpainting was proposed and called adaptive probabilistic PCA method. The procedure of the proposed model included 2 steps.After the leaf vein was identified and tracked based on tree, the vein inpainting was conducted by the probabilistic principal component analysis (PPCA) model, in which the iterations were adaptively selected according to the PSNR value of the restored images. To evaluate the effectiveness of the proposed model, a 3-step simulation test was invloved, and the evaluation criteria based on SNR and structural similarity image measurement(SSIM) was used to measure the degree of image distortion and similarity between the processed and the original image. Firstly, to determine the optimal iterations of the PPCA model, the inpainting results in different iterations were compared. Secondly, to test the image inpainting ability, the polluted images are simulated with different levels of noise. Finally, the proposed model had some comparison with the conventional filtering methods. The experiments showed that the iterations about 550 were appropriate while using the PPCA model for image inpainting. The restored image obtained by the proposed model was less residual noise and clearer textures than other filtering methods visually. The PSNR value of restored image was 26.8199dB, which was higher than using the wiener filtering, gauss filtering, average filtering and median filtering, by 9dB,7dB,6dB and 1dB, respectively. It was higher than the PSNR value of the noisy image by 14.48dB. The SSIM value of restored image was 0.9557, which was the largest among the above-mentioned methods. It indicated that the restored image using the proposed model was closer to the original image in the brightness, contrast and structure aspects. It could provide technical support to the subsequent processing of the color image.

Abstract:Non-destructive detection of chlorophyll content and drawing the chlorophyll distribution map of potato crop leaves could indicate crop growth and guide field management. In this paper, the hyperspectral imaging technique was used to diagnose the chlorophyll content index and help to describe chlorophyll distribution of potato leaf. The hyperspectral images of 65 potato leaves were collected and divided into 400 regions of interesting (ROI). Meanwhile, the SPAD values of these 400 ROI samples were measured. After extracting and calculating the average leaf spectrum of the chlorophyll measurement area, the 12 chlorophyll content sensitive wavelengths were chosen by the Monte Carlo uninformative variables elimination (MC-UVE) algorithm and the 23 chlorophyll content sensitive wavelengths were selected by the competitive adaptive reweighted sampling (CARS) algorithm. They were used to establish the partial least squares regression (PLSR) model of chlorophyll content index of potato leaves respectively. The results were as follows: 12 sensitive wavelengths selected by MC-UVE algorithm were 532.54nm, 534.27nm, 566.78nm, 737.60nm, 741.61nm, 742.51nm, 759.49nm, 772.92nm, 816.54nm, 880.88nm, 928.84nm, 943.88nm. The modeling determination coefficient was 0.79, and predictive determination coefficient was 0.73. Meanwhile, 23 sensitive wavelengths selected by the CARS algorithm were 394.01nm, 399.94nm, 492.03nm, 493.32nm, 494.18nm, 534.27nm, 536.86nm, 537.30nm, 537.73nm, 543.79nm, 544.22nm, 545.52nm, 547.25nm, 547.69nm, 548.12nm, 550.29nm, 550.72nm, 553.76nm, 555.49nm, 938.93nm, 986.36nm, 987.74nm, 1018.30nm.The modeling determination coefficient of the PLSR diagnostic model built with these wavelengths was 0.82, and predictive determination coefficient was 0.80. Thus, the chlorophyll content of potato leaves can be calculated by CARS-PLS model, and the visual distribution map of chlorophyll content in potato leaves was plotted by using pseudo-color drawing. It provides a method for the diagnosis of chlorophyll distribution in the future.

Abstract:Accurate prediction of wheat chlorophyll content is important for guiding precision management in the field. The canopy spectrum of winter wheat canopy was measured by ASD Field Spec Handheld 2, and the first-order differential processing method was conducted on the band of 400~900nm in the paper. In order to select the sensitive bands for the chlorophyll content detection of winter wheat, the Gram-Schmidt transformation algorithm was used in the research. The insignificant variables and the redundant information were identified and removed from the independent variables set. As a result, the orthogonal transformation data of first-order differential at 848nm, 620nm and 677nm were extracted. A representative set of wheat chlorophyll content of modeling samples was selected by using sample set partitioning based on joint x-y distance algorithm (SPXY). The results showed that multiple linear regression (MLR) prediction model based on Gram-Schmidt and SPXY algorithm is better than the random sampling method. The chlorophyll content of winter wheat were clustered respectively at intervals of 0.2mg/L, 0.3mg/L and 0.5mg/L. The modeling results showed that the optimal resolution was at 0.3mg/L, the determination coefficient R2c and the R2v of the calibration model which was built based on 620nm and 677nm sensitive bands were respectively 0.730 and 0.739. The study could help to evaluate the nutritional status of winter wheat and precision fertilization.

Abstract:The health of plants is directly related to the quality and quantity of agricultural products,therefore the disease detection of plants is an important research problem in agriculture. A multi-classification detection method based on kernel function support vector machine (SVM) was proposed for classification of healthy leaves and diseased leaves, and the detection of four diseases, including Alternaria alternata, Anthracnose, Bacterial Blight and Cercospora leaf spot. Because the image of diseased leaf was changeable, firstly, the contrast of diseased part and the healthy part was enhanced by the preprocessing,making the disease part more obvious. Then, leaf features were segmented and extracted on “a” and “b” component of the Lab color space. Using K-means clustering method, the clustering effect was enhanced. Finally, support vector machine (SVM) based on kernel function was used to identify and detect the four diseases. To improve the detection accuracy, 500 iterations were used to assess the maximum precision. Considering the influence of the coefficient of cross validation, 40% of the samples were used as validation data set, 60% were used as the training data set. Radial basis kernel function was adopted to carry out the training. In this method, traditional two kinds of leaf disease identification was extended to four kinds, and the experimental results proved the effectiveness of leaf classification of four kinds of diseases.And the recognition rate of the 4 diseases was the highest, reaching 89.5%, and the lowest was 70%.

Abstract:During the collection of agricultural images, noise often caused by environmental factors, and it often affects the final result of image processing. Thus, it is important to improve the quality of agricultural image. In recent years, the non-local means filter based on patch-ordering method has been applied to deal with Gaussian noise, which has obtained great success in denoising. However, the method suffers a shortcoming of long processing time and higher memory requirements, especially in large image processing. In order to improve the denoising effect, a block optimization algorithm was used in this paper. Firstly, the sampling image was split into several blocks, in which the number of the blocks was adapted to the image texture richness. After comparison with the speed of computer and the algorithm complexity, the segmented image blocks were obtained with an appropriate size to guarantee that they could be processed by the computer. Each image block was process separately. In view of the boundary effect caused by the combination of the processed image blocks, the method of image extension was applied to effectively eliminate the boundary influence and improve the image denoising effect. Experimental results show that, for general hardware devices, improved non-local means based on patch-ordering method could rapidly process the noise image commonly used in agriculture. For the size of the 512 pixels×512 pixels images, when the noise standard deviation was 50, the partition number was 16, the improved Non-local means based on patch-ordering method can effectively deal with the noise image, and the processing speed with 64 partitions was 1.89 times than 16 partitions.

Abstract:On internet agricultural technology platform, thousands of new questions are waiting to be answered by experts every day. It is generally doubted because of slowly response time and uncertain quality of the manual services. An intelligent response system based on agricultural technology knowledge base can help to answer some questions automatically. To build the knowledge base, it is necessary to recognize triples of “crop-disease-pesticide” named entities from mass of existing questions and answers data. However, fewer studies are reported on recognition methods for named entities of diseases and pesticides in Chinese, and accuracies of those for named entities of crops are low. Thus, a recognition method based on conditional random fields (CRF) was proposed, which recognized crops, diseases, and pesticides named entities from agricultural technology questions and answers data. In the method, question and answer texts was formatted and split to pieces of corpus. Each corpus piece was automatically annotated with several features, including whether it contained characteristic Chinese characters and characteristic radicals, whether it was numeral, whether it was the left or right bound of a compound word, and part of speech. A CRF model was trained with these annotated texts to classify pieces of corpus, including judging whether they were parts of crop, disease, or pesticide named entities and recognizing positions in named entities. With the trained model, three types of named entities could be accurately recognized and triples could be associated automatically. Recognition accuracies and time cost of model training were optimized by choosing input feature combinations and adjusting sizes of context windows in experiments. Accuracies of recognizing crops, diseases, and pesticides of this method were 97.72%, 87.63% and 98.05% respectively, which were significantly higher than existing methods.

Abstract:A strong correlation exists between fertilizer application and grain yield. Due to many factors affecting grain yield, the existing fitting methods of correlation between the two variables lead to large errors. Aiming at the data of fertilizer application and grain yield in Taihu Lake Basin, the back propagation (BP) neural network was used in this paper to model the correlation between the two variables accurately, which could guide to reduce use of fertilizer. This paper collected average fertilizer use and grain yield data per acre in 35 years i.e. from 1980 to 2014, in 16 counties and cities in Taihu Lake Basin. Missing items were filled automatically through a time series analysis approach called auto-regressive and moving average model (ARMA). For average grain yield data, ARMA(2, 6) model had higher accuracy with mean square error (MSE) less than 0.2 and R2 more than 0.85. For average fertilizer use, ARMA(3, 7) model had higher accuracy with MSE less than 0.02 and R2 more than 0.80. Then BP neural network with a single hidden layer (1-10-1) was established to fit correlation fertilizer use and grain yield data in each country. Goodness of the fit with BP neural network was better than other methods, with MSE less than 0.12 and R2 more than 0.80. Results indicate that there is a threshold for fertilizer use. When fertilizer is used less than the threshold, grain yield per acre is more, whereas when it is more than the threshold, grain yield per acre fluctuates and the average keeps invariant. The correlation implies excessive application of fertilizers can not achieve high yields.

Abstract:Carbon and nitrogen metabolism of plants reflect the physiological status and growth vigor, which is the physiological basis of peanut yield and quality formation. Therefore monitoring the ratio of sugar to nitrogen in real-times important for growth diagnosis and nitrogen management for peanut plant. The main purpose of this study was to establish a quantitative model for monitoring the ratio of sugar to nitrogen in peanut leaves, through analyzing the relationship between the ratio of sugar to nitrogen and the canopy hyper-spectral parameters. The experiment was carried out on peanut variety NO.1. Different nitrogen level treatments were applied, and canopy hyper-spectral parameters were collected in different growth stages of peanut. Then the regression model of canopy hyper-spectral parameters and leaf ration of sugar to nitrogen was established The results showed that peanut leaf sugar to nitrogen ratio was in the dynamic change pattern of “high-low-high-low” within the growth process. And from seedling to pod, peanut leaf sugar to nitrogen ratio for all N treatment group was lower than the control group, and in the peanut harvest period N treatment, it was the opposite. The most suitable period for monitoring hyper-spectral parameter of leaf sugar to nitrogen ratio is seedling stage to full fruit maturity stage. Flowering and knit stitch stages vegetation index and the ratio of sugar to nitrogen correlation reached a significant level, and in these two stages, the regression equation established by DVI was more reliable. The R2 value of it was 0.866~0.993, and SE value was 0.026~0.083. After the peanut pod, using the EVI to establish regression equation is more reliable. The R2 value was 0.893~0.927, and SE value is 0.054~0.082. It is found that the characteristics DVI and EVI have the best performance, which could be used to monitor the ratio of sugar to nitrogen in the early and late growth stages reliably.

Abstract:Turbidity of water refers to the content of suspended particles which are composed by large number of bacteria and pathogens. In aquaculture, the aquatic animals are influenced by the particulate matter concentration. It is necessary to monitor the turbidity. The measurement results of traditional turbidity detectors are easily influenced by environmental light, chromaticity, temperature factors. In order to solve the problems mentioned above, the intelligent optical fiber turbidity sensor was designed and developed based on IEEE1451.2 communication protocol. The sensor was composed of photoelectric detection module, signal transmission module and intelligent processing module. In the photoelectric detection module, a 880nm infrared light emitting diode was applied as the light source, and the scattered light was measured at vertical direction which could effectively reduce the influence of chroma. In the signal transmission module,the collected scattered light current signal was processed to obtain a linear DC voltage signalwith a linear relationship by the proportional circuit, I/V conversion, filtering and detecting circuit. In order to reduce the temperature influence and improve the detecting accuracy of the sensor, an intelligent processing module was designed based on IEEE1451.2 communication protocol. The temperature compensation algorithm was proposed and the TEDS parameter was calibrated to coducte the temperature compensation of the turbidity sensor. The performance of the sensor was tested. The experimental results showed that the accuracy of the sensor was within ±1.5%, the stability error was within ±1.0%, which satisfied the requirement of aquaculture turbidity measurement.

Abstract:In view of the river pollution control and water management, this study put forward a hybrid model of autoregressive moving average (ARIMA ) model and wavelet neural network combined with genetic algorithm, to predict the river water quality. For time series data of water quality parameters, it includes linear and nonlinear sequences. So using the least square method to estimate the ARIMA model parameters, ARIMA model was used to predict linear data. For the nonlinear relationship among the residual error data, prediction result, and original data, using genetic algorithm to optimize wavelet neural network (WNN) parameters, including selection, crossover and mutation operation, WNN was applied to obtain predicted data, which increased the traditional WNN prediction precision significantly. Experimental results show that the mean absolute error of ARIMA model, wavelet neural network ,genetic algorithm optimized wavelet neural network(GAWNN), or the hybrid model without genetic algorithm optimized model prediction results are 0.29%, 0.39%, 0.26% and 0.24% respectively. The mean absolute error of the combined model prediction is about 0.19%, which is the minimum, indicating that the prediction result is better than that of single model and the hybrid model without genetic algorithm optimized.

Abstract:Plant height is an essential phenotype parameter for assessing plant vigor and estimating plant biomass. In order to rapidly measure individual maize plant height, a method based on RGB-D (red, green, blue-depth) camera was proposed in this paper. Color images and depth images of maize at the jointing stage were captured using RGB-D camera in field. First, the color image of maize was processed by graying and binarizing. Then, morphological open operation was conducted within region of interest, and the largest region of maize image was extracted to remove weed and little leaves. Second, the optimized watershed algorithm was applied to the maize gray image and the boundary was generated, then the circle fitting was carried out for the boundary points. After that, the skeletonization operation was conducted for the maize binary image. There were crossing points at the contact points between leaves, and ending points at the end of leaves. The crossing points and ending points were searched and saved, and the distances between the center of the circle and each crossing point were calculated. Only the crossing point that was nearest to the center of circular was chosen as the maize center. Next, the Dijkstra algorithm was used to find the nearest paths between the maize center and each ending point. The color coordinates of the paths were saved and the corresponding point cloud data were generated based on the mapping relationship between color coordinate, depth coordinate and camera coordinate. Third, the differences between neighbor points of every path were calculated to determine the potential measurement points of the target maize and remove the point cloud data belong to nontarget maize leaves. All the paths were compared to find the highest point of maize. The histogram statistic method was applied for point cloud data that were around the highest point of maize to extract ground. Finally, the difference between the highest point of maize and ground was calculated to measure individual maize plant height. Samples were tested to verify the aforementioned method, and the results demonstrate that the method proposed in this paper has a good performance in measuring individual maize plant height. The mean errors and root mean square error (RMSE) of measuring plant height were 1.62cm and 1.86cm respectively, indicating that the proposed method can be applied to monitoring plant growth.

Abstract:Droplet adhesion on target is a common practice in pesticide spray. In order to accurately measure the droplet size and evaluate the distribution, adhesive droplets need to be identified and segmented by image processing. An improved method for judging the droplet adhesion and extracting the features based on droplet shape factor and area threshold of droplets was proposed in this paper. The adhesive droplets were counted by using ultimate erosion and iterative opening operation and segmented by using the watershed algorithm marked with the iterator open operation. The connection areas of the segmented droplets were marked and the shape was rounded. Experimental results showed that the method can effectively extract the characteristics of the adhesion droplets. The accuracy rate of judgment is 100% for weak adhesion and up to 97.2% for strong adhesion. The droplet size obtained by this image process is very close to that measured by laser particle size analyzer. Comparing with Deposit Scan software, this method may improve the size measurement accuracy about 7.67%. Based on the same sample, the comparative analysis showed that the proposed method may obtain the droplet numbers in a much faster speed than the artificial counting and achieved an accuracy of more than 97.06%. The research result also showed that compared with the laser particle size method, the image measurement method was more simple, effective, and suitable for measurement and statistics of the droplet parameters in the field spray experiment.

Abstract:Land ecological security is the foundation of resources, environment and ecosystem health. Focusing on the incompatibility of various evaluation indexes of the land ecological security system, this paper evaluated land eco-security using the theory of matter element analysis. An evaluation index system of land ecological security was established based on EES-PSR model. The land ecological security classification standard, the evaluation index and its characteristic value were regarded as matter-elements. Through the normalization of the evaluation index and the actual data to be evaluated, a comprehensive evaluation model of land ecological security was established after getting the classical domain, joint domain, weight coefficient and correlation degree of the matter element model. The evaluation model was applied to the land ecological security assessment in Heilongjiang, Harbin. With the support of GIS technology, the visual expression and analysis of the evaluation and evaluation of land ecological security in the study area were achieved. The results showed that during 2011 to 2015, the level of land ecological security in Harbin changed from “insecurity” to “security”, and the most prominent change occurred after 2013. The area of afforestation, the proportion of primary industry to GDP, the consumption of million yuan GDP energy (standard coal), the rate of industrial waste water treatment and the area of public green space per capita had been “ insecurity” for many years and were the restriction factor of land ecological security level in Harbin City. In 2015, the level of land ecological security was low in the middle and southwest of Harbin, but it was higher in the northern and southeastern regions. It is concluded that matter element model analysis had certain guiding significance for regional land ecological security assessment.

Abstract:Land ecological security is the core of human sustainable development. Due to the rapid growth of China’s economy and the rapid advance of urbanization, the irrational use of land resources has produced a series of land ecological security problems, such as soil erosion, land salinization, land desertification and other issues. The relationship of each factor in the land ecological system is complex, but the qualitative problem of the existing evaluation method is more serious, because it cannot reflect the relationship between the index and the system. In view of that problem, matterelement model was introduced into regional land ecological security evaluation. The evaluation results show that 87.45% land of ecological security level of Da’an County is at the safety level, nevertheless, the overall ecological security level still has room for improvement. Therefore, in order to improve and enhance the level of land ecological security, and to promote the sustainable development of urban ecology, groundwater exploitation needs to be controlled and reduced, the level of coal mining needs to be reduced, vegetation coverage needs to be increased, and energy conservation awareness needs to be enhanced, and the high-polluting industries relocation or ecological management policies needs to be resolutely implemented. In the face of unchangeable natural conditions and unavoidable meteorological disasters, the relevant departments of Da’an County should take appropriate measures to control the adverse effects, caused by disasters and reduce the damage caused by artificial disturbance to the ecological environment. Matter element analysis can reveal the information of individual evaluation index. Evaluation method of land ecological security based on the matter-element model was an objective and scientific comprehensive assessment method. This study could provide a reference for the sustainable development of regional land resources.

Abstract:Taihu Lake Basin was selected as the study area. The single factor pollution index and Nemerow comprehensive pollution index wereusedto evaluate pollution level of soil heavy metals. Descriptive statistical analysis, Pearson correlation analysis and factor analysis were used to analyze the sources of heavy metals in Taihu Lake Basin. The main conclusions were as follows: the comprehensive pollution level of soil heavy metals in Taihu Lake Basin was between safety to moderate pollution. Hg, Cu and Zn had great harm to Taihu Lake Basin’s soil environmental quality and the effect of Hg was the most profound and decisive. Cr was the only element that did not pose a hazard to its soil environmental quality. The source analysis results showed that Hg, Cu and Zn in the soil of Taihu Lake Basin were mainly derived from industrial sources. As, Pb and Cd were mainly derived from agricultural fertilizer and pesticide, among which Pb and Cd were also affected by natural factors. The source of Cr was relatively natural and was not affected by obvious human factors.

Abstract:Simulation of land use spatial pattern can reveal the geological regularity and identify the driving factors of regional land use change from multi-scales, which is an important way to clarify land use changes process and pattern for the future period. This paper took Shuangcheng District in Harbin City as a case and developed a simulation model based on Cellular automata (CA) combined with Artificial neural network (ANN). Totally 14 driving force factors were selected from four aspects which had significant impact on land use change, including the distance variables, number of adjacent land use type, unit natural attribute and socio-economic factors. They were used in the ANN-CA model to simulate land use change of Shuangcheng District from 2002 to 2013. The accuracy of simulation was verified by using the actual interpretation data in 2013 and it was 85.26%, which showed the model could be used to simulate the LUCC of Shuangcheng District. Furthermore, the land use pattern in 2024 was simulated under four scenarios: natural endowment scenario, rapid economic development scenario, basic farmland protection scenario and land use planning scenario. The results showed that land use pattern presented obvious spatial diversity under different scenarios, different driving factors resulted in diverse changes of various land-use types, and socio-economic factors played an important role in the conversion of land use types. The study outcomes could provide scientific decision support for the sustainable utilization of land resources in Shuangcheng District.

Abstract:The simulation of land use change was studied based on multi-agent model. This paper took Yinchuan Plain as a case study and chose the remote sensing images in 2002 and 2014 as data source. Based on the analysis of the land use pattern evolution and its driving actors, three kinds of agents: government, farmers and urban residents were selected as agents. The decision behavior of these three agents were analyzed combined with 5 factors selected from natural factors, social factors and economic factors and behavior rules of multi-agent were built. Finally, the basic data which were processed by ArcGIS, was imported into the model, and the simulation model of land use spatial pattern based on multi-agent model was established. Furthermore, in order to reveal the response relationship between land use pattern evolution and driving force factors, the study simulated land use pattern under three scenarios: natural increase scenario, rapid social economic development scenario, and basic farmland protection scenario. The simulation results can provide decision support for scientific planning and utilization of land in the study area.

Abstract:In order to overcome the influences of multi-collinearity and absorbance non-linearity in near-infrared spectroscopy on predicting soil total nitrogen content, the gray correlation-extreme learning machine method was used to select the combination wavebands with good prediction capability to establish high precision prediction model for soil total nitrogen content. First, the first derivative spectra was used to get the sensitive spectrum area. And then the grey correlation sensitive wavelength selection method was used to select wavelengths which were respectively 1007, 1128, 1360, 1596, 1696, 1836, 2149 and 2262nm. Finally, by using the above sensitive wavelengths as input data, a soil total nitrogen prediction model was established based on the method of extreme learning machine and multiple linear regression. As a comparison, while using the traditional correlation analysis method to select the sensitive wavelengths, the results showed that R2c of the soil total nitrogen forecast model established by using gray correlation-extreme learning machine was 0.9134, and the prediction R2v was 0.8787. Its accuracy was higher than that of the traditional modeling method. It indicated that the gray correlation-extreme learning machine method had more obvious advantages especially in the prediction of low soil total nitrogen content.

Abstract:In soil nitrate-nitrogen (NO-3-N) detection based on ion-selective electrode (ISE), coexisting chloride ion (Cl-) is the primary interference factor. Currently most detection devices detect only single ion and are off-line. In order to solve these problems, this paper aimed at developing a soil NO-3-N detection device with multiple parameters based on ISE. The device embedded a back propagation (BP) neural network model and achieved on-line and real-time detection. Five methods were adopted to improve the BP neural network model due to its shortcomings of slow convergence rate and easily falling into local minimum. Two correction methods were used to calibrate detection results of the device. Several judgement programs were applied to improve the stability of electric potential acquisition. Standard solution tests were conducted to validate the accuracy of device. The experiments of NO-3-N detection using 20 soil samples was conducted, and the detection results were compared with that of linear regression model and optical detection to validate the effect of reducing the interference of Cl- and the soil NO-3-N detection accuracy. The results showed that the deviation between the detection results of the device and that of an ion meter was less than 1.0mV, meeting the soil NO-3-N detection accuracy requirement. The average relative error between the soil NO-3-N detection results of the device and the results detected by optical method was 8.83%, while the average relative error between the results of linear regression model and the results detected by optical method was 12.17%. The fitting coefficients R2 were both greater than 0.97. It indicated that the device could effectively reduce the interference of Cl-, had high accuracy and could be used for on-line detecting soil NO-3-N.

Abstract:To monitor crop residue area, three ESA Sentinel-1A synthetic aperture radar (SAR) VV and VH polarization data were generated at 25m spatial resolution for Siping in Jilin Province, China, from September to November. In this study, we analyzed the backscattering characteristics of the residue area and other typical objects. The difference of the objects under different polarizations combination mode were compared. The experimental results show that a high recognition accuracy of crop residue area can be obtained using the support vector machine (SVM) method if appropriate phase is selected. Specifically, classification result obtained from VH and VV polarization radar images combinations has higher classification accuracy. In this combination, the identification accuracy of crop residue area is 90.26% and the overall identification accuracy is 86.15%.

Abstract:It is important to master the law of soil water infiltration for the rational development of irrigation scheme, setting irrigation parameters and improving irrigation technology. In order to explore the law of soil water infiltration under the condition of moistube-irrigation, HYDRUS-3D finite element model was used to simulate the soil water infiltration under micro-irrigation. The effects of initial pressure head and soil texture on soil water infiltration were discussed. The numerical simulation results showed that the wetting body was diffused around the micro-tube in the vertical section of soil water infiltration, and the diffusion rate was positively correlated with the initial pressure head. The average infiltration rate of soil water was 1.85cm/h in the first 5 hours, the average infiltration rate was 0.79cm/h between the 6th hour to 15th hour, and the average water infiltration rate in the period from 16th hour to 36th hour was 0.59cm/h. The maximum soil moisture content appeared around the micro fabricated tube and tended to decrease to the periphery. The migration distance of soil wetting peak increased with the increase of initial pressure head. The infiltration rate of micro-irrigation increased in three soil layers (sandy loam, loam, loam loam). When the pressure head was -180cm, the average water diffusion rates of the three soils were 0.69cm/h, 0.53cm/h and 0.46cm/h, respectively.

Abstract:To resolve the problems that the auto control parameters can not be modified according to the staggered demand or the manual control mode can not be remotely switched during the greenhouse environmental control process, a greenhouse environmental control system was developed based on embedded Web server. The system was divided into three levels: the local greenhouse measurement and control layer, the embedded Web server layer and the user adaptation layer. In the local greenhouse measurement and control layer, a wireless sensor network was used to collect air temperature, humidity, light and carbon dioxide. And a real-time video monitoring system was installed to enrich the information of local greenhouse. In the server layer, the embedded BOA server guaranteed the communication between the hardware and users. In the user adaptation layer, the administrator could directly check the greenhouse environment and control equipment manually through dynamic HTML page or the Android application. Besides, users could remotely access the webcam in greenhouse so that they could observe running status of the equipment and growth status of the crop in the greenhouse. To verify the reliability and practicability of the system, the greenhouse environmental control experiments were conducted. The process of greenhouse environmental control was simulated for cucumber during northern winter. Taking the temperature as main parameter and the humidity and the light as auxiliary parameters, the data collected from the trials showed that the process of auto measurement and control took 5s at least, and the process of manual measurement and control took 3s at least. It indicated that the system was reliable and stable for greenhouse environment monitoring and controlling.

Abstract:The current greenhouse data acquisition system is implemented in the way that data acquisition terminal uploads data to the host computer to manage the data or transfer them to cloud server. The network structure is relatively complex and the power consumption is large. In order to solve the above problems, a greenhouse environment monitoring and temperature prediction system was developed by using the Internet of Things, cloud services and WeChat platform. In this system, the data collection terminal directly connected the Internet to the cloud server through WiFi/GPRS to interact with the data, and the mobile terminal accessed the cloud server to obtain the data service through the WeChat public number. The temperature forecasting model adopted the differential time series model to solve the influence of seasonal periodicity in the temperature prediction process. The data analysis showed that the system effectively realized the lightweight and mobility of the data acquisition terminal. The relative error of temperature monitoring was less than 4.96%, and the relative error of temperature prediction was less than 3%. The prediction result has high precision and can meet the needs of daily production.

Abstract:This paper proposed an active greenhouse heating system which focuses on the influence of crop growth under extreme cold weather. At the beginning, the heat transfer process of heating system was analyzed based on thermal equilibrium theory. The temperature control system models were established based on radial and axial heat conduction mathematical analyses respectively. Then, the numerical simulation of two dimensional steady temperature distribution was carried out by means of finite element method (FEM) software. The simulation result indicates the control parameters in the greenhouse. When the effective operation range (later referred to as Qf) is defied with the soil temperature more than 15℃ at 20cm underground, the temperature of heating system should be opened no less than 28℃. The kinetics is changed in terms of the temperature dependence. If the air intake temperature of heating system is raised by 2℃, the axial effective operating range extends by 2.4~2.8m. Later, field test was conducted in greenhouses located in Changping District, Beijing. Compared with simulation data, the field testing data shows that the perturbation law of heating system heat transfer process on shallow and deep layer soil temperature is consistent with theoretical analysis. The simulation parameters could help to guide the greenhouse control.

Abstract:The collection of soil sampling information mostly relies on manual recording by far. In order to record and process soil information automatically, a new method for recording soil sampling information based on Android was proposed. Combined with the server and database architecture, it achieved wireless automatic sampling and soil information recording. Firstly, the system architecture and software function were designed. Then the cloud server and field information database management were set up and built. Based on JSP and MySQL database management system, the server imported the JDBC driver in the WEB project, which could make the WEB server connect with the database directly and guaranteed the interaction between the WEB server and MySQL database. The data in the DBMS was packaged into JSON format and then sent to the Android platform. The Android client was designed to implement the soil auto record system Android client interface using material design specifications, and to access and operate the MySQL database in the cloud server by analyzing the JSON data. Finally, the validity and robustness of the system were tested. The experiments showed that the system could effectively display the collected agricultural environmental information such as air temperature, humidity, latitude and longitude and soil nitrogen content in the position of the sampling point. The experiment results indicated that the effectiveness of this Android system in recording soil sampling information and the feasibility of the system.

Abstract:Phenotypic data of plant can reflect the growing of crops, which is one of great significance to guide the integrated water management. In order to study the effect of different fertilization strategy on growth of greenhouse tomato, the common small tomato (Jade) was selected as the research object due to suitable for greenhouse planting. Four kinds of different ratios of water soluble fertilizer were set up to affect the tomato growing. During the test, phenotypic parameters of tomato were collected respectively under the different fertilization levels, including plant height, stalk diameter, number of internodes, number of inflorescence, fruit production and so on. ANOVA single factor analysis method was conducted to analyze the experimental data. The results show that tomatoes with different ratios of fertilization have significant differences. Among which, the tomato plants with high fertilization level get the best production, and normal fertilization level is fit to plant growing with the largest single fruit weight and remarkable economic benefit. At the same time, the growth and fruit yield are more obviously distinguish than the relatively small amounts of fertilizer and no fertilizer. The multiple factors regression with fruit number, height, number of internodes, number of inflorescence and yield were introduced to analysis parameters effective and the best determination coefficient R2 is 0.965. The significant differences order is fruit number, plant height and number of internodes, number of inflorescent. The results of experiment show that the normal level of fertilization is best for growing, and the phenotypic data can provide reference for the establishment of irrigation strategies of the greenhouse tomato growing with fertilization equipment.

Abstract:Photosynthesis is the basis of plant growth and photosynthetic rate directly affecting the quality of fruit. The quantity and quality of tomato can be improved with the application of the appropriate amount of CO2, which is one of the principal raw material of photosynthesis. In this paper, photosynthetic rate prediction models under greenhouse condition in spring and autumn growth period were established respectively. The experimental data were collected during autumn of 2014 and spring of 2015. WSN was used to monitor greenhouse environmental parameters in real time, including air temperature, air humidity, CO2 concentration, soil temperature, soil moisture, and light intensity. An LI-6400XT portable photosynthesis analyzer was used to measure the photosynthetic rate of tomato plants, and the environmental information of leaves was controlled by small chamber environment. In order to verify the universality of the established model, three models using the data from both spring and autumn growth period, data only from spring growth period, and the data only from autumn growth period were established. The photosynthetic rate prediction models of single leaf were established based on the back propagation (BP) neural network. The environmental parameters were used as input neurons and the photosynthetic rate was taken as the output neuron. In order to improve the prediction accuracy of the model, the input neurons were standardized using Z score method and then processed by principal component analysis. Principal components were selected according to the principal components’ cumulative contribution rate. The photosynthetic rate prediction models of single leaf were established after principal components analysis and K-fold cross validation. The results indicated that the correlation coefficient of photosynthesis prediction model based on the data of spring 2015, autumn 2014 and the two seasons were 0.99, 0.95 and 0.85 respectively. The results of the models indicated that the universality of the model built using data from both seasons, and it has great potential for CO2 fertilizer control.

Abstract:CO2 is the plant’s raw material for photosynthesis. In order to realize the precise regulation of CO2 gas fertilizer under different nitrogen levels，a fitting model for the photosynthetic rate of tomato was establish throughout its growth cycle. In this paper, the experiments were conducted with 3 CO2 concentrations and 3 nutrient liquid nitrogen treatments, and the environmental information and the net photosynthetic rate of the leaf were collected by the LI-6400 portable photosynthetic rate instrument. The prediction model of photosynthetic rate in the whole tomato growth stage was established by multiple linear regression. The correlation coefficient is 0.885 and the adjusted coefficient of determination was 0.782. The experimental results show that the model has high prediction accuracy, and can be used to guide the accurate regulation of CO2 gas fertilizer under different nitrogen content.

Abstract:Closed soilless substrate cultivation is the direction of modern agricultural transformation, substrate water content is an important detection indicator of the cultivation process, but the study of the soil moisture sensor used for detecting water content of substrate is rear. To solve this problem, this paper evaluated the performance of a frequency domain soil moisture sensor FDS-100 in the substrate water content detection comprehensively. For different proportions of the substrate, compared and analyzed the standard water content obtained from the drying method and FDS-100 measurement. Besides, it also compared the measure performance of FDS-100 and ECH2O-5TE in a mixed substrate. The influence of compaction degree, EC value and temperature value on sensor is also tested. The results shows that measured values from FDS-100 measurement in different proportions substrate did not follow the same curve;the degree of compaction impacted the FDS-100 measurement, measured value increased with the degree of compaction, the measurement error is between -1% and 10%;EC had a certain influence on FDS-100 measurement, measurement error is between ±5%;temperature had less effect on the measurement of FDS-100, the measurement error is below 2%;the relative error between FDS-100 and ECH2O-5TE is 4.32%. In summary, FDS-100 can be used in substrate moisture detection, it should be calibrated for different substrates and different degrees of compaction in actual use.

Abstract:The control of pH value is an important part of water and fertilizer integrated nutrient solution circulation control system. In the process of water and fertilizer control, pH value in the optimal control is conducive to the development of root system and absorption of most minerals. In the process of nutrient solution regulation, due to the slow diffusion of the circulating line and acid fluid, the pH adjustment process has a large time lag, and the traditional PID is difficult to achieve good results. According to the characteristics of the controlled object, a mathematical model describing the process was established, and a pH control system with secondary mixed fertilizer was designed. Combining with the characteristics of parameter normal setting, the fuzzy PID does not need accurate mathematical model and the Smith predictor can perform the pure lag compensation, the parameter self-tuning fuzzy PID control was introduced into Smith prediction, which can overcome the influence of lag time on control system and improve the accuracy of the model. In order to verify the algorithm and the effectiveness and superiority of the system, the PID and Fuzzy-Smith control algorithm were simulated respectively, and the performance tests under different irrigation quantities were carried out. The experimental results show that the average maximum overshoot of pH value under Fuzzy-Smith control of different irrigation amounts is 0.83%. The average rise time of nutrient solution from pH value 8.0 to 6.0 is 157 s. It is better than the conventional PID control result which is 2.55% and 189 s. The Fuzzy-Smith control algorithm proposed in this paper has good control stability and dynamic performance, and can meet the requirements of automatic control of pH value of nutrient solution.

Abstract:Accurate prediction of indoor temperature in solar greenhouse is a precondition to accurately control the greenhouse. Because indoor temperature in solar greenhouse is affected by several outdoor environmental factors and the heat conduction mechanism is complex, indoor temperature changes severely in different time. Thus, it is difficult to establish an accurate physical model that describes how outdoor factors affect indoor temperature by mechanism analysis. The accuracy of existing prediction methods based on neural network is low. So，this paper proposed a two-steps method to predict indoor temperature in solar greenhouse based on twice cluster analysis and back propagation (BP) neural network. The first step of the method was two clustering. Similar days were classified to several categories according to clustering of outdoor temperature. Then a whole year training data were split to several continuous time frames.The frames were classified into different categories by clustering of similar days. In the second step, for different categories of time frames, different BP neural networks respectively modeled the relationships between input variables i.e. outdoor temperature, relative humidity, solar radiation, and wind speed and output variable i.e. indoor temperature. The models could be used to predict indoor temperature in solar greenhouse when the outdoor environment was detected. In experiments, two years data was collected from Zhuozhou. For the data, continuous time frames were split to 3 categories.Through the establishment of BP neural network and training respectively, the results show that the prediction error of this method is only 6.23%. Compared with the existing BP neural network prediction algorithm, this method can effectively improve the accuracy, and the average error is reduced by 5.4 percentage points.

Abstract:The daily activity of pigs can be used as an important data base for analyzing and evaluating the health status of pigs. In order to solve the problems of high cost and complexity by using the traditional method of monitoring swinery activity under the farm environment, a monitoring system of swinery activity based on LabVIEW was proposed. By using SRN-2000 passive infrared detector, the information of pig group activities was collected. The circuit, which had high accuracy, multi-channel and real-time operational performance, and took 24 bit ADS1256 chip as its A/D conversion and signal input channel, was designed for a data acquisition system. Based on LabVIEW software platform, real-time acquisition, display and storage of the data were achieved, and a model to monitor daily activity amount was established. Swinery activity amount had been approached by two different activity models: the single sinusoidal model and the double sinusoidal model. A suitable curve was applied and the swinery activity amount was predicted with the improved parameters. The correlation coefficient of single sinusoidal model was lower than the double sinusoidal model so that the later one was adopted. Test results for pigs which just entered the circle showed that the correlation coefficient of the daily activity amount model was 0.83. Through calibrating by using the data of the experimental station and the validation in the real farm, it was proved that the swinery activity amount could be fairly explained by the model.

Abstract:With the rapid development of computer technology, it is possible to process multi-type and mass data in real-time. In order to overcome the problems of distributed streaming computing in processing a large number of pig video streaming data when the delay was high and with poor scalability problems, a node resource scheduler algorithm was proposed, and a pluggable distributed real-time flow computation model was constructed. A system of video monitoring and analysis for pig breeding based on distributed flow calculation was developed. The system implemented the functions of pig video stream data acquisition, task scheduling, real-time calculation, pluggable expansion and result display. The test cluster consisted of a master node and three slave nodes. Under the cluster, the background refreshing method of improved hybrid Gaussian model was adopted to realize the multi-camera and multi-target detection in real-time. The average processing rate was 29.00% higher than the traditional mixed Gaussian model, the average detection rate was 79.00%, and the average false detection rate was 70.96% lower than that of the traditional mixed Gaussian model. The results showed that the pluggable distributed streaming real-time computing model had good scalability and low latency. The improved hybrid Gaussian model algorithm had high detection rate and low false detection rate.

Abstract:In order to clarify the distribution characteristics of dissolved oxygen in water in symbiosis system, the water quality parameters, meteorological data and biomass information of the system were extracted by using the multi-point sampling device and the networking sensor system. The influence of light intensity and feed intake on the spatiotemporal variation of dissolved oxygen in the symbiosis system was studied by correlation analysis method. On the one hand, the results showed that dissolved oxygen levels decreased with the increase of vertical depth in the fish pond, the correlation was between -0.9 and -0.7. The dissolved oxygen was changed at 9:00, and there was obvious correlation between the logarithm and vertical depth of dissolved oxygen with correlation coefficient value r=0.9894 and the value of the variance analysis F=126.94. On the other hand, the dissolved oxygen decreased with the horizontal increase of the water culture tank following the correlation coefficient from -0.9 to -0.8. A significant linear relationship reached 0.89 between the dissolved oxygen content and the horizontal distance. In addition, there was the logarithmic relationship between the dissolved oxygen content of hydroponic tank and light intensity, in which the correlation coefficient was 0.8158. According to the monitoring during different time, the diurnal variation of dissolved oxygen was affected by the spatial variation. The fluctuation values of dissolved oxygen content decreased during the daytime and increased during the nighttime which might be mainly related to the change of light intensity, feeding operation and consumption. As a result, the study provides a theoretical basis for the study of dissolved oxygen in the symbiosis system.

Abstract:The energy survey of two-harvests-a-year protected viticulture has a strong and typical significance, but traditional energy consumption analysis has some drawbacks: complicated artificial calculation, long period and low data quality. Aiming at solving above problems, an energy consumption evaluation system for two-harvests-a-year protected viticulture was developed based on MVC (mode, view, and controller) framework, B/S (browser/server) structure and Matlab Builder JA function. Firstly, a complex analysis of production process of two-harvest-a-year protected viticulture was conducted, an energy input and output system was established, and an energy efficiency evaluation model combined with DEA-BCC model and Malmquist index method was proposed. The model based on the input-oriented DEA-BCC model, through the three parameters: technical efficiency, pure technical efficiency and scale efficiency, to evaluate energy efficiency. At the same time, obtaining technical efficiency change, technical progress and scale efficiency change by Malmquist index decomposition, and conduct a comprehensive assessment of energy efficiency through the above two aspects.Then, the data from Guangxi province was chosen to test the system. Results showed that the system improved the computational efficiency, and can meet the needs of different users, as well as provided reliable data processing support for the study of energy consumption in two-harvests-a-year protected viticulture.

Abstract:The features of block chain technology, including decentralization, distributed storage, open, transparent, consensus mechanism, security, information encryption and anonymity, which provides an important way to improve the value of modern agricultural products supply chain. Through analyzing the concepts and connotations of agricultural products, logistics, agricultural products logistics, supply chain and agricultural product supply chain, this paper further defines the supply chain of agricultural products and its business logic. The problems existing in the circulation of agricultural products were analyzed and pointed out. With the demand of agricultural product supply chain, this paper analyzes the concept, technical characteristics and structure of block chain technology, and put forward the logic structure of agricultural product supply chain based on block chain. Furthermore, the information flow and capital flow in the processed architecture were discussed. This study provides useful inspiration and reference for the study of supply chain of agricultural products based on block chain.

Abstract:The paper designed a wireless real-time monitoring system with 433 transmission frequency for fresh cut branches of Ilex verticillata in cold chain logistics, which can monitor the temperature, relative humidity, CO2 and ethylene gas. The objective is to ensure and improve the quality, traceability and transparency for fresh cut branches of Ilex verticillata in cold chain logistics. The system includes master sensor node which is used to build and maintenance of wireless sensor network and data acquisition, slave sensor nodes which contain temperature, humidity, O2, CO2 and ethylene sensors and remote monitoring system for remote monitoring and controlling. At the same time, the sensor calibration, accuracy and power consumption are tested, and the system was applied to the North American holly cold chain monitoring. The results proved that the wireless real-time monitoring system can effective applied in cold chain logistics for fresh cut branches of Ilex verticillata, which was stable and accurate for monitoring the critical parameters and quality change for fresh cut branches of Ilex verticillata in cold chain logistics.

Abstract:Bitter melons are rich in active substances, such as saponins, polysaccharides, peptides andflavonoids, which have a high edible and medicinal value. Drying the bitter melons can extend the shelf life. There are some commonly drying techniques, such as sun drying, hot air drying, microwave drying and spray drying. But the active substances are easily to be damaged during the drying process due to high temperatures. Vacuum freeze-drying technology completely freezes the material and then heats it in a vacuum to sublimate the water in the material. So it can be as much as possible to retain the material’s color, shape, nutritional composition and the products have a good quality.In order to improve the quality of dried bitter melon, the vacuum freeze-drying technology was used to dry the bitter melon slices to get the best process parameters in this paper. Because the eutectic temperature referred to the temperature which could make the material completely frozen during the freezing process, the eutectic temperature of fresh bitter melon was first measured.The eutectic point temperature of bitter melon was -19℃. It was determined that the pre-freezing temperature of bitter melon was -30℃ and the pre-freezing time was set to 2h. On the basis of single factor test, the quadratic polynomial regression model between inspection indexes (moisture content, rehydration ratio) and drying parameters (slice thickness, temperature, working pressure, drying time) of the bitter melon slices were established by response surface methodology (RSM).Under the condition that the moisture content is less than the safe water content (10%), the optimum process parameters are obtained by optimizing the rehydration ratio using the genetic algorithm in Matlab. The results showed that the regression equation had a good fitting degree (R2=0.9371,R2=0.8548), the model was significant and the obtained process parameters were reasonable.The thickness, pressure and drying time had significant effect on the moisture content during the drying process. They also had significant effect on the rehydration ratio. The interaction between the partition temperature and the pressure had significant effect on the two indexes. The results were optimized by genetic algorithm. After being verified, the optimum technological parameters were slice thickness of 4mm, heating plate temperature of 46℃, absolute pressure of 73Pa and drying time of 8.7h. In this condition the moisture content was 6.23% and the rehydration ratio was 11.75. The study provides reference for vacuum freeze drying bitter melon slices and other materials.

Abstract:In order to establish a rapid and nondestructive method for the detection of protein in walnut kernel, a prediction model for walnut protein content was established based on near infrared spectroscopy. Near infrared band selection method was also studied for optimizing protein content predicting model. Firstly, near infrared spectroscopy information of three different sized walnut samples in the full range of 1040~2560nm was collected, and the pretreatment was completed by multiple scattering correction and standard normal method. Then the spectral characteristic bands were selected by interval partial least square algorithm，and the partial least squares prediction models were established at full bands ranging from 1040nm to 2560nm and selected feature bands, respectively. The spectral analysis results showed that the walnut size has no significant effect on the walnut protein content prediction using near infrared spectrum method. The root mean square error and the correlation coefficients of the optimized model were 0.021 and 0.913 in the whole walnut sample validation set, which indicated that the application of interval partial least algorithm could optimize the model quality and reduce the complexity.

Abstract:Near-infrared spectrum analysis as a rapidly developing technique has been applied in recognition analysis because of their simplicity, promptness and low cost. It was used to build an effective model to qualitatively analyze the corn. To evaluate the analysis results, an innovative grading evaluation index, defined with the relative distance of inter-species, was proposed for optimizing the near-infrared spectrum analysis process. It was applied to analyze the effect on optimizing the performance of the near-infrared spectrum qualitative analysis of corn. Firstly, two group spectral data were measured including the transmittance of 6 corn species sampled in Beijing (group A) and the reflectance of 6 corn species sampled in Hainan province (group B). The sampling data were processed involving original spectral data, the spectral data after pre-processing, and the spectral data after feature extraction from the group A and B experimental data. The relative distances of inter-species were calculated by using correlation, Euclidean distance, and entropy respectively. The result of contrast analysis showed that Euclidean distance was an effective calculation method for varieties recognition with good performance both in group A and B. Secondly, the reflectance of 6 corn species sampled in Henan province (group C) was measured. The Euclidean distance method was used to calculate the inter-specific relative distance between process steps as mentioned above. As a result, after the adjustment of the pretreatment algorithm, the relative distance between species increased from 0.6582 to 1.2972, and the correct recognition rate increased from 40.86% to 70.08%. By optimizing the feature extraction algorithm, the relative distance between species increased from 1.3102 to 2.4910, and the correct recognition rate increased from 68.32% to 93.27%. It was indicated that the correct recognition rate could be improved by the evaluation of the data analysis process.

Abstract:Quantitative detection of five kinds of raw materials (peanut oil, sesame oil, rapeseed oil, soybean oil and corn oil) in blended edible oil was realized by laser Raman spectroscopy combined with Partial Least Squares method. Firstly, the 169.58 ~ 1813.61cm-1 spectral region with abundant fatty acid information was selected, and the spectral denoising and purification of the spectral region were carried out by first derivative + Norris 3 preconditioning. Then, Raman quantitative detection model was established for raw material components of blended edible oil by using Partial Least Squares. The correlation coefficients of calibration set of peanut oil, sesame oil, rapeseed oil, soybean oil and corn oil were 0.9998, 0.9418, 0.9988, 0.9998,0.9961, respectively. And the correlation coefficients of the verification set were 0.9435, 0.8593, 0.9542, 0.9676,0.9429, respectively. The root mean square error (RMSE) were 0.117, 0.218, 0.128, 0.125,0.179, respectively. The results showed that laser Raman spectroscopy combined with chemometric resolution method can be used to determine the contents of raw material components in blended edible oil quickly and accurately. The good prediction ability can provide a theoretical basis for rapid detection of the content of blended edible oil.

Abstract:In the conventional near infrared qualitative identification, the maize seed kernel epidermis was not treated with seed coating agent. However, in the actual agricultural production, in order to resist the invasion of diseases and insect pests, improve the germination rate, and achieve the effect of maintaining and increasing yield, maize seeds often need to be coated with seed coating agents. In reality, on the market, it is usually necessary to model maize seed kernels without seed coating to identify the ones with seed coating, so as to achieve the purpose of cracking down fake and shoddy products. The maize seeds coating usually consist of a mixture of insecticides, fungicides, fertilizer, plant growth regulators and other ingredients. Their types are diverse and the components are different. These components contain hydrogen group organic compounds, which have certain absorption to near infrared spectrum. Therefore, the seed coating agent had an interference effect on near infrared spectroscopy qualitative identification, which reduced the performance of some conventional shallow learning model. According to the effects of seed coating on maize variety authenticity identification accuracy, a method of near infrared spectroscopy qualitative modeling based on stacked autoencoder (SAE) neural networks has been proposed. Firstly, taking maize seed spectrum without seed coating agent as the training set, a qualitative analysis model was constructed through SAE unsupervised learning algorithm and Softmax classifier. Then, based on this model, the authenticity of maize seeds with seed coating agents was identified. The experimental results showed that, by using the method based on SAE, the effect of seed coating on maize varietal authenticity recognition rate was controlled within 3%.