Abstract:The precision of curve path tracking affects the efficiency of unmanned rice transplanter by affecting the efficiency of line changing. In order to improve the curve path tracking accuracy of the unmanned rice transplanter when turning from the headland, in view of the poor adaptability of the traditional linear quadratic regulator (LQR) path tracking controller with fixed error weight matrix to the changes of the longitudinal speed, lateral deviation and heading angle deviation of the rice transplanter, an optimization method of path tracking controller by adjusting the error weight matrix of linear quadratic regulator in real time through fuzzy control was presented. The method took the longitudinal speed, lateral deviation and heading angle deviation as the input, and the error weight corresponding to the lateral deviation and heading angle deviation as the output, and a fuzzy control model was established to adjust the error weight matrix of the linear quadratic regulator in real time. In order to verify the accuracy and feasibility of the curve path tracking control of the proposed algorithm, the refitted Yangma VP6E unmanned rice transplanter was taken as the object, and Carsim and Simulink joint simulation tests and real vehicle tests were carried out. The simulation test results showed that when the rice transplanter was controlled to track a quarter arc path with radius of 2m, the average value of absolute value of lateral deviation under the control of the proposed algorithm was 0.014m, the maximum value was 0.032m, and 100% of those were less than 0.04m, the average value of the absolute value of heading angle deviation was 1.67°, and the maximum value was 4.94°. Compared with the traditional linear quadratic regulator with feedforward control, the average value of absolute value of lateral deviation was reduced by 50%, the average absolute value of heading angle deviation was decreased by 23%. The real vehicle test results showed that when the rice transplanter tracked a quarter arc path with radius of 2m, the average value of the absolute value of lateral deviation under control of proposed algorithm was 0.027m, the maximum value was 0.048m, and 62% of those were less than 0.04m, the average value of absolute value of heading angle deviation was 1.86°, and the maximum value was 4.94°. Compared with the traditional linear quadratic regulator with feedforward control, the average value of absolute value of lateral deviation was reduced by 40%. The average absolute value of heading angle deviation was decreased by 4.1%. The method improved the curve path tracking control accuracy of the unmanned rice transplanter, and provided a reference for the curve path tracking control of the unmanned rice transplanter.

Abstract:In order to improve the precision of unmanned navigation path of combine harvester, a real-time extraction method of crop harvesting navigation line was proposed based on LiDAR. A point cloud data acquisition system was built, and the installation height and angle of LiDAR were obtained by plane fitting method. The transformation of point cloud data from LiDAR coordinate system to vehicle coordinate system was realized by using three-dimensional LiDAR to scan the crop point cloud data in front of the harvester and combining the attitude information fed back by IMU inertial sensor. The coordinates of the ROI were obtained based on the scanning field angle, installation height and installation angle of LiDAR, and the ROI was filtered through and statistically to remove impurities such as dust and straw powder, so as to eliminate invalid points and outliers in point cloud data. A fast recognition algorithm of navigation line based on the elevation difference of eight neighboring grids was proposed. The coordinate value in the Z-axis direction after the point cloud rasterized was taken as the detection basis, and the difference between grid and its eight neighboring grids in the Z-axis coordinate was defined as the elevation difference. The grid was traversed and compared according to the set threshold, so as to effectively extract the harvesting boundary points. The least square algorithm was used to fit the harvesting boundary points, and the harvesting navigation line was obtained. Experiments showed that the algorithm had good robustness, and can maintain high accuracy when crops were scarce and there were many weeds, among which the average error angle of forward direction was 0.872°, the lateral deviation of header was 0.104m, and the correct rate of harvesting navigation line was 93.5%, which can provide auxiliary navigation for combine harvester and improve the accuracy of unmanned driving.

Abstract:Aiming at the complexity and diversity of the characteristics of field crop rows, the lack of robustness of the traditional crop row detection method, and the difficulty of parameter adjustment, a field crop row detection method based on feature engineering was proposed. Taking the seedling cotton crop row canopy as the recognition object, the crop row canopy characteristics were analyzed, and the feature expression model of the canopy of cotton crop was established with RGB image and depth image as the data source. The key feature parameters of crop row canopy were extracted by using feature dimensionality reduction method to reduce the amount of computation. A crop canopy feature segmentation model was established based on support vector machine technology to extract crop feature points. The method of crop row centerline detection was established by combining random sample consensus algorithm and principal component analysis. Using cotton crop row images with different illumination, weed and camera positions as test data, SVM classifiers with linear, RBF, and polynomial kernels were employed to conduct crop row canopy segmentation experiments. The performance of typical Hough transform, linear square method and the established crop row centerline detection method was compared and analyzed. The results showed that the RBF classifier had the best segmentation accuracy and robustness. The accuracy and speed of the established crop row centerline detection method were the best. The mean value of heading angle deviation was 0.80° and the standard deviation was 0.73°;the mean value of lateral position deviation was 0.90 pixels and the standard deviation was 0.76 pixels;the mean value of centerline fitting time was 55.74ms/f and the standard deviation was 4.31ms/f. The research results can improve the adaptability of crop row detection model, reduce the workload of parameter adjustment, and provide accurate navigation parameters for navigation system.

Abstract:In order to solve the problems of large quantity of corn stover and insufficient depth of returning machine and tools used in production in China, which affected subsequent sowing, emergence quality, high incidence of diseases and insect pests, and “carbon hunger” in deep soil, a kind of straw crushing centralized and full deep returning machine was designed. The machine can complete the operations of picking up, crushing, conveying, deep loosening of trenches, centralized escape and injection burying and crushing of crushed soil at one time. Under the effective reduction of soil disturbance, multiple rows of standing or crushed straw can be buried in a trench with depth of 380~400mm below the surface. The overall structure and working principle of the straw crushing and centralized full quantity deep return machine were described, and key components such as the straw screw conveyor, the straw discharge fan, the deep loosening and trenching device, and the transmission device were designed and calculated. The rotational speed of the screw conveyor and the straw discharge fan, the size of the discharge port of the straw discharge fan, and the cross-sectional shape parameters of the injection device were preliminarily determined. The flow field inside the straw exhaust fan and the injection chamber was simulated, analyzed and verified. According to the design requirements, the straw chopping centralized full depth return machine was trial-manufactured, and the machine performance and field positioning test were carried out. The results showed that at operation speed of 3km/h, the straw stalk collecting rate was 90.8%, trenching depth and average straw buried depth was 394mm and 200mm,respectively, trenching depth and straw burying depth stability coefficient were 97.4% and 92.5%, respectively, the indicators were in conformity with design value, and satisfied the agricultural requirements. At the same time, the results of field positioning test showed that compared with traditional field, the water storage of 0~200cm soil was increased by 29mm, and the contents of organic carbon, total nitrogen and available phosphorus in 20~40cm soil layer was increased by 7.14g/kg, 0.59g/kg and 1.53mg/kg, respectively. The soil nutrient content and soil fertility and water storage capacity were increased effectively. The method results can provide a scheme and technical reference for the promotion of the technical model of centralized deep returning of straw to the field.

Abstract:To reduce the influence of vibration on the performance of the finger clip seed catcher, a corn finger clamp seed feeder with auxiliary clamping structure was designed, and the working principle was described. Besides, the key components of the auxiliary clamp were optimized, and the dynamic model of finger clip seed catcher was established. RecruDyn software was utilized to establish the virtual prototype model of seed discharge device, and corn grain particles were constructed in EDEM software based on the characteristics of corn grains. The process of seed filling, seed carrying and seed discharge in the seed discharge device under vibration conditions was simulated by EDEM-RecurDyn coupling simulation, and the main influencing factors on the performance of finger clip seed dispenser were analyzed. Finally, combined with bench tests, the operation speed, vibration amplitude and vibration frequency of the seed feeder were selected as the test factors, and the qualified index and leakage index served as the test indexes to verify the seed performance. Results showed that the simulation result showed that when the finger clip was opened at 28.65°, the corn grains were gripped, which was consistent with the motion process of the finger clip seed separator under vibration. The results of coupling simulation were basically consistent with the actual motion process of the finger clamp seed feeder. The influence of vibration on the holding performance of the finger clamp can be effectively reduced under the cooperation of the finger clamp and the seed tray, and the design requirements can be met. When the operating speed was 3.8km/h, the vibration amplitude was 5mm, and the vibration frequency was 32.52Hz, the qualified index and missing sowing index of the seed feeder were 91.0% and 6.68%, respectively. It was combined with further tests that the parameter combination metering qualified index was 90.0% and the missing sowing index was 7.1%, which proved the theory of optimal value was close to the actual tests. It was illustrated that the auxiliary clamping structure can maintain good stability in the planting operation, which made the device have a good seeding effect, and met the requirements of precision seeding operation. It provided a reference for the improved design of new corn precision seeding.

Abstract:Compared with traditional mechanical seed metering, pneumatic seed collectors are beneficial for the configuration and efficiency improvement of seeders, and have become the main trend in the development of seed collectors both domestically and internationally. The pressurization pipe of the pneumatic collector of the seeder is an important component of the seed particle conveying process, and its structure directly affects the operational performance of the seeder. A combination of theoretical analysis and EDEM Fluent coupling was used to study the transport characteristics of seed flow in a pressurized pipe. After theoretical analysis, it was found that the main factors affecting seed particles included the structure of booster tube, the conveying airflow, and the material of device. Due to the external factors of device materials and conveying airflow, rapeseed seeds were mainly taken as an example to study the influence of pressure pipe structure on the motion state of seed particles. Coupling analysis showed that the commonly used V-shaped corrugated booster pipe had a good disturbance and dispersion effect on the seed flow. When the airflow speed was 16m/s, the mass of the outlet seed flow was relatively uniform when the diameter of the booster pipe was 30mm, the length was 100mm, the width was 10mm, and the depth was 2mm, with a coefficient of variation of 17.32%. The correctness of coupling parameters and model selection was verified through bench tests.

Abstract:In order to solve the problem in vegetables seed-metering device that the seeding quality decreased due to multiple grains in one hole during seed suction, a four-stage sawtooth for seed cleaning was proposed and a four-stage seed cleaning mechanism was designed. The reasons for the phenomenon of resuction were expounded, and a mathematical model was established to determine the key parameters such as the installation position and tooth profile angle of the seed clearing mechanism. The distance between the top of the tooth and the mold hole, the working speed of the seed dispenser and the negative pressure of the chamber were selected as the main factors to carry out the quadratic regression orthogonal rotation combination test. The significance and regression analysis of the test results were carried out. The optimal parameter combination was obtained as the distance between the top of the tooth and the mold hole of the 4th seed dispenser was 0.41mm, the negative pressure was 1.22kPa and the working speed of the seed dispenser was 5.85r/min. The verification test and comparison test of the optimal working parameters were carried out. The verification test showed that when the working speed was 4.5~9.5r/min, the leakage rate was not more than 1.50%, and the overclearing rate was not more than 1.88%, which was basically consistent with the calculated value of the regression equation. The comparative test showed that when the rotation speed was 4.5r/min, 7.0r/min and 9.5r/min, the leakage rate of the four-stage seed clearing mechanism was reduced by 3.2 percentage points, 4.0 percentage points and 5.0 percentage points, the overclearing rate was reduced by 0.19 percentage points, 0.83 percentage points and 1.45 percentage points, and the qualified rate of seed discharge was increased by 3.42 percentage points, 4.83 percentage points and 6.45 percentage points, respectively. The four-stage seed cleaning mechanism can effectively reduce the leakage rate of seed cleaning link and improve the working quality of seed discharging device.

Abstract:Aiming at the problem that the seed quantity of the high speed precision seed discharging device can not maintain the best working state due to the fluctuation of the seed quantity in the device, a technical idea was put forward to improve the structure of seed feeding device and optimize seed feeding control model. An intelligent seed supply system was designed, which adopted a seed supply wheel structure with replaceable wheels and a wave variable supply mode based on sine function. The fluctuation amplitude of seed supply speed was adjusted in real time according to the seed supply effect, so as to keep the seed quantity in the seed discharge device in a good range and ensure the continuous work of the seed discharge device under high operating performance. The structural parameters of the key components were designed, and the advantageous operating intervals of thetype seed feeding wheel with different structures were obtained and verified by discrete element simulation and bench test. The experimental results showed that the type of seed feeding wheel could maintain the variation coefficient of seed feeding speed below 20% in the range of 250~1500g/min by changing its own structure. After optimization, the type of fluctuation variable seed supply control model would make the seed supply rate change smoothly in the way of fluctuation variable seed supply, and carry out the fluctuation seed supply in the range of 500g/min to 1500g/min around the supply demand. The research result can meet the technical requirements of corn precision sowing.

Abstract:Ridging is beneficial for increasing soil permeability and water retention capacity, and is an important part of potato cultivation. A two-stage combined ridging device suitable for Shandong region was proposed to address the problems of single function, low efficiency, and poor operation effect of the current potato seeder ridging device. By combining the soil covering disc and wing shovel, the device can simultaneously complete the soil covering, ridging, and shaping operations, which was conducive to stable rooting and developed growth of potatoes, and improved the efficiency and quality of potato seeder operations. By analyzing the operation process of the soil covering disc and wing shovel, as well as the movement process of the soil, the overall size of the shovel and the main components such as the shovel tip, shovel surface, and wing plate were designed with parameters. Using EDEM simulation software to establish a joint simulation model of soil combined two-stage combined ridging device, and conducting experiments, the simulation curve of ridging amount of the device was analyzed. It was found that compared with relying solely on the covered disc for ridging operation, the operation quality of the two-stage combined ridging device was significantly improved. Field experiments showed that when the potato seeder advanced at a speed of 0.25m/s, the average height of the soil ridge raised by the two-stage combined ridging device was 251mm, the average width of the ridge bottom was 698mm, the average ridge distance was 902mm, and the average soil compaction of the ridge can reach 390.33kPa. The operation effect was good, meeting the agricultural requirements of the potato ridge raising in Shandong.

Abstract:Aiming at the phenomena of imperfect filling of substrate leading to shallow and exposed root system during potting and transplanting of non-woven containers, and slow transplanting efficiency due to low automation degree of transplanting machine, a fully automated secondary transplanting device based on non-woven containers was designed in combination with the growth characteristics of young forest seedlings. The mechanical claw installed on the transplanting arm realized XYZ three-dimensional space transplanting operation without dead angle, and the transplanting efficiency was improved by alternately picking up seedlings through the two-way assembly line platform. Through the design and force analysis of the key components of flexible gripper, the influence of deformation coefficients of seedling substrate under different gripper pin spacings was explored, and the motion state of transplanting arm was optimized through kinematic analysis by combining the RecurDyn rigid-flexible coupling simulation. Considering the key factors affecting the success rate of transplanting, the secondary filling force rate, moisture content of seedling substrate and clamping pin spacing were determined as the test factors. Taking 60pin capillary seedlings as the test object, the relevant parameters were derived from a one-factor test, and a multifactor orthogonal test was conducted to determine the optimal operating parameters of the automatic secondary transplanting device for non-woven container seedlings. The test results showed that when the force rate of the secondary filling mechanism was 83.51%, the needle spacing was 19.66mm, and the water content of the seedling substrate was 33.45%, the transplanting success rate was 95.29%, which was basically the same as the results of validation test (95.33%), and the requirements of high-speed and high-efficiency transplanting were met, which provided a reference for the field of secondary transplanting of non-woven containerized seedling in forest trees.

Abstract:In view of the present wheat seeder in complex field exists in the process of deep consistency and stability is difficult to control, through adjust the volume of overburden soil from aspects determine wheat sowing depth control perspective, a fuzzy PID control method was proposed based on sowing depth feedback, the designed wheat seeder sowing depth control system had high precision, which realized the automatic control of sowing depth, and the uniformity of sowing depth of wheat was ensured. The system was mainly composed of four parts: vehicle terminal, sowing depth detection module, pre sowing suppression roller detection module and pre sowing suppression roller adjustment mechanism, which can realize real-time sowing depth detection and adjustment of wheat sowing machine. The real-time sowing depth was obtained through the sowing depth detection module and used as feedback input. Combined with the preset sowing depth value, PID parameters were adjusted online according to the expert fuzzy rule and Mamdani reasoning method to obtain the control output. The driver was controlled to adjust the position of the suppression roll before sowing and the amount of soil cover during the operation constantly, so as to achieve real-time and accurate control of sowing depth and ensure consistency of sowing depth. The results of field experiment showed that the sowing depth fluctuated in a small range during the sowing operation. When the sowing depth was set to be 30mm, the average sowing depth was 30.13mm, planter speed was 3~5km/h, the standard deviation of sowing depth was 0.18mm, the average passing rate of sowing depth was 93%, and the mean coefficient of variation of sowing depth was 2.93%. The system realized real-time adaptive control of uniform sowing depth of wheat planter.

Abstract:At present, the hole formed by the pit-type transplanter has poor stability, which is easy to collapse, and the soil backflow is serious. Aiming at the transplanting characteristics of the water flushing well cellar type multifunctional transplanting machine and the agronomic requirements of well cellar transplanting technology, in order to improve the quality of the hole and reduce the rate of seedling leakage, a hole-forming drill was designed, and its main structure and work were expounded. Mechanism and perform performance tests were carried out. The burrowing process was divided into three stages. Through the research methods of dynamics and kinematics, the force of soil particles in each stage was analyzed, the key structural parameters and operation parameters were determined, and the relationship between the key operation parameters and the burrowing quality was clearly analyzed, and the burrowing was improved. Taking the number of soil particles in the grid bin group as the response index, a discrete element simulation experiment was carried out, and the value range of each factor was determined by statistical analysis of the least significant difference;a quadratic orthogonal rotation combination experiment was arranged, and the Design-Expert software was used to analyze the test data, and the order of the influence of each factor on the index value was obtained. The results showed that the drill bit diameter was 100mm, the pitch was 75mm, and the rotational speed was 350r/min. The device had the best performance. The depth of the hole and the diameter of the hole were 182mm and 80.7mm, respectively. The optimization factors were tested and verified, and the test and optimization results were basically consistent, meeting the agronomic requirements of well cellar transplanting.

Abstract:In order to solve the problems of jacking seedling failure, compression deformation and damage when the lump of mud is clamped, taking jacking and clamping components as the research object. Mechanical analysis models of jacking and clamping the lump of mud were established. Needle diameter, needle length, velocity, and compression distance were the main factors affecting the effect of seedling taking extraction by mechanical analysis. The related physical properties of the lump of mud were measured by experiments. The EEPA contact model used in the EDEM software was used as the contact model between particles, between particles and tray, and between particles and seedling taking parts. And the particle model was established to simulate the real lump of mud. EDEM software was used to simulate the process of jacking and clamping seedlings. Box-Behnken design method and univariate control method were used to design the simulation experiment of jacking and clamping seedlings. The optimized parameter combination that was obtained by using Design-Expert software were needle diameter of 1.9mm, needle length of 18mm, jacking velocity of 0.3m/s and compression distance of 4mm. The seedling taking performance verification tests was carried out, and the seedling taking frequency was 100 plants/min. The seedling taking success rate of the optimized end-effector was 93.25%, and the lump of mud of the bowl was good. The results of this experiment met the requirement of automatic dry land transplanting of leaf vegetable pothole seedlings.

Abstract:Aiming at the operation of driven stubble breaking and blocking no-tillage planter in Southwest China, due to poor flatness caused by the large slope, no-tillage surface straw and stubble, stubble breaking and blocking operation of the machine produced vibration as a whole, which resulted in poor performance of the fertilizer discharger and fertilizer guide tube, based on the principle of spiral conveying, a shaftless spiral fertilizer discharging and fertilizer conveying device was designed. Through the calculation of shaftless spiral conveying volume per unit time, the factors affecting the device discharge volume were analyzed. According to agronomic requirements, the amount of fertilizer required per unit time was calculated. And through the critical conveying speed and flow characteristics of material, the working parameters of shaftless spiral fertilizer conveying and fertilizer transfer device range were determined. Then the optimal size of the spiral blade fertilizer filling as well as the range of rotational speeds was obtained. The quadratic regression orthogonal rotation test was carried out by using EDEM simulation, and the response surface method was used to analyze the working parameters of the shaftless spiral fertilizer discharge and transfer device. Then the optimal parameters were obtained as follows: the shaftless radius was 3mm, the radius of the spiral outer axis was 12.8mm, the rotational speed of the blade was 319r/min, and the rotational pitch of the spiral was 24.5mm. Based on data above, the device was manufactured and its field performance was tested. The results showed that when the average value of the surface flatness was 8.9cm and the slope of the surface was 16.1°, the fertilizer discharge performance of the shaftless spiral fertilizer transfer device was better than that of the external grooved wheel fertilizer discharger, and the error of fertilizer application accuracy and the coefficient of variation of uniformity were 1.87% and 2.52%, respectively, under the operating speed of 1.5m/s. At 0.5m/s and 1m/s operation speed, the fertilizer application precision error and uniformity coefficient of variation all met the national fertilizer application standard. The designed shaftless spiral fertilizer discharging and conveying device met the requirements of no-tillage sowing and fertilizer application, and it can provide reference for the design and improvement of fertilizer discharging device and fertilizer conduit under the conditions of poor surface flatness and long-distance fertilizer conveying.

Abstract:The application of water-fertilizer integration technology in cotton, wheat, tomato and other field crops planting scenarios is gradually increasing. However, the current research on control algorithms that can quickly and effectively adjust the fertilizer flow in the water-fertilizer integration system for field crops is relatively limited. The water-fertilizer integration system has the characteristics of time-varying, hysteresis and nonlinearity, and the common PID and BP-PID control algorithms cannot obtain the expected control effect. To solve these problems, a BP neural network PID controller based on bat algorithm (BA) optimization was designed. By using BA to optimize the initial weights of the BP neural network, the self-learning speed of the BP neural network was accelerated to achieve fast and accurate control of the fertilizer flow rate in the water-fertilizer integration system, which reduced the amount of overshooting and improved the response speed. At the same time, a water-fertilizer integration flow regulation test platform was built based on STM32 microcontroller, and the performance of the controller was experimentally verified. The results showed that compared with the conventional PID controller and the BP neural network-based PID controller, the designed controller had higher control accuracy and robustness, and reduced the effects caused by time lag, nonlinearity and other factors. The average maximum overshoot was 4.78% and the average regulation time was 41.24s. Especially when the fertilizer application flow rate was 0.6m3/h, the controller showed the best comprehensive control performance and achieved the effect of precise fertilizer application.

Abstract:China is one of the origins of forest fruits, and picking is an important part of forest fruit industry production. Mechanized picking is one of the important means to achieve rapid harvesting of forest fruits instead of manual work, which is an inevitable direction driven by the development of modern agriculture. Mechanical vibration picking is currently the most widely used technology, and the working performance of forest fruit picking machinery is mainly determined by the vibration effect. Therefore, through the research and analysis of various types of forest fruit mechanical vibratory picking equipment at home and abroad, four key mechanical vibratory picking theories were summarized, such as factors affecting vibratory picking, three-dimensional reconstruction and equivalence model of fruit trees, fruit vibratory shedding characteristics as well as vibratory energy transfer and dissipation, and the current deficiencies of each theory were elucidated. On this basis, the research status and progress of mechanized picking equipment for forest fruits at home and abroad were described, the scope of application, working principle and classification of mechanical pickers for forest fruits were analyzed, the models and technical parameters of typical vibratory and contact pickers were summarized, and the practical problems arising from the application operation of each model were pointed out. Combined with the application scenarios and development requirements of the forest fruit industry, the main problems faced by China’s forest fruit mechanized picking were analyzed, and it was believed that the lack of systematic theoretical research and the slow progress of key technology research were the key constraints to the development of forest fruit mechanization, and it was put forward that the technical focus of the future forest fruit mechanized picking would be on high-efficiency, precise, and low-loss picking, and that ultimately, it would be necessary to achieve the automation of forest fruit and intelligent picking proposals.

Abstract:In view of the problems of small shape difference, large degree of mixture, and difficulty in sorting the flax exudate after segmented harvesting, a flax threshing sorting device was designed. In order to improve the operation efficiency of the flax threshing sorting device, the mechanism of airflow cleaning of flax threshing materials was explored, and the CFD model of the cleaning system and the DEM model of flax detachment were established with the airflow sorting system of the device as the research object. Using CFD-DEM coupling simulation technology, the separation law of flax exudate in the sorting system was obtained by studying the motion trajectory and spatial position distribution of the detached materials of each component, and verification tests were carried out to verify the reliability of the simulation model. The simulation test showed that the flax threshing particles showed a good separation and sorting effect under the action of the air flow field in the cleaning system, and at the same time, the variation curve of the number and average velocity of flax threshing particles obtained by the simulation test was analyzed, and the change law of the average speed and quantity of flax threshing materials in the process of separation and sorting was detected. The verification test showed that the removal loss rate of flax grain after operation under the best working condition of the device was 2.78%, and the impurity rate was 2.23%, which was only 0.73 and 0.67 percentage points different from the simulated flax grain loss rate (2.05%) and impurity content rate (1.56%), respectively, and the actual test results were in high agreement with the simulation results, which verified the reliability of the model.

Abstract:Intelligent control technology of operating speed based on feeding rate is an important means to optimize the efficiency and quality of combine harvester operations. Aiming at the obvious time delay of the traditional feeding rate automatic control technology and the inability to adapt to the actual situation in time when the feeding rate is adjusted. By analyzing the influencing factors of the feeding rate, an imagebased deep learning method was used to carry out a research on the classification and recognition method of wheat plant density in the mature stage. By sensing crop density in advance, the operating parameters of the combine harvester can be automatically adjusted. Firstly, a multi-variety and multi-region mature stage wheat plant image dataset was constructed based on vehicle-mounted cameras and UAV images, and classified it into four categories: low density, medium density, high density, and very high density. Next, a density classification recognition model was built based on the lightweight MobileViT-XS network, and trained and tested the model by using the established dataset. Finally, it was compared with VGG16, GoogLeNet, and ResNet. The results showed that the overall recognition accuracy of the MobileViT-XS model reached 91.03%, and the inference time for a single image was 29.5ms. Compared with VGG16 and ResNet networks, the overall recognition accuracy was 3.51 percentage points and 2.34 percentage points higher respectively. The MobileViT-XS model can effectively accomplish the classification recognition of wheat at different density levels, providing technical support for real-time prediction of wheat feeding rate.

Abstract:China is the original country of Chinese milk vetch (Astragalus sinicus L.), and also the country with the earliest utilization and cultivation of Chinese milk vetch and the largest planting area in the world. Chinese milk vetch, also known as grass seed, Ziyunying and so on, is one of the main winter green manure crops in paddy fields of central and southern China. To enhance the machanization level of separating and cleaning of Chinese milk vetch combine harvest material, according to the complex component compositions and small difference confounding characteristics of Chinese milk vetch combine harvest materials, the operation mode of “screening first, and then air flow cleaning” was put forward, and a separating and cleaning machine for Chinese milk vetch combined harvest material was designed. Through design selection and parameter calculation of key components of the prototype, the important working parameters of the feeding hopper outlet material layer thickness adjustment mechanism, the screening device and impurity-collection and dust-removal device were determined. Based on the DEM-CFD coupled numerical simulation method, the reasonable value ranges of the main influential factors, such as the material layer adjustment thickness, the vibration amplitude of the screening device, and the air volume adjustment handle gear of the impurity-absorbing duct were confirmed, an orthogonal experimental design was conducted by using Minitab, with the grain cleaning rate and the total entrained loss rate as the response values, the optimal parameter combination that affected the operation quality of separating and cleaning machine for Chinese milk vetch combined harvest material was obtained. The prototype field validation test showed that, when the material layer adjustment thickness was 16.8mm, the screening device vibration amplitude was 35mm, and the air volume adjustment handle gear of the impurity-absorbing duct was 5, after the operation of separating and cleaning machine for Chinese milk vetch combine harvest material，the average grain cleaning rate was 98.07%, and the total entrained loss rate was 2.96%, the experimental results met the design requirements of operation machine.

Abstract:In order to solve the problems of high impurity rate and high damage rate of sugar beet root in the complicated separation and conveying process of sugar beet combined harvester, a six-row sugar beet combined harvester with three-stage separation and conveying device was designed.The main structure and working principle of the device were described, and the key parameters were determined. Through the kinematic analysis of soil and sugar beet in the separation and transportation process and energy analysis of sugar beet in the collision process, the main factors affecting the impurity rate and damage rate of sugar beet were determined. The rotation speed of the dial plate, the line speed of the rod-type chain screen and the inclination angle of the rubber tail screen were taken as the test factors, and the hybrid rate and the damage rate were taken as the test indexes. The quadratic regression orthogonal rotation combination test was carried out, and the results were analyzed by using Design-Expert 8.0.6 software, and the regression equation of the test factors and each index was obtained. The influence of each factor on the evaluation index was analyzed intuitively by response surface. The optimized parameters were obtained as follows: the rotation speed of the dial plate was 100.0r/min, the linear speed of the rod-type chain screen was 1.4m/s and the inclination angle of the rubber tail screen was 39.0°. The verification test was carried out. The results showed that the rate of breakage was 3.4% and the damage rate was 2.6%.

Abstract:Aiming at the problems of difficult root-soil separate and low transportation efficiency in mechanized harvesting of Panax notoginseng in hilly and mountainous areas, experimental research on the operation mechanism and parameter optimization of the conveying and separating device of the Panax notoginseng harvester was conducted. Firstly, the dynamics model of Panax notoginseng root-soil composite during the process of transportation was established through theoretical analysis;secondly,high-speed photography was used to obtain the movement trajectory of the Panax notoginseng root-soil composite, and determine the operation mechanism of Panax notoginseng transportation, root-soil separation, and fibrous root fracture, etc. Meanwhile, based on EDEM-RecurDyn coupling, the joint simulation of the conveying and separation of Panax notoginseng root-soil composite was carried out, which verified the reliability of the conveying separation operation joint simulation model. The conveying and separation laws of Panax notoginseng root-soil composite was clarified, and the main operating parameters that affected the separation of Panax notoginseng root-soil were determined to be lifting speed, lifting inclination angle, vibration amplitude and vibration frequency. Finally, bench tests were conducted and Design-Expert software was used to analyze and find the optimal operating parameters. The results showed that when the optimal operation parameters were as follows: the lifting inclination angle was 21°, the vibration amplitude was 44mm, the lifting speed was 0.9m/s, and the vibration frequency was 1.6Hz, the conveying rate and the screening rate of Panax notoginseng were 93.60% and 92.64%, respectively, which met the requirements of the conveying and separation operation of the Panax notoginseng harvester.The research result can provide a theoretical basis and reference for the harvesting of rhizome crops in hilly and mountainous areas.

Abstract:Accurate identification and classification of pests under intelligent insect monitoring and reporting lights are the prerequisite for realizing early warning of rice insect situation. In order to solve the problems in image recognition of rice pests, such as dense distribution, small body size and susceptibility to background interference, the recognition accuracy is not high.A lightweight MS-YOLO v7 (Multi-Scale-YOLO v7) based classification method for rice fly identification was proposed.Firstly, a rice planthopper pest collection platform was built with a migratory pest trapping device, and the images of rice planthopper were obtained to form the ImageNet dataset. Then the MS-YOLO v7 object detection algorithm used GhostConv lightweight convolution as the backbone network to reduce the number of parameters for model operation. CBAM attention mechanism module was added to Neck to effectively emphasize the highly differentiated feature channels of rice planthopper, suppress redundant and useless features, accurately extract key features of rice planthopper images, and dynamically adjust the weights of different channels in the feature map. SPPCSPS spatial pyramid module was replaced by SPPFS pyramid module to improve the feature extraction ability of the network model. At the same time, SiLU activation function was replaced by Mish activation function in YOLO v7 model to enhance the nonlinear ability of the network. The test results showed that the mean average precision (mAP), precision (96.4%) and recall (94.2%) of the improved MS-YOLO v7 on the test set were 95.7%, 96.4% and 94.2%, respectively.Compared with that of Faster R-CNN, SSD, YOLO v5 and YOLO v7 network models, mAP was improved by 2.1 percentage points, 3.4 percentage points, 2.3 percentage points and 1.6 percentage points, respectively, and the balance score F1 was improved by 2.7 percentage points, 4.1 percentage points, 2.5 percentage points and 1.4 percentage points, respectively.The memory occupation, number of parameters, and number of floating-point operations of the improved model were 63.7 MB, 2.85×107, and 7.84×1010, respectively, which were scaled down by 12.5%, 21.7%, and 25.4% compared with that of the YOLO v7 model. The MS-YOLO v7 network model can realize high-precision identification and classification of interspecific pests of rice fly, with good robustness, and it can be used to realize the technical support for early warning of rice fly pest in paddy fields.

Abstract:In order to effectively lightweight the leaf disease detection model under the premise of ensuring the recognition performance, a model lightweight method was constructed based on trunk replacement, model pruning and knowledge distillation technology, and a lightweight test was carried out on the leaf yellow leaf curl disease detection model based on YOLO v5s. Firstly, the main body of the model was reduced by replacing the YOLO v5s trunk with the common lightweight convolutional neural networks (LCNN) with excellent performance. Then, the unimportant channels were screened and deleted by using the sparse training of the model and the distribution of the scaling factors in the batch normalization layer. Finally, by fine-tuning retraining and knowledge distillation, the model accuracy was adjusted to a level close to that before pruning. The experimental results showed that the accuracy, recall and mean average accuracy of the lightweight model were 91.3%, 87.4% and 92.7%, respectively. The memory consumption of the model was 1.4MB, and the detection frame rate of the desktop was 81.0f/s. The detection frame rate of the mobile terminal was 1.2f/s. Compared with the original YOLO v5s leaf disease detection model, the accuracy, recall and average accuracy were reduced by 3.7 percentage points, 4.6 percentage points and 2.7 percentage points, and the memory consumption was only 10% of that before processing. The frame rate of the desktop and mobile terminal detection was increased by nearly 27% and 33%, respectively. The proposed method can effectively reduce the weight of the model under the premise of keeping the performance, which provided a theoretical basis for the deployment of mobile leaf disease detection.

Abstract:Rapid and accurate identification of crop diseases in the early stage is an important guarantee to reduce crop economic losses. In view of the actual production environment, crop yellow leaf curl virus (YLCV) cannot be accurately and quickly identified by color or texture features by traditional image processing algorithms in the early stage of disease, and the YOLO v5s general model has poor recognition effect and low efficiency in complex environments. The dataset was made from two sources: the images of single diseased leaves in the public dataset of Plant Village and the canopy images of diseased crop taken by mobile phones in the actual production, and manually labeled the diseased leaves in the images to achieve the correct identification of targets in complex terrain background and leaf occlusion, that was, to accurately identify all diseased leaves in healthy leaves, diseased leaves, withered leaves, weeds and soil. In addition, a smartphone was used to shoot images at the production site, there would be a variety of factors such as mobile phone resolution, light, shooting angle, etc., which would lead to problems such as reduced recognition accuracy, and it was necessary to preprocess data and enhance the collected images to improve the model recognition rate, and enhance the extraction ability of YOLO algorithm to key information by repeatedly increasing the CA attention mechanism module (coordinate attention) for many times on the YOLO v5s original model backbone network. The weighted bidirectional feature pyramid network (BiFPN) was used to enhance the fusion ability of different feature layers of the model, thereby improving the generalization ability of the model, replacing the loss function EIoU (Efficient IoU loss), further optimizing the algorithm model, and realizing the comprehensive improvement of the target recognition performance of the system after multi-method superposition optimization. Under the same experimental conditions, compared with the original YOLO v5, YOLO v8, Faster R-CNN, SSD and other models, the precision rate P, recall rate R, average recognition accuracy mAP0.5, mAP0.5:0.95 reached 97.40%, 94.20%, 97.20% and 79.10%, respectively, and the proposed algorithm maintained a high operation speed while improving the accuracy and average accuracy. It met the requirements of accuracy and timeliness of the detection of crop yellowing leaf curvature virus disease, and provided a theoretical basis for the intelligent identification of crop leaf diseases on mobile terminals.

Abstract:To address the challenges of detecting cotton leaf diseases in natural environments and the difficulty of manually designing feature extractors that capture similar feature expressions as those of cotton leaf diseases, an improved attention mechanism YOLO v7 algorithm (CBAM-YOLO v7) was proposed. Building upon the YOLO v7 model, the approach integrated the convolutional block attention module (CBAM) into the backbone and head of the model and incorporated a four times downsampling step within the head. The CBAM-YOLO v7 model was employed for the identification of cotton leaf diseases in Southern Xinjiang, and comparative experiments were conducted against YOLO v5 and YOLO v7. Experimental results revealed that in terms of aphid and normal leaf detection, YOLO v7 achieved favorable detection outcomes. Notably, CBAM-YOLO v7 demonstrated higher accuracy in detecting diseases like Fusarium wilt, cotton mirid bugs, and red spider mites when compared with other models. CBAM-YOLO v7 achieved a mean average precision (mAP) of 85.5%, representing a 21 percentage points increase over YOLO v5 and a 4.9 percentage points increase over YOLO v7. Moreover, the detection time for a single image was 29.26ms, offering a theoretical foundation for online monitoring of cotton leaf diseases.

Abstract:Accurate and rapid identification of cow manure morphology is of great significance for monitoring and precise management of cow gastrointestinal health. In response to the problems of strong artificial dependence and difficulty in identification in current cow manure recognition methods, a method for identifying cow thin, loose, hard, and normal manure was proposed based on the VGG-ST (VGG-Swin Transformer) model. Firstly, a total of 879 images of the four different forms of manures was collected from lactating Holstein cows and augmented to 5580 images using operations such as flipping and rotation as the dataset. Then, five typical deep learning image classification models, namely Swin Transformer, AlexNet, ResNet-34, ShuffleNet and MobileNet, were selected for cow manure classification research. Through comparative analysis, Swin Transformer was determined to be the optimal base classification model. Finally, the VGG-ST model combined the VGG model with the Swin Transformer model. The VGG model was utilized to capture local features of cow manure, while the Swin Transformer model extracted global self-attention features. After feature concatenation, the cow manure images were classified. The experimental results showed that the Swin Transformer model achieved a classification accuracy of 85.9% on the testing set, which was 1.8 percentage points, 4.0 percentage points, 12.8 percentage points, and 23.4 percentage points higher than that of ShuffleNet, ResNet-34, MobileNet, and AlexNet, respectively. The classification accuracy of the VGG-ST model was 89.5%, which was 3.6 percentage points higher than that of the original Swin Transformer model. The research result provided a method reference for the development of automatic inspection robots for cow manure morphology.

Abstract:The Holstein cow individual recognition network has the problems of high parameter adjustment cost, poor generalization and low efficiency, and it is difficult to achieve accurate recognition under partial occlusion conditions.An adaptive network parameter optimization identification algorithm (NAS-Res) was proposed based on ResNet framework and neural network architecture search (NAS). Firstly, a hyperparameter network was constructed by designing an operation set, including CBR_K1, CBR_K3, CBR_K5, and SkipConnect, together with dense connection paths. Then the search strategy based on gradient descent strengthened the design of a low-cost model under the constraint of multi-objective optimization composite loss function. The results showed that NAS-Res only took 6.18 GPU hours to obtain the best architecture.On the PO-Cows dataset, which contained side images of 168 cows, NAS-Res achieved 90.18% Top-1 Acc. Compared with ResNet-18, ResNet-34, and ResNet-50, the accuracy was improved by 5.04 percentage points, 3.02 percentage points, and 14.92 percentage points, respectively, while the parameters were reduced by 5.9×105, 1.069×107, and 1.317×107, respectively.It achieved 99.25% accuracy on the Cows2021 dataset, which contained 174 back images of cows. In addition, NAS-Res can ignore the influence of the scale change of the PO-Cows dataset, and when the number of cattle was changed between 50 and 168, the change range of Top-1 Acc and Top-5 Acc was only 1.51 percentage points and 1.01 percentage points, which showed strong applicability. In general, the NAS-Res algorithm achieved accurate individual identification of partially occluded cows, and the research result can provide technical reference for individual identification of livestock and poultry under complex background.

Abstract:Individual identification is the foundation for achieving digital management of cattle. In order to achieve non-contact and high-precision individual identification, a dairy cow face recognition method based on RGB-D information fusion was proposed. Totally 108 Holstein cows aged 28 months to 30 months were selected as the research subjects, and 2334 color/depth images of cattle faces were collected by using the Intel RealSense D455 depth camera as the original dataset. Firstly, image preprocessing was carried out by using redundant image elimination and adaptive threshold background separation algorithms. After enhancement, a total of 8344 cattle face images was obtained as the dataset. Then, three feature extraction networks, including Inception ResNet v1, Inception ResNet v2, and SqueezeNet, were selected to extract the facial features of the cattle face. The optimal backbone feature extraction network of the FaceNet model was determined through comparative analysis. Finally, the extracted dairy cow face image features were L2 regularization and mapped to the same feature space. A classifier was trained to achieve individual classification of dairy cows. The test results showed that using Inception ResNet v2 as the backbone feature extraction network of the FaceNet model had the best performance. After testing the cow face recognition accuracy on the preprocessed dataset with background separation, the accuracy reached 98.6%, the verification rate was 81.9%, and the misidentification rate was 0.10%. Compared with that of Inception ResNet v1 and SqueezeNet networks, the accuracy was improved by 1 percentage points and 2.9 percentage points, respectively. Compared with that of the dataset without background separation, the accuracy was improved by 2.3 percentage points. The research result can provide a method for dairy cow face recognition.

Abstract:The efficiency of pig body temperature measurement based on thermal infrared technology is low in the process of large-scale pig breeding. Temperature detection method in herd of pigs based on improved YOLO v7 was proposed, and an automatic pig head detection model was constructed. The VoV-GSCSP structure was introduced at the Head layer to reduce the complexity of the network structure. The content-aware reassembly of features (CARAFE) was used to replace the original up-sampling operator of the model to improve the quality of the feature map after zooming in, and strengthen the effective features in the head region of the pig;the receptive field enhancement module (RFE) was introduced to enhance the extraction capability of the feature pyramid on the head region of the pig. RFE was applied to enhance the extraction capability of the feature pyramid for the head region of pigs. The improved YOLO v7 algorithm had a detection accuracy of 87.9%, recall rate of 92.5%, and mean average precision (mAP) of 94.7% for the pig head. Compared with the original YOLO v7, the accuracy was increased by 3.6 percentage points, the recall was increased by 7.0 percentage points, and the mAP was increased by 3.6 percentage points. The average absolute error of temperature extraction of this method was only 0.16℃, and the detection speed was 222 frames/s, which realized the real-time accurate detection of body temperature of group pigs. Comprehensive results of the above experiments showed that the method can automatically localize the head region of pigs, meet the requirements of high efficiency and high precision for the determination of body temperature of pigs, and provide effective technical support for the automatic detection of body temperature in herd of pigs.

Abstract:In recent years, with the increasing scale of pig farming in the world, farms are in urgent need of automated livestock information management systems to ensure animal welfare. As one of the significant growing information of pigs, the weight of pigs can help farmers to grasp the healthy status of pigs. The traditional methods manually measure pig weight, which are time-consuming and laborious. With the development of image processing technology, the estimation of pig weight by analyzing images has opened up a way for intelligent determination of pig weight. However, many recent studies usually considered only one image modality, either RGB or depth, which ignored the complementary information between the two modalities. To address the above issues, a cross-modality feature fusion model CFF-ResNet was proposed, which made full use of the complementary between texture contour information of RGB images and spatial structure information of depth images, for realizing the intelligent estimation of pig weight without human contact in a group farming environment. Firstly, RGB and depth images of the piggery in top view were acquired, and the correspondence between the pixel coordinates of the two different modalities were used to achieve alignment. Then the EdgeFlow algorithm was used to segment each target individual pig in the coarse-to-fine pixel level, while filtering out irrelevant background information. A two-stream architecture model was constructed based on the ResNet50 network, and a bidirectional connection was formed by inserting internal gates to effectively combine the features of RGB and depth streams for cross-modal feature fusion. Finally, the two streams were regressed separately to produce pig weight predictions, and the final weight estimation values were obtained by averaging. In the experiment, the data was collected from a commercial pig farm in Henan, and a dataset with 9842 pairs of aligned RGB and depth images was constructed, including 6909 pairs of training images and 2933 pairs of test images. The experimental results showed that the mean absolute error of the proposed model on the test set was 3.019kg, which was reduced by 18.095% and 12.569% compared with the RGB and depth-based single-stream benchmark models, respectively. The average accuracy of proposed method reached 96.132%, which was very promising. Noting that, the model did not add additional training parameters when compared with the direct use of two single-stream models to process RGB and depth images separately. The mean absolute error of the model was reduced by 46.272%, 14.403%, 8.847%, and 11.414% compared with other existing methods: the conventional method, the improved EfficientNetV2 model, the improved DenseNet201 model, and the BotNet+DBRB+PFC model, respectively. In addition, to verify the effectiveness of cross-modal feature fusion, a series of ablation experiments were also designed to explore different alternatives for two stream connections, including unidirectional or bidirectional additive or multiplicative connections. The experimental results showed that the model with a bidirectional additive connection obtained the best performance among all alternatives. All the above experimental results showed that the proposed model can effectively learn the cross-modal features and meet the requirements of accurate pig weight measurement, which can provide effective technical support for pig weight measurement in group farming environment.

Abstract:With the rapid development and expansion of global aquaculture and the diversification of farming models, the scale, intelligence, and informatization of the aquaculture industry have become trends in its development. Fish behavior recognition is of significant importance in ecology, aquaculture, and fisheries resource management. It enables the assessment of fish growth, developmental status, and activity levels based on their behavioral patterns, indirectly evaluating the impact of environmental factors. This can help reduce stress responses in fish growth, improve resource utilization efficiency, and lay the foundation for intelligent development in aquaculture. Traditional fish behavior identification mainly relies on manual observation and recording, which consumes a considerable amount of time and effort and is subject to subjectivity and uncertainty. In recent years, fish behavior recognition methods based on artificial intelligence get extensive attention, is lossless, such as low cost advantage. The fish behavior recognition technologies were reviewed based on artificial intelligence over the past five years, including convolutional neural networks, recurrent neural networks, and two-stream convolutional neural networks. It also provided a summary and analysis of fish behavior recognition methods and datasets. Based on these foundations, an outlook on future research directions was discussed and provided.

Abstract:Focusing on the industrial development needs of accurate underwater target recognition in aquaculture, and aiming at the problems of low target recognition accuracy of small samples and poor adaptability of model algorithm to scenarios, a small sample aquaculture fish recognition method based on improved cycle constraint adversarial network (CycleGAN) sample amplification and attention enhancement transfer learning was proposed. Firstly, the underwater sampling equipment was used to collect the images of the actual and controllable breeding scenes of Larimichthys crocea, and the controllable scene images were used as the auxiliary sample set. CycleGAN was used as the basic framework to realize the migration of auxiliary samples to the actual breeding scene images. In particular, an optimization method of the loss function of the migration model based on the maximum mean discrepancy (MMD) was proposed. Then in the transfer learning phase, ResNet50 was used as the basic framework, and SK-Net (selective kernel network) attention mechanism optimization model was introduced to improve the perception ability of different receptive field targets, so as to improve the recognition accuracy of the model for unconstrained fish targets. The experimental results showed that the method proposed effectively improved the recognition ability of fish small sample targets, with a recall rate of 94.33% of fish recognition, and an mAP of 96.67%, providing effective technical support for the next step of fish behavior tracking and phenotype measurement.

Abstract:The number of dairy cows in China has been steadily increasing over the years. Monitoring their physiological parameters and behavior is crucial to improve their production performance, economic benefits, and enhance the competitiveness of China’s dairy industry. However, traditional monitoring methods, such as using thermometers to monitor physiological parameters, can lead to stress reactions in cows. Additionally, behavioral monitoring relies mainly on manual observation, which is prone to errors due to differences in experience and low efficiency. Fortunately, with the advancement of artificial intelligence technology, equipment and technology for cow health monitoring are constantly evolving. The methods provided convenient and accurate means for cow health monitoring, overcoming the limitations of traditional methods. The advantages of the methods included being non-contact, stress-free, and highly efficient. The research progress in equipment and technology for monitoring the health of dairy cows both domestically and internationally were systematically analyzed and summarized. Specifically, it discussed the monitoring of physiological parameters such as body temperature, body size, weight, and respiratory frequency, as well as behavioral monitoring, including basic behaviors (such as standing, walking, lying down), rumination, and limping. In terms of physiological parameter detection, three methods for monitoring cow body temperature were compared. Implantation and contact methods had advantages such as high stability and low cost, while non-contact methods had a wide range of temperature measurement and can be used for non-invasive temperature measurement. The methods were analyzed based on two-dimensional and three-dimensional images, as well as the combination of two-dimensional and three-dimensional images, and concluded that the combination of two-dimensional and three-dimensional images can more accurately measure the size of cows. The relational model used in cow weight estimation was compared and analyzed, and the conclusion that multiple regression model can accurately estimate cow weight was drawn. The characteristics of sensors, cameras, infrared thermal imagers and other equipment used for monitoring cow respiration were analyzed, and their applicability was compared. In terms of cow behavior monitoring, the application principles, advantages and disadvantages of contact sensor monitoring and non-contact visual monitoring in cow behavior monitoring were summarized. Sensor monitoring had more advantages in monitoring accuracy, while visual monitoring had advantages such as high efficiency and non-invasive. In addition, typical health monitoring application systems in smart aquaculture both domestically and internationally were summarized and analyzed, and the characteristics and application scenarios of the application systems were introduced. Finally, in view of the problems faced by cow health monitoring methods, the need to further optimize data analysis and processing methods while improving equipment and technology was proposed, providing efficient tools that combine software and hardware for cow health monitoring, and providing sustainable development methods and reference ideas for cow health management and the breeding industry.

Abstract:Quantitative measurement is the basic work of biological research and breeding management, and its results are of great significance to the production efficiency, cost control of animal breeding and assessment of economic benefits. In recent years, with the development of image acquisition equipment, image processing technology and computer vision algorithms, the research on animal counting based on computer vision has also made great progress. Artificial counting often needs to rely on breeding personnel to observe and count the animals one by one, which is not only prone to omissions and errors, but also requires a lot of time and human resources. Computer vision-based counting methods can realize automated counting, which to a certain extent reduces the workload of breeding personnel and improves the breeding efficiency. The research related to farm animal counting in the past ten years was counted, and the farm animal counting algorithms were analyzed and discussed from both traditional machine learning and deep learning. Among them, the traditional machine learning method mainly relied on manually extracted features for recognition and counting, with fast computation speed and small resource consumption, but lacked the understanding of the global semantic information of the image;counting algorithms based on deep learning had a stronger generalization ability to complex scenes, and achieved better results in the counting task for farmed animals, which was the mainstream direction of the current research. In addition, the applications of farmed animal counting in the fields of aquaculture, livestock and poultry farming and special animal farming were sorted out and summarized. At the same time, the current publicly released farmed animal counting datasets were summarized. Finally, the main challenges of farmed animal counting research were analyzed and discussed in terms of datasets, application scenarios and counting methods, and the future development trend was outlooked. Specifically, by constructing larger and richer public datasets, improving the accuracy and generalization ability of counting algorithms, and expanding the counting models in specific scenarios to a wider range of application scenarios, the research on farmed animal counting would make greater progress and development, so as to truly play its role in supporting agricultural production.

Abstract:There is a high similarity among different levels of the koi for beauty pageant, and beauty grading for koi is currently done manually. To solve these problems of low efficiency, strong subjectivity and high cost of manual beauty pageants, a sorting method for koi beauty pageant was proposed based on transfer learning and improved ResNeXt50 residual network. Firstly, a rank dataset was constructed for the beauty pageant on Kohaku, Taisho and Showa koi. Secondly, the transfer learning strategy was adopted to improve the training speed and improve the ResNeXt50 model from three aspects of SE attention module, Hardswish activation function and Ranger optimizer, further a SH-ResNeXt50 classification model was proposed and constructed for koi pageant. The experimental results showed that the SH-ResNeXt50 model effectively improved the sorting ability for koi beauty pageant, with an accuracy of 95.6% and a loss value of only 0.074, which was better than the commonly used AlexNet, GoogLeNet, ResNet50 and ResNeXt50 network models. Finally, the interpretability of SH-ResNeXt50 model was analyzed by Grad-CAM, and the results showed that the regions of interest of SHResNeXt50 model was basically consistent with those recognized by the humans. The approach proposed realized the intelligent sorting of different levels of koi beauty pageant with high similarity, which had reference significance for other biological level sorting with high similarity.

Abstract:Vinasse is a by-product in the process of wine production. Vinasse were used as the research object. The effects of different substrate total solids (TS) concentrations (0.5%, 1%, 1.5%, and 2%), inoculation and vinasse ratios (0.25, 0.5, 0.75, 1, and 1.5), and temperatures (25℃, 37℃, and 50℃) on performance of methane production for vinasse AD were investigated. The results showed that methane production was increased gradually with the increase of TS concentration, and the maximum cumulative biogas production (532.8mL/g) and methane production (294.7mL/g) were obtained at 2% TS. The inoculation ratio was an important factor affecting the anaerobic digestion system, with the increase of inoculation ratio, the cumulative biogas and methane production of the system showed an increasing and then decreasing trend, the system collapsed and did not produce methane at inoculation ratios of 0.25 and 0.5, and the maximum cumulative biogas and methane production were obtained when the inoculation ratio was 1. When TS was 2%, and inoculation ratio was 1.5, the cumulative biogas production obtained at a fermentation temperature of 50℃ was 559.4mL/g, which was 10.2% higher than that at 37℃, and the cumulative methane production obtained was 284.0mL/g, which was not significantly higher than that at 37℃ (P>0.05). After gas production kinetic analysis by using the modified Gompertz model, it was found that the higher the TS was, the higher the inoculum ratio and the temperature were, and the shorter the methane production latency period was. The microbial communities in the AD system at different temperatures were also analyzed, and it was found that Bacteroidetes and Firmicutes were the dominant phyla, and the methanogenic bacteria were gradually replaced by hydrogenotrophic instead of acetoclastic methanogens with the increase of temperature. Therefore, TS concentration, inoculation ratio and temperature were important factors for AD, and it was initially determined that the methanogenic performance of vinasse mesophilic AD with 2%TS and an inoculation ratio of 1 was relatively good, and the research result can provide data support for the application of vinasse AD technology.

Abstract:Biomass pellet fuel made from crop straw has the advantages of convenient storage and transportation, high combustion heat efficiency, and can be used to replace coal as the main energy source for cooking and winter heating in rural households. However, the combustion of biomass particles will emit a large amount of nitrogen oxides. Currently, biomass denitrification is mainly used in large industrial furnaces, and there are basically no nitrogen oxide treatment devices in rural household biomass boilers. Based on the combustion characteristics of biomass fuel, the structure and key combustion technologies of a special stove for straw briquette fuel were firstly introduced, and then combustion experiments were conducted on corn straw briquette fuel. KM9106 flue gas analyzer was used to measure NO, NO2 and oxygen content in flue gas during combustion. The results indicated that by changing the air volume ratio, intake position, and inflation angle of the two winds, different air pressure jets were formed within the furnace to strengthen heat and mass transfer, ensure burnout effect, and effectively control the generation and emission of nitrogen oxides during the combustion process. Analysis of experimental data showed that when the ratio of primary to secondary air flow was 3∶1, when distributing secondary air by air pressure jet formed by disturbing flow, the straw shaped fuel had the highest combustion temperature, sufficient combustion, which can meet the requirements of efficient and clean combustion, and had high promotion and utilization value.

Abstract:Grain storage pests can reduce the weight, quality and nutritional health index of grain and its products, and the way of grain pest detection in China is still dominated by manual detection. To meet the needs of modern detection of grain storage pests, gas chromatography-mass spectrometry (GC-MS) was used to obtain the main specific volatile organic compounds (VOCs) of Tribolium castaneum (Herbst), screened multiple metal oxide gas sensors with the obtained compounds as reference. Then the air chamber for sensor response was optimally designed and an electronic nose detection device for grain storage pests was developed based on the composed sensor array. The device collected odor information from three experimental subjects, including T.castaneum, the flour during infestation of T.castaneum and the flour during infestation of Latheticus oryzae Waterhouse. Relative change values and relative integration values of the response curve corresponding to each sensor as the original feature matrix (10×2). Principal component analysis (PCA) and partial least squares regression algorithm (PLSR) were used to optimize the original feature matrix. Finally, a predictive regression model was built to forecast the population density of T.castaneum and Latheticus oryzae Waterhouse in flour. Two GC-MS studies were carried out for the purpose to collect 12 distinct volatile chemical compounds of T.castaneum that were not found in other grain insects or stored grains. The number of sensors was reduced from 10 to 8, and the contribution of the two principal components of the T.castaneum samples was increased to 79.4%. The odor of the flour itself would be a great interference to the electronic nose detection of T.castaneum, and under the condition of no flour interference, the electronic nose device can discriminate between samples with varying insect population densities. The PLSR-based prediction model was highly effective in predicting the number of T.castaneum in flour (correction set: correlation coefficient r=0.88, root mean square error (RMSE) was 8.09;validation set: correlation coefficient r=0.89,RMSE was 7.75);the prediction model was also highly effective in predicting the number of Latheticus oryzae Waterhouse in flour (correction set: correlation coefficient r=0.94, RMSE was 5.85;validation set: r=0.94, RMSE was 6.08). The research results indicated that the device can meet the needs of distinguishing samples with different insect densities in stored grains and had reliable stability. This method also provided a method reference for detecting other pests in stored grain.

Abstract:In order to verify the drying uniformity of the device and study the drying characteristics of paddy, through the multi-factor and multi-level test, taking drying temperature, drum inclination angle, drum speed as the influencing factors, and drying time and drying rate as the evaluation indexes, the influence of the indexes on the drying characteristics of paddy was investigated, and the influence of different drying processes on the paddy crackle ratio was analyzed. The test results showed that the order of the primary and secondary factors affecting the drying time and drying rate of the paddy was as follows: drying temperature, drum inclination angle, and drum speed. The optimal drying process for the drying temperature was 55℃, drum inclination angle was 2°, and drum speed was 40r/min. Verification test through the moisture content uniformity of the K to determine the optimal drying process parameters for the drying temperature of 55℃, drum inclination angle of 2°, and drum speed of 60r/min. Under these conditions, the drying time of the paddy was 191min, the drying rate was 0.036%/min, and the uniformity of moisture content of paddy was 99.6%. The research results can provide certain reference value for the development and process formulation of paddy variable temperature homogeneous drying equipment.

Abstract:In the process of pepper drying, the airflow intensity and uniformity in the mesh-belt pepper dryer will affect the quality of pepper after drying. The results of pepper parameters were measured by experiments to set the parameters of porous media. The flow field model of the drying chamber was constructed by ANSYS Fluent and the reliability of the model was verified. The average velocity and non-uniformity coefficient M were used to characterize the airflow intensity and uniformity of the drying chamber. The porous media model was used to simulate the flow field of the meshbelt pepper dryer before and after structural optimization. The results showed that the adjacent air inlets of the original structure of the mesh-belt dryer were arranged in opposite directions, which caused the airflow to disperse and converge to produce eddy current. The pressure at the airflow confluence was increased, the airflow flow was blocked, and the flow velocity was decreased. The flow velocity at the center of the eddy current was extremely low, resulting in the unevenness of the overall flow field. According to the airflow distribution characteristics of the mesh-belt dryer, a structural optimization scheme for increasing the deflector was proposed. The change of the deflector angle increased the average velocity of the airflow and reduced the nonuniformity coefficient. When the deflector angle was 0 °~7.5°, the drying chamber had the best airflow intensity and uniformity, so the deflector deflection angles were set to be 0°, 2.5°, 5°, and 7.5°. The influence of four angles of the deflector on the airflow in the drying chamber was analyzed. When the angle of the deflector was 2.5°, the airflow distribution was optimal. Compared with the original structure, the average velocity increments on the XY, XZ, and YZ planes of the dryer were 6.8%, 10.8% and 5.2%, respectively, and the non-uniformity coefficient was reduced by 8.7%, 8.5% and 2.7% on average.

Abstract:In order to improve the uniformity of airflow field in integrated silo for corn drying and storage, the computational fluid dynamics and orthogonal test method was used in numerical simulation and parameter optimization of airflow field distribution in integrated silo for corn drying and storage. By single factor experiment design, the influence of three factors, including the position of horizontal air inlet pipe, the diameter of vertical ventilation cage, and the unit ventilation volume on the ventilation uniformity of the integrated silo for corn drying and storage was studied. And through a series of numerical simulations and orthogonal experiments, the ventilation structure and parameters of the integrated silo for corn drying and storage were optimized and designed. Based on the analysis and comparison of the data and velocity cloud images obtained from the single factor experiment, it was found that the average velocity of air flow in the integrated silo for corn drying and storage was not obviously affected by the position of air inlet pipe at different levels, which was decreased gradually with the increase of diameter of ventilation cage and increased with the increase of unit ventilation volume. And the velocity non-uniformity coefficient was decreased first and then increased with the position of air inlet pipe from top to bottom, and first sharply and then slowly decreased with the increase of ventilation cage diameter, however, the overall trend of it was increased with the increase of unit ventilation volume. According to the results of single factor experiment, it was preliminarily determined that the position range of air inlet pipe was -0.34~0.34mm,the diameter range of ventilation cage was 200~400mm, and the range of unit ventilation volume was 20~40m3/(h·t). Then the orthogonal experiment of L9 (34) was designed and carried out to analyze and optimize the parameters affecting the flow field uniformity of integrated silo for corn drying and storage. Based on the comparison of the comprehensive weighted score of airflow velocity nonuniformity coefficient and velocity cloud image as well as the range analysis under different combinations of the orthogonal experiment, it was concluded that the diameter of ventilation cage was the most significant influence factor on the airflow field uniformity of integrated silo for corn drying and storage, followed by the position of horizontal air inlet pipe and the unit ventilation volume. The optimized ventilation structure and parameters of the integrated silo for corn drying and storage were as follows: the position of horizontal air inlet pipe was -0.34m, the diameter of vertical ventilation cage was 400mm, and the unit ventilation volume was 20m3/(h·t). Under this scheme, the comprehensive weighted score of velocity non-uniformity coefficient of the integrated silo for corn drying and storage was 77.4% higher than that of the initial plan, indicating the feasibility and practical application value of the optimized scheme. The research results can provide theoretical guidance and technical support for the ventilation structure and parameter optimization design of the integrated silo for corn drying and storage.

Abstract:To enhance the field turning efficiency of agricultural robot tractors, a dual-steering control system based on active disturbance rejection control was developed for the agricultural robot tractor. The system was designed to meet the agricultural technology requirements and driving environment. The composition and main technical parameters of the agricultural robot tractor were determined, and the hardware system was assembled, and the component selection was made. A 4-degree-of-freedom dynamics model of the agricultural robot tractor was established, and the state space equation of the turning efficiency was determined. A dual-steering control strategy based on active disturbance rejection control was proposed, and a Simulink dynamic simulation model was established, and the turning simulation was carried out. The simulation results indicated that the angular velocity of the active disturbance rejection dual-steering control model was 0.241rad/s, the turning radius was 1.96m, and the disturbance recovery time was 1.04s. Compared with the Ackermann turning control model, the active disturbance rejection dual-steering control model had a larger angular velocity, smaller turning radius, and faster recovery time. The field experimental results showed that the average lateral displacement of the agricultural robot tractor was 18.5cm, the slip rate was 4.84%, and the small-radius turning test showed that the turning radius of the double turning control agricultural robot tractor was reduced by about 0.60m and 0.57m compared with that of the Ackermann turning control, and the average turning time was reduced by 4.70s and 3.41s. In the large-radius turning test, the turning radius of the double turning control agricultural robot tractor was reduced by about 0.52m and 0.49m compared with that of the Ackermann turning control, and the average turning time was reduced by 10.27s and 8.22s.

Abstract:Accurate estimation of the slip ratio for tractor driving wheels is of great significance to improve the working efficiency and safety of tractors and realize the antiskid control of tractors.A multi-innovation parallel extended Kalman filter algorithm was proposed, then through online statistics of multi-sensor innovations, D-S evidence theory was introduced to make decisions and correct the measurement noise matrix, which fused the information of multiple sensors, including machine vision, and then realized the accurate estimation of tractor driving wheels’ slip ratio.The simulation results showed that compared with the common Kalman filter algorithm, the proposed fusion algorithm had higher accuracy in estimating the slip ratio, the root mean square error of the estimated slip ratio was reduced from 2.34% to 1.45%, and the proposed algorithm was insensitive to interference signals. The test results showed that under two different working conditions, the average absolute error and root mean square error of the proposed multi-sensor information fusion algorithm were lower than those of single vision method or radar method. It was verified that the proposed algorithm can accurately estimate the slip ratio for tractor drive wheels, and then provide research help for the follow-up realization of tractor drive anti-skid control.

Abstract:To solve the problem that the traditional power supply methods were difficult to adapt to the needs of modern agricultural sensor networks, an electromagnetic self-powered tractor wheel speed sensor based on printed winding and Halbach array was designed. The device adopted the structural form of an intermediate stator and two rotors on both sides. Based on Faraday’s electromagnetic induction law, the output induced electromotive force of the device was theoretically derived and its influence parameters were determined. By using finite element simulation software, the magnet arrangement, coil turns and air gap which affected the output performance of the device were simulated and analyzed, and the best parameters of stator coil shape and permanent magnets pole number were determined. The back-end circuit was designed, which can rectify, measure and transmit the output voltage of the device, so as to calculate the speed, and can be directly supplied by the device. The prototype of the device was made, and the bench calibration test and actual vehicle application test were carried out respectively. The results showed that the linear fitting results of the device speed and voltage were good, the R2 was 0.99991, and the rectified DC voltage at 400r/min was about 3.16V, which can meet the demand of supplying power to the back-end circuit. Compared with the commercial encoder, the measurement accuracy of the speed in the bench test was kept within 1%, and the accuracy in the actual vehicle test was basically kept within 5%. It can meet the requirements of actual use.

Abstract:Aiming at the problem of unclear mechanism of fuel consumption factors of heavyduty fertilizer spreader under ‘human-vehicle-soil’ operating system and the problem of ‘power supply abandonment’ caused by short single operation period and low frequency of operation cycle of agricultural machinery, the interaction mechanism of fuel consumption structural equation model of ‘human-vehicle-soil’ system was analyzed and the gain potential of ‘one source for multi-purpose’ mode of pure electric tractor power supply to ‘life-ecology’ was predicted. Firstly, a comprehensive field test of heavyduty tractor spreading fertilizer operation was carried out, and the data of operation factors such as driver fatigue, tractor fuel consumption and plough layer soil characterization were collected, and the fuel consumption structural equation model of human-vehicle-soil layer was established. Then based on the causal relationship of structural equation, the correlation between agricultural machinery fuel consumption factors was explored. Finally, based on the equivalent conversion theory of fuel consumption, the equivalent power demand and carbon emissions of fuel agricultural machinery were calculated, and the contribution potential of the ‘one source multi-purpose’ mode of pure electric tractors to the exchange and sharing of production and living power sources and emission reduction and carbon reduction was predicted. The results showed that the soil compaction was the main fuel consumption factor of tractor. The driver’s operating fatigue had an indirect effect on fuel consumption, which can indirectly affect the physical properties of the plough layer and the carbon emissions and operating efficiency of the tractor by manipulating the tractor. The electric tractor power supply ‘one source multi-purpose’ had certain potential for power supply to small power facilities at the end of the agricultural network. During the mechanized field management period, the equivalent power consumption was about 45kW·h/hm2, reducing carbon emissions by 12kg/hm2(CO2). Small and medium-sized farms can reduce the carbon emissions of agricultural machinery by about 2464 tons of CO2, saving 9.24×106kW·h electricity, and providing about 110 tons of electricity required for corn drying when the power supply was idle, or the power required for ventilation for 165 sets of ventilation equipment at the same time throughout the day, or 5500 street lights for night lighting, or the annual electricity demand for 11 duty rooms.