Abstract:With the continuous development of mobile robot technology, odometry technology has become a key technology for mobile robots to realize environmental perception, and its development level is of great significance to improving the autonomy and intelligence of robots. Firstly, the current development status of laser simultaneous localization and mapping (SLAM) and visual SLAM in simultaneous localization and mapping was systematically explained. The classic SLAM framework and its mathematical description were expounded, and the camera models of three common types of cameras and their mathematical description of visual odometry were briefly introduced. Secondly, the research progress of traditional visual odometry and deep learning odometry were systematically elaborated. The advantages and disadvantages of various mileage calculation methods in the past ten years were compared and analyzed. In addition, the performance of seven commonly used data sets was comparatively analyzed. Finally, the problems faced by odometry technology were summarized from the aspects of accuracy, robustness, data sets, and multi-modality, and five development trends of visual odometry were prospected from the aspects of improving the real-time performance and robustness of the algorithm. For the development of more intelligent and miniaturized new sensors, the integration with unsupervised learning, the improvement of semantic expression technology and the development of cluster robot collaboration technology were introduced.

Abstract:The autonomous navigation operation of agricultural machinery has become an inevitable trend in domestic and international development，and path tracking control is the key to improving the control accuracy of autonomous navigation systems. In response to the problem of low tracking accuracy of conventional pure tracking algorithms when turning in complex agricultural environments, an novel path tracking control algorithm for the four-wheel synchronous turning agricultural machinery was proposed based on the improved pure tracking model. A kinematic model and a pure tracking model based on four-wheel synchronous steering agricultural machinery were established. On this basis, an improved pure tracking model was obtained by considering heading error. RTK positioning coordinates were modified, and the optimal forward looking distance was derived from the optimal target points in the forward-looking area according to the evaluation function of the quantitative error. The proposed algorithm can obtain the appropriate forward looking distance in real-time for the improved pure tracking model of four-wheel synchronous steering agricultural machinery, which minimized heading and lateral errors and achieved adaptive optimization of target point. The simulation results showed that the average absolute lateral error of the method during turning was reduced to 0.035m, and the average absolute heading error was reduced to 0.212°. The paddy field experiment showed that when the operating speed of the four-wheel synchronous steering agricultural machine was 3.6km/h, the average absolute lateral error of the four-wheel steering agricultural machine trajectory tracking was reduced to 0.109m, the average absolute heading error was reduced to 2.799°, and the turning tracking accuracy was significantly improved.

Abstract:Hilly mountainous areas of tea plantation soil crust, gravel is more, the use of traditional rotary or mobile furrowing machine will appear to hit the stone jump knife, the knife can not enter the soil, the furrow is not deep, the operation has the problem of high resistance. In view of the above problems, according to the artificial shoveling has the spontaneity to complete the optimal operating path with the lowest power consumption characteristics, the design of crank linkage mechanism device imitating the action of artificial shoveling, and the development of small-scale tea plantation reciprocating trenching and loosening machine were done. Through the analysis of artificial shovel shoveling action, the establishment of the soil into the soil, cut the soil, throw the soil movement model was carried out, based on Matlab software analysis to get the artificial shovel tip trajectory fitting equation, taking this equation as a benchmark, the crank linkage mechanism of the objective function, combined with the constraints of the crank rocker mechanism was established to solve the structural parameters, and at the same time on the trenching shovel for the analysis of the resistance to trenching, to determine the structural parameters of the trenching shovel. Coupled Recurdyn and EDEM furrowing shovel-soil interaction simulation model was established, and a three-factor, three-level orthogonal test was conducted to optimize the operational and structural parameters, and the optimal parameter combinations were obtained as follows: the forward speed v of the implement was 0.06m/s, the rotational speed of the crank was 42r/min, and the inclination angle of the soil entry φ was 80°. The field test showed that the average furrowing depth of reciprocating furrowing and loosening machine operation in tea plantation was 211.5mm, the power consumption of furrowing was 0.119kW, and the coefficient of stability of trench depth was 90.9%, which reduced the power consumption of furrowing by 6.3% and the coefficient of stability of furrow depth was increased by 3.1 percentage points and the quality of the whole machine operation satisfied the agronomic requirements.

Abstract:To meet the no-till sowing pattern of wide and narrow rows in the drip irrigation area of Xinjiang, the sowing of traditional large maize no-till planter has problems of touching stubble, failing to constrain path accurately and having a low qualified rate of maize kernel spacing. Combining the agronomic requirements of no-tillage seeding, the navigation and positioning technology was utilized to obtain information about the deviation between the current position of the seeder and the target path. The deviation automatic adjustment system was designed based on the no-tillage seeding stubble avoidance device. The system included a stubble avoidance device, a hydraulic actuating system, and a hydraulic steering control system. Through establishing the mechanical model of machine, stubble avoidance device and hydraulic actuating system were analyzed in terms of movement and force. The key structural parameters of stubble avoidance and hydraulic steering were determined. Meanwhile, the optimal hook-up length of stubble avoidance device and the maximum driving force of hydraulic actuating system were obtained. Moreover, the hydraulic steering control system was optimized to realize function of automatic adjustment for stubble avoidance devices and acceptance of feedback information. The results showed that the maximum steady state error of desired adjustment angle was 0.932° for neural network PID, the overshooting amount was less than 1%, and the average response steady state error was less than 0.9°, which met expectations. When the tractor operating speed was no more than 1.0m/s and amount of straw covered between rows was no more than 1.0kg/m2, the field test showed that the stubble avoidance rate was no less than 85%, vertical adjustment distance was no more than 8.6m, and coefficient of variation for qualifying grain spacing in maize was no more than 21.63%. Seeder had the best effect of offsetting and stubble avoidance and met the agronomic index requirements of maize no-tillage planter.

Abstract:A air-suction double-row staggered precision seed metering device was designed to address the problems of short filling time and poor airflow stability caused in the high-speed operation of the traditional air-suction seeder's single row seed tray, which made it difficult to achieve high-speed and precise seeding in the soybean corn compound dense planting mode. The structure and working principle of the seeder were explained, theoretical analysis of the seeding process and key components, construction of seed mechanics models for filling and seeding stages, determination of key structural parameters such as the arrangement of inner and outer ring holes in the seeding disc, seeding wheel, and air chamber, and analysis of negative pressure distribution and airflow characteristics inside single and double air passages were done. Based on the DEM-CFD coupling method, the seeding process of the seeder was simulated and analyzed with operating speed, air chamber structure and negative pressure as experimental factors, and filling qualification rate, refilling rate and leakage rate as evaluation indicators, the optimal chamber structure was selected. A comparative test was conducted on the seeding performance of different air-suction seeders through bench tests. The test results showed that the air-suction double-row staggered dense planting precision seeder had a seeding qualification index greater than 88.7% under high-speed dense seeding conditions with operating speeds of 5~10km/h. Moreover, compared with the commonly used single ring air-suction seeder at operating speed of 10km/h, the qualification index was increased by 5.5 percentage points and the missed seeding indexwas decreased by 5.6 percentage points. The field test results showed that under the working condition of 5km/h, the qualified sowing index was 95.7%, the replanting index was 1.6%, and the missed sowing index was 2.8%. The proposed air-suction double-row staggered corn dense planting precision seeder had good seeding performance in high-speed operation and could meet the requirements of high-speed precision seeding of soybeans and maize.

Abstract:In order to solve the problem that peanut seed filling effect is poor under the condition of high speed operation in the process of peanut planting to precision and high speed, a kind of pneumatic assisted seed-filling precision seed metering device was designed, specially, the structures of planter plate and pneumatic assisted seed-filling device were mainly designed. For peanut seeds with large particle size and mass, by analyzing the phenomenon of peanut seed accumulation in the seed metering device and the filling time, it was concluded that the peanut high-speed seed discharging and filling process needed to enhance the filling performance, so as to improve the filling efficiency. Through analyzing the principle of peanut seed filling, the relationship between the movement and force of seeds and seed metering device in the seed filling process of peanut seeds was clarified, and the influencing factors of the seedfilling process were analyzed. The key structural parameters of the seed suction holes and seed guide slot of the seed metering device and the parameters and arrangement of the auxiliary seed blowing holes were analyzed and calculated by designing the seed metering device with seed guide slot and the auxiliary seed filling structure with auxiliary seed guide holes. Taking the qualified rate of seed filling and the leakage rate of seed filling as indexes, a three-factor and three-level combination test was carried out, and the test results were subjected to multivariate regression analysis to optimize with the optimal objective, and the optimal parameter combinations of seed planter plate were determined to be the negative pressure of seed suction in the seed metering device of 5.156kPa, the forward speed of the high-speed planter for peanut was 8.007km/h, and the positive pressure of disturbed blowing of seeds was 1.149kPa. In this case, the qualified rate of peanut seed filling was 95.84%, and the leakage rate of peanut seed filling was 4.06%, which could realize the effective seed filling of peanut seeds.

Abstract:A high-speed no-tillage seeder seeding depth monitoring system based on the improved wild horse optimizer-extended Kalman filter (IWHO-EKF) was proposed. The system addressed the mechanical vibration issues caused by uneven terrain during operation, which led to a decrease in accuracy of the seeding depth monitoring. Additionally, it improved the poor reliability of a single monitoring sensor. Firstly, a mathematical model for monitoring seeding depth was established by using laser, ultrasonic, and angle sensors as the multi-sensor monitoring unit. Secondly, a Kalman filtering algorithm was implemented to filter the measurements from the three individual sensors separately. Lastly, the IWHO proposed the use of the Levy flight and Gaussian mutation algorithms to optimize the key parameters of the EKF for data fusion. Qsigma, Rsigma1, Rsigma2, and Rsigma3 were the parameters that were optimized for the fusion of filtered measurements from the three sensors. Technical term abbreviations such as EKF were explained when first used. The aim was to reduce interference from mechanical vibration, decrease sensor measurement errors and ensure accurate and reliable real-time seeding depth monitoring during high-speed operation of the no-till seeder. To ascertain the effectiveness of the proposed method, simulation experiments and field validation experiments were conducted, comparing the IWHO-EKF with original sensor measurements, filtered seeding depth values and the WHO-EKF. The results from simulation experiments demonstrated that the IWHO-EKF algorithm had a mean absolute error (MAE) and root mean squared error (RMSE) of 0.073cm and 0.090cm, respectively, with a high correlation coefficient (R) of 0.983. This suggested a high level of accuracy and significant improvements in precision compared with measurements from the original sensor and filtered values, as well as the WHO-EKF. Technical term abbreviations were explained when it was firstly used. Field validation tests confirmed that the IWHO-EKF for seeding depth monitoring system in high-speed no-till seeders reduced the average MAE and RMSE by 0.063cm and 0.067cm, respectively, when compared with data from the three sensors. Additionally, the average R was increased by 0.027. This system offerred improved, accurate, and dependable monitoring values for seeding depth. The research result can provide lessons and references for high precision seeding depth monitoring during high-speed seeding.

Abstract:Aiming at the problems of large seed amount, large variation coefficient of sowing amount and uneven sowing, a spoon-tongue precise point planter for flax was designed based on the physical characteristics of flax seeds and the requirements of planting agronomy. By analyzing the working principle of the burrow planter, the composition of the burrow planter, the range of structural parameters and the number of installations were determined. The mechanical analysis of the process of spooning seed and clearing seed was carried out to determine the rotational speed range of the burrow planter. According to the EDEM simulation process, the pattern inner wall chute can not only improve the simulation efficiency, but also increase the seed fluidity, which was convenient for the scoop filling. With the rotation speed of the burrow planter, the radius of the transition angle at the top of the scoop and the height of the isolation plate of the seed chamber as the test factors, and the qualification rate of seed exclusion, the leakage rate and the replay rate of the burrow planter as the test indicators, the quadratic rotation orthogonal combination test was carried out by using EDEM discrete element simulation software. The results showed that the optimal parameter combination was as follows: the rotating speed of the burrow planter was 2.9rad/s, the tip of the scoop was 2.5mm, and the height of the isolation plate was 6.8mm. The seed discharge performance of the spoon was verified by 3D printing. The average pass rate, miss rate and repeat rate of the spoon were 87.00%, 6.33% and 6.67%, respectively. Field experiments showed that the qualified rate of the spooning spoon was 88.33%, the missed rate was 6.67%, and the replay rate was 5.00%. The average planting density of flax was 50 plants/m2, the results of bench test and field test were basically consistent, and the performance met the agronomic requirements of fine sowing flax.

Abstract:In order to lower the damage to the root system and pot soil of the seedling caused by the picking seedling pot mechanism in the tomato pot seedling transplanting process, and avoid the optimization problem of special seedling-picking trajectory and posture design of the mechanical pot seedling transplanting mechanism, the extensible type of picking seedling pot mechanism that can be used in conjunction with a series of transplanting mechanisms was proposed. It was achievable that the seedling needles of the mechanism at key positions of the picking seedling completed the process of protruding into the seedling pot, moving and delivering seedlings, and retrieving and pushing seedlings at a fixed angle. Based on the analysis and design requirements of the process of transplanting and picking seedling pot, the mechanical analysis model of the picking seedling pot mechanism was established, and the factors affecting the minimum driving force on the driving rod when the seedling needles protruding into the seedling pot were obtained. A computer-aided analysis and design software for the picking seedling pot mechanism was developed based on the Matlab App Designer platform, obtaining a set of picking seedling pot mechanism design parameters that met the requirements of tomato pot seedling transplantation. Using the three-factor five-level quadratic regression orthogonal rotation center combination test method, taking the angle of inclined rods, moisture content of seedling pot and depth into the pot as the test factors, and taking the integrity rate of the pot and the success rate of taking seedlings as the evaluation indexes, the prototype was trial manufactured and the bench was built to implement the parameter combination optimization and verification test. The results showed that the extensible type of picking seedling pot mechanism could effectively cooperate with the seedling picking mechanism to complete the performance requirements. The results showed that the operation effect was the best when the combination parameters included angle of the driving inclined rods was 112°, the moisture content of the seedling pot was 57.5%, and the depth into the pot was 28.4mm. The integrity rate of the seedling pot was 96.44%, and the success rate of picking seedlings was 97.06%, which met the performance of pot seedling transplanting.

Abstract:The plant factory rock wool block seedling transplantation equipment has a low penetration rate at present, and most of the operation still relies on manual labor, which is labor-intensive and inefficient. A high-speed transplantation mechanism combining secondary variable spacing for hydroponic leafy vegetable seedlings in plant factories was designed, and it was also equipped with a finger-type planting cup separation mechanism. The stress analysis of rock wool block seedlings in the process of transplanting hand picking seedlings was carried out to provide basis for the design of transplanting hand. The drop cup test of the finger-type planting cup separation mechanism was carried out to lay the foundation for the subsequent planting of seedlings to the planting cup of the cultivation slot. The transplanting mechanism test bench was built, and the transplanting success rate was tested by five-factor three-level orthogonal test with the grasping seedling depth, water content of rock wool block, overall traverse speed, lifting speed, and clamping spacing as the test factors. The influence of each factor on the success rate index was analyzed by variance analysis. The test showed that the seedling depth was 24mm, rock wool block water content was 90%, overall traverse speed was 0.8m/s, lifting speed was 0.24m/s, clamping spacing was 14mm, the transplantation success rate of the mechanism was 97.9%, and the transplant speed reached 3.132 plants/h. which can meet the technical requirements of high-speed, high-efficiency and stable mechanized operation of rock wool block leaf vegetable seedling transplanting in plant factory.

Abstract:Pepper is an important domestic economic crop that has a high economic benefit, and its large-scale mechanized planting can further reduce its production cost. However, as a crop planted in various regions in either field or greenhouse, pepper transplantation with the semi-automatic transplanter remains labour intensive and with low efficiency, and research on the automatic seedling picking and throwing device is in urgency. Hence, a double row stem clipping type automatic picking and throwing device for commercial 2ZBX-2 vegetable semi-automatic transplanter was designed. The cyclic picking and throwing assignments were realized for pepper pot seedlings through the two rows of clipping jaws that equipped on two face to face type mechanical arms, and orderly and smooth operation and the picking and throwing device was guaranteed by the PLC control system. The key components of clipping jaw were the soft materials and the spring steel sheet that fitted tightly, and accurate positioning of clipping jaw was realized with the horizontal and vertical motion mechanisms. Effect of soft material thickness, age of pepper seedling and cylinder pressure of clipping jaw on picking failure rate, delivering failure rate and throwing failure rate was analyzed by single test, and an orthogonal test was conducted, with the picking-throwing success rate as an optimal object, to determine the optimal working parameters. Under the optimal working parameter combination that soft material thickness was 10.0mm, age of pepper seedling was 51d, and cylinder pressure of clipping jaw was 0.40MPa, the test showed that the average real pickingthrowing success rate reached 94.6%, which basically met the technical requirements of vegetable transplanting operation.

Abstract:In order to improve the transfer efficiency of seedling tray in vegetable seedling raising line, and aiming at the problems of high labor intensity and low uniformity that exist in manual tray stacking, a multi-adaptive automatic tray stacking device for vegetable floating seedling was designed. The device was composed of frame, seedling tray conveying mechanism, tray stacking mechanism and the control system. The device took 200Smart PLC as the control core, and used the main and auxiliary conveyor belts to realize the transportation of the seedling trays, and combined the photoelectric sensor to realize the positioning of the seedling trays, which can complete the automatic stacking of foam seedling trays of different sizes, and the designed horizontal adjustment mechanism and vibration reduction mechanism can achieve the horizontal position adjustment of the seedling trays on the stacking mechanism and reduce the vibration impact of the stacking process. The test results showed that under the production rate of 450 plates/h of seedling line, the best effect of automatic stacking device was achieved when the wire diameter of the damping spring was 1.5mm, the speed gap between the main and auxiliary conveyor belts was 0.1m/s and the rising and falling speed of the electric cylinder was 0.13m/s for tray of 200 holes, and the success rate of tray stacking was 100%, and the deviation variance of tray stacking uniformity was 2.32mm2. At the same time, the amplitudes of X axis, Y axis and Z axis of the seedling trays did not exceed 0.8mm in the vibration detection test. And when 135 holes and 160 holes of foam seedling trays were replaced in the test, the success rate of stacking were 100%, and the deviation variance of tray stacking uniformity were 3.94mm2 and 5.98mm2, which showed that the device met the requirements of multi-adaptability. Under the optimal operating parameters, when the productivity of seedling line was increased to 900 plates/h, the maximum amplitude of the seedling trays was 0.709mm, and there was no significant change in the amplitude, which indicated that the device had good operational stability. The research results can provide reference for improving the stacking effect and reducing the vibration of the seedling line.

Abstract:In response to the difficulties in resource utilization of machineharvested residual film mixtures, and the fact that the residual film mixtures processed by the existing shredding and kneading devices do not satisfy the palatability of larvae of white-starred golden tortoise, a residual film mixture crushing and kneading device was designed, processing of residual film mixtures to meet the palatability of white-starred golden tortoise larvae using crushing and kneading techniques. The device was mainly composed of crushing device, conveying device and silk kneading device, etc. The structural and working parameters of each component were determined through kinematic and dynamic analysis of the operation process of the residual film mixture crushing and silk kneading device. In order to verify the operational performance of the residual film mixture crushing and kneading device, a three-factor, three-level quadratic regression response surface experiment was conducted by using the grinding roller speed, kneading roller speed, and kneading roller gap as experimental factors, and the residual film crushing qualification rate, cotton stem crushing length qualification rate, and cotton stem kneading rate as experimental indicators. A regression model was established to analyze the impact of each factor on the operational performance of the residual film mixture crushing and kneading device, and parameter optimization and experimental verification were carried out. The experimental results showed that the main and secondary factors affecting the qualified rate of residual film crushing and the qualified rate of cotton straw crushing length were the speed of the crushing roller, the gap between the kneading rollers, and the speed of the kneading roller. The main and secondary order of factors affecting the cotton straw kneading rate was the gap between kneading rollers, speed of kneading rollers, and speed of crushing rollers. The optimized optimal working parameters were: crushing roller speed of 13.0r/min, kneading roller speed of 60.0r/min, kneading gap of 1.6mm, and the average values of residual film crushing qualification rate, cotton stalk crushing length qualification rate and cotton stalk kneading rate were 90.4%, 92.6% and 92.2%, respectively, which were the same as the theoretical optimization value, with the relative error of no more than 2.0%, the research results can provide reference for the design of residual film mixture crushing and kneading device.

Abstract:A cotton chemical topping system based on top bud intelligent recognition was designed to achieve precise operation, rational and efficient use of cotton chemical topping agents, and reduce environmental pollution caused by excessive use of chemical topping agents. The system mainly consisted of cotton top bud recognition system, control system, and spraying system. A cotton top bud recognition model was constructed by using the YOLO v5s algorithm. The control system adopted STM32F407 microcontroller, which was responsible for receiving signals from the recognition system and controlling various cotton topping agent pipelines. At the same time, the display interface can display real-time parameters such as the driving speed of the equipment, the flow rate of the medicine, and the liquid level of the topping agent. The experimental results showed that in the field all day light experiment, the morning and afternoon time periods had the best recognition performance. At a speed of 0.4m/s, the average recognition rate was about 94%. When the signal transmission interval was 100mm, the success rate of successfully sending signals to the lower computer reached 92%. The field target spraying experiment showed that the effective spraying rate was 94%, which met the spraying requirements.

Abstract:In order to reveal the mechanism and advantages of ultrasonic technology in cutting and harvesting clustered tomatoes, the cutting force and removal characteristics of clustered tomato stem materials in conventional cutting and ultrasonic cutting were compared. Firstly, the parameters of clustered tomato fruit stalks and self-made ultrasonic cutting knife were measured, and then the fruit stalks were fiberized based on Abaqus simulation. The stress and removal mechanism during conventional cutting and ultrasonic cutting were compared in macro and micro simulation. Finally, the self-made test bench was used to measure the cutting force by changing the excitation frequency, input voltage, cutting speed and cutting angle of the ultrasonic cutting knife, and the Box-Behnken of response surface method was used to analyze the four factors and three levels, and then the crosssection morphology of the fruit stalk was observed. The results showed that within the working frequency (35~37kHz) and voltage (340~380V) of the self-made ultrasonic scalpel, the cutting speed and angle had the most significant influence on the cutting effect, and the cutting effect was the best when the excitation frequency and input voltage were near a specific value. Under the conditions of 36kHz, 360V, 0.125cm/s, and 0°, the ultrasonic cutting time in the simulation was about 8s, and the average maximum cutting force was 0.635N, which was 37.7% lower than that of the conventional cutting (1.019N). In the experiment, the ultrasonic cutting took about 5.3s, and the maximum cutting force required was 0.543N, which was 46.6% lower than that of the conventional maximum cutting force (1.017N), and the surface roughness was 20.9% lower. The error between the experimental and finite element simulation results was 8.9%, which was basically consistent. Ultrasonic cutting can reduce the cutting force, shorten the cutting time, improve the section quality, reduce the damage of fruit stem tissue and water loss, which was of certain significance to prolong the fruit preservation time.

Abstract:As a liquid residual pressure energy recovery device, hydraulic turbine is widely used in the field of small hydropower construction and energy recovery, but its internal energy loss characteristics are unclear. The two-stage radial hydraulic turbine was taken as the research object. Based on the entropy production theory, the energy loss in each flow component was quantitatively analyzed, and the energy dissipation mechanism in the turbine was further revealed by combining the Omega vortex identification criterion and flow field distribution. The results showed that velocity pulsation and wall effect were the primary sources of energy dissipation. The total proportion of the two was 98.03% under the design condition. The impeller and the guide vane were the main areas of energy dissipation in the turbine; the impeller loss accounted for a higher percentage in the small flow condition, while the guide vane loss accounted for a higher percentage in the large flow condition. The energy loss in the impeller originated from the unstable flow phenomena such as vortex separation at the leading edge of the blade, return vortex at the suction surface, and vortex at the trailing edge of the blade, and the matching of the relative liquid flow angle and the angle of placement of the inlet of the blade was the fundamental reason for the unstable flow in the impeller; in the guide vane Ⅰ and the guide vane Ⅱ-anti-guide vane, the factors leading to the dissipation of their energy at different flow rates were basically the same, and the poor flow such as the stagnation vortex at the leading edge of the blade and the flow separation. The momentum exchange caused by the blade leading edge stall vortex and flow separation was the main cause of energy loss. Due to the asymmetry of the flow inside the annular suction chamber, the entropy yield distribution in each channel of the guide vane Ⅰ was not uniform, while the guide vane Ⅱ-anti-guide vane reduced the shock effect through the rectification of the positive guide vane, and the entropy yield distribution in each channel was uniform and the high entropy area was small.

Abstract:In order to improve the accuracy of winter wheat yield estimation and the phenomena of underestimation of high yield and overestimation of low yield that exist in yield estimation models, the Guanzhong Plain in Shaanxi Province, China was taken as the study area, and the vegetation temperature condition index (VTCI), leaf area index (LAI) and fraction of photosynthetically active radiation (FPAR) at the ten-day interval were selected as remotely sensed parameters, and a deep learning model was proposed for estimating winter wheat yield by combining the local feature extraction capability of convolutional neural network (CNN) and the global information extraction capability of Transformer network based on the mechanism of self-attention. Compared with the Transformer model (R2 was 0.64, RMSE was 465.40kg/hm2, MAPE was 8.04%), the CNN-Transformer model had higher accuracy in estimating winter wheat yield (R2 was 0.70, RMSE was 420.39kg/hm2, MAPE was 7.65%), which can extract more yield-related information from the multiple remotely sensed parameters, and improved the underestimation of high yield and overestimation of low yield which existed in the Transformer model. The robustness and generalization ability of the CNN-Transformer model were further validated based on the five-fold cross-validation method and the leave-one-out method. In addition, based on the CNN-Transformer model, the cumulative effect of the winter wheat growth process was revealed, the impact of gradually accumulating ten-day scale input information on yield estimation was analyzed, and the ability of the model to characterize the accumulation process of winter wheat at different growth stages was assessed. The results showed that the model can effectively capture the critical period of winter wheat growth, which was from late March to early May.

Abstract:In recent years, although some scholars have achieved satisfactory research results on hyperspectral image (HSI) classification, they often fail to achieve ideal classification results when facing small sample learning. Aiming at this problem, a hyperspectral image classification method was proposed by the organic combination of multi-attention mechanism fusion, compiled graph neural network and convolutional neural network. Firstly, a type of multiple mixed attention convolutional neural network (MCNN) and compiled graph neural network (CGNN) was designed, which can effectively retain the spectral and spatial information of HSI with limited learning samples; secondly, the introduced graph encoder and graph decoder can effectively map irregular HSI feature information; finally, the designed multi-attention mechanism can focus on some important HSI feature categories. In addition, the effect of different training samples on different algorithms for learning example classification was also investigated. Experiments on the public dataset Botswana (BS) showed that the proposed method improved the overall classification accuracy (OA) by 2.72 percentage points and 3.86 percentage points compared with the current state-of-the-art algorithms (CNN-enhanced graph convolutional network, CEGCN; weighted feature fusion of convolutional neural network, WFCG).Similarly, the experimental results on the IndianPines (IP) dataset with only 3% of the training sample data showed that the method also improved the OA of the current state-of-the-art algorithms (CEGCN and WFCG) by 0.44 percentage points and 1.42 percentage points, respectively. This demonstrated that the proposed method not only had good spatial and spectral information perception for HSI, but also showed strong classification accuracy with small learning data.

Abstract:In order to accurately obtain spatial distribution information of citrus orchards and achieve adjustments in citrus cultivation structure, yield estimation, and resource management, focusing on three main citrus-producing regions in southern Jiangxi: Xinfeng County, Anyuan County, and Xunwu County, in addressing the challenge posed by frequent cloud cover and rainfall in the southern region, resulting in a scarcity of traditional optical images, Sentinel series data and the PIE-Engine platform were employed. Spectral features, vegetation water body index features, red edge band features, and texture 〖JP3〗features were constructed and optimized. Furthermore, the backscatter coefficients of time-series Sentinel-1 synthetic aperture radar (SAR) data were incorporated to collectively explore the recognition and extraction effects of different feature combinations on citrus plantations. Based on the random forest algorithm and the fusion of Sentinel-2 and temporal Sentinel-1 SAR feature recognition, the citrus planting area in Gannan was extracted. The results indicated that the average backscatter coefficient separation between citrus plantations and other ground features was most pronounced in May, September, and November, which were the critical periods for citrus identification and extraction. The involvement of index features and texture features in classification proved advantageous for classification effectiveness and enhanced classification accuracy. In comparison with single SAR features, as well as index and texture features, the overall accuracy of the classification results with the inclusion of temporal SAR features was 90.084%, with Kappa coefficient of 0.863. misclassification and leakage errors were relatively small, aligning with the actual distribution of land objects, signifying the availability and practicality of temporal SAR features. The research result can provide reference for the identification and extraction of citrus orchards in the cloudy and rainy southern regions, and it had certain application potential.

Abstract:The intelligence of agricultural monitoring requires real-time, efficient and reliable video data processing and transmission solutions. In order to solve the problems of low rate and poor real-time performance of traditional systems based on CPU and GPU architecture, a high-speed transmission system based on Zynq architecture with PCIe cascaded network interfaces was designed. For the development of PCIe interface, at the hardware level, XDMA IP core parameters were optimized, the interface data transceiver engine was designed, and DDR cache area was optimized based on MIG 7 IP core. At the software level, the PCIe driver was used to schedule VLC software for video data read operations, enabling fast data transmission and smooth playback between the board and the host computer. For the implementation of the network interface protocol stack, the ARM programmable feature was utilized to schedule the lightweight LWIP protocol stack and develop the TCP protocol, achieving fast data transmission for the network interface and avoiding the delay and computational overhead of the host computer CPU directly processing network protocols. Additionally, the AXI protocol was scheduled to establish high-speed connectivity between the PCIe interface and the network interface. Moreover, the rate and reliability of the system were also measured by transmitting video files on Zynq MZ7030FA platform. The results showed that the transmission rate of the network interface was 800Mb/s, that was, gigabit ethernet was basically realized; the maximum transmission rate of PCIe interface was 816MB/s, which was close to the maximum speed of hardware PCIe 2.0 x2, and the whole system achieved reliable transmission at the application layer. The research result can provide an efficient and reliable solution for the application of agricultural monitoring video transmission, and the system had good scalability and generalization.

Abstract:Targeting the continuous ripening process of green mature tomatoes after harvest, timely temperature adjustment plays a pivotal role in meeting the appropriate storage and transportation temperature requirements for tomatoes at different stages of ripeness. Meanwhile, automatic recognition and dynamic prediction of fruit ripeness serve as fundamental prerequisites for achieving temperature control at the right time. A tomato ripeness recognition and temporal dynamic prediction model was proposed based on Swin Transformer and improved GRU. Firstly, by fusing the images of both sides of tomatoes, the overall redness proportion as a visual feature was obtained and a dataset of tomato images at different ripeness stages was constructed. Through transfer learning, the initial weight configuration of the Swin Transformer model was optimized to achieve tomato ripeness classification. Secondly, tomato image data at different storage temperatures (4℃, 9℃ and 14℃) was periodically collected, and the initial color features of tomatoes were combined with storage environment information to build a tomato ripeness temporal prediction model based on Swin Transformer and GRU. Furthermore, a time attention module was incorporated to enhance the prediction accuracy of the model. Lastly, the prediction results of different models were compared and analyzed to validate the accuracy and superiority of the proposed model. The results demonstrated a correct recognition rate of 95.783% for tomato ripeness classification, with respective improvements of 2.83%, 3.35%, and 12.34% compared with that of the VGG16, AlexNet, and ResNet50 models. The mean square error (MSE) for tomato ripeness temporal prediction was 0.225, representing a maximum reduction of 29.46% compared with that of the original GRU, LSTM, and BiGRU models. The research result can provide a key theoretical basis for the flexible and timely regulation of storage temperature considering tomato maturity.

Abstract:Dense mushroom clusters can significantly impact mushroom quality and the success rate of automated harvesting. To address this issue，a spatiotemporal prediction algorithm for mushroom growth status based on historical time series growth images was proposed，which can facilitate early bud thinning to prevent the formation of dense mushroom clusters. The algorithm employed a sequence-to-sequence structure, comprising an encoder and a predictor. In the input, historical image sequences were transformed into 3D tensor sequences and sent to encoder. Within the encoder network, a three-layer long short term memory (LSTM) model was utilized. Here, convolution was fused into LSTM cell to extract spatiotemporal correlation features of mushroom growth. Meanwhile, a diffusion model was introduced into the predictor to address the blurriness issue in predicting images. Furthermore, a mushroom area difference loss function was designed and incorporated into the loss function to further reduce the shape and positional deviations between the predicted and actual mushrooms. The experimental results indicated that the proposed spatiotemporal prediction algorithm for mushroom growth status achieved a peak signal-to-noise ratio of 35.611dB, a multiscale structure similarity of 0.927, and a high mushroom mean intersection over union of 0.93, which represented improvements of 36%, 33% and 24%, respectively, over that of the ConvLSTM（Converlution LSTM）spatiotemporal prediction algorithm. This showed the proposed algorithm can effectively enhance the quality and accuracy of mushroom growth status image prediction, providing a approach for precise forecasting of edible mushroom growth.

Abstract:In the task of apple recognition in natural orchard environments, it is difficult for traditional object detection algorithms to achieve a balance between the accuracy, speed, and lightweight of the detection model. Therefore, a lightweight apple detection model based on improved YOLO v7 model was proposed. Firstly, partial convolution (PConv) was introduced in the multi branch stacking module to replace regular convolution, in order to reduce the parameter and computation of the model. Then a lightweight efficient channel attention (ECA) module was introduced to enhance the feature extraction ability and improve the problem of false and missed detection of occluded targets in complex environments. Finally, a learning rate optimization strategy based on sparrow search algorithm (SSA) was adopted in model training to further increase the detection accuracy of the model. The experimental results showed that compared with the original YOLO v7 model, the precision, recall, and average accuracy of the improved model was raised by 4.15 percentage points, 0.38 percentage points and 1.39 percentage points respectively; the number of parameters and computations were decreased by 2293% and 27.41%, respectively; and the average time to detect each image under GPU and CPU was decreased by 0.003s and 0.014s, respectively. The results indicated that the improved model can quickly and accurately detect apple fruits in natural orchard environments, and the number of parameters and computations were less, which was suitable to be deployed on the embedded devices of apple harvesting robots, and laying the foundation for unmanned intelligent apple picking.

Abstract:Pepper trees yield is a substantial quantity of fruits, characterized by crisscrossed branches and dense foliage, resulting insignificant challenges for automated peppercorn picking. Therefore, a fast identification and localization method of pepper clusters in complex environment based on improved YOLO v5 was proposed. By adding efficient channel attention (ECA) after the CSPLayer of the backbone extraction network CSPDarknet and the upsampling layer of Neck to simplify the computation of the CSPLayer layer and improve the feature extraction capability. In the downsampling layer, coordinate attention (CA) was added to reduce the loss of information in the downsampling process, strengthen the spatial information of features, and cooperate with the heat map (Grad-CAM) and the depth map of the point cloud to complete the spatial localization of pepper clusters. The test results showed that the improved network over the original YOLO v5 reduced the residual computation to 1 time, which ensured the model was lightweight and the efficiency was improved. Under the same frame number interval, the accuracy of the improved network was 96.27%, comparing with three similar feature extraction networks YOLO v5, YOLO v5-tiny, and Faster R-CNN, the precision of the improved network was improved by 5.37 percentage points, 3.35 percentage points, and 15.37 percentage points, respectively, and the ability of separating and recognizing the pepper clusters of the successive plants was greatly improved. The experimental results showed that the average checking accuracy of the system in the natural environment was 81.60%, and the leakage rate was 18.39%, which can satisfy the pepper cluster recognition, and build the foundation for mobile deployment.

Abstract:The classification of traditional tea diseases mainly relies on manual categorization. Such methods are labor-intensive and time-consuming.Furthermore, insufficient availability of tea disease samples hampers the adequate training of existing machine learning models, resulting in decreased accuracy in disease classification. To address this problem, a tea disease classification method was proposed for four types of tea diseases, including tea anthracnose, tea black rot, and others. This method was based on a dual node-dual edge graph neural network.Firstly, RGB tea disease features and grayscale tea disease features were extracted by using two branches of convolutional neural networks, both branches employed ResNet12 as the backbone network, with independent parameters.The two types of features acted as two sub-nodes within the graph neural network, aiming to obtain disease information from different domains. Secondly, two types of edges, including relative metric edges and similarity edges, were created to improve the aggregation capability of disease features from neighboring nodes.Finally, with the dual node and dual edge feature updating modules, a dual-node and dual-edge alternate updating process was achieved. This process aimed to enhance the accuracy of edge features in measuring node distances. This resulted in achieving accurate classification of tea diseases, even when training samples were limited. Comparative experiments were conducted between the proposed methods, which were based on small-sample learning method. The results indicated that the proposed method achieved superior accuracy in tea disease classification. Specifically, on the miniImageNet and PlantVillage datasets, the proposed method achieved the accuracy of 69.30% and 88.42% in the 5way-1shot, respectively. In the 5way-5shot, the accuracy was improved to 82.48% and 93.04% on the miniImageNet and PlantVillage datasets. Furthermore, on the TeaD-5 tea dataset, the accuracy of the proposed method reached 84.74% in the 5way-1shot and 86.34% in the 5way-5shot.

Abstract:Aiming at the adhesion problem in the process of apple orchard pest identification, a pest adhesion image segmentation method was proposed based on shape and color screening. Firstly, the apple orchard pest images were collected, focusing on the feathered pests. Pests have completed most of their growth and development during the feathering process, and their external morphology, color, and texture are more stable and significant. Therefore, based on the analysis of the shape and color feature information of different kinds of pests, the pest HSV segmentation threshold and template outline were obtained. Secondly, the shape factor was used to determine the segmentation of adherent regions, and the segmentation of non-inter-species and inter-species adherent pests was achieved by the color segmentation method and the contour localization segmentation method. Finally, the collected pest images of apple orchard were experimentally analyzed, and the segmentation method based on shape-color screening was used to segment individual pests, and the results showed that the average segmentation rate, average segmentation error rate, and average segmentation efficiency of the proposed method were 101%, 3.14% and 96.86%, respectively, and the segmentation effect was superior to that of traditional image segmentation methods. In addition, with predefined color thresholds, the method achieved accurate classification of cotton bollworm, peach borer and corn borer, with average classification accuracies of 97.77%, 96.75% and 96.83%, respectively. At the same time, the Mask R-CNN model was used as the recognition model, and the average recognition accuracy was used as the evaluation index, and the recognition test was carried out on the pest images that were segmented by the proposed method and those that were not segmented by the proposed method, respectively. The results showed that the average recognition accuracies of cotton bollworm, peach borer and corn borer pest images that were segmented with the proposed method were 96.55%, 94.80% and 95.51%, respectively, and the average recognition accuracies were improved by 16.42, 16.59 and 16.46 percentage points, respectively, which indicated that the proposed method can provide a theoretical and methodological basis for accurate identification of orchard pests.

Abstract:In order to realize accurate classification of tea diseases, a disease image classification method based on SimAM-ConvNeXt-FL model of migration learning was proposed to address the small sample problem and uneven distribution of categories in tea disease image classification. Firstly, an SimAM module was added to the ConvNeXt model to enhance the extraction of complex features. Secondly, to address the problem of uneven sample distribution, the Focal Loss function was used as the loss function in the training process, and the effect of uneven sample distribution was reduced by increasing the weights of a smaller number of samples. Finally, the SimAM-ConvNeXt-FL model was used to train the Plant Village dataset, and the parameters obtained from the training were migrated to the measured tea leaf disease images and fine-tuned to reduce the impact of overfitting, and ablation experiments were set up to prove the validity of the model improvement, and comparison experiments were carried out with the different classification models AlexNet, VGG16, and ResNet34 models comparison experiments were conducted respectively. The experimental results showed that the SimAM-ConvNeXt-FL model had the best recognition effect, with an accuracy of 9648%, and the F1 values of the SimAM-ConvNeXt-FL model compared with the original ConvNeXt model for tea coal disease, tea phoma, tea anthracnose, healthy leaves, and tea white star disease were improved by 4.46 percentage points, 3.76 percentage points, 0.43 percentage points, 0.22 percentage points, and 5.23 percentage points respectively. The results showed that the model proposed had high classification accuracy and strong generalizability, which can promote the development of tea disease classification.

Abstract:The fine segmentation of cow body parts has significant applications in research fields such as cow body condition scoring, posture estimation, behavior recognition, and body measurement. Due to the limited practicality of existing segmentation methods for different cow body parts, an improved YOLO v8n-seg model named YOLO v8n-seg-FCA-BiFPN was proposed for cow body part segmentation tasks. The improved model added FCA channel attention mechanism to the YOLO v8n backbone feature extraction network to better extract the geometric feature information of small targets, and used repeated weighted bidirectional features in the network feature fusion layer. The BiFPN was used to achieve the purpose of increasing the coupling of features at each scale. In order to validate the model performance, side-view images of cows at the channel were collected for network training. To ensure the quality of the dataset, the structural similarity algorithm was used to remove similar redundant images, resulting in a total of 1452 images. LabelMe software was used to label the target cows, which were divided into eight parts, forelimbs, hindlimbs, udders, tails, belly, head, neck, and trunk, and was sent to the training model. The test results showed that the precision was 96.6%, the recall was 94.6% and the mean average precision was 97.1%, the parameters number was 3.3×106, and the detection speed was 6.2f/s. The precision of each part was from 90.3% to 98.2%, and the mean average precision was 96.3%. The YOLO v8n-seg-FCA-BiFPN network could realize accurate segmentation of various parts of dairy cows. Compared with the original YOLO v8n, the precision, recall and mean average precision of YOLO v8n-seg-FCA-BiFPN were 3.2 percentages points, 2.6 percentages points and 3.1 percentages points higher than that of YOLO v8n-seg, respectively. The precision under occlusion was 93.8%, the recall value was 91.67%, and the mean average precision was 93.15%. The volume of the improved model remained unchanged and had strong robustness. Under occlusion, the precision was 93.8%, the recall was 91.67%, and the mean average precision was 93.15%. The overall results showed that the research can provide necessary technical support for precise segmentation of dairy cows' body parts.

Abstract:In aquaculture, there is an inseparable mutual mapping relationship between water quality and fish behaviors. In the past, monitoring was more biased towards one-way mapping, which generally indicated the water quality through fish behaviors. In order to solve the problem of misjudgment and lag only by fish behaviors, a bidirectional mapping model between fish behaviors and water quality was constructed based on random forest. The bidirectional mapping model can not only provide more information to improve the accuracy of prediction, but also improve the reliability of the model through mutual verification. Firstly, YOLO v7 was improved by introducing a deformable convolution module, and the position of fish in the video was detected by using the improved model, and then the swimming parameters of fish were quantified by the coordinates of the front and back frames. Then, the collected fish swimming parameters and the corresponding water quality parameters were taken as inputs, the random forest model was used for classification and regression, and the specific numerical values of fish swimming parameters and water quality parameters and the abnormal level of indicators were predicted respectively, so as to obtain a bidirectional mapping relationship. In order to show the generalization ability of the model, experiments were carried out under two data sets: Li'an Port and Xincun Port Fishing Ground. The experimental results showed that the proposed method can realize the bidirectional mapping between fish behaviors and water quality. Among them, the average accuracy of classification experiment can reach 90.947%, and the average value of regression experiment determination coefficient R2 can reach 0.801.

Abstract:There is a complex relationship of interdependence and mutual constraint in the water-energy-food system, and quantifying the interrelationships of the water-energy-food system is of great significance for the rational allocation and coordinated utilization of water, energy and food resources, as well as for promoting sustainable socio-economic development in the region. The models of energy and food production water footprint, competition index, decoupling theory and Theil index were applied to calculate and analyze the spatiotemporal variation characteristics of the water footprint of energy and food production in Sichuan Province from 2009 to 2019. It also explored the competition relationship between energy and food production for water and investigated the inherent connection between water footprint and gross domestic product (GDP). The research findings were as follows: both the water footprint of food production and the water footprint of energy production showed an increasing trend. The annual average blue water footprint of energy production was 5.45×109m3, accounting for 93.76% of the water footprint of energy production. The annual average blue water footprint of food production was 1.06×1010m3, accounting for 26.02% of the water footprint of food production, and the contribution of the green water footprint exceeded the sum of blue water footprint and grey water footprint, accounting for 53.06%. The competition index of water for energy and food production in Sichuan Province was on the rise, with an increase of 42.69%. Due to the consideration of both food and hydropower production in areas such as the Ecological Demonstration Zone in Northwest Sichuan and the Panxi Economic Zone, their competition index was significantly higher than that in other areas. The decoupling relationship between the water footprint and GDP of Sichuan Province was weakly decoupling and in a state of relatively coordinated development, indicating that economic development had a relatively good state of control over water consumption. The total difference in water footprint intensity showed an expanding trend, indicating that the unevenness of water use efficiency in Sichuan Province was gradually expanding, but it declined in the later period of the study, with a gradual convergence of development. The research result can provide suggestions for water resource allocation, energy development and food production in Sichuan Province, as well as for optimizing the economic development approach and the coordinated development of water use efficiency.

Abstract:Agricultural non-point source pollution is an important problem for water environmental protection in China, and risk assessment is of great significance for the prevention and control of agricultural non-point source pollution. Taking Henan Province as the research area, a multi factor comprehensive evaluation model for nitrogen agricultural non-point source pollution was constructed, the hierarchical assignment method of Analytic Hierarchy Process was used, the weights of each influencing factor were determined based on entropy method and expert scoring, the risk index of nitrogen agricultural non-point source pollution in Henan Province was calculated, and it was verified at the basin scale. The risk level of nitrogen agricultural non-point source pollution was divided and critical control areas were identified. The results showed that planting, breeding and domestic sources contributed 31.52%, 38.47% and 30.01% of the nitrogen loss load, respectively. The nitrogen loss load was low in the west and high in the middle and east of Henan Province. In Henan Province, 39.429km2 area was at moderate risk, accounting for 23.61% of the total area of Henan Province, and 17.318km2 area was at high risk, accounting for 10.37% of the total area of Henan Province. The medium and high-risk zones within 2km from the river were classified as general control areas and key control areas, with the general control area of 10.982km2 and the key control area of 9.285km2. Correlation analysis was carried out between the model and the data of automatic water quality monitoring stations in Henan Province in 2021 at the basin scale, and the fitting degree R2=0.82, indicating that the model had a high accuracy. The results showed that the multi-factor comprehensive evaluation model established was scientific and accurate, which can be used for the identification of nitrogen agricultural non-point source pollution.

Abstract:Farmland drainage is the main contributor to regional non-point source pollution in the rice-wheat rotation area in the lower reaches of the Yangtze River, China, it is of great importance to examine the characteristics of farmland drainage and nitrogen (N) losses for sustainable development of regional agricultural production and ecological environment protection. In the existing studies on drainage and nitrogen losses from rice-wheat rotation fields, nitrogen losses were generally estimated by observing surface runoff and deep percolation with lysimeters or soil column test, which was different from the actual situation that soil moisture and nitrogen entered the drainage system mainly through the lateral flow path. A simulation study on nitrogen losses with field drainage was presented by using the DRAINMOD-NⅡ model; drainage and nitrogen losses from rice-wheat rotation fields under different rainfall conditions were predicted after validating the model with 4-year field monitoring data. The results showed that the predicted average annual nitrogen loss in drainage was 28.4kg/hm2, accounting for 6.0% of the fertilizer application rate; most N losses occurred in the rice growing season with the average value of 25.6kg/hm2. The predicted nitrogen losses in the wheat growing season was only 2.8kg/hm2. When compared with the report values in the existing literature, the total drainage volume was 35.4% higher, and the total nitrogen loss was 44.6% lower. The differences were mainly from wheat growing season, the average reported N losses (31.8kg/hm2) was 11 times of this research. The predicted field drainage and nitrogen losses were significantly correlated to the rainfall pattern (coefficient of determination R2>0.5), the probabilities of the relative increment of the three variables were distributed nearly the same; the predicted nitrogen losses were relatively stable and maintained at 0.8~1.2 times of the average value in normal years with the rainfall return period of less than five years. In the drier or wetter years (i.e., with rainfall return period greater than five years), the predicted nitrogen losses appeared to be more variable. Hence, drainage reduction during the rice growing period through proper control measures was critical to overall reduction of drainage and nitrogen losses from the rice-wheat rotation fields.

Abstract:The feasibility of electrokineticassisted phytoremediation (EKPR) with polarity reversal to remove cadmium (Cd) contaminants from paddy soils under controlled irrigation was investigated. The experiments were carried out in cuboid pots powered by solar energy. The rice crops were planted near the electrodes under controlled irrigation, and the Scirpus tabernaemontani was planted as hyperaccumulator between two rows of rice plants. The results showed that H+ and OH-produced by electrolysis reactions could be directly neutralized by the polarity reversal operation, thus effectively avoiding the polarization of soil pH value. The soil current was varied from 0.08A to 0.36A, indicating that polarity exchange and high soil water content effectively ensured the number and mobility of free-moving ions in soils, which could drive heavy metal migration. Compared with CK treatment, EKPR treatment significantly increased the root dry biomass of Scirpus tabernaemontani by 34.93%. EKPR treatment significantly reduced the root and brown rice dry biomasses by 17.21%~30.16% and 16.18%~22.28% respectively, while significantly increased the leaf and stem dry biomasses by 3.82%~13.17% and 7.59%~30.91%, respectively. EKPR treatment increased the Cd content in root and ground parts by 15.49%~22.45% and 33.30%~35.45%, respectively. The content of Cd in rice root and brown rice was decreased by 14.48%~35.06% and 39.04%~57.43%, respectively. The electrokineticassisted phytoremediation technology could increase the Cd enrichment of Scirpus tabernaemontani and decrease the Cd enrichment of rice crops. Compared with CK treatment, the bioenrichment of Cd in Scirpus tabernaemontani under EKPR treatment was significantly increased by 46.48%, and that in rice crops was significantly decreased by 24.75%. At the end of EKR experiments, the soil Cd content near rice roots was decreased by 16.33%~19.14%.The results showed that under controlled irrigation, the electrokineticassisted phytoremediation with polarity reversal was a feasible method for cadmium contaminants in paddy soils. It had a good application prospect and practical significance to use electrokineticassisted phytoremediation to achieve soil purification in crop production.

Abstract:In agricultural production within arid and semiarid regions, common practices involve rhizobium bacteria inoculation and nitrogen application to promote soybean growth and increase seed yields. However, there has been limited research on the interaction between rhizobium inoculation and nitrogen application and their impact on soybean growth and yield. This twoyear field experiment aimed to address this gap by investigating four nitrogen application levels (N0: 0kg/hm2, N1: 60kg/hm2, N2: 120kg/hm2, N3: 180kg/hm2) and two inoculation levels: rhizobium inoculation (R) and water mixed with no inoculation (unmarked). Various parameters related to soybean growth, including nodule number, nodule dry weight, leaf area index (LAI), biomass accumulation and root characteristics, were measured at different growth stages. Additionally, physiological indicators such as chlorophyll content, photosynthetic parameters, fluorescence parameters and nitrogen uptake, along with nitrogen use efficiency calculations were assessed. The results demonstrated that soybean growth reached its optimum under the RN2 treatment, with maximum nodule number of 241.47, maximum nodule dry weight of 1.30g,maximum root length density of 15.00cm/cm3, maximum LAI of 5.44cm2/cm2, maximum biomass accumulation of 17530.51kg/hm2, maximum chlorophyll content of 53.55, maximum net photosynthesis rate of 32.75μmol/(m2·s), and maximum seed yield of 4659.4kg/hm2. In conclusion, reducing nitrogen fertilizer application (N2) while concurrently inoculating with rhizobium (R) was essential for enhancing the physiological growth of soybeans in the Guanzhong Plain, improving nitrogen use efficiency and increasing soybean yields. The research result can provide both theoretical underpinning and practical experience to elevate soybean production in arid and semi-arid regions.

Abstract:At present, there are some problems such as low energy saving efficiency and large indoor temperature fluctuation in the control methods of mushroom air conditioning in factory production. An energy saving control method based on convolutional neural network (CNN), gated recurrent unit neural network (GRU) and self-attention mechanism was proposed. The CNN-GRU-Attention combined neural network was used as the prediction model, and the prediction error compensation and the dynamic updating mechanism of the prediction model data set were combined to achieve accurate prediction of indoor temperature in mushroom houses. The control quantity of air conditioning was established as the objective function of state quantity, and the weight coefficient of the objective function was defined by entropy weight method and subjective method, respectively. The optimal control sequence of air conditioning in the control time domain was solved by non-dominated sorting genetic algorithm Ⅱ (NSGA-Ⅱ), and the rolling optimization and feedback mechanism were integrated to realize the accurate and energy-saving control of the greenhouse environment. The experimental results showed that the CNN-GRU-Attention indoor temperature prediction model proposed in mushroom house showed that the previous 30min data had the best effect in predicting the indoor temperature in the future 10min. On a typical intra-day the maximum root mean square error of prediction accuracy was 0.122℃, the minimum coefficient of determination was 0.807, and the maximum mean absolute percentage error was 0.611%. The model predictive control method of mushroom air conditioning had a good anti-interference ability in weather fluctuation. Compared with threshold switching method and PID method, the energy consumption of air conditioning can be saved by 21% and 14%, respectively. In terms of temperature control accuracy, the root mean square error was decreased by 72% and 46%, respectively.

Abstract:With the rapid development of blockchain-based agricultural product traceability systems, blockchain query capabilities face great challenges. For supply chain participants, most of the data stored in the blockchain are coded or serialized data, which makes the process of multi-condition query such as audit and supervision of supply chain participants very difficult. In general, native blockchains do not provide a query method to satisfy multi-condition queries. Therefore, in order to realize multi-condition query and improve query efficiency, an optimization method for agricultural product traceability data was proposed. Firstly, the method used an optimized Merkle tree structure (n-Tree) to reconstruct the transaction information, so as to provide more efficient conditional verification ability. Secondly, the adaptive multi-condition block Bloom filter was used to judge the existence of query conditions in the transaction information, and then the blocks were quickly filtered. Finally, an index construction method using time weight and transaction number based heap structure was proposed, and the block number index list related to the main condition was constructed in the order of block weight. The process of querying product data included traversing the block index list, filtering irrelevant blocks, and validating specific query conditions to obtain conditional query results. The experimental results showed that the query method proposed can effectively solve the problem of conditional query in the supply chain of agricultural products. At the same time, the query time consumption was maintained at about 15ms, and the query efficiency was improved by 60.9% compared with Merkle semantic trie method and 87.7% compared with original traverse method.

Abstract:Spontaneous fermentation dry red wines of Cabernet sauvignon collected from five sub-producing areas (Yinchuan, Qingtongxia, Hongsibao, Shizuishan and Nongken) in the eastern foothills of Helan Mountain of Ningxia were selected as the materials. The basic physical and chemical indicators and electrical properties were measured, the differences of electrical properties among different producing regions were analyzed, and the characteristic frequency and effective electrical parameters to identify wines from different producing regions were screened. The aim was to explore the ability to identify wines from different sub-production areas in the eastern foothills of Helan Mountain in Ningxia based on electrical characteristics, in order to provide a method for simple, fast and effective identification of wine production regions. The results showed that there were significant differences in the physiochemical indicators of wines among the five sub-regions of Helan Mountain in Ningxia. Wine from Nongken region had the highest titratable acid content and the lowest reducing sugar content. The wine from Yinchuan had the lowest volatile acid sugar content, while the wine from Shizuishan had the highest alcohol content. Through correlation analysis, analysis of variance and multiple comparisons, the characteristic frequency of 0.1kHz and the effective electrical parameters Z, Lp, X, Cp and Q were selected to distinguish wines from different regions at 2V. The results of principal component analysis and discriminant analysis showed that the five sub-production regions of the eastern foothills of Helan Mountains could be clearly distinguished by the electrical parameters of wine. The prediction model established by Fisher-LDS had the correct rates of return test and cross validation of 100%. Therefore, it was feasible to identify wine producing region by wine electrical characteristics.

Abstract:During the process from harvest to sales, apples are susceptible to mechanical damage, which can have detrimental effects on their quality and lead to rotting. Detecting and removing this damage in a timely manner is crucial to prevent further deterioration. However, early-stage mechanical damage in apples often manifests as subtle color changes, making it challenging to detect. To address this issue, an apple implicit damage detection method was presented based on structured-illumination reflectance imaging (SIRI) and convolutional neural network (CNN). By building an SIRI system to acquire modulated structured light images of the measured apples, and utilizing three-phase demodulation method to extract the alternating current component, the image contrast of the apple implicit damage can be enhanced. The dataset of apple implicit damages was produced by using the images of alternating current components. Several CNNbased semantic segmentation networks, including FCN, UNet, HRNet, PSPNet, DeepLabv3+, LRASPP, and SegNet were employed to train the damage detection models, respectively. Several groups of experimental results demonstrated that these models can effectively detect the apple implicit damages in different situations. In contrast, the precision (P), recall (R), F1 score, and mean intersection over union (MIoU) of the HRNet model were respectively 97.96%, 97.52%, 97.74% and 97.58%. However, its detection speed was only 60 frames per second. The PSPNet model had a faster detection speed, reaching up to 217 frames per second. However, it had slightly lower detection accuracy, with precision (P), recall (R), F1 score, and mean intersection over union (MIoU) of 97.10%, 94.57%, 95.82%, and 95.90%, respectively.

Abstract:In order to improve the atomization characteristics of quick-frozen blueberries, a fan-shaped nozzle with multi-outlet was designed and optimized. It can spray a fan-shaped dry ice particle jet, which can ultimately freeze blueberries more evenly and quickly. A physical model of a multi-outlet fan-shaped nozzle and a computational model of the dry ice jet flow field for quick-frozen blueberries were established. Using Fluent software and the gas-solid two-phase dynamics model DPS, Realizable,k-ε turbulence model, a numerical simulation study was conducted on the process of quick-freezing blueberries by using a multi-outlet fan-shaped nozzle for dry ice particle spraying. The different V-shaped groove angles (60°, 70°, 80° and 90°) at the outlet of the fan-shaped nozzle were investigated, the effects of different angles on the flow field distribution of dry ice particles in quick-freezing chamber, as well as the freezing rate of blueberry and the freezing uniformity were studied under the same inlet flow rate and outlet aperture. The results showed that as the angle of the V-shaped groove was increased, the width of the fan-shaped impinging jet was decreased, and the flow velocity in the core region of the impinging jet was increased. When the V-shaped groove angle of the multi-outlet fan-shaped nozzle outlet was 70°, compared with 60°, 80°, and 90°, the freezing completion time distribution of the whole plate of blueberries in the quick-freezing chamber was the most concentrated, the overall freezing speed was fast, and the flow field was the most uniform. Therefore, it was the optimal outlet parameter for this nozzle model (inlet diameter of 30mm, inlet velocity of 0.25m/s, outlet was a combination of circularly arranged diameter of 5.2mm×6（there were six outlets with a hole diameter of 5.2mm）and centrally arranged diameter of 2mm×4). The optimal result of the simulation was then tested experimentally. The results showed that the whole plate of blueberries completed quick-freezing in 119s, with a freezing rate of 0.50cm/min. The error between the experimental and simulated cooling curves was 4.3%. Tests were conducted on anthocyanins content, soluble solids, mass fraction, and water loss rate of frozen blueberries after quick freezing. The results showed that the sensory quality of dry ice quick-frozen blueberries during storage was better than that of national standard quick-frozen blueberries.

Abstract:The distributed drive system allows for independent control of each wheel, providing greater maneuverability and adaptability to various terrains and working conditions. Additionally, when combined with electric technology, the distributed drive system can reduce emissions, decrease reliance on fossil fuels, and improve sustainability. These advantages position distributed drive electric tractor (DDET) as having broad potential for applications in agriculture and industry. Aiming at the low traction efficiency and high energy consumption of the DDET, a distributed drive system parameter optimization design and verification method based on the multi-island genetic algorithm (MIGA) was proposed. According to the working conditions of plowing operations, a 7-DOF coupled dynamics model of the tractor distributed drive system and a tire-soil interaction model were established. The parameter design and matching selection of key components in the drive system were completed. An MIGA-based optimization strategy for the front and rear wheel side transmission ratios (WTR) was proposed, taking WTR as the decision variable, minimizing energy losses in the drive system as the optimization objective, and with constraints on the power and speed of the drive motor. This effectively prevented the algorithm from prematurely falling into local optima during the optimization process, improving the efficiency and reliability in obtaining the globally optimal. A Matlab/Simulink-NI PXI joint simulation platform was built to verify the correctness and real-time executability of the parameter optimization strategy. The joint simulation results showed that the distributed drive system optimized based on MIGA achieved effective performance improvements. Under cyclic plowing conditions, the average traction of the tractor was 10.610N with maximum traction power of 31.25kW. The average efficiency was increased by 0.38% and energy consumption of the drive motor was decreased by 7.53%. The research result can provide theoretical foundations and verification methodologies for the optimal design and system control of distributed drive electric tractors.

Abstract:In response to the sensitivity of greenhouse small agricultural machinery to ground flatness, small ground undulations can cause the equipment to pitch. Based on the greenhouse electric tractor developed by the research group, the method based on time series analysis was introduced into angle prediction and feedforward PID control (APF-PID) to solve the problems of poor responsiveness, unstable tillage depth, and sudden power changes caused by equipment pitch in greenhouse rotary tillage operations. Firstly, a power model for greenhouse electric tractor rotary tillage operation was established, and a conversion matrix between pitch angle and tillage depth was established to obtain the conversion value of the actual tillage depth of the rotary tillage system. Secondly, time series analysis was used to predict the pitch angle of the aircraft body and serve as disturbance input for the rotary tillage system. Then, combining the conversion value of tillage depth and the predicted disturbance, the APF-PID controller was used to adjust the lifting mechanism of the rotary tillage system, and the rotary tiller was maintained at the target tillage depth. Finally, actual vehicle experiments were conducted on two types of plots in a greenhouse, one without rotary tillage and the other with rotary tillage. The results showed that the correlation coefficient of the pitch angle time series prediction model can reach 0.9832, the control performance of APF-PID control was superior to that of PID control. In the test road surface with a target tillage depth of 6cm, the average tillage depth of APF-PID on two test plots was 6.47cm and 6.44cm, with root mean square errors of 0.80cm and 0.72cm, absolute average errors of 0.67cm and 0.58cm, and tillage depth stability coefficients of 89.95% and 91.30%, respectively. The total energy consumption was reduced by 4.18% and 19.13% compared with that of PID control, which effectively achieved the stability control of greenhouse electric tractor rotary tillage and met the requirements of greenhouse operations.

Abstract:Kinematics analysis is the basis of kinematics performance evaluation and structure size optimization of parallel mechanism. The existing kinematics analysis methods of parallel robots have the problem of separating geometric modeling from geometric calculation. A method for inverse kinematics analysis of parallel robots was proposed by taking advantage of conformal geometric algebra (CGA), which integrated geometric representation and calculation. Firstly, the rigid body transformation of any point on the moving platform was realized by the geometric product under the framework of conformal geometry algebra, and then the conformal geometric expression of any point in the process of motion was obtained. Then the kinematic equation of the mechanism was established by using the inner product operation in combination with the dimension and geometry constraints of the mechanism. Finally, according to the kinematics equation, the inverse kinematics was calculated and the velocity was analyzed. A 3-RPS parallel robot of three degrees of freedom and a 6-UPS parallel robot of six degrees of freedom were taken as examples to verify the correctness of the proposed method, and the results of inverse kinematics were compared with those obtained by simulation software. This method combined the geometric objects such as space vector and rotation representation with the calculation methods such as matrix multiplication and vector outer product, so that the spatial geometry problems of parallel mechanism were handled in a unified algebraic system, so the analysis process was geometrically intuitive, and the analysis and calculation process of inverse kinematics were simplified.

Abstract:In order to solve the problem of stable control in the gait transition process of quadruped robots by using model-based control methods, a quadruped robot prototype platform was designed based on bionics and mechanics, and the robot's single leg kinematic model was derived. The robot's leg height and step length were planned within the reachable workspace at the foot end. By using an ideal compound cycloid trajectory and controlling the gait period reasonably, a transition period variable control method was proposed, which achieved fixed speed control and variable step length control before and after gait transition, ensuring that the speed remained unchanged or variable before and after gait transition. In order to verify the correctness and stability of the proposed algorithm, single leg foot trajectory experiments and whole machine gait transformation experiments were conducted, respectively. On the basis of completing the overall motion control, the application of model-based control and central pattern generator based control in the gait transition process of quadruped robots was compared. The simulation and experimental results showed that under the model-based control algorithm, the quadruped robot can achieve smooth gait transformation, and the speed can be adjusted with changes in step size and period, meeting the walking requirements at different speeds, providing a reference for the motion control of quadruped robots.

Abstract:In order to improve the gradual deterioration of the dynamic performance of the spatial parallel mechanism caused by the lubrication clearance effect, a dynamic optimization method of the spatial parallel mechanism considering the lubrication clearance effect was proposed by taking the 3-R[TXX-]RPaR redundant parallel mechanism as the research object. Firstly, the kinematic model of the lubrication clearance of the revolute pair and the oil film bearing capacity model were established. The transition force model of the contact state was derived, and the dynamic model of the 3-R[TXX-]RPaR redundant parallel mechanism considering the lubrication clearance of the revolute pair was established. Then, the objective function was set up to optimize the dynamic response error of the end-effector and the constraint reaction force at the clearance joint. By optimizing the quality of the end-effector and the moment of inertia, the degradation effect caused by the clearance of the kinematic pair was alleviated, and the dynamic optimization model of the 3-R[TXX-]RPaR redundant parallel mechanism considering the lubrication clearance effect was established. Finally, the effectiveness of the established dynamic model was experimentally verified, the influence of the two objective functions on the optimization effect was compared and analyzed to select the best optimization method, and the dynamic characteristics of the spatial parallel mechanism considering the lubrication clearance effect before and after optimization were analyzed. The results showed that the optimization reduced the peak value of the constraint reaction force at the lubrication clearance revolute joint by 16.16%. The research result can provide theoretical support for improving the dynamic performance of spatial parallel mechanism by improving the clearance effect.

Abstract:Due to the use of offline parameter identification methods, the existing hysteresis models are difficult to characterize the time-varying and load dependent properties from the hysteresis of pneumatic muscle (PM), which was easy to generate significant modeling errors. In order to accurately characterize the hysteresis characteristics of PM, the Prandtl-Ishlinskii (PI) model was used to describe the lengt h-pressure hysteresis characteristics of PM, and the forgetting factor recursive least squares (FFRLS) was used to identify parameters of the PI model online. Compared with offline identification, the online identification method can effectively improve the modeling accuracy of PI models. Then the feedforward online compensation controller was designed based on the PI inverse model, and a composite controller was established by combining with feedback control. This composite control approach was used to realize the motion control of PM. At the same time, corresponding experimental equipment was built and hysteresis modeling and motion control experiments of PM were conducted to compare and analyze the trajectory tracking effects of offline identification and online identification under different loads. The experimental results showed that the PI model using online parameter identification method can effectively describe the load-dependence of hysteresis and greatly reduce control errors caused by load variation.