Abstract:The data acquisition of cultivated land quality index (CLQI) is the basis of cultivated land quality (CLQ) evaluation. Remote sensing (RS) provides a new method for rapid and large-scale monitoring of CLQ data. Firstly, the connotation and function of CLQ were analyzed. On this basis, bibliometric method was used to summarize the research of CLQI in recent five years. Combined with the research status of CLQ and soil quality, CLQI system was established based on RS, which was divided into three dimensions: topographic conditions, soil properties and field utilization status. Secondly, the research status of various index acquisition methods in different dimensions was analyzed, and the commonly used RS analysis methods and corresponding technical principles were summarized. The basic data can be effectively obtained for field slope, field condition, field road accessibility, forest network degree and other indicators, and the large-scale acquisition method of soil properties were needed further research. Finally, aiming at the problems to be solved in RS monitoring of CLQ, the following suggestions and prospects were put forward: mining the remote sensing characteristics of different scales of CLQI; strengthening the research on automatic extraction of remote sensing information of CLQI; and building a remote sensing big data platform for CLQ evaluation, so as to promote the application of RS in CLQ evaluation.

Abstract:In the current Leader-Follower AGVs cooperative operation, in addition to obtaining environmental information, the positioning and navigation of the following AGV also need to observe the position and attitude of the leading AGV for path tracking, which has higher requirements for accuracy and stability of navigation and location. In order to improve the navigation accuracy of following AGV, an integrated navigation method combining inertial navigation and multi-vision was proposed. Aiming at the problem of multi-sensor data fusion, an optimal pose estimation method based on adaptive unscented Kalman filter was proposed. The output signal of inertial navigation sensor was used to follow the AGV attitude prediction;the path tracking navigation and RGB-D navigation constituted the multi-vision navigation, which was used as the system observation to correct the accumulated offset of inertial navigation. The experimental results showed that the compound navigation scheme had faster convergence speed, more stable path tracking state and formation maintenance.This method improved the real-time performance and robustness of the two AGVs cooperative handling system.

Abstract:In response to the current needs of farmland obstacle detection in the automatic driving of agricultural machinery, a method of using threedimensional LiDAR to detect field obstacles was proposed. Firstly, the collected environmental point cloud was preprocessed. The voxel grid down-sampling method was used to filter the dense point cloud without losing feature information. A bounding box was used to segment the region of interest for fast calculation. The random sample consensus algorithm (RANSAC) was used to detect the farmland ground, and the ground point cloud was removed from the whole point data so that the obstacle points were extracted. Then the obstacle point cloud was clustered by Euclidean distance based on the K-d tree, and the distance threshold of clustering was 0.6m in this test. Finally, the size of the cluster and the volume of the circumscribed cuboid were judged, and invalid clusters that were too large or too small were filtered out to obtain obstacles. A LiDAR with 32 channels was used to collect field obstacle point cloud at National Experiment Station for Precision Agriculture in Beijing Xiaotangshan. The algorithm was used to detect agricultural implement, haystack, field ridge, low houses, roadside trees, and field pedestrian. The test showed that the algorithm was suitable for the field common obstacles detection. When detecting pedestrians in the field, the people crossed the front view of the LiDAR and the distances from the LiDAR respectively were 5m, 10m, 15m, 20m, 25m and 30m to test the effect of the algorithm at different distances. The results showed that the average detection rate of dynamically walking people in the field within 30m was 96.11%. This algorithm can be used to detect obstacles in the field environment and can provide a basis for the research of obstacle avoidance strategies in agricultural machinery autonomous driving.

Abstract:Aiming to automatic guidance operation of the combined direct seeding unit for rapeseed in the scene of unmanned farm, a generally applicable path planning algorithm for rapeseed direct seeding operation in convex polygon boundary field was put forward. Based on the basic requirements of path planning algorithm for rapeseed direct seeding operation, the operation area was divided into internal direction parallel operation area and external contour parallel operation area, which were called direction parallel operation path and contour parallel operation path. The operation path was further divided into working path and non-working path, and the generated principles of the latter, including interline connection path, area connection path, corner connection path, entry path and exit path were introduced in detail. The optimal operation direction of the direction parallel path was obtained by minimum span method, and the operation sequence of the direction parallel path was optimized by greedy algorithm or Google OR-Tools. Considering the complete length of operation path, algorithm time consuming, repeated and missed area, aiming at the no-tillage or rotary tillage seeding operation of the combined direct seeding unit for rapeseed, taking Dongfanghong-LX804 tractor + 2BFQ-6 no-tillage direct seeder and other units as the object, the performance of algorithm was tested in four convex polygon boundary fields with different operation parameters, ground turning strategies and scheduling strategies. The test results showed that the coverage of rapeseed was more than or equal to 95.14%, the repeat rate was less than or equal to 2.63%, and the percentage of effective working path was more than or equal to 57.39%, the algorithm took time less than or equal to 8.003ms, which was proved stable and efficient, and satisfied the basic requirements of path planning for rape direct seeding operation.

Abstract:With the popularization and development of solar greenhouse in recent years, more and more crops began to be planted in solar greenhouse. At present, in the greenhouse garden, picking tomatoes mainly relies on fruit farmers to complete manually, which not only consumes a lot of work force, but also wastes a lot of time. Taking tomato as an example, a tomato picking robot was designed which can be applied to solar greenhouse garden, patrol between greenhouse and ridge road and automatically identify ripe tomato in real time, and complete picking and collecting after identifying successful tomato. STM32 microcontroller was the main core of this design. The Mecanum wheel omni directional mobile platform was used as the mobile chassis of the robot, and the depth camera driven by Raspberry Pi 4B controller was used as the recognition device of mature tomatoes. The platform on the chassis was equipped with a wind compensation fan and a 6-DOF horizontal sliding mechanical arm. The arm was equipped with a flexible grip attached to a thin-film pressure sensor. The whole upper platform was driven by the vertical lifting structure installed on the platform under the chassis. The sliding mechanism and the lifting mechanism were equipped with ranging sensors. Through the recognition and capture of the camera driven by Raspberry Pi 4B, the coordinate data of the ripe tomato was transmitted to the STM32 controller through the serial bus. The STM32 controller drove the joint action of the mechanical arm, slide rail and lifting platform through inverse kinematics analysis of the mechanical arm, and can pick tomatoes at different heights and depths with the action of the end joint of the mechanical arm.

Abstract:Under the background of conservation tillage, the advantages of less tillage and no-tillage cultivation techniques, such as water storage and soil conservation, have been emphasized. In order to solve the shortcomings of full tillage, a belt tillage machine was designed for seedbed preparation, and its key components were studied. According to the existing wide and narrow planting row spacing of rice, belt land preparation with 30cm bandwidth was realized by selecting rotary blade and designing blade arrangement. Meanwhile, no-tillage operation was carried out in the adjacent 30cm area to realize the purpose of fine land preparation between paddy beds. After one operation, seedling transplanting could be carried out in the water field. The orthogonal rotation test method was used to simulate the actual field operation through EDEM discrete element simulation, and the optimal arrangement of strip rotary blades was obtained. At this time, the simulation results were as follows: the soil crushing rate was 93.26%, and the surface flatness was 17.3mm. On this basis, field experiments were carried out, and the results showed that the soil crushing rate, surface flatness, single tillage bandwidth, and tillage bandwidth stability of multi-functional combined tillage machine were 92.97%, 18.4mm, 300.9mm, 95.3%, respectively. The stability of tillage depth, shallow subsoil depth, tillage depth and deep subsoil depth were 151.3mm, 221.4mm, 93.5% and 92.3%, respectively, which fully met the needs of largescale rice seedbed soil preparation. The research was focused on determining the structural parameters of strip rotary tillage, ensuring the stability of tillage bandwidth and good operation results. It provided reference for the design and research of less and no-tillage equipment for paddy field.

Abstract:Aiming at the problem of lack of accurate and reliable discrete element simulation parameters for the research on the interaction mechanism of soil-rotary tillage components with sloping soil on the Loess Plateau and the optimized design of special rotary tillage equipment, taking the typical slope clay loam soil with common moisture content of 13.4%±1% on the Loess Plateau as the research object, the relevant parameters were calibrated based on Hertz-Mindlin with JKR Cohesion contact model in EDEM. Firstly, the contact parameters between soil particles were calibrated. The simulated accumulation angle of soil particles was used as the response value, the regression model of soil accumulation angle was obtained based on the Box-Behnken optimization method in the Design-Expert software. The regression model was optimized by using the measured accumulation angle as the target, and the optimization results of restitution coefficient, static friction coefficient, rolling friction coefficient and surface energy parameters between soil particles were 0.15, 0.33, 0.05 and 9.04J/m2, respectively. Under the optimal parameter combination, the simulation value of soil accumulation angle was 41.59°, and the relative error with the measured value was 3.8%. Then, the contact parameters between soil and rotary tillage knife material 65Mn steel were calibrated. The numerical range of the above three parameters between soil and 65Mn steel was obtained by static friction test, oblique plate test and impact test. Based on this, the static sliding friction angle of soil on 65Mn steel was obtained based on Box-Behnken optimization method. The regression model was optimized by using the measured static sliding friction angle as the target, and the optimization results of static friction coefficient, rolling friction coefficient and restitution coefficient between soil particles and 65Mn steel were 0.50, 0.06 and 0.18, respectively. Under the optimal parameter combination, the simulation value of static sliding friction angle was 24.0°, and the relative error with the measured value was 1.7%. Finally, through the comparative analysis of field test and simulation test of slope rotary tillage, it was concluded that the maximum relative errors of horizontal and lateral displacement of soil particles were 4.3% and 5.1% respectively, within the acceptable range. It showed that the calibration results and research methods of discrete element simulation parameters were accurate and reliable.

Abstract:Under the corn-soybean rotation no-tillage planting mode, the corn root-soil combination is large and the ridge distribution is uneven, which hinders the ridge movement of the double-row sowing opener on the soybean ridge, resulting in the uniformity of the seeding depth and the consistency of the row spacing. For poor problems, a double-row sowing opener on a sawtooth soybean ridge was designed. Based on the description of the structure and working principle of the double-row sowing opener on the zigzag soybean ridge, the rake curve was designed, and the rake curve equation was deduced;the mechanical model of the side cutting root system was analyzed, and the side cutting was determined to be sawtooth. The agronomic measurement of the corn root-soil combination determined the key parameters of the sawtooth side edge;according to the agronomic requirements of soybean seeding, it was clear that the seed guide tube in the main body of the opener was arranged in a staggered manner, and the sawtooth opening was determined. The gap angle of the ditch device was 5°. The results of the discrete element simulation verification test showed that the sawtooth opener can effectively cut the corn root system and had better working stability, which verified the rationality of the design idea. Under the conditions of a forward speed of 7km/h and a trenching depth of 50mm, the field performance test was conducted with the root cutting rate as the main indicator. The results showed that the serrated side edge had a strong sawing ability, and the average root cutting rate was 97.25%. Field comparison experiments with double-disc openers showed that compared with double-disc openers, the coefficient of variation of trenching depth was reduced by 43.33%, and the coefficient of variation of transverse seed spacing was reduced by 60.81%. Both the simulation test and the field test showed that the sawtooth opener met the agronomic requirements of double-row sowing on no-till soybean ridges.

Abstract:Improving the prediction accuracy of working state of no-tillage seeder and cleaning the invalid data timely will improve the data quality and reduce the cache pressure of monitoring equipment. However, as the agricultural machinery moved back and forth in the farmland, monitoring equipment captured a large number of invalid images at both ends of the farmland or when the vehicle stopped. These images affected the accuracy of farmland operation quality evaluation and created congestion in transmission network. A data cleaning method based on multi-condition time series, mainly vehicle speed, seeding rate and instantaneous area, was proposed to deal with the periodic change of long time series of agricultural machinery in the farmland. The model included multiple long-short term memory (LSTM) and spatiotemporal feature channel fusion (CONCAT connect) to maintain the individual difference under multi-condition. The current time sequence state of the agricultural machinery working condition can be predicted, and the real-time cleaning state of the image capture system can be indirectly acquired. Due to screen and capture valid image from captured image every three minutes by cleaning state, the system achieved the maximum efficiency in transmission channel and memory space. The comparison results of different models after 40 iterations showed that the prediction accuracy of this method for both valid and invalid samples was over 85% and the average accuracy of image cleaning was 92.4%. The data cleaning results showed that about 63% of the redundant data was removed after data cleaning. Therefore, the research method took the working condition of no-tillage seeder as the basis of image cleaning was effective, which had high research value and application prospect.

Abstract:Considering the problems that the existing drill or large row spacing hole-drop planter could not meet the agronomy requirements of American ginseng for narrow-row-dense-planting precision planting, and the mechanized planting of American ginseng lacked suitable planting machinery, a narrow-row-dense-planting precision planter for American ginseng was designed, which adopted multi-row parallel pneumatic needle seed metering device and planet-wheel dibbling seed guiding device. The structure principle of the planter and seed metering device was expounded and the transmission scheme of the whole machine was designed. The planet-wheel dibbling full-constrained seed guiding device was designed emphatically. The structural parameters of planetary gear systems were determined through mechanism analysis. The theoretical analysis and high-speed photography experiment of seed unloading process were carried out to determine the horizontal displacement and influencing factors of seed unloading trajectory, and determine the opening size of inserting apparatus of seed guiding device and the appropriate pressure of seed unloading. The theory and simulation analysis of seed guiding trajectory were carried out to determine the motion law of inserting apparatus and the low-position zero-speed dropping condition, and determine the profile of dropping control cam of seed guiding device. The results of field validation experiment showed that when the negative pressure of seed suction was -4.5kPa, the positive pressure of seed unloading was 3.0kPa and the working speed was 0.54km/h, the qualified rate of seed per hole was 86.2%, the multiple rate was 4.4%, the seed-missing rate was 9.4%, the qualified rate of seed depth was 92.8%, and the qualified rate of hole distance was 93.9%, which met the agronomy requirements of American ginseng planting.

Abstract:The sparse planting and transplanting of plug seedlings is a key step in the cultivation of seedlings in facility agriculture, which can provide an excellent growth environment for seedlings and increase production and income. Aiming at the problems of low automation degree of adjustable plant spacing equipment, inefficient manual work, vulnerable to seedling damage and injury in the process of thinning and transplanting, a plant spacing adjustable seedling end-effector was designed based on cylindrical cam, which can realize thinning and transplanting between different plant spacings. First of all, the overall structure of the end-effector was designed and its working principle was determined. Secondly, through theoretical analysis, the key parameters of the cylindrical cam and the finger were determined, and the stress condition under operation was analyzed. Thirdly, EDEM and Recur Dyn were used to establish a composite model of seeding bowl root-soil for coupling simulation single factor simulation test, and the factor range of subsequent orthogonal test was determined. Finally, a transplanting test platform for plug seedlings was built, and orthogonal tests were carried out with the seedling needle included angle, penetrating angle, seedling needle spacing and variable pitch speed as test factors, and the seeding bowl deformation quantity and the transplanting success rate as test indexes. The experimental results showed that the primary and secondary influence order of test factors on the seeding bowl deformation quantity were penetrating angle, variable pitch speed, seedling needle included angle, seedling needle spacing. And the primary and secondary influence order of test factors on the transplanting success rate were seedling needle spacing, penetrating angle, pitch change speed, seedling needle included angle. The verification test was carried out under the optimal parameter combination with seedling needle angle of 10°, penetrating angle of 4°, seedling needle spacing of 8mm, and pitch change speed of 5mm/s, and the plug seedlings were transplanted from 128cell to 72cell and from 72cell to 50cell respectively. The average seeding bowl deformation quantity of 128cell was (1.13±0.68)mm, and the average seeding bowl deformation quantity of 72cell was (1.51±0.64)mm. The total success rate of transplant was 93.33%, and the transplanting efficiency of the whole machine was 22 plants/min. It satisfied the needs of different plug tray specifications and had strong applicability. The research result can provide theoretical basis and technical support for the research of the thinning and transplanting operation device.

Abstract:In order to solve the problems such as easy winding and hanging of the reel of rape windrower, a double crank plane five-bar reel mechanism was proposed, and the kinematic analysis model of the reel mechanism was established. According to the constraints of the double crank plane five-bar mechanism, based on the compiled reel finger movement analysis program, a set of structural parameter combinations of the reel mechanism meeting the requirements of rape cutting reel was obtained by human-computer interaction. The movement requirements of the reel’s finger for low-speed entry, accelerated push, fast delivery and finger upturned was realized. Compared with the eccentric reel, the dynamic trajectory of the reel was increased at the same turning radius and revolving speed. Height and width of the reel buckle of the reel finger dynamic trajectory were increased. The influence of the initial structure angle of the reel and the advancement speed of the machine on the motion state, speed, acceleration and trajectory of the reel finger was analyzed. The analysis showed that the speed of the reel finger was changed at a non-uniform speed during a rotation period, the reel finger speed was changed slowly during the reeling process, and there was no sudden change in acceleration. By reasonably matching the reel rotation speed and the forward speed of the machine, zero-speed lifting of the grains could be realized, the impact on the rape head was small, and the twine was reduced. The comparative analysis showed that the ratio of the reel linear velocity of the double crank plane five-bar mechanism reel can be obtained by approximately calculating the circumferential linear velocity of the finger tip static trajectory with the reel axis as the center and the minimum inscribed circle radius as the length at a constant rotated speed. A comparative test of double crank plane five-bar mechanism reel and cam-action reel for rape swathing was carried out. Field experiments showed that the laying angle and angle difference was respectively less than or equal to 25°and 8°of the windrower with double crank planar five-bar mechanism reel under the same test conditions. The effect of the reel speed on the laying angle was not significant （P=0.165>0.1), and the difference of the reel type had an extremely significant effect on the laying angle （P=0.002<0.05). The reel speed had no significant influence on the laying angle difference （P=0.750>0.1), the reel difference had a significant influence on the laying angle difference（P=0.015＜0.1). There was no hanging and entanglement of the double crank plane five-bar mechanism reel in the whole test. The research result can provide the basis for the improvement of swathing laying regularity and reducing entanglement and hanging of rape windrower.

Abstract:In order to study a suitable intelligent root cutting device for garlic combined harvesting, a non-contact bulb root cutting method with machine vision was proposed, and a garlic root cutting test bench based on a deep convolutional neural network was designed afterwards. Specially, the test bench adopted a deep learning theory to perform target detection on the collected images, through using the APP software in Matlab to complete the human-computer interaction. Then, the results presented that the deep convolutional neural network could determine the cutting position of the garlic root, and the motor control system could adjust the position of the double disc cutting automatically, ensuring the root cutting process completed by the root knife. Target comparison tests showed that bulb (availability rate was 94.79%, confidence score was 0.97697) was suitable for detecting, among the three kinds of bulb, root plate and garlic root. Comparison tests of detection models performed with ten models based on Faster R-CNN, SSD, YOLO v2, YOLO v3 and YOLO v4. The improved YOLO v2 model combined the detection speed and accuracy (the detection time in the test program was 0.0523s, and the confidence score was 0.96849), where ResNet50 was selected as the feature extraction network;by using the improved YOLO v2 model, the root cutting test took bulbs as the targets (the confidence score was 0.97099, the availability rate was 96.67%, the qualified rate of cutting roots was 95.33%, and the detection time in the APP was 0.0887s), can meet the requirements of garlic combined harvesting and cutting roots.

Abstract:Aiming to solve the problem of that the high-stalk grass harvester was prone to tipping when operating on the southern mountain slopes, an anti-tip warning system was designed. The system was mainly composed of a host computer, a controller and sensors. Through the joint monitoring of the inclination sensor and the pressure sensor, the anti-tip warning function was realized. The center of mass of the whole machine was tested. The horizontal distance between the center of mass of the whole machine and the drive wheel shaft was 944.5mm. The height of the center of mass was 1080.4mm. The distance between the center of mass and the longitudinal center plane was 27.9mm. The position of the center of mass was determined. The software RecurDyn was used to perform the dynamic simulation analysis of the whole machine under typical working conditions at a speed of 7.2km/h. The critical tipping angles for horizontal straight, turning and longitudinal up and down slopes were 24°, 11°, 33° and 31°. Under the turning conditions, the critical tipping lateral load transfer ratio was 0.49. A static tipping test was performed on the whole machine. The critical tipping angles for horizontal straight and longitudinal up and down slopes were 24.7°, 34.2° and 31.8°. Based on the size of the whole machine and the position of the center of mass, a scale model was built. A dynamic tipping test was performed under a fixed steering path. The test results showed that the critical tipping angle of steering was 14° and the critical tipping lateral load transfer ratio was 0.54. The research can provide theoretical support for high-stalk grass harvesters applied to operations under complex working conditions.

Abstract:Breaking of walnut shell with less damage remains to be a challenge because walnut kernel is crisp. The existing equipments which were used to break shell and take kernel cannot meet the high quality and mass requirements of walnut kernel. Aiming to resolve the above gaps, the cam rocker bidirectional extrusion self-positioning walnut shell breaking device were designed. By analyzing the critical condition of crack fracture, the mechanism of crack propagation and the displacement of breaking walnut shell, the double-screw quantitative feeding mechanism and cam rocker bidirectional extrusion breaking mechanism were designed. The size of the extrusion/blanking U-shaped block structure, the motion angle and displacement of the cam mechanism were determined, the rocker displacement equation was determined and the cam profile curve was designed. L9(34) orthogonal test was carried out with extrusion clearance, camshaft rotation speed and walnut circumference size as test factors, and one kernel rate, two kernel rate, nibs rate and hulling rate as test indexes. The results showed that the primary and secondary order of the various affecting factors were the walnut circumference size, extrusion clearance and camshaft rotation speed, and the optimized combination was as follows: the walnut circumference size was 34~37mm, extrusion clearance was 7mm, camshaft rotation speed was 53r/min. Taking the optimal combination for verification experiment, the results showed that one kernel rate was 61.39%, two kernel rate was 23.30%, nibs rate was 15.31% and hulling rate was 92.36%. Compared with the traditional walnut sheller and kernel extractor, the high kernel rate was increased by 14.69 percentage points, which met the walnut industry primary processing needs.

Abstract:For the automatic control of the traditional unmanned autogyro, a combination of one-degree-of-freedom PID and optimization algorithm is used, because the onedegreeoffreedom PID cannot solve the problem of fast response and interference suppression at the same time due to the response lag characteristics of autogyro. A two-degree-of-freedom PID controller based on the particle swarm algorithm and the simulated annealing algorithm is proposed, and a noise reduction method of feedforward controller is designed to achieve effective control of unmanned autogyro. The feasibility is verified by building a Matlab simulation model. The simulation results show that the response speed of the algorithm in this study is faster than that of the traditional one-degree-of-freedom PID control. The feedforward controller, with strong robustness and system response can effectively suppress external disturbances. As the result, the speed is fast and the response time is increased by about 11%. The control system is more stable with higher response accuracy and slighter convergence error value. The possibility of error is about 1/6 lower than the traditional PID. At the same time, the actual flight experiments in two different flight environments verified the feasibility of the two-degree-of-freedom PID controller based on PSO-SA, which can provide a theoretical basis for the application of unmanned autogyro in the field of agricultural aviation.

Abstract:There are few studies on the design and optimization of liquid fertilizer sprayer, a trailed liquid fertilizer sprayer was designed, which mainly composed of tank, suction system and discharge system. The theoretical model of spraying operation was established, and the key parameters such as vacuum pump displacement, tank structure and nozzle outlet flow rate were designed and calculated. The key factors such as the shape of scattering baffle, orientation of scattering baffle, shape of nozzle and flow velocity of nozzle outlet were simulated and analyzed. The analysis showed that the ideal spraying effect can be achieved by selecting fan-shaped scattering baffle, positive angle between scattering baffle and horizontal plane, circular nozzle and outlet velocity of about 13m/s. Taking the radiation width, uniformity variation coefficient and average thickness of spraying as the spraying performance indexes, the spraying operation was optimized and simulated by orthogonal experiment. It was concluded that the significant factor affecting the spraying width was the longitudinal axis length of the scattering baffle, and the significant factor affecting the horizontal average thickness of the ground was the inclination angle of the scattering baffle, each factor had little effect on the coefficient of variation of spraying uniformity. Combined with the design goal, when the traveling speed was 5km/h, the included angle between the scattering baffle and the horizontal was 35°, the longitudinal axis length of the scattering baffle was 32cm, the outlet flow rate was 13m/s and the nozzle height was 1m. The test spraying width was 11m, the average ground thickness was 1.65mm, and the spraying uniformity variation coefficient CV was 34.86%. The simulation analysis was basically consistent with the test results, which verified the accuracy and reliability of the design. The design goal was achieved, the application requirements were met, and some data support was provided for the spraying research of liquid fertilizer.

Abstract:In order to further explore the effectiveness potential of the inclined opposite disk side throwing device in the aspects of throwing uniformity and throwing distance, the shooting flow shape was described qualitatively and quantitatively by using the parameters such as the angle of shooting, scattering angle, and effective throw distance, etc. Through the theoretical analysis, the mathematical models between each parameter and the structural parameters of the throwing device were established. Then, through these models, it was obtained that factors such as the speed of the discs, the inclination angle of the discs and the shape of the scraper had an influence on the shooting flow shape. Based on the simulation results, the wide bottom extended inclined quadrilateral fertilizer scraper, throw-off angle control baffle and other parts were innovatively designed, and the ejection effect of the shooting flow was tested by the prototype test. The results showed that when the height of the throw-off angle control baffle was 140mm and the leading out angle of the upper deflector was 40°, the throw-off angle of the shooting flow can be controlled at about 30°, the innovative design of wide bottom extended inclined quadrilateral fertilizer scraper can reduce the scattering angle to 25°, a narrow shooting flow was obtained to reduce the upwind side of the shooting flow, and the effective throw distance was 11.6m. Under the joint action of this part and the throw-off angle control baffle, a flexible cylinder without guide obstruction was formed, and the throwing effect was better. The research effectively controlled the shooting flow of the spreader by innovating components and optimizing structural parameters, it provided theoretical basis and basic data for improving the operation ability of the side throwing device with inclined opposite discs.

Abstract:In order to research the hydraulic stability of axial flow pump in the double hump region under small flow conditions, the relationship between the head-flow curve of the axial flow pump and the characteristics of the internal flow field was discussed, and a test and research on an axial flow pump with a specific speed of 825 was carried out. The internal and external hydraulic characteristics of the impeller under small flow conditions were obtained through the characteristic energy curve test, the internal flow field tracer particle trajectory was captured by a high-speed camera, the wall pressure pulsation collection, and the impeller inlet and outlet section velocity measurement by LDV equipment. According to the results of energy characteristics, it was found that there was a double hump phenomenon in the head-flow curve of the axial flow pump. The first hump appeared between 0.5QBEP and 0.6QBEP, corresponding to the tangential fluctuation of the speed on the side of the inlet flange, the rapid increase of the velocity circulation, and the extreme peak-to-peak pressure pulsation. The test results showed that the first hump was related to the generation of the impeller inlet reflux. The second hump occurred between 0.33QBEP and 0.4QBEP, the test results showed that the second hump was related to the drastic changes in the flow field of the impeller outlet. The impeller inlet reflux was generated during the second hump formation, although the strength was increasing and the influence range was wider, the pressure pulsation peak-to-peak value on the wall of the inlet side of the impeller chamber was no longer increased. The pressure pulsation peak-to-peak value on the wall of the outlet side of the impeller chamber was increased rapidly, and the unevenness of the velocity distribution between the outlet hub and the blades was obviously deteriorated, the reflux area of impeller outlet was increased, which indicated that the second hump was related to the drastic changes in the flow field of the impeller outlet.

Abstract:Precision control of the concentration of fertilizer is an important part of the integrated control system of water and fertilizer. It is the foundation to realize precision fertilization by detecting the electric conductivity (EC) value of fertilizer and controlling it in a reasonable range. Based on Internet of things technology, the remote control system was designed and developed, self-tuning fuzzy PID control algorithm was introduced into the equipment in the terminal, and the fuzzy PID control algorithm was used to control the local end frequency of variable frequency pump, and in turn to realize the precise control of EC value, and the local end PID and remote end fuzzy PID control algorithm were validated through the contrast test. The results showed that the larger the target EC value was, the more accurate the steady-state EC value was, but the steady-state time and overshoot were increased. Compared with the traditional local PID control, the response speed of the system was fast, the EC fluctuation range was small and stable. When the target EC value was 2.5mS/cm, the steady-state time and overshoot reached the maximum, which were 120s and 20.8%, respectively. The mixing time and measured EC value could meet the actual demand of water and fertilizer control. The research not only realized the remote fuzzy PID control of EC value, but also realized the irrigation data monitoring and switching control of the computer, mobile phone and WeChat multi-terminal of the irrigation and fertilization system.

Abstract:Due to the unbalanced distribution of cultivated land resources in the Beijing-Tianjin-Hebei region, the regional differences of social and economic development are prominent. It is of great significance to clarify the spatial and temporal heterogeneity of cultivated land for the coordinated and sustainable development of the region. Based on the Beijing-Tianjin-Hebei region, the evolution of cultivated land spatial distribution and its influencing mechanism during the period of 1990—2020 were studied by using the dynamic change model of cultivated land and geo-detector methods. The results showed that Beijing took the lead in the construction of industrialized cities, followed by Tianjin, and Hebei lagged behind. The relatively high dynamic value of cultivated land change also appeared one after another. The distribution of cultivated land had a significant agglomeration characteristic, showing a kernel density pattern of “dense in southeast and sparse in northwest”. The characteristics of land use structure change were obvious. Construction land, woodland and water area were the main areas of net increase, while cultivated land, grassland and unused land were the main areas of net decrease. The loss of cultivated land mainly showed the aggregation characteristics around Beijing and Bohai Sea, and the heat of loss in the east was higher than that in the west, and that in the south was higher than that in the north. The influence of socio-economic and policy factors on cultivated land use showed an upward trend, while the influence of natural factors showed a downward trend. The results can provide a scientific basis for farmland protection and regional coordinated development in Beijing-Tianjin-Hebei region.

Abstract:In order to improve the prediction accuracy of winter wheat yield in large scale region, taking remote sensing data, meteorological data, soil moisture data of Henan Province from 2005 to 2019 as characteristic variables, the correlation between them and wheat yield was analyzed. The importance of characteristic variables was analyzed based on random forest algorithm. And a wheat yield prediction model was established by means of fusing multi-source spatio-temporal data. The results showed that enhanced vegetation index (EVI), solar-induced chlorophyll fluorescence (SIF) and elevation was an important factor for remote sensing estimation of wheat yield, which was highly positively correlated with wheat yield. The importance of EVI, SIF and elevation to wheat yield exceeded 0.45, far greater than soil moisture, rainfall, maximum temperature, minimum temperature and other factors. The yield prediction model based on random forest algorithm and constructed with the wheat growth stage from October to next May and October to next April as the characteristic variables had higher accuracy, coefficient of determination (R2) were 0.85 and 0.84, and respectively, the root mean square error (RMSE) were 821.55kg/hm2 and 832.01kg/hm2. The prediction relative errors in hills and mountains of western and southern Henan was higher than that in plain areas. The research results provided a reference for large-scale crop yield.

Abstract:Topography is an important factor that affects soil erosion, which is usually measured by slope gradient and slope length (LS) in erosion estimation models, and extracted based on digital elevation model (DEM) on a vast area. SRTM, as currently high-quality and easily accessible elevation data on a vast area, has been applied in global soil erosion evaluation. However, the traditional algorithm for topographic factor extraction requires that the unit of elevation identical with cell size (usually meters), which makes SRTM need perform coordinate transformation before extraction. Aiming at the problem of high cost in performing coordinate transformation on SRTM data in a large area, an algorithm for extracting terrain factors directly was proposed based on SRTM (LSA-SRTM). The longitude and latitude information of the geographic coordinate system was used to directly calculate the cell size and the unit slope length. D8 method was used to acquire slope gradient and flow direction matrix. Then, the slope gradient cutoff was calculated according to the slope gradient result, the catchment area was calculated and the channel network cutoff was set, furtherly, the cumulative slope length was obtained by “forward-reverse traversal”. Finally, the LS factor was calculated according to the slope gradient, cumulative slope length and the segmentation formula of CSLE. Using Himmelblau-Orlandini mathematical surface and 1″SRTM of five typical samples in China as the data source, the LSA-SRTM method, the projected coordinate system DEM-based LS extraction algorithm (LSA-DEM) and manual measurement method were compared. On the mathematical surface and the typical sample area, the R2 of the slope length of LSA-SRTM method and the measured value were 0.8552, 0.7788, 0.7269, 0.7024, 0.6909 and 0.7255. The R2 of the LS and the measured value were 0.8907, 0.8209, 0.8213, 0.7142, 0.7145 and 0.8212. In terms of execution time, the LSA-SRTM method had high efficiency. The experiment results showed that the LSA-SRTM had higher accuracy and efficiency, which can provide a support for the study of topographic factor extraction in vast areas.

Abstract:Anoplophora glabripennis is widely distributed in China as the main pest of forest trees, and the distribution area is constantly expanding. It is important to study the spatial correlation of insect infestation and analyze the factors that affect its distribution. Using geodetector, the effects of environmental factors on the Anoplophora glabripennis disaster were analyzed. The results showed that the most seriously damaged areas were in the northern part of Shanxi Province, while the incidence rates in Henan, Shandong, and southern Hebei were relatively low. The main meteorological factors affecting the spatial distribution of the incidence rate were precipitation and temperature, and the main social and economic factors were population and regional GDP. The effect of temperature on the value of the first and second industries was significantly different from that of other factors. Economic development had significant positive effect on the control of local insect pests. The interaction detection showed that the combination of any two variables can explain the spatial heterogeneity of Anoplophora glabripennis disaster more effectively. The interpretation ability of incidence rate was increased significantly after interaction between precipitation and control rate.

Abstract:Plastic-film-mulching has made an outstanding contribution to agricultural production and food security in China, but also caused many serious environmental problems. It is very important to quickly and accurately obtain the spatial distribution information of plastic-mulched farmlands. In order to establish a framework for remote sensing recognition of plastic-film-mulched farmland, the Tuanjie Town of Dingxi City in Gansu Province was chosen as the research area, which was a typical dry farming agricultural area with heavy plastic film application on the Loess Plateau. Based on the Google Earth Engine, Landsat-8 reflectance data was used to analyze the importance of different features and select the optimal textural features. Then, the random forest (RF) algorithm with optimized parameters was used to extract the plastic-film-mulched farmland area and select the best feature combination. Finally, based on the best feature combination, the performance of RF algorithm was evaluated through comparison between the classification results based on the other algorithms of support vector machines (SVM), decision tree (DT), and minimum distance classifier (MDC), respectively. The results showed that the optimized parameters of RF algorithm could greatly improve the classification accuracy. Among the schemes based on single kind of features, the accuracy of scheme based on spectral features was the highest. The addition of index and textural features could also improve the overall identification accuracy to some extent. The performance of the selected optimal texture features was better than that of all texture features. The recognition result based on the combination of ‘spectral + index + optimal textural features’ was the best, whose overall accuracy and Kappa coefficient were 95.05% and 0.94, respectively. The overall accuracy of RF algorithm was 3.10 percentage points, 7.74 percentage points and 50.78 percentage points higher than the algorithms of SVM, DT and MDC, respectively, which proved the RF algorithm had some obvious advantages in recognition of plasticfilmmulched farmlands. The research realized an accurate identification of plasticfilmmulched farmlands in areas with complex terrain features in China. The results can provide theory and technology supports for the studies related to spatial variations and sustainability production with plastic film mulching in the near future.

Abstract:In view of the inefficiency and lag of the traditional artificial diagnosis and classification methods for grape downy mildew, an improved residual network model for grape downy mildew identification and disease degree classification was proposed. The images of downy mildew in the prophase, metaphase, anaphase and healthy leaves were collected in the field, and the effects of the factors of weather, shooting angle and equipment noise were simulated to increase the data capacity. Based on the characteristics of high similarity and difficult to distinguish between leaves with different disease degrees, by using the optimized ResNet-50 model, a 3×3 maximum pool layer with step size of 2 was added into the Base Block of Conv3, Conv4 and Conv5 (the residual body composed of several residual blocks) to solve the problem of serious information loss of the shortcut branch and the insufficient feature extraction ability of the main branch, so as to achieve dimensionality reduction of retaining important information. The main branch structure of the residual block in the ID Block was improved, and the 1×1 dimensionality reduction convolution layer in the first layer was replaced by 3×3 dimensionality reduction convolution layer with a step of 1;a newly full connection layer was designed, in which the global average pooling and 3 layer full connection layer network were used to replace the original model full connection layer structure, and the Dropout (random inactivation) layer was added to avoid the model over fitting. The experimental results of the original data set and the expanded data set showed that when the momentum factor m was 0.60 and the learning rate α was 0.001, the improved ResNet-50 network model had the best recognition effect compared with ResNet-34/50/101, AlexNet, VGG-16 and GoogLeNet. The improved residual block enhanced the feature extraction ability of the network. On the basis of optimizing the super parameters, the accuracy of the improved residual block was 2.31 percentage points higher than that of the original model. Different data augmentation methods had certain contribution to improve the recognition accuracy of the model. The recognition accuracy of the improved residual network model was 4.68 percentage points higher than that of the original model, reached 99.92%, which provided a real-time and accurate solution for automatic classification of grape downy mildew disease degree in complex environment.

Abstract:To allow fast and automatic detection of the same semantic agriculture-related questions, a method based on BERT-Attention-DenseGRU (gated recurrent unit) was proposed. According to the agriculture question characteristics, twelve layers of the Chinese BERT model method were applied to process and analyze the text data and compare it with the Word2Vec, Glove, and TF-IDF methods, effectively solving the problem of high dimension and sparse data in the agriculture-related text. Each network layer employed the connection information of features and all previous recursive layers’ hidden features. To alleviate the problem of feature vector size increasing due to dense splicing, an autoencoder was used after dense concatenation. The experimental results showed that agriculture-related question similarity matching based on BERT-Attention-DenseBiGRU can improve the utilization of text features, reduce the loss of features, and achieve fast and accurate similarity matching of the agriculture-related question dataset. The precision and F1 values of the proposed model were 97.2% and 97.6%. Compared with six other kinds of question similarity matching models, a state-of-the-art method with the agriculture-related question dataset was presented.

Abstract:Phenotyping aims to measure traits of interest and a key part of this requires the accurate identification of defined parts of the organism. Instance segmentation of organs, such as leaves, is a crucial prerequisite for plant phenotyping. Firstly, whether deep learning methods (such as Mask-RCNN) had generality for leaf and stem segmentation was evaluated. Training was conducted using four datasets about three plants, a public Arabidopsis dataset (CVPPP2014), and three developmental multi-view datasets from Arabidopsis, maize, and physalis. Multi-view images of given plants were collected at different developmental periods. The Arabidopsis datasets contained only leaf, and the maize and physalis datasets were different from the Arabidopsis datasets, having clearly distinct leaf, stems, and petioles. The results showed that the mean accuracy precision (mAP0.5) of the Mask-RCNN model for Arabidopsis in the public datasets which was in the same growth period reached 85.3% and the mean intersection over union (mIOU) was 73.4%. The mean accuracy precision was more than 70.0% across different growth periods of Arabidopsis, maize, and physalis. The mean intersection over union was more than 60.0% across different growth periods of Arabidopsis, which indicated that Mask-RCNN displayed satisfying versatility for plant phenotyping and had high value for plant phenotyping. The results showed that the model had competitive advantage compared with previous plant segmentation algorithms. Furthermore, taking advantage of multi-view images, a leaf tracking method was presented to solve the problem of plant occlusions. It was helpful for the leaf counting and leaf area calculation of plants. The results showed that the proposed methods had a superior performance compared with other existing plant segmentation algorithms, and was promising to build a dynamic modeling for various plants during their entire growth cycles.

Abstract:Air quality plays an important role in mutton sheep breeding environment, in order to reduce the stress response of CO2 to the growth of large-scale mutton sheep and ensure the healthy growth of mutton sheep in the appropriate environment, the key is to accurately control the CO2 in the mutton sheep breeding environment. A CO2 prediction model of mutton sheep breeding environment was proposed based on light gradient boosting machine (LightGBM), sparrow search algorithm (SSA) and extreme learning machine (ELM). Firstly, LightGBM was used to screen out the important characteristics of carbon dioxide concentration and reduce the input dimension of the prediction model. Then, ELM neural network algorithm with single hidden layer with strong nonlinear processing ability was used to build the CO2 prediction model. Finally, through the sparrow intelligent optimization algorithm, the super parameters needed in ELM model were optimized to obtain the best prediction model. The prediction model was applied to a large-scale mutton sheep breeding base in Manas County, Changji Hui Autonomous Prefecture, Xinjiang Uygur Autonomous Region, and good prediction results were obtained. The experimental results showed that the prediction model had good prediction effect, and the root mean square error (RMSE) of ELM was higher than that of SVR, BPNN, LSTM, GRU and LightGBM. The RMSE, mean absolute error (MAE) and R2 were 0.0213mg/L, 0.0136mg/L and 0.9886, respectively. The results showed that the combined model can not only achieve accurate control of carbon dioxide in sheep house, but also meet the needs of fine decision-making for mutton sheep breeding. It also can help farmers make decisions and reduce farming risks.

Abstract:The objective was to clarify the effects of the snowmelt on cropland soil moisture in spring of the high latitude cold region in China. The Hulan River Basin, which locates at the middle Heilongjiang Province, was adopted as the study area. In consideration of the deficiencies of spatial and temporal resolution of monitoring station data, remote sensing data and land surface model assimilation data, the hydrological processes such as snowfall, snow cover, and snow melting were simulated using soil and water assessment tool (SWAT). Research on spatial and temporal characteristics of snowfall, snow cover and snowmelt and their effects on the cropland soil moisture in spring of the Hulan River Basin were carried out. The results indicated that the spatial distributions of the average snowfall, the maximum daily snow cover, snowmelt within the Hulan River Basin showed a similar characteristic that gradually increased from west to east;the change rates of snow cover, snowfall and snowmelt exhibited the similar spatial distribution characteristics and these change rates showed increasing trend at most areas within the basin;the close relationships existed among the series of annual snowfall, snow cover and snowmelt, that were similar to the interannual variation trend of soil moisture, and the interannual variation of cropland soil moisture was greatly affected by precipitation. A related significant correlation existed between the snow cover in a period from the last November to the current March, the snowfall in the last November, snowmelt in current April, and the cropland soil moisture in April and May of the current year. The snowmelt water recharged the soil moisture from the late March to the early April, which could promote the shortterm increase of soil moisture, and the amount of snow cover determined the replenishment of snowmelt water to cropland soil moisture;the change trends of soil moisture were gently decreased before snowmelt, whereas significantly decreased after snowmelt ended.

Abstract:Aiming at the problem of soil salinization in salinized irrigation areas，taking Wulate Irrigation Area in the lower reaches of the Hetao Irrigation District, Inner Mongolia as the research area，through the combination of field test and indoor test analysis, the redundancy analysis method was used to explore the correlation between salt segregation elements and total salt, pH value, and nutrients in the root layer soil (0~20cm, 20~40cm) in spring and autumn on the salinized improved cultivated land and wasteland，and its change characteristics and quantitative relationship were clarified, and the salt migration of 1m soil between the improved cultivated land and the wasteland in the experimental area was estimated. The results showed that the anions of the improved farmland and wasteland soil were mainly Cl-, accounting for 45.27% and 58.78% of the total anions, respectively, and the cations were mainly Na+, accounting for 60.67% and 53.94% of the total cations respectively. The total salt content of wasteland exceeded 7.0g/kg on average, and the degree of soil salinization was relatively serious. Redundancy analysis showed that SO2-4, Cl- and Ca2+ played the leading role in improving the total salt content of cultivated soil, and Cl-, Mg2+ and Na+ played the leading role in the total salt content of wasteland soil. The change of soil pH value was mainly related to HCO-3. Improved cultivated land had a significant negative correlation between available phosphorus and Na+ (P<0.05), wasteland available potassium had a significant positive correlation with K+ (P<0.05), and available phosphorus had a significant negative correlation with Na+ (P<0.05). In the first two ranking axes, the cumulative proportion of changes in the relationship between the cultivated land research object and the environment was 71.38% and 71.65%, respectively. The cumulative proportions of changes in the relationship between wasteland research objects and the environment were 89.02% and 89.16%, respectively, reflecting the linear relationship between the research objects and environmental variables of the improved cultivated land and wasteland. In cultivated land, Na+ had the most important impact on total salt, pH value, and soil nutrients, followed by Ca2+;in wasteland, Na+ had the most important impact on total salt, pH value, and soil nutrients, followed by K+. In one growth cycle of crops, the amount of salt accumulated in 1m of wasteland soil was 939.842kg/hm2. Groundwater replenishment brought salt into the soil, accounting for 70.06% of the salt accumulation. The salt transferred from the cultivated land groundwater to the wasteland during the irrigation period was concentrated in the 60~100cm soil and accounted for 22.83% of the salt accumulation. The amount of salt that the cultivated land penetrated into the wasteland was 7.11% of the salt accumulation.

Abstract:In order to study the effects of biochar application on soil structure, water holding capacity, maize yield and sustainability of sloping farmland in black soil area, four consecutive years of biochar effect experiments were carried out in the runoff plot of 3° sloping farmland in black soil area from 2015. A total of five biochar application rates C0(0t/hm2), C25(25t/hm2), C50(50t/hm2), C75(75t/hm2) and C100(100t/hm2) were set. The results showed that soil bulk density was decreased with the increase of biochar in the four years, and porosity was increased gradually. Appropriate amount of biochar can effectively reduce the proportion of soil solid phase and improve the proportion of gas phase and liquid phase. In addition to 2015, the generalized soil structure index (GSSI) was increased first and then decreased with the increase of biochar application rate for three consecutive years, and the three-phase structure distance index (STPSD) was decreased first and then increased with the increase of biochar application rate, reaching the optimal values (99.96,0.63) in the third year of C50 treatment. At the same time, the soil three-phase ratio R was the smallest (1.03), and the threephase ratio was the closest to the ideal state. For four consecutive years, the aggregate content R0.25, mean mass diameter (MWD) and geometric mean diameter (GMD) were increased first and then decreased with the increase of biochar application rate. Adding 50t/hm2 biochar for three consecutive years had the best effect on improving soil stability. The saturated moisture content was positively correlated with carbon application rate for four consecutive years. In addition to 2015, the field water holding rate was increased first and then decreased with the increase of carbon application rate, and reached the optimal treatment at C100, C50, C50 and C25, respectively. The peak value was 37.33% at C25 in 2018. The change rule of soil effective moisture content and field water holding capacity was the same. Soil water storage in 0~60cm soil layer of maize at different growth stages was increased first and then decreased. Soil water storage in 60~100cm soil layer was negatively correlated with carbon application rate. The maize yield sustainability index (SYI) reached the maximum at C50 (0.954), and the coefficient of variation (CV) was the lowest at C100 (0.022). Annual application of 50t/hm2 biochar to promote maize yield stability and sustainability effect was the most obvious.

Abstract:Chemical amendment is an effective measure to improve soil water and salt environment under saline alkali stress and alleviate land secondary salinization in arid areas of Northwest China. The water and salt transport characteristics and infiltration model parameters of medium saline-alkali soil at different biochemical fulvic acid (BFA) application rates (0g/kg, 1g/kg, 2g/kg, 4g/kg and 8g/kg) were investigated through the one-dimensional vertical infiltration test. The results showed that after the infiltration, compared with no application of BFA, the cumulative infiltration amount at the BFA application rates of 1g/kg, 2g/kg, 4g/kg and 8g/kg was increased by 1.00%, 4.67%, 7.14% and 3.44%. The soil water infiltration rate was decreased first and then increased with the increase of BFA application, respectively, the average volumetric water content of soil profile was increased by 8.90%, 17.70%, 20.41% and 11.67%, respectively. In addition, at the 0~20cm soil layer, the average desalination rates were 5.29%, 27.04%, 42.77% and 14.74%, respectively. At the same time, the Philip model, Green-Ampt model and algebraic model could well simulate the soil water infiltration regulation at different BFA application rates, and there was a functional relationship between model parameters and BFA application rates. Compared with no application of BFA, the soil infiltration rate S and the soil saturated hydraulic conductivity Ks were decreased, and showed a trend of first decrease and then increase with the increase of the BFA application amount, and the minimum value was obtained at application rate of 4g/kg. The suction Sf at the wetting front and comprehensive shape factor α were increased first and then decreased, and the maximum value was obtained at application rate of 4g/kg. Therefore, the application amount of BFA significantly affected the characteristics of soil water and salt transport, and the application amount of 4g/kg had obvious effects on water conservation and desalination of medium salinealkali soil in Xinjiang.

Abstract:Basic greenhouse is a kind of primary greenhouse without operation equipment. At present, more than 90% of greenhouse in China belongs to basic greenhouse. In order to solve the problem of evapotranspiration simulation in basic greenhouse, Penman-Monteith (PM) model, which is the most widely used mechanism model in greenhouse, was used to simulate the evapotranspiration in basic greenhouse. Bayesian method was used to estimate the parameters and analyze the uncertainty of six PM models composed of two key parameters ra (aerodynamics resistance) and rc (canopy resistance). Three precision indexes, namely mean relative error (MAE), determination coefficient (R2) and Wilmot consistency index (D), were used to evaluate the model accuracy for the model calibration year (2018) and test year (2019). The results showed that the Bayesian parameter estimation method made the PM evapotranspiration model with high accuracy in the basic greenhouse day scale estimation, and five of the six PM models met the three accuracy indexes in the model calibration year and model test year. The Bayesian parameter estimation method can effectively reduce the uncertainty of some parameters in the basic greenhouse day scale PM model. In the model ra2-rc3, the posterior distribution interval of parameter a2 was reduced by 97.65%, 92.38%, 93.31% and 98.24%, respectively in the four growth stages compared with the prior distribution interval. In the model ra2-rc2, the posterior distributions interval of parameters D50, KQ, Q50 and gmax were reduced by 96.44%, 56.08%, 97.78% and 99.75%, respectively compared with the prior distribution interval. Two PM models suitable for daily scale evapotranspiration simulation of basic greenhouse were screened out: ra2-rc3 and ra3-rc3, the optimal model was ra2-rc3. The research result can provide theoretical basis for improving crop irrigation and water use efficiency in basic greenhouse.

Abstract:The cosmic-ray neutron sensing method is a mesoscale and non-contact method for measuring soil moisture, which has been widely studied and applied. However, whether the accuracy and the applicability of the measurement results was consistent under the influence of typical meteorological conditions have not been clearly confirmed. The seasons were divided into time scales and typical meteorological periods were selected to study the difference of continuous observation effect between cosmic-ray neutron sensing (CRNS) and frequency domain reflection(FDR) under mountainous terrain. The result indicated that the seasonal variation of CRNS horizontal footprint was relatively stable under mountainous terrain, while the vertical footprint was fluctuated within the tillage layer. Different precipitation levels in different seasons was the main factor causing CRNS measurement deviation, during the period of severe soil moisture change, such as the continuous loss of soil moisture caused by high temperature and little rain and the process of precipitation of larger magnitude in summer and autumn, the results of CRNS and FDR were consistent. The spring precipitation was also smaller magnitude, canopy interception results in slightly different consistency between CRNS and FDR. In winter, precipitation and evapotranspiration were basically balanced, and the single precipitation was smaller magnitude, CRNS and FDR had a minimal deviation. The variation of CRNS measurement accuracy under typical soil moisture conditions was analyzed to further explain the influence of precipitation on the accuracy of CRNS, while the consistency was good in the period of continuous high temperature and drought, the root mean square error（RMSE）, Nash-Sutcliffe efficiency coefficient（NSE）and Kling-Gupta efficiency coefficient（KGE）was 0.014m3/m3, 0.925 and 0.919, respectively. The research results can provide a scientific basis for the selection of application scenarios and data quality control of CRNS.

Abstract:Aiming to reveal the synergistic effect of various factors on cropland evapotranspiration (ET) and net CO2 exchange (NEE) in the Huang-Huai-Hai Plain, a field experiment based on an eddy covariance technique was conducted at a typical winter wheat-summer maize field in Yucheng City, Shandong Province. Based on daily observation data of water and carbon fluxes from 2003 to 2010 and the structural equation model, the characteristics and influencing factors of ET and NEE were systematically analyzed. The results showed that the ET and NEE of different photosynthetic plants (e.g., wheat and maize) had significant differences. Seasonal variations of ET generally showed bimodal curves, with the first peak (in May) higher than the second one (in August) in each year. The ET in the wheat season fluctuated between 315.56mm and 499.54mm, with a multi-year average of 398.63mm, whereas it fluctuated from 221.68mm to 314.95mm in the maize season, with a multi-year average of 256.59mm. The significant trends of ET were found in these two different growing seasons (P<0.05). Compared with the ET, the NEE also showed a double-peak curves, with the first peak (in April) lower than the second one (in August) in each year. The NEE in the wheat season fluctuated between -367.84g/m2 and -149.93g/m2, with a multiyear average of -272.57g/m2, whereas it fluctuated from -469.63g/m2 to -118.79g/m2 in the maize season, with a multi-year average of -293.57g/m2. The net radiation was an important factor affecting seasonal changes of the ET and NEE in this field, which were mainly found to impose direct effects on the ET and NEE. As for indirect effects, radiation and temperature had great indirect effects on the ET and NEE through the leaf area index in the wheat season, whereas the indirect effect of VPD were larger in the maize season. Additionally, other environmental factors, e.g., the soil water content and wind speed had different effects on the ET and NEE in different growing seasons.

Abstract:The agricultural drought vulnerability dynamic assessment is of great significance for the sustainable development of regional agriculture. In order to analyze the development trend of agricultural drought vulnerability evaluation subsystem and indexes, and identify the main indexes that affect the change of regional agricultural drought vulnerability, an adjoint function of connection number, which was the five-element semipartial subtraction set pair potential was constructed. Then, a dynamic evaluation model of regional agricultural drought vulnerability was established, application results of Bengbu City showed that, in the agricultural drought vulnerability evaluation system of Bengbu City, the overall trend of exposure subsystem from 2001 to 2010 was fluctuated between the symmetrical potential and the partial inverse potential, and it was developed towards the reduction of set pair potential. The agricultural drought vulnerability showed an increasing trend, the main index that had a great impact on the vulnerability in Bengbu City was multiple cropping, and it displayed a trend to the inverse potential. Furthermore, the development of damage sensitivity subsystem was fluctuated greatly, from 2001 to 2010, it was gradually developed from the partial inverse potential to the partial identical potential, and the evaluation results showed a positive trend. The decrease of unit added value for agricultural water consumption made an obvious improvement of agricultural drought vulnerability in Bengbu City. Moreover, in 2001—2009, the disaster prevention ability subsystem was in the partial inverse potential. After 2009, it was gradually changed to the symmetrical potential. Meanwhile, the vulnerability showed a slow improvement trend, and the primary influence factors were water-saving irrigation rate and per capita income of farmers. The research provided a scientific decision basis for improving the regional agricultural drought disasters vulnerability, and also provided an effective way to deal with the multi-level problems of complex systems.

Abstract:It is vital to promote the comprehensive utilization of straw for enhancing the ability of agriculture and rural areas to reduce carbon emissions and fix carbon. Based on the IPCC Guidelines (2006) greenhouse gas (GHG) emission accounting theoretical framework, and in view of the problems of unclear base and unclear bases of GHG emission factors in the comprehensive utilization of straw, an evaluation method for the comprehensive utilization of straw was constructed, the scope and boundary of the evaluation was clarified, the GHG emission factors were scientifically calculated and the carbon-reduction base of different straw utilization technologies were evaluated. Based on different scenarios, the emission reduction and carbon sequestration potential of comprehensive utilization of straw by 2030 and 2060 was predicted. The results showed that the GHG emission reduction contribution of comprehensive utilization of straw in 2020 was 70 million tCO2e. Among them, the GHG emission from open burning and natural decomposition was 5.6×107tCO2e, and the GHG emission reduction from comprehensive utilization of straw was 1.26×108tCO2e. The emission reduction of straw fertilizer and energy utilization were the highest, with emission reduction of 7.9×107tCO2e and 3.8×107tCO2e, respectively. It was estimated that the GHG emission reduction and carbon sequestration contribution potential of comprehensive utilization of straw by 2030 and 2060 would be 1.52×108~1.72×108tCO2e and 2.20×108~2.73×108tCO2e, respectively. The contribution of straw fertilizer to emission reduction in comprehensive utilization was the highest and continued to increase steadily, and it was estimated that the contribution to emission reduction in 2030 and 2060 would account for 40.7% and 31.9%, respectively. Straw energy utilization had the greatest emission reduction potential, and it was estimated that the contribution to emission reduction in 2030 and 2060 would be 55.2% and 62.8%, respectively. Therefore, the dual carbon reduction effects of soil/forest carbon sinks and fossil energy substitution should be paid attention in comprehensive utilization of straw. Adhere to the principle of agriculture-oriented and diversified utilization, increasing efforts to promote full-scale utilization, reducing GHG emissions from incineration, and making full use of straw in reducing carbon emissions, can provide support for achieving the goal of carbon neutrality.

Abstract:Photosynthesis directly affects the quality and growth of blueberries, and the rate of crop photons is mainly influenced by temperature and photon yield density. At present, most greenhouse control did not consider the coordination of light temperature, and the actual energy consumption of greenhouses, not only resulting in meaningless waste of energy, but also creating a greenhouse small-climate environment which reduced the efficiency of blueberry photosynthesis. In order to solve the above problems, considering the photosynthemum and greenhouse cooling energy consumption during the spring and summer when blueberry was in flower fruit period, and the temperature and lighting control value of greenhouse were optimized by multi-target optimization algorithm. Firstly, the blueberry photosynthing rate model with temperature correction was established, which was based on the results of the temperature and photon pass density nesting test. Using a right-angled bi-curve correction model with temperature correction to model the blueberry photosynthetic rate. The model fitting results had a coefficient of determination (R2) of 0.9836, an average square root error of 0.5701μmol/(m2·s), and an average relative error of 3.86%, which can better reflect the relationship between blueberry photosynthing rate and temperature and light. Then a greenhouse energy consumption model was established, and the optimal solution of Pareto was solved by using NSGA-Ⅱ multi-objective optimization algorithm with greater net photosynthing rate and energy saving as the optimization goal. In order to further illustrate the optimization effect, different selection strategies were adopted for the optimization solution, which can reduce the energy consumption by about 21.3% while maintaining the photosynthetic rate of blueberries basically unchanged;under the premise of giving priority to planting benefits, the energy consumption can be reduced by 8.6% while the average increase of the photosynthetic rate by about 28.9%. The results can provide a theoretical basis for analyzing the physiological characteristics of crops and optimizing the greenhouse light temperature regulation setting. Greenhouse decision makers or control algorithms can use this method to set greenhouse temperature, light regulation settings. At the same time, the research method can also be applied to the setting value optimization of other crops which missing yield models in greenhouse production.

Abstract:Nearly two-thirds of China’s population uses rice as their staple food. The safety of the rice supply chain is directly related to the health of the people and the prosperity of the country. Rice supply chain has the characteristics of long life cycle, complex participation roles of main links, many harmful species, multidimensional and multi-source heterogeneity of information and so on. The realization of dynamic supervision of rice supply chain is conducive to the guarantee and traceability of rice quality and safety. Driven by blockchain smart contract, a dynamic model suitable for the information flow characteristics of rice supply chain was constructed, and contractual implementation and prototype verification were carried out. Firstly, based on the analysis and classification of rice supply chain supervision information, a dynamic supervision model framework of rice supply chain was constructed based on blockchain smart contract. Secondly, under the logical framework of the regulatory model, six types of smart contracts, including initialization smart contract, data calling smart contract, model verification smart contract, smart contract for data transmission, contribution evaluation smart contracts, and credit assessment smart contract, were custom-designed. And the smooth operation of the model was analyzed. Finally, based on the dynamic supervision model and smart contract, the rice supply chain supervision prototype system was designed and developed. And simulation analysis and case verification were carried out. The results showed that the dynamic supervision model and prototype system constructed can solve the whole process real-time management of rice supply chain business information, hazard information and personnel information, and realize the dynamic and credible supervision of rice supply chain in the whole life cycle at the information level. The research presented a generation of information technology in the efficient coordination and resource sharing of food supply chain, and provided ideas for the digital transformation of the grain industry.

Abstract:In order to solve the problems of export mixed internal quality and not easily to detect of walnuts in China, X-ray imaging technology combined with convolution neural network was proposed to quickly detect the internal quality of walnut. Using X-ray transmittance, X-ray images containing internal information were obtained. Firstly, X-ray images of walnut were preprocessed and data expanded. Then, four transfer learning models, including GoogLeNet, ResNet 101, MobileNet v2 and VGG 19, were used to construct convolutional neural networks to train walnut data sets. The model was analyzed through prediction set accuracy, loss value, test set accuracy and running time, and the model parameters were optimized. Finally, the walnut internal quality detection and sorting system was developed and applied to model verification. The results showed that among the four different transfer learning models, GoogLeNet model had the highest prediction accuracy. When the learning rate of GoogLeNet model was set to 0.001 and the epoch was set to 25, the prediction effect was the best, and the prediction accuracy was 96.67%. The results of system verification showed that the discriminant accuracy of shell walnut reached 100%, and the average discriminant accuracy was 96.39%. The system could realize the non-destructive testing and sorting of walnut internal quality, and provide further theoretical basis and technical reference for the equipment research and development.

Abstract:Computer vision is one of the commonly used technical methods in the field of cotton seed detection. It has been widely used in the field of non-destructive inspection of agricultural products. However, in most cases, it is difficult for researchers to use computer vision to detect small-sized objects such as cotton seeds on both sides. The detection effect is not good. Aiming at this problem, a type of cotton seed detection and sorting device was designed, which used the transparent characteristics of the acrylic plate under strong light and white background to slide the cotton seed into the groove of the transparent acrylic plate through the feeding device. With the rotation of the turntable, the front and back images of the same batch of cotton were collected by two CCD cameras at different positions. The improved YOLO v4 target detection algorithm was used to detect damaged cotton seeds. The experimental results showed that the model established by this method can detect damaged and intact cotton seeds in the population cotton seeds with an accuracy of 95.33%, recall rate of 96.31%, and missed detection rate of 0. The detection effect was better than that of the original YOLO v4 network, respectively. The proposed method realized the identification of the damage of double-sided group cotton seed, and provided technical support for the subsequent research and development of related delinted cotton seed intelligent detection equipment.

Abstract:With the aim to reveal the interaction between soybean protein isolates (SPI) and dextran (Dex) in the mixed system and the effect of dextran concentration (1%, 3%, 5% and 7%) on the structure and functional properties of SPI, fluorescence spectrum, ultraviolet spectrum and Fourier transform infrared spectrum were used to characterize the conformational changes of the SPI-Dex non-covalent polymer, and the effect of dextran concentration on the functional properties of SPI was analyzed through particle size, surface hydrophobicity, turbidity, solubility, emulsification activity, emulsification stability and antioxidant activity. The results showed that SPI and Dex can interact through two noncovalent forces: hydrophobic interaction and hydrogen bond under neutral conditions, thereby changed the structure and functional properties of SPI. The addition of Dex can prevent the exposure of tryptophan and tyrosine residues, and form a tighter tertiary structure compared with SPI alone. When the addition of Dex in the mixed system was less than 5%, with the increase of Dex, the particle size, surface hydrophobicity, and turbidity of SPI-Dex polymer were decreased significantly, and the solubility, emulsification, and antioxidant activity of SPI-Dex polymer were improved significantly. When the concentration of Dex in the mixed system was 5%, the improvement effect on the functional properties of SPI was the most significant, the solubility, emulsifying activity index and antioxidant activity of SPI were increased by 16.35%, 18.71% and 11.30%, respectively.

Abstract:Most tractors in the world still use manual mechanical shifting transmissions. Such transmissions face problems for instance power interruption and cumbersome operation during gear shifts, which is extremely unfavorable for vehicles with load operations such as construction machinery, which not only increases the fatigue intensity of the driver, but also reduces productivity. To ensure the maximum flexibility of use at each speed and exploit the maximum engine power available under all working conditions. A series hydraulic and mechanical hybrid transmission system (HMD) and its supporting tractor overall design method were proposed, including functional requirements, transmission route design, component characteristics analysis, performance parameter calculation, simulation model building, and operation performance comparison. The system scheme was based on the series combination of hydraulic torque converter (TC) and dual clutch transmission (DCT). Based on the proposed design method, the Dongfanghong 1804 tractor was selected as the research object. According to the effectiveness of stepless speed regulation module, the traction performance and matching performance test were designed. Reasonable engine matching made the maximum efficiency of hydraulic transmission reach 0.894, and the stepless variable speed can be realized in heavy and light load sections, and the adaptability of working conditions was excellent. The comparison results showed that the traction characteristic field of HMD was located between the original tractor and HMCVT, and the peak amplitude of tractive force in each gear varied smoothly and the area of deep valley was significantly reduced. Under ploughing tillage, traction coverage was increased by an average of 15.3%, and the load operation speed was increased by an average of 8.2%. Under rotary tillage, the take-off power was increased by 1.7% on average. The research can provide a technical route reference for the overall design of multimodal infinitely variable speed tractors.

Abstract:Based on the mountain crawler tractor (referred to as mountain tractor ) developed by our team, a collaborative control system for vehicle body and farm tool attitude was designed. In view of the different attitude adjustment requirements of body leveling and farm tools profiling during the contour operation of mountain tractors, the control strategy of the whole system was constructed on the basis of establishing the kinematics model of body and farm tools, and the collaborative control system was designed. Among them, PID algorithm and double closed-loop fuzzy PID algorithm were used for the control of body and farm tools, respectively. The control algorithm was simulated and analyzed based on Simulink. The results showed that the PID algorithm can meet the basic leveling requirements. The control effect of using double closed-loop fuzzy PID algorithm to realize the attitude adjustment of farm tools was better than that of PID algorithm. Finally, the static and dynamic tests of the coordinated adjustment of the unit attitude were carried out. The results showed that when the static adjustment was carried out on the slope of ±15°, the maximum error of the body roll angle was 1.10°, and the maximum error of the farm tool roll angle was 0.46°. In the dynamic operation on the slope of 14°~16°, the maximum error of the roll angle of the body was 1.90°, and the maximum error of the roll angle of agricultural machinery was 0.93°. It can be seen that the accuracy and stability of the attitude cooperative control system can meet the requirements of slope contour operation in hilly and mountainous areas.

Abstract:Lifting with a single arm is a common upper-limb action but it is easy to cause the rotator cuff injury. An upper-limb exoskeleton was designed to provide assistant force/torque to the human for accomplishing the lifting task on the premise of safety. The proposed prototype was comprised of two parts: 1-DOF elbow exoskeleton and a pneumatic shoulder pillow. The proposed structure was combined with the benefit of power transmission in the rigid exoskeleton and flexible double weave shoulder movement in the bionic exosuit. Such devices can provide necessary force assistance for the elbow flexion/extension and shoulder abduction of the wearer and improve the kinematic compatibility between the human arm and the exoskeleton. Further on, to verify the performance of the exoskeleton, a subject was asked to lifting 1kg and 5kg weight of loads with and without the robot-assisted, respectively. From the results, the mean absolute values of electromyography on biceps and supraspinatus were reduced to 70.97% and 80.67% of the non-wearable state during the 1kg weight lifting task, and 69.92% and 81.51% in the 5kg weight lifting task. The root mean squares of the relative muscle electromyography were reduced to 78.85% and 78.02% in 1kg, and 66.46% and 89.22% in 5kg. By using the pneumatic abduction shoulder pillow, the abduction trajectory of the shoulder was similar to the natural state during the lifting tasks. The heart rate of the subject was also recorded in the experiment. From the statistical analysis, the heart rate was not obviously changed (p>0.01). These results illustrated the potential risk of rotator cuff injury can be reduced in the robot-assisted lifting task.

Abstract:Magnetorheological (MR) dampers with fixed damping gaps and short damping lengths have the disadvantages of small damping force, single dynamic performance and low adaptability. To overcome these shortcomings, a MR damper with adjustable damping gaps was proposed and designed. The MR damper had annular and conical fluid flow channels, and four segments of effective damping lengths in the annular and conical fluid flow channels. By adjusting the position of the valve core controlled by the locking nut, the relative positions between the valve core and the left and right yokes would be changed, so the damping gaps in the conical fluid flow channel can be adjusted continuously. The magnetic circuit of the proposed MR damper was studied and its mechanical model was also established. ANSYS was used to simulate the change of the electromagnetic field in the MR damper. A prototype was manufactured, and its dynamic performance was tested experimentally. The results showed that the damping force of the proposed MR damper was increased with the decrease of the damping gaps, and the maximum damping force can reach 7.2kN at the damping gap of 0.6mm. Moreover, the adjustable coefficient was decreased with the decrease of damping gaps, and the maximum adjustable coefficient can reach 13.6 at the damping gap of 1.6mm. In addition, the damping force was ranged from 0.2kN to 7.2kN with different damping gaps, and the maximum adjustable coefficient can reach 33. The dynamic performance of the developed MR damper can be greatly enhanced, and the traditional MR dampers with its application in fixed situations can be possibly replaced.

Abstract:In order to improve the lubrication performance of the cylinder/valve-plate interface of a small axial piston pump, an oil film lubrication model was established based on the force analysis of the cylinder block. With leakage, cylinder overturning angle and viscous friction torque as the optimization objectives, the structural parameters of the valveplate seal ring were optimized by multi-objective non-dominated sorting genetic algorithm (NSGA-Ⅱ)， including the radial dimensions of valve-plate sealring R1,R2,R3,R4 and the starting angle θ of the waist groove.Considering the micro tilt motion and macro rotation motion of the cylinder block, the dynamic change process of the wedge oil film was simulated.The pressure distribution of the oil film on the cylinder/valve-plate interface was obtained by discretizing the Reynolds equation by the finite volume method. The results showed that the size of the outermost edge of the seal ring had little influence on the lubrication performance of the oil film, while the starting angle of the waist groove and the size of the inner edge of the seal ring had large influence on the overturning angle of the cylinder. After the optimization of the structure of valve plate, the comprehensive lubrication performance of the oil film was improved by 5.4%, and the overturning angle and leakage rate were decreased by 3.8% and 29.6%, respectively.

Abstract:The feature relationship of the motion state of CNC machine tools is very complex. Realizing the prediction of the future operation state of CNC machine tools can tap the potential abnormal emergencies of machine tools and enhance the stability of machine tool processing. In view of the problem of poor adaptability and low accuracy of prediction under dynamic label of machine tool state and differential distribution data, an adaptive hybrid deep learning model was established to predict machine tool state by combining time series feature relationship and model fusion method. Firstly, by combining the nearest neighbor classifier, an adaptive updating rule based on weight accumulation was designed, and a state prediction model with data adaptability was established. On this basis, an optimization strategy of feature distance metric based on center loss function was proposed, and a comprehensive decision loss function was constructed to ensure model fusion effectively. Based on a combination convergence criterion, the BBPT method was used to train the model, and the test data was verified . The experimental results showed that the model can adapt dynamic label and differential distribution data. The prediction of the state category of CNC machine tools had strong antiinterference, fast response and high accuracy, and can better meet the requirements of machine tool state classification and prediction. The prediction accuracy and real-time performance were significantly compared with BP and LSTM classification networks, and the shortest prediction time was only 100ms in GPU mode.