Abstract:As the three main grains in China, rice/paddy, wheat and corn play an important role in the food structure of Chinese residents. Quality monitoring is an indispensable and important link in the industrial chain of grain production, processing, storage and transportation. In particular, efficient, nondestructive, objective and real-time optical quality detection is of great significance to the healthy development of the grain industry. Firstly, the optical characteristics of visible/near-infrared spectrum, Raman spectrum and fluorescence spectrum of three main grains as well as the optical detection mechanism of internal quality were compared and analyzed. The application and research status of optical detection technology for internal quality of grain at home and abroad was summarized and analyzed. The application scope and research status of machine vision, hyperspectral and other grain appearance quality detection technologies was discussed. Secondly, combined with the specific quality detection needs of the three main grains, the research and development status of optical detection devices for grain internal quality at home and abroad was summarized and analyzed. The research status of hardware composition and spatial arrangement of appearance quality detection devices was emphatically discussed, and the commercialization and application of optical detection technology related devices were analyzed. Finally, from the bottleneck of optical detection technology of grain quality, the problems and development trend of fast optical detection technology and its equipment were prospected.

Abstract:To reduce the navigation path tracking steering control frequency of crawler-type combine harvester and improve the stability of the control system, an aiming-tangent local tracking path dynamic planning algorithm was proposed. The planned local tracking path consisted of two smoothly connected arcs, the first arc defining the aiming point on the line of 1/2 lateral deviation from the current position of the harvester, and the second arc defining the geometry of the desired path from the actual positioning of the harvester on the line of 1/2 lateral deviation. An adapted steering control model was established based on the actual steering motion characteristics of the harvester, with R2 of 0.978 and 0.980 fitted to the left-turn and right-turn control models, respectively. The straight-line navigation tracking comparison test in the field showed that the standard deviation of lateral deviation was 0.0489m and 0.0507m, the standard deviation of heading deviation was 3.94°and 4.66°, and the number of steering control was 19 and 12 when the vehicle speed was 0.4m/s and 0.8m/s, correspondingly. Compared with the conventional aiming pursuit algorithm, the standard deviation of lateral deviation was reduced by 19.04% and 31.30%, the standard deviation of heading deviation was reduced by 25.94% and 9.16%, and the number of steering control was reduced by 47.22% and 42.86%, respectively. The results can provide a reference for crawler-type agricultural vehicles navigation controllers.

Abstract:In order to improve the higher frequency control of path tracking of automatic rice transplanter, a path tracking control method was proposed based on model prediction. The automatic driving controller was designed based on the model prediction algorithm. By simplifying the model of the agricultural machinery vehicle, linearizing the kinematic equation and formulating the constraint quantity, the current state quantity p=(x,y,θ) can predict the vehicle state at the next time and control the automatic rice transplanter to walk along the reference path. The feasibility of the controller was verified by establishing a simulation model in Matlab. The results showed that the lateral deviation of the straight-line path tracking was less than 0.02m, the heading deviation was less than 0.08°, the average value of the lateral deviation of the curve path was 0.022m, and the average value of the heading deviation was 0.699°, which can be used for the actual vehicle test. In addition, taking the rice transplanter as the test platform, the robustness of the algorithm was verified by setting different vehicle speeds. The average horizontal and heading deviations of straight-line path tracking were 0.021m and 6.187°, respectively, and the average horizontal and heading deviations of curve path tracking were 0.450m and 10.107°, respectively, which can meet the needs of the automatic driving rice transplanter for path tracking accuracy and real-time performance, and provide a reference for the research of agricultural machinery path tracking control.

Abstract:In order to improve the autonomy, safety and efficiency of agricultural robots in orchards, effective and reasonable motion planning methods are essential. Aiming at the problems of the traditional RRT* (rapidly exploring random tree star) global path planning algorithm in the continuous corridor environment, such as low search efficiency, low utilization of sampling points, and large corners of the generated path. The Ackerman chassis spray robot was used as the motion model, and an improved bidirectional RRT* algorithm was proposed. Firstly, a two-dimensional plane map of the orchard was established based on lidar, and the fruit trees and obstacles were regarded as obstacle areas. The obstacles were expanded with the kinematic constraints of the spray robot. Then, the improved bidirectional RRT* algorithm was used to search the path. In the process of searching the path, the dynamic terminal node guidance and potential field guidance were combined to conduct bias sampling, and the paths generated initially were de-redundant and adjacent broken line segment angle constraint processing. Finally, the third-order quasi-uniform B-spline curve was used to optimize the trajectory of the processed path points, and the collision detection and the curvature constraint of the spray robot were mainly considered in the optimization process. Experimental results showed that compared with the traditional bidirectional RRT* algorithm, the proposed improved algorithm reduced the planning time by 57.5% on average, improved the sampling point utilization by 28.55 percentage points on average, and shorted the final path by 7.14% on average. The trajectory obtained by the third-order quasi-uniform B-spline curve optimization satisfied the maximum curvature constraint of the spray robot in both environments with and without obstacles, and only turns occurred at line breaks and obstacles, which conformed to the operating trajectory conditions of the spray robot, and improved the work efficiency and autonomy of the spray robot.

Abstract:To further improve the quality of film mulching soil on whole plastic film mulching on double ridges, rationally cover soil on the film surface, reduce dust, and explore the interaction between soil on the film surface and airflow, the annual mean wind speed of 1.32m/s, annual mean maximum wind speed of 18.07m/s, and monthly mean top wind speed of 26.5m/s in 52986 meteorological observation station of Lintao County, Dingxi City, Gansu Province in recent 30 years were used as simulation data source. Based on the experience of farmers, the minimum, middle and maximum values in the range of 0°~90° were the planting bed directions of whole plastic film mulching on double ridges, and three whole plastic-film planting bed models, T1(0°), T2(45°) and T3(90°), were established, respectively. Using CFD-DEM gas-solid coupling technology, the interaction mechanism of the wide large airflow field at different wind speeds and direction of membrane double dealt with all kinds of bed was got, and combining with the influence of solar radiation energy, cultivated land utilization of double-membrane dealt with all kinds of bed on the building, the building method was optimized, finally, the field validation test was carried out. The analysis result of the soil surface flow field showed that when the air velocity was constant, the maximum air velocity on the surface of the horizontal belt was T3, T1, and T2 in descending order, and the difference between the air velocity and the standard air velocity on the prominent ridge surface was T3, T1, and T2 in descending order. The analysis of soil covering the process of seedbed showed that when the air velocity was constant, the influence degree of seedbed and soil particles on airflow was T3, T1, and T2 in descending order, and the influence degree of airflow on particles was T3, T1, and T2 in descending order. Therefore, it can be seen that the air velocity of the T3 model seedbed and soil covering surface was the largest, which was affected by airflow, and the movement distance of film covering surface was the largest, which was easy to form dust. At the same time, the film covering the intersection point on a prominent ridge surface was easy to penetrate the airflow, causing film uncovering phenomenon in solid wind, affecting crop growth and endangering economic benefits. The optimized seedbed construction method should follow the principles of minimum soil cover displacement, maximum solar radiation energy, the fastest construction efficiency, the south slope (sunny slope) cultivated land priority, northsouth direction cultivated land priority, the priority model was T1, T2 and T3. Field test results showed that when the air velocity was 2.77m/s and the wind direction was north, the average pass rate of seedbed was T2, T1, and T3 from large to small, the film mulling efficiency, the utilization rate of cultivated land and the occupancy rate of lighting area were T1, T3, and T2 from large to small. The test results were entirely consistent with the simulation results, which mutually verified the reliability of the model.

Abstract:In order to reduce torsion and consumption in rotary tillage operation, a self-excited vibration rotary blade device was designed based on the national standard IT245 rotary blade, and its working principle was described. Through the analysis of motion force, the selection of large spring parameters and the design of waist hole of spring spindle were completed. Based on DEM-MBD technology, the simulation model of soil rotary tiller interaction was established, and the variation law of three-dimensional resistance and torque of national standard and self-excited vibration rotary tiller under five knife shaft speeds was analyzed. In the simulation test, the effect of drag reduction and torque reduction was not obvious at the low speed when the cutter shaft speed was 150r/min and 200r/min. When the rotating speed was 250r/min and 300r/min, the self-excited vibration rotary blade had better drag and torque reduction effect than the national standard rotary blade. The vertical resistance was reduced by 6.96% and 10.41%, respectively, and the average torque reduction rate was larger, which were 9.80% and 19.63%, respectively. When it reached 350r/min, the drag and torque reduction effect was reduced. By analyzing the average torque of two kinds of rotary tiller simulation and soil tank test, the correlation coefficients of the national standard and the average torque curve of self-excited vibration rotary tiller were obtained, which were 0.997 and 0.998, respectively, which basically verified the accuracy of DEM-MBD coupling simulation model. The frequency domain analysis of the vibration acceleration data in the Y-direction collected in the soil trough test showed that with the increase of the rotating speed of the cutter shaft, the amplitude of the Y-direction power spectral density generally showed an upward trend. When the rotating speed reached 300r/min, the excitation frequency reached near the natural frequency of the installation in the Y-direction. At this time, resonance occurred, and the amplitude of the Y-direction power spectral density reached the maximum value. At this time, the rotary tiller obtained the maximum energy. The torque reduction was the largest, and the effect of reducing torque and consumption was the best.

Abstract:Aiming at the problems of poor seed-filling performance and seeds were easily stuck in the seeding wheels of metering device for small particle size seeds，seeding wheels with inclined parabolic holes and stirring structure were designed which could plant 2±1 seeds of rapeseed sesame and pakchoi in one hole. Mechanical models for seed-filling and seed-casting were constructed. The range of main structural parameters of seeding wheels and how the parameters influence on seed-filling and seed-casting were analyzed, based on the mechanical and physical properties and precision hole-seeding requirements of rapeseed sesame and pakchoi. The influence of the main structural parameters on seed-filling was verified by using EDEM software, and high-speed camera. The optimal value of distance from parabolic vertex to circular center, focal distance, parabolic tilt angle, width coefficient, side tilt angle and churning structure height were determined. The critical conditions for avoiding seeds stuck in holes or dragged by churning structure were determined. The empirical formulas for calculating the optimal structure parameters by physical properties of seeds were presented. The JPS-12 test-bed was used to study the seeding performance of seeding wheels with the optimal structure for Huayouza 62, Hangtianxinzhi T31-8 and Wuyueman. The qualified rates of seeds per hole were 92.00%, 90.00% and 90.67%, and the qualified rates of hole spacing were 83.67%, 81.83% and 82.50%, respectively. Field tests showed that the average number of Huayouza 62 seedlings per hole was 1.16, the qualified rate of 2±1 seedlings per hole was 89.67%, and the qualified rate of hole spacing was 81.54%;the average number of Hangtianxinzhi T31-8 seedlings per hole was 1.15, the qualified rate of 2±1 seedlings per hole was 85.77%, and the qualified rate of hole spacing was 75.51%. The metering device could meet the requirements of precision hole-seeding for rapeseed and sesame. The research result can provide a reference for the design and research of seeding wheels of metering device for small particle size seeds.

Abstract:There are great difficulties in sowing green manure in paddy fields before rice harvest and in hilly and mountainous areas. These problems mainly include sowing equipment cannot work in the field and the labor intensity of manual sowing is high. Based on the agricultural multi rotor unmanned aerial vehicle platform, a centrifugal disc type green manure seed broadcast sowing device was developed, which can be used to sow seeds such as Chinese milk vetch and Orychophragmus violaceus. The device can be easily and quickly assembled and connected with unmanned aerial vehicle platform. It was mainly composed of hitch mechanism, seed box, seed metering mechanism, broadcast sowing mechanism and sowing automatic control system. The screw conveying seed metering mechanism was used to achieve continuous and stable quantitative seed metering. After the broadcast sowing mechanism was optimized, the seeding was more uniform and smooth. The control system can follow the unmanned aerial vehicle flight speed to control the seed amount of the seed metering mechanism, and set the rotation speed of the seed-rotating disc for broadcast sowing mechanism according to different varieties of green manure seeds, so as to complete the quantitative seeding and uniform sowing of different varieties of green manure. The seed of milk vetch, a typical green manure variety, was selected as the test object. Three-factor and three-level orthogonal performance test was carried out with setting variation coefficient of sowing uniformity Y1 and relative error of sowing rate Y2 as the evaluation indexes, rotation speed of the auger for seed metering mechanism A, rotation speed of the seed-rotating disc for broadcast sowing mechanism B and flight speed C as the influence factors. According to the results of orthogonal test, rotation speed of the auger for seed metering mechanism A and rotation speed of the seed-rotating disc for broadcast sowing mechanism B had extremely significant influence on the two evaluation indexes, flight speed C had significant influence on the two evaluation indexes. The importance order of the factors which affected the Y1 was B，A and C, and affected the Y2 was A，B and C, the optimal combination of working parameters was A2BC2, A was 190r/min, B was 1700r/min, C was 5m/s, and Y1 was 28.47% , Y2 was 11.81%. The field experiment under the optimal combination of working parameters showed that the seedling emergence was good. The research result provided a theoretical basis for improving the green manure broadcast sowing device based on unmanned aerial vehicle platform, and provided equipment support for large-scale promotion of green manure planting.

Abstract:Considering at the problem that the narrow-row close planting, high sowing uniformity and lack of suitable sowing technology and equipment for small-size vegetable seeds such as spinach, a pneumatic needle planetary gear train multi-row parallel low-drop precision metering device suitable for close planting precision sowing of small-size vegetable seeds such as spinach was designed. The working principle of seed metering device was expounded, and the seed mechanical models of seed suction and seed feeding were constructed, and the main structural parameters of seed metering device were determined. ADAMS software was used to simulate and analyze the static trajectory and dynamic trajectory of the suction needle of the planetary gear train seeding mechanism, and the low zero-speed seeding conditions were clarified. The performance test of seed metering device was carried out. The results of seed-metering test showed that the primary and secondary order of affecting the qualified index was rotation speed of seeding, suction negative pressure and unloading positive pressure. The best combination of parameters was seed metering speed of 19.56r/min, suction negative pressure of 2.05kPa, and unloading positive pressure of 1.00kPa. Through bench test verification, the performance indexes were as follows: the average qualified index was 91.48%, the average missing index was 4.28%, and the average replay index was 4.24%. The results of seeding test showed that when the seed pressure was 0.8~1.0kPa, the working speed was 18~20r/min and the seed height was no more than 200mm, the coefficient of variation of grain spacing was not more than 13.2%, and the working performance was better. The research result can provide a reference for the design of vegetable narrow row close planting precision seeder.

Abstract:The mechanisms governing the precision of seeding depth and the related agronomic outcomes are site-specific. Regional soil mechanics in the Yangtze River Basin plays a role in affecting the relationship between the seeding unit and the soil, which is a key consideration for machine design. Using four technical objectives for evaluation, a field bench experiment was conducted in the field using a market available seeding unit (2BMYFQ) to illustrate the tool-soil interactions. Two tillage treatments (i.e. no-till and rotary till), three depth settings (i.e. 2.5cm, 4.0cm and 6.0cm) and three downward forces (i.e. 0.6kN, 1.0kN and 1.2kN) were adopted in the experiment. Seeding depth, soil properties after seeding and seedling establishment rate were measured. Results showed that the interactions between the seeding unit and soil mechanics affected seeding depth significantly. The maximum seeding depth variation was 37.61%. Results showed that linear elastic force depth control assembly plus the poorly managed seedbed made it impossible for precision seeding depth control. The mechanisms leading to the poorly controlled seeding depth were identified, including both the void soil support and the over sinkage of the ground wheel. Meanwhile, the seeding unit affected soil mechanics significantly, which in due resulted into non-uniform seedling establishment rate. Results indicated that the combination of the no-till, 4cm depth setting and 1.2kN downforce provided the best precision of the seeding depth. While in the tilled soil, 4cm depth setting and 1.0kN led to the best result. Overall, the depth control performance in the tilled seedbed condition was higher than that in no-tilled soil. The research result indicated that suitable downforce selection was inherently related to both soil mechanics and agronomically defined seeding depth. Inter-relationship between the seeding unit and soil mechanics as well as on-line soil monitoring system for downforce control were key measures for precision seeding in a given agricultural zone.

Abstract:Aiming at the problem that the precision and performance of maize mechanical seed-metering device was decreased and unstable when it was at high-speed sowing of the seeder. The technical idea of adjusting and controlling maize seed filling posture by using posture adjustment teeth and unit hole was put forward. A precision metering device with posture control and driving was designed. It adopted the structural layout of double sided seed discs opposed and single row seeding, so as to reduce the rotation speed and improve the uniformity of seed-metering. The structural parameters of the key components were designed. The principle of seed posture adjustment was analyzed. Through single factor test and combination test, the optimal parameters of the seed-metering device was obtained. Subsequently, the seed-metering performance comparison test were carried out. The results showed that when the type of the posture adjustment teeth was linear, the effect of improving the qualified index was the best, which can be increased by 29.1 percentage points compared with that of no posture adjustment teeth. When the rotation speed was 16.7r/min, the inclination angle of the hole outer wall was 46.9°and the fillet radius of the hole was 4.5mm, the qualified index, missed index and multiple index were 91.6%, 2.8% and 5.6%, respectively. Within the range of operating speed of 8~14km/h, the qualified index was over 90%, the miss index was below 3%, the multiples index was below 8%, the damage index was below 0.5%, and the variation coefficient of grain spacing uniformity was below 19%. Moreover, the seed-metering effect was better than that of no posture control seed-metering device and scoop type seed-metering device, which met the technical requirements of maize precision sowing.

Abstract:The garlic aligning bud seeder, composed of single seed picking device, scale bud direction control device, vertical planting device, transmission system, frame, ground wheel and other parts, was designed in order to solve the problem of garlic aligning bud seeding. It can complete seed picking, reversing, vertical planting and suppression operations at one time. According to the dimension parameters of garlic scale bud, the key parts of the planter were optimized, mainly including the development of large, medium and small seed picking spoons according to the dimension distribution of garlic scale bud. A curved opening commutator was designed to make the bud tip of garlic scale bend and expose the commutator as much as possible. The vertical planting mechanism with the middle shaft rotating simultaneously with the driving disk was designed to realize the simultaneous stable operation of the duck tip in 11 rows, and to realize the normal bud with the bud tip no shorter than 6mm with the arc commutator. Taking the ‘Cangshan’ garlic and ‘Jinxiang’ hybrid garlic as experimental subjects, the field seeding performance test was carried out on the seeder. The results showed that when the walking speed was in the range of 0.14~0.19m/s, the normal bud rate of ‘Jinxiang’ hybrid garlic was about 85%, while the ‘Cangshan’ garlic was about 90%, and the single grain rate was more than 93%, which met the agronomic requirements of garlic sowing.

Abstract:China is the largest vegetable producer and consumer in the world, and seedling transplanting is the main method of vegetable cultivation. To address the shortcomings of the two types of automatic vegetable transplanters, the stalk clamping type and the pot ejection type, which are used to pick up seedlings, an alternating seedling picking and dropping mechanism was designd based on the combined ejecting-clamping picking method for symmetrically arranged bendable seedling trays. The mechanism was driven by the combination of crank rocker and cylinder. After the seedling ejecting device released the seedling from the cavity wall of the seedling tray, the pneumatic seedling clip clamped the seedling from the seedling picking position to the seedling throwing position to complete the seedling picking and throwing operation, and the seedling picking and throwing were carried out alternately on the front and rear sides of the seedling tray, thus enhancing the efficiency. The working principle, key point motion trajectory and structural composition of the mechanism were analyzed. The influence of key factors on the trajectory of seedling clamping point was analyzed and the following values were selected. The drive crank speed was 10r/min, the length of the seedling throwing rocker was 309mm, the drive cylinder extension speed was 25mm/s, the seedling pulling was completed within 0.8s, and the extension moment was 0.4s ahead. Under this combination of parameters, the maximum lateral displacement of seedling clamping point in the seedling pulling stage was 9.6mm, the accumulated lateral displacement was 0mm, and the theoretical lifting height was 44mm, which met the theoretical requirements of seedling picking and dropping operation. The performance test of picking and dropping seedlings under the planting frequency of 70~120 seedlings/min in a single row was carried out with 45d growth time of pepper seedlings as the operation object. The test results showed that the mechanism can achieve 93% success rate of seedling picking, 95% success rate of seedling dropping and 88% overall success rate at planting frequency was 100 plants/min in a single row, which met the requirements of seedling picking and dropping operation. The feasibility of the seedling picking and dropping mechanism was verified.

Abstract:An air-suction substrates removal device was designed to solve the problem of low removal rate in the hole of the plug. Firstly, the missing seedling holes were detected and located by the deep learning model, and then the seedlings were transported to the substrates removal module. The linear module drived the air suction port to above the missing seedling holes. Finally, the negative pressure adsorption method was used to complete the task of removing the substrates. The effects of nine suction port structures were compared and analyzed by using the DEM-CFD coupled simulation method. The results showed that when the diameter of the circular tube at the suction port was 30mm and the height of the shrinking tube was 50mm, the optimal performance of high substrates removal rate and more uniform delivery was exhibited. A test platform for air-suction substrates removal of missing seedling holes was built, and a multi-factor orthogonal test study was carried out. The results showed that the optimal parameter combination was air pressure of 0.5MPa, substrates moisture content of 50%~55%, and air-suction time of 3.0s, with silicone pad. The performance verification test was carried out, and the results showed that the mean average recognition precision was 96.1%, the average positioning success rate was 95.45%, and the average substrates removal rate was over 90%, the working efficiency of the whole machine was 57s/disk, which met the actual requirements of removing seedlings.

Abstract:Based on the problem of the uncertainty of fertilization position of existing liquid fertilizer deep application machine, combined with mechanical structure design and automatic control technology, a fertilizing device for deep application of liquid fertilizer with target fertilizing system was proposed, which included a liquid fertilizer deep application trencher and a liquid fertilizer target application control system. The furrow opener for deep application of liquid fertilizer was innovatively designed, the mechanical contact model between the furrow opener and soil was constructed, the dynamatic model of soil above the furrow opener was constructed, the structural parameters of the furrow opener were determined, the soil disturbed and falling principle was analyzed, and EDEM software was used to build the DEM simulation model of furrow opener-soil, the feasibility of the furrow opener structure for deep application of liquid fertilizer was verified. A target fertilizing system was developed with microcontroller as the core, using photoelectric sensor and solenoid valve synergy, the photoelectric sensor sensed the crop plant position, the speed measurement module measured the device operation speed in real time, and the microcontroller combined the crop plant position information and operation speed to control the opening and closing of the solenoid valve to complete the liquid fertilizer target fertilizing operation. The fertilizing device for deep application of liquid fertilizer with target fertilizing system was verified through field experiments. At the operation speed of 0.4~1.0m/s, the average depth of return soil was 52.8mm, the average target application rate was 84.03%, and the return performance and target application performance of the device were stable, which could meet the agronomic requirements of liquid fertilizer deep application.

Abstract:In view of the inaccuracy of artificial fertilizer application amount and uneven fertilization of orchard organic fertilizer, an organic fertilizer strip spreading rotary tillage mixed fertilizer application machine was designed according to the requirements of fertilization agronomic. The device used scraper drainage fertilizer, through the ring chain to drive the scraper to the front row of fertilizer, and spread organic manure in strips on the surface, then the rotary tillage device would be mixed with the soil. First of all, through the design calculation, the maximum opening height of the fertilizer device, fertilizer box volume and other structural parameters were determined, and the upper and lower organic fertilizer discharge process was analyzed. Then, discrete element simulation test was carried out with the opening height of fertilizer discharge port, forward speed, sprocket speed and scraper spacing as test factors. The parameters were solved with the relative error and variation coefficient of organic fertilizer as evaluation indexes, and the working parameters of fertilizer discharge process were optimized and solved. The optimal parameter combination was obtained as follows: the opening height was 53.17mm, the forward speed was 2.8km/h, the sprocket speed was 15.96r/min, and the scraper spacing was 160mm. The experimental verification under the optimal working conditions showed that the average organic fertilizer discharge was 5.099kg/m2, the relative error was 4.5% and the coefficient of variation was 8.8%. This showed that the simulation optimization results were reliable, the fertilizer discharge was accurate and the uniformity of fertilizer discharge was good, and the fertilization performance of the fertilization device was good. Finally, in the rotary tillage mixing test, the mixing ratio of upper organic fertilizer was 11.83%, and that of lower organic fertilizer was 6.29%, indicating that after rotary tillage, soil and fertilizer mixing effect could be achieved, and the mixing ratio of upper soil and fertilizer was higher than that of the lower layer.

Abstract:The performance of soil-breaking and root-cutting cutters with different structural forms is quite different. To better break the soil compaction structure of the grassland, the structural design and parameter optimization of soil-breaking and root-cutting cutters were carried out. The discrete element method was applied to construct the grassland soil model, and the parameters of the model were calibrated by direct shear test. A three-factor, five-level quadratic orthogonal rotational combination design test was conducted with cutting edge angle, sliding angle, and cutting tooth angle as test factors, and tillage resistance, soil disturbance area, and specific resistance as target parameters, and a grassland verification test was conducted for the optimal parameter combination. The experimental results showed that when the cutting edge angle was 37.8°, the sliding angle was 33.6°, and the cutting tooth angle was 51.8°, and the operation effect was the best. Grassland tests showed that compared with the triangular soil-breaking and root-cutting cutter, the reduction rate of the optimized soil-breaking and root-cutting cutter was 11.8% and 12.8%, respectively in grassland soil of different firmness, and no significant overturning was produced after the operation, which was more in line with the agronomic requirements of grassland operation. The research results could provide a theoretical basis for the standardized design of soil-breaking and root-cutting cutter.

Abstract:In silage harvesting process, the clearance between moving and fixed blades has an important influence on the shearing performance of silage harvester cutting device, which directly affects the quality of silage cutting. Although the development of domestic silage harvester markets increases mature, the key technology of clearance automatic adjusting is still lacking for silage harvester. At present, the manual regulation is still relied on, and the complicated regulating process seriously increases labor intensity, and even delays the farming time. In order to improve the mechanization and automation level of clearance adjusting for silage harvester, an electric drive rocker arm eccentric clearance automatic adjustment device was designed. One end of the rocker arm was driven by the motor to rotate with the thread shaft, and the other end was connected with the fixed blade seat. The fixed blade rotated around the rotating shaft under the drive of the rocker arm, so as to realize the adjustment of the appropriate gap. A clearance control system based on vibration acceleration sensor was developed, and the contact state was judged by vibration acceleration signal when the fixed blade and moving blade contacted. For checking the rationality of the clearance automatic adjustment device structure and the accuracy of the control system, indoor tests were carried out at three clearance measured values of 0.2mm, 0.6mm and 1.0mm, and three chopping cylinder rotating speeds of 500r/min, 800r/min and 1100r/min. Test results showed that the clearance measured value and chopping cylinder rotating speed had a very significant effect on the clearance uniformity between the moving blade and fixed blade by the analysis of variance. With the increase of rotating speed of chopping cylinder, the variation coefficient of the clearance uniformity between the moving blade and fixed blade was increased. With the increase of clearance measured value, the variation coefficient of the clearance uniformity was decreased. The precision of left and right clearance synchronization adjusting was high, while the highest error was only 0.12%(＜1%).When the rotating speed of chopping cylinder was 500r/min, and the clearance measured value was 0.2mm, 0.6mm and 1.0mm, respectively, the variation coefficient of the clearance uniformity between the moving blade and fixed blade was 6.03%, 5.78% and 5.36%, respectively. The device and its control system realized precise clearance regulation and control, which provided technical support for the intelligent control of domestic silage harvesting device.

Abstract:The non-contact evaluation of pesticide inline mixing uniformity based on image processing can promote the development and performance evaluation of the mixers in direct nozzle injection spraying systems (DNIS). In view of the phenomenon that the uniformity results obtained by image processing cannot be directly matched to the traditionally widely accepted and referenced numerical simulation results, the linear models to map the image processing results with numerical simulation results was constructed as inline injecting and mixing viscous water-soluble pesticides and water in a long transparent detection tube, and tested by a jet mixer in a DNIS. Results showed that differing image methods (HSM, OAU, PCA) corresponded to varying optimal linear fitting orders. The optimal order was 4 and the fit goodness was higher than 0.95 when each single image method of them was applied. When each combination of two image methods of them and all the three methods were applied, the order for them can be reduced to 3 and 2, respectively, and the goodness of fit can increase to about 0.98. Based on the above image processing methods and linear models, the uniformity performance of the mixer can be predicted with the error (Mean absolute error, MAE) universally less than 0.05 as the carrier flow rates (Q) were in the range of 800~2000mL/min and the mixing ratios (P) were in the range of 0.01~0.10. Also, the use of univariate and bivariate linear models reduced the average prediction error by 84.1% and 79.8%, respectively, and reduced the variations in prediction results between different algorithms by 31.6% and 78.0%, respectively. The MAE can be limited within 0.03 when univariate models based on the PCA algorithm or the OAU algorithm were applied alone for prediction, and their accuracy was higher than the prediction results of the combinations of different algorithms, indicating the rationality of uniformity prediction using the linear models based on image processing. Though the MAE for the bivariate model based on HSM-PCA algorithm combination was only slightly higher than 0.03, it may have the advantages of avoiding inaccuracy risks of prediction caused by using a single indicator. The research established the relationship between image processing and numerical simulation, thus further improving the feasibility of inline pesticide uniformity assessment inside pipelines based on experiments.

Abstract:In order to study the effect and optimization feasibility of different pressure-reducing structures in the upper crown flow channel of medium and high head Francis turbines, taking the No. 4 unit of Hongshanzui First-Stage Hydropower Station as an example, four different depressurization structure models of upper crown channel established by UG were taken as the research object. Based on computational fluid dynamics（CFD）technology, shear stress transport (SST) turbulence model was used to simulate a total of 28 calculation conditions of four different upper crown drainage structures under seven flow rates. The research indicators were the flow distribution characteristics of leakage water, the lower side pressure of main shaft seal, the axial water thrust of the upper crown and the sealing performance of the comb ring. The results showed that there were some differences in the leakage water flow regime in different drainage and depressurization structures. In order to improve the sealing performance of the turbine main shaft, a combined drainage and depressurization structure with a runner pump can be adopted. Compared with other structures, this structure had significant effects on reducing the main shaft sealing pressure, the axial water thrust of upper crown and the leakage of upper crown clearance. Adjusting the geometric parameters of the pump blades or pump cover of the pressure-reducing structure of the runner pump can achieve the optimization purpose. In view of the leakage problem of the main shaft seal of the power station, the combined drainage and pressure reduction structure with runner pump can reduce the lower side pressure of the main shaft seal by about 15.98% on average and the axial water thrust of the upper crown by about 52.99% on average, which can greatly improve the operation efficiency of the power station. A runner pump was added on the basis of the traditional single drainage and pressure-reduction structure, which provided a reference for obtaining the best comprehensive benefit and its transformation and optimization of a medium-high head Francis turbine.

Abstract:The relationship between the quantity increase and decrease of production-living-ecological space (PLES) and the evolution process of spatial pattern are the premise of realizing the sustainable development and utilization of national space. Land use dynamic index, geo-information graphic, gravity analysis and bivariate spatial auto-correlation analysis were used to study the quantity change and spatial pattern evolution process of production-living-ecological space in the Yangtze River economic belt (YREB) from 1980 to 2020. The results were as follows: from 1980 to 2020, the main types of land space in the YREB were production space and ecological space, and the quantity changes of different land space types were obviously different. The agricultural production space and ecological space were decreased by 39403km2 and 248km2, respectively. The non-agricultural production space and living space were increased by 14804km2 and 27271km2, respectively. From 1980 to 2020, the spatial orientation and regional differences of land spatial pattern change in YREB were significant. From 1980 to 2020, the interaction between different land spatial types in the YREB had significant differences and obvious spatial heterogeneity.

Abstract:Owing to the all day and all weather advantages of radar remote sensing and the strong penetrability of microwaves, information from radar may be supplementary to that of optical sensors, and thus facilitating the research of soil salinization using both radar and optical images. However, at present, few quantitative studies on soil salinization have been carried out by using polarimetric synthetic aperture radar (PolSAR) image and polarization characteristic parameters. Moreover, different variables extracted from optical and radar images as well as DEM data have been adopted to retrieve soil salinity by previous scholars. As to their retrieval efficiencies and comparative advantages, there are still some uncertainties and confusions which should be explored comprehensively to locate those variables with strong universality. Taking Juyanze, which is located at southeastern Ejina Banner in Inner Mongolia, as the study area, six types of variables including band reflectance, vegetation index, salinity index, polarimetric SAR parameter, land surface temperature and topographic factor were extracted based on Sentinel-2, Radarsat-2, Landsat-8 and SRTM DEM data. Variable optimization strategy was adopted to screen the optimal variable of each variable type and their combinations, and then multiple random forest (RF) and support vector machine (SVM) soil salinity prediction models were established and evaluated. The optimal model was used to predict soil salinity in Juyanze area, which was expected to provide practical reference for soil salinity monitoring in arid area. The results showed that variables such as short-wave infrared band (B11), canopy response salinity index (CRSI), extended ratio vegetation index (ERVI), salinity index Ⅱ rededge3 (S2re3), single scattering (FOdd), land surface temperature (LST) and total catchment area (CA) had high universality for soil salinity monitoring. For single variable models, the salt prediction accuracies were ranked in descending order as topographic factor, polarimetric SAR parameter, land surface temperature, salinity index, vegetation index and band reflectance. Multi-variable combination can effectively improve the model accuracy and stability. With the addition of environmental variables, R2 of the optimal model was increased by 0.117 and the corresponding RMSE was decreased by 2.556 percentage points when all six types of variables were involved in the model. RF model was more suitable for soil salt inversion in arid areas than SVM, and the RF model based on the optimal total variable group had the highest accuracy. The inversion results showed that the soil was mild salinized in northeast part and areas around Swan Lake, while in southwest paleolake basin, severe soil salinization was generally occurred.

Abstract:Desert steppe with features of sparse vegetation and fragmented bare soil distribution, required for high spatial resolution and spectral resolution of remote sensing data. There were some problems with over calculation and time-consuming according to present situation of deep learning use for remote sensing. Firstly, multiple hidden layers with complex structure were common in remote sensing scenes application. Secondly, inherent characteristics of remote sensing data were lack of consideration when some classical models were applied directly. A low altitude unmanned aerial vehicle (UAV) platform was established with a hyperspectral remote sensing sensor on it, which gave full play to the strengths of spatial and spectral resolutions. A simplified learning classification model were proposed by using three-dimensional convolutional network (3D-CNN) in desert steppe with hyper parameters of learning rate, batch size, number and size of convolutional kernels optimized for the classification of vegetation, bared ground and indicators. The highest overall accuracy (OA) of the model was evaluated to be 99.746% after optimized. The results suggested that the optimization of simplified learning classification model should build on constantly adjusting hyper parameters and sufficiently comparing with classification results of various combinations for higher precision, shorter time-consuming and more reliable stability. These results demonstrated that the simplified learning classification model based on UAV hyperspectral remote sensing had good performance in classifying ground target in desert steppe.

Abstract:Mikania micrantha is one of the top ten harmful weeds in the world, and its flooding will have a great impact on the ecosystem. Establishing a high spatial resolution and global scale early warning and assessment method for Mikania micrantha is one of the key measures to control Mikania micrantha. At present, Mikania micrantha is mainly monitored by manual survey and satellite remote sensing, but the former is inefficient and the latter is not accurate enough. Unmanned aerial vehicle (UAV) was used as the carrier to collect Mikania micrantha color images in the area to be monitored, the Otsu-K-means, RGB, HSV color space threshold segmentation algorithm and K-means-RGB, K-means-HSV, K-means-RGB-HSV fusion algorithm and MobileNetV3 deep learning algorithm were used for recognition. The recognition results were evaluated by three evaluation indexes: recall rate, accuracy rate and average F1-score value. The experimental results showed that K-means-RGB-HSV algorithm had the best overall recognition effect on Mikania micrantha in full bloom. On this basis, based on the recognition results, an early warning evaluation system of Mikania micrantha was constructed by applying fuzzy analytic hierarchy process and coverage formula, and five Mikania micrantha invasion hazard grades were divided. According to the different monitoring accuracies, grids with different sizes and radiation radius were set, and the accurate distribution heat map of Mikania micrantha invasion was drawn, which could clearly and accurately reflect the harm degree of Mikania micrantha invasion in different areas. Accurate monitoring of Mikania micrantha in full bloom based on UAV remote sensing was achieved with centimeter-level resolution accuracy, which provided strong support for monitoring, early warning and accurate prevention of Mikania micrantha invasion.

Abstract:Real-time detection of orchard environment is an important prerequisite to ensure the accurate operation of orchard spray robot. An improved DeepLab V3+ semantic segmentation model was proposed for multi-category segmentation in orchard scene. For deployment on the orchard spray robot, the lightweight MobileNet V2 network was used to replace the original Xception network to reduce the network parameters, and ReLU6 activation function was applied in atrous spatial pyramid pooling (ASPP) module to reduce the loss of accuracy when deployed in mobile devices. In addition, hybrid dilated convolution (HDC) was combined to replace the void convolution in the original network. The dilated rates in ASPP were prime to each other to reduce the grid effect of dilated convolution. The RGB images of orchard scene were collected by using visual sensor, and eight common targets were selected to make the dataset, such as fruit trees, pedestrians and sky. On this dataset, DeepLab V3+ before and after improvement was trained, verified and tested based on Pytorch. The results showed that the mean pixel accuracy and mean intersection over union of the improved Deeplab V3+ model were 62.81% and 56.64%, respectively, which were 5.52 percentage points and 8.75 percentage points higher than before improvement. Compared with the original model, the parameters were reduced by 88.67%. The segmentation time of a single image was 0.08s, which was 0.09s less than the original model. In particular, the accuracy of tree segmentation reached 95.61%, which was 1.31 percentage points higher than before improvement. This method can provide an effective decision for precision spraying and safe operation of the spraying robot, and it was practical.

Abstract:In order to meet the real-time detection requirements under the limited resources of embedded devices, a passion fruit detection model based on improved YOLO v5 lightweight network (MbECA-v5) was proposed. Firstly, MobileNetV3 was used to replace the feature extraction network, the depth separable convolution was used to replace the traditional convolution to reduce the number of model parameters. Secondly, the effective channel attention network (ECANet) was embedded to focus on the whole passion fruit. Point-by-point convolution connection feature extraction network and feature fusion network were introduced to improve the feature extraction ability and fitting ability of the network for passion fruit images. Finally, the transfer learning strategy combined with cross-domain and within-domain multi-training was used to improve the network detection accuracy. Experimental results showed that the accuracy and recall of the improved model were 95.3% and 88.1%, respectively. The mAP value of 88.3%,compared with the model before the improvement, it was increased by 0.2 percentage points. And the number of calculations was 6.6 GFLOPs. The model volume was only 6.41MB, which was about half of the improved model. The real-time detection speed in embedded device was 10.92f/s, the detection speed in embedded device was about 14 times,39 times and 1.7 times of SSD, Faster RCNN and YOLO v5s. Therefore, the lightweight model based on improved YOLO v5 greatly reduced the amount of calculation and model volume, and it can detect passion fruit in complex orchard environment efficiently on embedded devices, which was of great significance to improve the intelligent level of orchard production.

Abstract:There are several challenges for locust recognition, i.e., sample collection, small sample targets and multi-scale transformation in grassland locust images. A multi-scale grasshopper target detection and recognition model was proposed under complex background based on YOLO v5 network, which was used to recognize common grasshoppers in Ningxia grassland. To address the difficulty in sample collection, CycleGAN was used to expand the locust data set. Then, ConvNeXt was adopted to preserve the characteristics of small target locusts. Finally, Bi-FPN was utilized for neck feature fusion to enhance the capability of extracting locust features, which effectively solved the problem of large-scale transformation of locust photos. The experimental results showed that the best accuracy of the proposed model YOLO v5-CB was 98.6%, the mean average accuracy of the proposed scheme was 96.8%, and the F1 was 98%, which performed better than the Faster R-CNN, YOLO v3, YOLO v4 and YOLO v5. Using the improved model YOLO v5-CB, combined with the ecological environment collection equipment installed in Yanchi and Dashuikeng in Ningxia, a Web-based locust identification and detection platform was established, which had already been applied to grassland ecological environment data collection in Ningxia Yanchi Dashuikeng, Huangji Farm and Mahuang Mountain. This platform performed real-time tracking of locust in desert steppe of Ningxia, which can be further used for locust control in Ningxia. 

Abstract:Citrus Huanglongbing is known as the “cancer” of citrus, which seriously affects the yield and quality of citrus. Therefore, accurate detection of citrus Huanglongbing is of great significance for timely protection and management of citrus. However, in the natural background, there are problems of mutual occlusion and large size changes among citrus leaves, which makes the occlusion and small-sized leaves of Huanglongbing easy to miss. In addition, because the color and texture characteristics of the leaves of Huanglongbing are very similar to other diseases of citrus, there is a problem of false detection. Therefore, when the background is complex, it is difficult for the existing algorithms to accurately detect and identify the leaves of Huanglongbing. In response to the above problems, a natural background citrus Huanglongbing detection method was proposed based on shearing mixed splicing and two-way feature fusion. The method proposed used Cascade RCNN as the baseline network and used LabelImg to manually label the Huanglongbing samples in training and validation images. Firstly, in order to reduce the impact of complex background on the detection of Huanglongbing, the training set and validation set were augmented with the shearing mixed splicing method, mirror flips and rotations, which increased the number and diversity of background objects in the training set and validation set images. Secondly, deformable convolution was used to replace all standard convolutions in the backbone network Conv3~Conv5 to reduce the influence of irregular leaf shape and increase the effective receptive field and adaptively change the local sampling points in the detection of citrus Huanglongbing. Thirdly, in order to reduce the influence of the natural background on the detection results of citrus Huanglongbing and enhance the ability of the backbone network to extract the detailed features of the citrus Huanglongbing disease area, the global context block was used to enhance the feature map output by Conv3~Conv5 to establish an effective long-term distance dependence, so that the network can better learn the global context information. Finally, in order to reduce the influence of large changes in the size of the leaves of Huanglongbing on the detection results, two-way fusion feature pyramid networks was used to improve the information exchange path between shallow features and deep features, thereby improving the detection accuracy of small-sized blades. To verify the rationality and effectiveness of the method, in the training phase, the stochastic gradient descent strategy was adopted to train the network model. The initial learning rate was 0.02, the momentum was 0.9, the weight decay was 0.0001, and the number of iterations was 500. During the testing phase, the method proposed achieved 85.0% recall, 86.4% precision, and 84.8% average precision on the test set. The proposed method was compared with other detection algorithms (SSD, RetinaNet, YOLO v3, YOLO v5s, Faster RCNN, Cascade RCNN). Comparative experiments showed that the mean average precision of this method was 30.5 percentage points higher than that of SSD, 21.9 percentage points higher than that of RetinaNet, 13.2 percentage points higher than that of YOLO v3, 6.8 percentage points higher than that of YOLO v5s, and 20.1 percentage points higher than that of Faster RCNN, which was 3.2 percentage points higher than that of Cascade RCNN, and the detection result of this method was better than other classical deep learning methods.

Abstract:Rice disease is one of the important factors affecting rice yield. Early prediction of rice disease is very important for rice disease prevention. In order to realize the prediction of rice bacterial leaf blight disease, hyperspectral images of leaves under the stress of bacterial leaf blight disease were collected continuously for seven days from inoculation to early onset. The Savitzky-Golay algorithm was used to preprocess hyperspectral images, and the principal component analysis (PCA) and random forest (RF) algorithms were used to extract spectral features. The prediction model of multi-task learning (MTL) and long-short term memory (LSTM) network fusion was constructed to predict the incidence rate and incubation period of rice diseases. The MTL-LSTM model was optimized by using the whale optimization algorithm (WOA). The experimental results showed that PCA and RF can effectively extract spectral features from hyperspectral and reduce the dimension of hyperspectral images, and the performance of the prediction model based on spectral features was better than that of the prediction model based on full spectra. The modeling time of the former was about 98% lower than that of the latter. The prediction model constructed based on time series hyperspectral achieved the expected results in the prediction of the incidence probability and latency. The WOA-MTL-LSTM model, constructed based on the first ten characteristic wavelengths, achieved the best prediction performance. The R2 of the test set for the prediction of the incidence probability and latency was 0.93 and 0.85, the RMSE was 0.34 and 2.12, and the RE was 0.33% and 1.21%, respectively. The prediction performance of MTL-LSTM can be improved by WOA algorithm, and the R2 of disease probability and incubation period was increased by 0.05. The results indicated that RF extracted characteristic wavelengths can effectively characterize the full spectrum. The WOA-MTL-LSTM model based on time-series hyperspectral can accurately predict the incidence rate and incubation period of bacterial leaf blight disease, which provided technical support for the prevention of rice bacterial leaf blight disease.

Abstract:In the process of planting sandalwood trees on a large scale, there are problems such as low efficiency, high cost, and difficulty in the supervision of manual ranking of missing seedlings, and the necessary companion plants for each sandalwood tree and other crops interspersed between the trees, further deepening the difficulty of checking and replenishing. For these problems, a seedling deficiency detection and precise localization method in complex environment was proposed based on YOLOv4 algorithm and double regression strategy. Firstly, the YOLOv4 target detection model was used to achieve sandalwood plant detection from remote sensing images collected by UAV. Then the missing seedling localization algorithm (MSL) was constructed based on the double linear regression and extended column line fixing strategy: arbitrary sandalwood trees were selected as the benchmark, column regions were divided according to the pixel coordinates, and column lines were fitted to the sandalwood trees in each column region by using linear regression;for the omitted sandalwood trees that were not classified into columns after fitting, the attribution was judged again with the extended regression line strategy, and the column lines were optimized by linear regression again. Finally, the missing seedlings were calculated and localized according to the spacing at the time of planting. The results showed that the precision was 86.82%, the recall was 82.25%, the F1-score was 84.47%, and the running time was 8.19s, respectively. In summary, this method combined the rapidity of DJI UAV remote sensing image acquisition system, the accuracy of YOLOv4 algorithm and double regression strategy, which can be used to achieve realtime intelligent seedling deficiency detection and accurate localization of sandalwood trees under complex growth conditions.

Abstract:The size and depth of fruit tree roots can reflect the growth and health of fruit trees and affect the profits of the orchardist. However, the roots are more difficult to observe and sample than the subaerial parts of fruit trees, such as the tree trunk, branches, and crown. Ground penetrating radar (GPR), as an emerging non-destructive testing technology, has the advantages of simple operation and convenient carrying. However, using GPR to quantify the radius of the roots is still a challenging task. To that extent, a prediction method for tree root radius and depth was proposed based on GPR and convolutional neural networks. Firstly, the simulated one-dimensional data of ground penetrating radar (A-Scan) was used as the data set to train the model. Secondly, the attention mechanism allocated more weights to essential features, highlighting key features and speeding up convergence. Finally, the feature information was extracted through the convolutional layer. The local features learned by the previous convolutional layer were integrated into the global features of the A-Scan data through the fully connected layer to predict the root radius and depth accurately. The model was tested on simulation data and real data. In the simulation experiment, the maximum error of root radius prediction was 2.9mm, the coefficient of determination value was 0.990, the root mean square error was 0.00068m, the maximum error of root depth prediction was 11.2mm, the coefficient of determination value was 0.999, and the root mean square error was 0.0020m. In the field experiment, the maximum error of sample roots radius prediction was 1.56mm. The maximum error of sample roots depth prediction was 9.90mm. The total average relative error was 5.83%, indicating the proposed method’s efficacy for estimating the radius and depth of roots.

Abstract:In order to construct a rice fertilizer knowledge structure, based on the existing rice fertilizer unstructured data information, a rice fertilizer knowledge graph entity and relationship knowledge structure was proposed and designed, through which the existing rice fertilizer information in the network was stored in the knowledge graph as structured data;in order to extract a large amount of information to be stored in the knowledge graph, and at the same time, for the information extraction i.e., the existence of the overlapping triad problem, a rice fertilizer information extraction model based on RoBERTa-wwm coding + improved CASREL decoding was proposed, and the model was improved according to the characteristics of rice fertilizer data, and relevant experimental comparisons were conducted in coding and decoding, respectively. The results showed that the F1 value of this rice fertilizer information extraction model reached 91.86%, which was a significant improvement in extraction effect compared with the comparison model. Therefore, it can be concluded that the information extraction model based on the improved RoBERTa-wwm-CASERL can effectively improve the extraction effect of rice fertilizer information, which provided a basis for the next step of constructing rice fertilizer knowledge map and rice fertilizer decision system.

Abstract:In order to improve the quality of rice seed and eliminate weedy rice seeds, an adhesion segmentation algorithm based on concave point matching was proposed, and an online shape and color double choice rice seed recognition platform was built. The platform consisted of seed metering system, image acquisition system, transmission system and motor drive system. The algorithm of the platform was based on the concave point segmentation method of ECMM. Firstly, the collected image was preprocessed, and the adhesion contour with morphological factor less than 0.4 was extracted. The edge of the extracted contour was smoothed by one-dimensional Gaussian convolution kernel, and the curvature and mean curvature of the smooth contour curve were calculated. Several points that were different from the mean curvature were found as corners. Secondly, according to the positive and negative of the vector triangle area to determine whether the corner was a real concave point, the angle range (0°~180°) was found between the concave point and the normal direction composed of the preceding point and the successor point, and the matching concave point pairs in this angle range was found to complete the adhesion segmentation. The average accuracy of the algorithm was 92.90%, which was 19.82 percentage points higher than that of the limit corrosion method and 12.85 percentage points higher than that of the watershed algorithm. Finally, the length of seeds in each contour of the segmented image and the proportion of R channel pixels were calculated to identify weedy rice seeds. Through the identification platform test, the average time of 100 seeds per identification was 0.95s, and the average recognition accuracy was 97.50%.

Abstract:Aiming at the imprecise identification and positioning of crop seedlings and weeds, which would cause the problems of weeding robots unclean weeding, harming seedlings and affecting yield, a multi-stage image recognition method based on skeleton extraction algorithm was proposed, which realized the accurate identification and location of crop stem center through multi-level progressive fusion of different image algorithms. Firstly, the collected color images were converted to HSV color space for background segmentation. Then, the corrosion algorithm was used to corrode the image, which corroded the weed image information to obtain the image information only containing crops. Finally, the Zhang-Suen thinning algorithm was used to extract the skeleton of the crop image, and the skeleton intersection point was calculated and analyzed to identify and locate the center of the crop stem, so as to achieve accurate identification and positioning of crops. Experimental tests were carried out on 100 images collected at seedling stage. The results showed that the accuracy error of identification and positioning of stem center of crop seedlings was less than 12mm. The method presented can accurately identify the seedlings and weeds in real time and accurately locate the seedlings, providing an accurate and reliable method for crop identification and location for realizing the mechanization of agricultural plant protection operations such as weeding in the field.

Abstract:To realize the efficient, accurate and rapid automatic picking of brown mushroom, the identification, size measurement and positioning of mushroom are the key to the robot selective picking operation. An integrated method for in situ identification, measurement and location of brown mushroom was proposed based on YOLO v5 transfer learning (YOLO v5-TL) and dynamic diameter estimation based on 3D edge information. Firstly, YOLO v5-TL algorithm was used to realize rapid identification of brown mushroom under complex mycelia background. Then, the image enhancement algorithm, denoising, adaptive binarization algorithm, morphological processing and contour fitting algorithm were used to locate the edge of the mushroom image in the anchor frame area, meanwhile, the pixel coordinates of the edge point and the center point were extracted. Finally, the dynamic diameter estimation method based on 3D edge information was used to accurately measure the size and locate the center point of the mushroom. The experimental results showed that the average processing time of single frame image was 50ms. The average success rate of picking object recognition under low, medium and high light intensity was 91.67%, and the recognition rate reached 100% under high light intensity. The average measurement accuracy of mushroom cover was 97.28%. The results showed that the proposed YOLO v5-TL method combined with 3D edge information diameter dynamic estimation method can realize the integration of identification, measurement and location of brown mushroom under factory planting, which met the demand of automatic picking of brown mushroom by robot.

Abstract:Soil organic matter (SOM) is the leading factor of soil fertility and quality. Revealing the spatiotemporal variation and driving factors of SOM is the basis for studying sustainable use of cultivated land and food security. Geographic detector, geostatistics and center of gravity shift methods were applied to analyze the spatiotemporal variation distribution pattern and identify the driving factors of SOM content in Shaanxi Province. The results showed that the distribution of SOM showed a pattern of high in the south and low in the north in Shaanxi Province in 2017 overall, with an average content of 15.63g/kg, an increase of 8.61% compared with that of 2007. Spatially, the center of gravity of SOM content shifted southwest, southern Shaanxi shifted westward, Guanzhong shifted eastward, and northern Shaanxi shifted southwest in 2017 compared with that in 2007;STN content (q=0.74) was the leading driving factor of SOM content spatial variation, followed by county administrative division, municipal administrative division, annual precipitation, annual mean temperature, soil subtypes, and soil types, which q values were greater than 0.3 in 2017;during 2007—2017, the driving force of soil total nitrogen (STN) content, annual average temperature and total mechanical power on SOM content variation was increased significantly, q value was increased by 0.39, 0.21 and 0.18, respectively;from 2007 to 2017, natural and human factors jointly drove the spatiotemporal variation of SOM content, but human activities had an important impact on both factors.

Abstract:Soil oxygen content (SOC) is one of the important soil environmental factors that affect crop growth. It has the characteristics of time series, instability and nonlinearity. It can accurately predict the change trend of oxygen content in the soil environment, which is helpful to formulate a more reasonable soil aeration and oxygenation program. A prediction model based on the sparrow search algorithm (SSA) and long and short-term memory (LSTM) neural network was proposed, the meteorological environment and soil environment record data during the corn planting period were to recorded by using the equipment at the National Soil Quality Zhanjiang Observation and Experimental Station. The SSA-LSTM model predicted and analyzed the SOC changes, and it was compared with the traditional BP prediction model, LSTM prediction model, GA-LSTM prediction model and PSO-LSTM prediction model. The test results showed that the correlation between SOC and rainfall, soil water content, soil temperature and air-filled porosity was extremely significant, the correlation coefficient was higher than 0.8, the correlation with atmospheric temperature and wind speed was significant, and the correlation with atmospheric humidity and soil respiration rate was relatively significant. The prediction accuracy of the SSA-LSTM model was significantly higher than that of the other four groups of control prediction models. The R2 reached 0.95979, the RMSE was only 0.4917%, the MAPE was 3.7331%, and the MAE was 0.3620%. The degree of fit between the predicted value and the experimental value was high. The research result can provide theoretical support and scientific basis for the accurate prediction of soil oxygen content changes and the application and promotion of soil aeration and oxygenation technology.

Abstract:It is important to analyze the efficiency of agricultural water and land resources utilization in macro-regions from the perspective of synergistic inputs and “economic-social-ecological” benefits for the sustainability of agricultural production. The connotation of agricultural water and land resources use efficiency was clarified by combining the broad concept of water resources and the characteristics of “multiple inputs-multiple outputs” in agricultural production, the super slacks based measure (Super-SBM) model and super undesirable slacks based measure (Super-Undesirable-SBM) model were constructed by using data envelopment analysis to measure the production allocation efficiency. The Super-SBM model and Super-Undesirable-SBM model were used to measure the efficiency of agricultural water and land resources utilization of the concept. Agricultural water and land resources utilization efficiency without considering ecological benefits (WLUE), agricultural water and land resources utilization efficiency with considering ecological benefits (WLUEE), water resource utilization efficiency loss (WUEL) and arable land resource utilization efficiency loss (LUEL) were measured for 51 counties in the study area, taking the Shandong Yellow Diversion Irrigation District as an example. By comparing and analyzing the WLUE and WLUEE measurement results, WUEL and LUEL decomposition results, the characteristics of agricultural water and soil resource utilization and the size difference of the two resource utilization efficiency losses in each county of the study area were revealed, and the counties of the study area were classified into four types: green and efficient production type, ordinary efficient production type, green and inefficient production type and ordinary inefficient production type. The targeted improvement measures for agricultural soil and water resource use efficiency in each county and a perspective for the study of agricultural soil and water resource utilization efficiency were proposed. The research results were conducive to promoting the sustainable development of agricultural production in the study area.

Abstract:Soil organic matter (SOM) is an important part of soil fertility and the main nutrient source for crop growth. In order to explore the inversion effect of fractional-order derivatives (FOD) combined with spectral optimization index on SOM in low fertility areas, taking Yinchuan Plain as the study object, the original data of hyperspectral reflectance of field were processed by 0~2 order FOD (with an interval of 0.2 order) after log reciprocal transformation, the spectral optimization indices DI/RDI, DI/NDI, NDI/RDI, RDI/NDI, DI/GDI and RI/GDI were constructed, the two-dimensional correlation between each index and SOM content was analyzed, the optimal spectral optimization index was selected, and a support vector machine (SVM) model was established to inverse the SOM content. The results showed that the content of SOM in Yinchuan Plain was generally low, of which 93.05% was at the level of 4~6 class. There were obvious differences in the absorption characteristics of the original spectral reflectance of soil in the field, with obvious absorption peaks at 1400nm and 1900nm. With the increasing fractional order, the spectral reflectance was approaching 0. The maximum absolute correlation coefficient (MACC) values of soil DI/NDI, DI/GDI, RI/GDI, NDI/RDI and RDI/NDI were all less than 0.80 in order 0~2. The MACC values of DI/RDI in order 0.2~2.0 were ranged from 0.9965 to 0.9986, and their sensitive bands were mainly concentrated in 1450~1750nm and 2100~2400nm. The model inversion accuracy based on DI/RDI-SVM model was the best at order 0.2, modeling determination coefficient (R2c) and verification determination coefficient (R2p) were 0.98 and 0.99, and residual predictive derivation (RPD) got 4.31. The results can provide scientific basis for rapid and accurate estimation and mapping of SOM in areas with low organic matter content.

Abstract:Aiming to study the effects of different biochar species and their application amounts on soil water infiltration in newly reclaimed area, a vertical one-dimensional water infiltration experiment was conducted with two biochar species (corn stalk biochar A and rice husk biochar B) and three application amounts gradients (2%, 4% and 8%) and no biochar application (CK) in seven treatments. The results showed that except for low-application amount of rice husk biochar treatment (B2), the addition of biochar delayed the process of soil water infiltration in the newly reclaimed area, and corn stalk biochar was superior to rice husk biochar. The infiltration time of the treatments (A2, A4 and A8) with the addition of 2%, 4% and 8% corn straw stalk biochar was gradually increased with the application rate, and compared with that of CK, the infiltration time was increased by 35.0%, 46.0% and 59.1%, respectively. However, only the 4% application amount treatment (B4) in the rice husk biochar group delayed water infiltration, and the infiltration time was increased by 28.5% compared with that of CK. Meanwhile, the addition of biochar reduced the initial infiltration rate of soil and the migration distance and cumulative infiltration of wetting fronts within the same infiltration time, and the effects of biochar species and its application amounts on these three indicators were similar to the effects on infiltration time. Both biochar additions increased the soil surface water content by percentages ranging from about 2.2% to 20.3%, and the soil water retention capacity of both biochar was significantly better under the high application amount treatment conditions than the medium and low application amounts. The distance and time of wetting front migration were in accordance with the power function, and the Philip model can better simulate the water infiltration process of newly reclaimed soils under different species and application amounts of biochar treatment. In general, the treatment of corn stalk biochar addition at 8% was beneficial to improve the problem of rapid soil water infiltration and weak water retention capacity in newly reclaimed areas, which was a more recommended choice for rapid maturation and utilization of newly reclaimed areas.

Abstract:Corn straw cellulose triacetate (CTA) was used as raw material. Self-made ionic liquid phosphotungstate [PyPS]3PW12O40 (ILP) was used as hydrolysis catalyst. A green and efficient process for the preparation of cellulose diacetate (CDA) was proposed. Taking the degree of substitution and mass fraction of CDA as evaluation indexes, the effects of water addition, ILP addition and reaction time on the degree of substitution and mass fraction of hydrolysate were analyzed, and the physicochemical properties and structure of the products were characterized. The results showed that when CTA was 0.6g and reaction temperature was 110℃, the optimal hydrolysis conditions of CDA were as follows: water addition was 0.3g and ILP addition was 0.1g, that was, the mass ratio of ILP to water in hydrolysate was 1∶3, the mass ratio of raw materials to hydrolysate was 3∶2, and the reaction time was 60min. The degree of substitution of CDA was 2.62 and the mass fraction was 69.33%. The degree of polymerization CDA was 66.54, which can be dissolved in acetone, glacial acetic acid, dichloromethane, 1,4-dioxane and dimethyl sulfoxide. The physical and chemical properties of the product were characterized by Fourier transform infrared spectroscopy, scanning electron microscopy and thermogravimetric analysis.The results of scanning electron microscope (SEM) showed that the microstructure of CDA was rough and scattered, and the surface damage was serious. Fourier transform infrared spectroscopy (FT-IR) and thermogravimetric analysis showed that CTA was successfully hydrolyzed into CDA. The research can provide an idea for the preparation process of CDA. When corn straw was used as raw material, the mass conversion rate of raw material reached 47.72%, which was of great significance to the diversified utilization of corn straw.

Abstract:The internal and external quality of green plum has an important impact on its processing process. Conventional manual sorting not only has low classification efficiency, but also is difficult to realize standardized operation due to personal subjective factors, which can not meet the market requirements. In the aspect of defect classification, based on deep learning technology the vision transformer network was used in machine vision system, which introduced multihead self-attention to improve the global feature representation ability, and reduce the gradient through the softmax function to realize the detection and sorting of multiple categories (rot, crack, scar, spot and normal) on the surface of green plum. The results showed that the discrimination accuracy of rot, scar, crack and normal plum images reached 100%, spot reached 97.38%, the average discrimination accuracy was 99.16%, and the average test time of each group was 100.59ms. The discrimination accuracy and average discrimination accuracy of this network were significantly better than VGG and ResNet-18 network. In terms of internal quality (SSC) prediction of green plum, based on hyperspectral imaging technology, the LRTR-SCAE-PLSR prediction model of green plum was constructed by combining the denoising advantages of LRTR and the dimensionality reduction advantages of SCAE. The results showed that when the network scale was 119-90-55-36, RP was 0.9654 and RMSEP was 0.5827%. By comparing the two dimensionality reduction models of SCAE and LRTR-SCAE, LRTR-SCAE model not only had lower dimensions, but also significantly improved the correlation coefficient of prediction set, which verified the dimensionality reduction and denoising advantages of LRTR-SCAE model. An intelligent equipment for nondestructive sorting of internal and external quality of green plum was designed and built. The whole machine had small size and simple structure. The sorting results met the requirements of green plum deep processing.

Abstract:The quality and safety of grain and oil is directly related to people’s life and health and national security and stability. The grain and oil supply chain is characterized by complex subjects, multiple risks, cross domain supply network, and difficulty in getting through the information chain. New generation information technologies, such as blockchain, providing new solutions and application models for food quality safety assurance and traceability, but also introducing new systematic risks, and there are security challenges. Based on the analysis of the risks and information characteristics of the grain and oil quality and safety blockchain, the general-purpose blockchain structure in the existing completely untrusted execution scenario was improved and optimized. At the network layer, a special blockchain network structure suitable for grain and oil quality and safety in non-completely trusted execution scenarios was proposed, and a Kafka consensus optimization algorithm P-Kafka based on PBFT improved Byzantine fault tolerance and in line with the characteristics of grain and oil quality and safety blockchain was proposed at the consensus layer. The performance of P-Kafka was compared with the traditional consensus algorithm from the perspectives of correctness and decentralization, security, scalability, consensus efficiency and consistency. Through analysis and comparison, the network node partition and sub chain partition proposed saved the operation cost of blockchain system and improved the privacy security of nodes to a certain extent. The improved P-Kafka consensus algorithm had Byzantine fault tolerance and inherited the high throughput characteristics of Kafka partition optimization, making it more suitable for grain and oil quality and security application scenarios.

Abstract:In order to reveal the airflow characteristics in the leaf collecting pipe of a riding tea picker, the gas-solid two-phase flow in the pipe was numerically simulated by computational fluid dynamics (CFD) and discrete element method (DEM). The numerical calculation model of machine-picked fresh leaves was established by the multi-sphere polymerization method. On the basis of analyzing the movement law of fresh leaf particles, the different changes of inlet wind speed, fresh leaf particle size and bend structure were simulated and analyzed respectively. Through the numerical model the blade collecting effect and the optimal wind speed and feeding quantity of the riser pipe could be predicted. In the optimal inlet wind speed range, the larger the fresh leaf particles were, the more residual particles were in the pipeline, which was easy to produce deposition. Fresh leaf particle flow formed a bend curve when it passed through the vertical pipe to the bend. The bend structure had a certain influence on the movement of fresh leaf particles. The average velocity of flow field was decreased firstly and then increased. A rounded elbow with a radius of 0.04m was selected as the blade collector elbow structure, and the transverse pipe length was reduced. The inner length was 0.03m to avoid deposition caused by gravity action of fresh leaf particles and ensure the smoothness of blade collector. Through mathematical model simulation and experimental results, it was verified that the blade collecting pipe with rounded corner bend can meet the requirements of blade collecting by reducing the length of horizontal and straight pipe. The pipe with round corner elbow structure had a penetration rate of more than 86.8%. The fresh blade particle modeling method proposed was used for discrete element simulation analysis and structural optimization of the interaction between the leaf collecting pipe and the fresh blade flow. The research results provided a theoretical basis for the optimization of optimization design of the fresh leaf collecting pipeline.

Abstract:In order to realize the fast and accurate detection of chicken freshness, an integrated detection device based on electronic nose and vision technology was designed. The structure of the device was divided into three parts, including the control system, the vision system, and the electronic nose system. It simultaneously detected the concentration of gas emitted from chicken samples through the sensor array of the electronic nose, and obtained the visual images of chicken samples by the camera. The gas concentration data were firstly transmitted from the control board to the Jetson Nano board, and then was fused with the visual images for feature extraction and further analysis. Computational fluid dynamics techniques were used to simulate the velocity cloud and velocity vector diagrams of the device under suction conditions to verify the feasibility of gas flow. Based on the gas concentration and image data of chicken samples of different freshness obtained by the device, principal component analysis method was adopted for dimensionality reduction, and chicken freshness grading model was established using support vector machine method with an accuracy rate of 98.7%. The device has the characteristics of high accuracy, portability and stability, which can provide technical support for meat freshness detection.

Abstract:To achieve accurate prediction of rotary tillage quality based on tractor multi-sensor load data, a tractor rotary tillage quality identification model based on GAF-DenseNet was proposed, rotary tillage quality grading standard was designed, and field tests of rotary tillage were carried out, and model accuracy verification and performance analysis were conducted. The Gramian angular field (GAF) algorithm uniquely encoded the time series data while preserving the time dependence of the original load sequence. The DenseNet network deeply mined the load information embedded in the image array, and significantly improved the computing efficiency of this network while ensuring the depth of feature extraction through feature reuse, model compression, and other technical aspects. The analysis results showed that the model performance was reduced by either too large or too small a resampling sliding window size and the experimental effect of Gramian angular difference field (GADF) was stronger than Gramian angular summation field (GASF), and the experimental data showed that the model performance was optimal when the resampling sliding window size was 250 and the GADF algorithm was selected. The growth rate k tended to be positively correlated with the overall performance of the model, but too large a value of k reduced the real-time performance of the model and had limited improvement in accuracy, and the growth rate k was set to 24 in the experimental scenario to better meet the actual demand. The GAF-DenseNet model achieved accuracy and F1 value of 96.816% and 96.136%, respectively. It had good performance in real-time capability, and the interfence time can be as low as 16s. The overall performance of this model was better than the control group analysis results in the comparison tests with other intelligent algorithms.

Abstract:Six-degrees of freedom parallel mechanisms driven by linear motors can realize high-precision and wide-band movements, and it had broad application prospects in inertial unit calibration, vibration testing and other fields. In order to reduce the amplitude attenuation of 6-DOF parallel mechanism caused by linear motor drive, dynamic feedforward control was analyzed for the mechanism. Firstly, the parameter model of the 6-DOF parallel mechanism was determined, and then the vector method was used to analyze the kinematic of the mechanism. Secondly, the dynamics model of the parallel mechanism used Newton-Euler principle. The driving force relationship of the mechanism was obtained by simplifying the dynamics equation. The driving force simulation curve was obtained by numerical analysis, and the experimental platform was built to obtain the experimental curve of the device driving force, which verified the accuracy of the dynamic model. Based on the classical motion closed loop control system and the dynamics model, a dynamic feedforward control method was designed to reduce the motion error of a given trajectory. Finally, experimental analysis of mechanism’s motion error in traditional kinematic control was done, and the motion error of the mechanism was compared. The experiment results showed that after added the dynamic feedforward control to the 6-DOF parallel mechanism driven by linear motor, the kinematic errors of the mechanism were reduced by 55.5%, 54.2% and 59.8%, when the mechanism carried out sinusoidal motion in X, Y and Z axes.