Abstract:Aiming to address the challenge where canopy occlusion and repetitive features compromise unmanned aerial vehicle (UAV) pose estimation systems in orchard environments, a multi-feature fusion-based pose estimation method for UAVs in orchards was designed by introducing visual-inertial odometry technology and incorporating geometric constraints from point and line features. Firstly, the EDLines algorithm replaced the traditional LSD for extracting line features, while optical flow enabled rapid tracking and matching of feature points and lines across consecutive frames, with feature poses obtained through 3D feature reconstruction. Secondly, a tightly coupled pose estimation model was constructed to fuse inertial and visual information, within a local sliding window framework, a jointly minimized global cost function was established, the accurate position and attitude information of the orchard UAV was obtained by solving the cost function through optimization methods. Finally, comparative experiments were conducted in fruit-bearing apple orchards and grape greenhouses, with absolute trajectory error and relative trajectory error serving as evaluation metrics to validate the method’s effectiveness. Experimental results demonstrated that compared with traditional pose estimation methods utilizing LSD algorithm-extracted line features, the proposed method reduced the average absolute trajectory error by 10% and the average relative trajectory error by 27%. This approach effectively enhanced the accuracy and robustness of orchard drone navigation systems, providing reliable support for ensuring the safety of orchard drone operations.

Abstract:Weeds compete with crops for nutrients and negatively affect agricultural yield. Traditional chemical weeding, though efficient, tends to cause environmental pollution and herbicide resistance. To achieve green, high-precision, and non-contact weeding operations, a target-oriented control system for a laser weeding robot was designed based on visual perception and path control. The system consisted of a depth camera, galvanometer scanner, and high-power CO2 laser, enabling weed detection, spatial localization, and laser operation. A Teensy 3.2 microcontroller-based integrated control scheme for the galvanometer and laser was proposed, which combined with the XY2-100 communication protocol and TTL triggering mechanism, allowed precise galvanometer control and laser switching without the need for an additional control card. A camera-galvanometer extrinsic calibration method combining manual measurement and refinement optimization was established, and an attitude-error compensation algorithm based on geometric inverse solving was proposed to achieve automatic correction of aiming and firing under different pose conditions, followed by verification experiments. The results showed that the average aiming error of the system remained below 1cm at different heights and tilt angles, and the laser weeding hit rate reached 98.6%. The optimal operating parameters were determined as a working height of 80cm and a laser exposure time of 0.5s. This system can enable high-precision, low-energy autonomous weeding and provide a reference for the application of laser weeding technology in complex field environments.

Abstract:Aiming to address the issue that high-speed operation induces excessive vibrations of no-till seeding units, thereby deteriorating seeding quality, a dual-regulation vibration-reduction no-till seeding unit was developed based on downforce stabilization control and furrow-depth attenuation control. By integrating multiple sensors, including an inclinometer, a pin-shaft force sensor, and an angular displacement sensor, the system enabled real-time acquisition of the profiling angle, downforce cylinder thrust, and depth-limiting wheel arm rotation. Consequently, closed-loop regulation of the downforce hydraulic cylinder pressure and the damping cylinder displacement was achieved. Stable downforce output and impact-limiting damping control effectively suppressed unit vibration, enabling active regulation of downforce and seeding depth. A coupled co-simulation model of the dual-regulation hydraulic control system was constructed by using AMESim and Simulink. Simulation results indicated that PID, fuzzy-PID, and sliding-mode control (SMC) exhibited comparable performance in regulating the downforce cylinder force. However, for damping-cylinder displacement control, SMC demonstrated clear advantages over conventional PID and fuzzy-PID control, reducing the maximum overshoot by 5.56 percentage points and 2.01 percentage points, and shortening the settling time by 1.27s and 1.45s, respectively. A dual-regulation vibration-reduction test bench for the no-till seeding unit was built, and a downforce measurement model integrating profiling angle and hydraulic pressure was established. Experimental validation of the proposed control strategy showed that under different hydraulic pressures and four-bar linkage inclination settings, the coefficient of determination reached 0.94463 with an adjusted value of 0.93819, demonstrating high control accuracy of the seeding downforce. The findings can provide a theoretical foundation for vibration-reduction control of no-till seeding units.

Abstract:Aiming to address the limitations in plant spacing adjustment capability and the issues of low seeding precision and poor uniformity caused by high-speed slippage in traditional ground-wheel-driven potato planters, an intelligent precision seeding control system for potatoes was developed. The system replaced the mechanical ground-wheel drive with a hydraulic motor drive, achieving independent control of two seeding units through a dual-circuit hydraulic integrated valve block. A PLC controller integrating a fuzzy PID algorithm was developed, establishing a closed-loop speed regulation system based on dual feedback signals from the travel speed and hydraulic motor speed. Parameters were dynamically adjusted according to the error value (e) and the error change rate (ec), significantly enhancing plant spacing uniformity and enabling stepless adjustment capability. AMESim-Simulink co-simulation results demonstrated that: under step signal input (70r/min) and sudden load application (100N·m) conditions, the system exhibited excellent synchronization between the dual motors (with no overshoot), a response time under 0.5s, and a rotational speed fluctuation amplitude below 1.5%, converging within 0.3s. Laboratory bench tests and field trials indicated that: across a plant spacing range of 0.16~0.24m and operational speeds of 2~8km/h, the actual seeding plant spacing exhibited a relative deviation not exceeding 8%, with a qualification rate reaching or exceeding 92%. These performance metrics complied with relevant industry standards and fulfilled the high-reliability requirements for precision seeding under diverse operating conditions.

Abstract:To address the problems of seed bouncing and non-uniform distribution during the seed delivery process of wide belt seed wheat planters, a pneumatic wide belt seed delivery device based on the action of a slit-induced adsorption force field was designed. A slit structure was introduced into the device to generate a localized adsorption force field, enabling uniform division of the falling seed population. Discrete element and finite element methods were employed to model and optimize key parameters, including the slit structure and the length of the intermediate section, thereby clarifying the effects of structural parameters on seed-guiding behavior and the uniformity of the pneumatic action zone. The pneumatic seed delivery process was simulated using a CFD-DEM coupling method. Air pressure, seeding rate, intermediate seed-guiding channel width, and slit structure were selected as influencing factors, while the longitudinal seeding rate consistency coefficient of variation among rows and the transverse in-row coefficient of variation were used as evaluation indices. Regression models were established using Minitab, and response surface analysis was conducted to determine the optimal parameter combination. The results indicate that seeding rate, slit structure, and intermediate seed-guiding channel width have significant effects on the coefficient of variation of longitudinal seeding rate consistency among rows, whereas seeding rate, air pressure, and channel width significantly affect the in-row transverse coefficient of variation. When the air pressure was -0.8kPa, the seeding rate was 150kg·hm-2, the seed-guiding channel width was 11mm, and the slit structure was rectangular, both coefficients of variation reached their lowest overall levels, with the longitudinal seeding rate consistency coefficient of variation at 2.16% and the in-row transverse coefficient of variation at 8.97%. Bench tests further verified the simulation results. With increasing positive pressure, the coefficient of variation of seeding rate consistency among rows first decreased and then increased under low seeding rates, while it continuously decreased under high seeding rates;a backflow phenomenon occurred when the air pressure reached approximately 2.25kPa. The in-row transverse coefficient of variation remained generally within the range of 8%~15% with no significant variation, meeting the operational requirements for precision and uniform wheat sowing.

Abstract:In response to the problems as multiple interference factors, and long computation time in processing field information using current radar perception, an automatic alignment system that combined laser radar and Beidou positioning for tobacco harvesters. The control algorithm fit the point cloud processing centerline based on the historical positioning coordinates of the aircraft body, effectively solving problems such as deviate guide lines and misalignment caused by aircraft body deviation. Based on the collected point cloud information of tobacco plants, the current row spacing and plant height of the field were calculated as the decision basis for extracting the region of interest (ROI). To shorten the processing time of point clouds, the length, width, and height range of ROI space were dynamically adjusted, and interference point clouds were filtered out. A crop row classification method was proposed based on point cloud density, and the initial clustering center of K-means algorithm was determined through vertical projection and sliding window method to improve the accuracy of tobacco row navigation line fitting. The optimal preview algorithm was used to track and drive the obtained forward tobacco navigation line. Automatic parallel testing was conducted in three tobacco fields with different row spacing. The maximum lateral error between the center of the machine and the actual centerline of the tobacco row was 0.107m, with an average lateral error of 0.074m. The accuracy of the harvest navigation line was 94.8%, indicating that the system can effectively ensure automatic parallel driving of the harvester. The proposed dynamic ROI point cloud processing algorithm reduced the average point cloud processing time from 536.2ms to 148.95ms compared with existing algorithms. The tobacco harvester automatic alignment system and dynamic ROI point cloud processing algorithm proposed can provide technical references for unmanned autonomous operation of field planting crop equipment.

Abstract:Aiming at the problem that the online detection of moisture regain of seed cotton during cotton picking operations is significantly interfered by environmental temperature and contact pressure, a moisture regain sensor for seed cotton in cotton pickers was designed based on the resistive method, integrating temperature and pressure compensation mechanisms. By optimizing the resistance detection circuit to broaden the measurement range and combining temperature and pressure sensing units, a sensor hardware system capable of synchronously collecting environmental temperature, contact pressure, and seed cotton resistance was designed, and it was calibrated and performance tested. An experimental platform was built to analyze the influence of temperature, pressure, and moisture regain on the conductive characteristics of seed cotton, and a moisture regain prediction model incorporating multi-parameter compensation was established. The results showed that the designed hardware circuit had an average absolute error of less than 0.4℃ for temperature measurement, an average relative error of less than 0.2% for pressure measurement, and an average relative error of less than 3% for resistance measurement. In the moisture regain prediction model, the BPNN algorithm performed best, with a coefficient of determination (R2) of 0.986 and a root mean square error (RMSE) of 0.377%. To verify the reliability of the moisture regain sensor, indoor static tests and field harvesting tests were conducted. The indoor static test results indicated that the sensor’s detection range was from 4% to 15%, with an average absolute error of 0.22% and an average relative error of 2.32%. The field harvesting test results showed that the absolute error of the detection results was no greater than 0.5%, and the relative error was no greater than 4.32%. The tests demonstrated good accuracy and practicality, providing effective technical support for the online detection of moisture regain of seed cotton in cotton pickers.

Abstract:The seed cotton cleaning machine, as the core of cotton processing, faces operational challenges that limit overall efficiency. These include the inability to monitor equipment status in real time, delayed and passive fault alerts, and loosely defined operation and maintenance strategies during the impurity removal process. Such limitations have constrained further improvements in cotton processing quality, production efficiency, and overall enterprise profitability. To address these issues, digital twin technology was applied to create a virtual replica of the physical seed cotton cleaning machine. Based on a detailed analysis of the machine’s operational principles and mechanical structure, a high-fidelity digital twin model was constructed. This model established a dynamic, bidirectional mapping mechanism between the physical machine and its virtual counterpart, enabling seamless data exchange and state synchronization. Using the Unity platform, a comprehensive digital twin monitoring system was developed for the seed cotton cleaning machine. This system integrated real-time data acquisition, simulation, and analysis capabilities. It allowed for real-time monitoring of the machine’s operational status, facilitated proactive fault warnings through predictive analytics, and supported dynamic optimization of process decisions based on simulated scenarios. Performance evaluations of the system demonstrated strong stability and reliability with key metrics, including a data packet loss rate of 0, a CPU usage rate of approximately 5%, an average GPU memory occupancy of around 4%, and an average motion simulation frame time of 20.416ms. The system was verified to possess excellent stability, reliability and robustness.

Abstract:Aiming to achieve the synergistic optimization of impurity removal efficiency for the inclined seed cotton cleaner and the quality of cotton fibers post-cleaning under varying seed cotton moisture regain levels, the design of a device that adjusted the rotation angle of elliptical rods to modify the gap between adjacent grids was presented. It provided a systematic account of the device’s key structure and operating principle, and established the adjustable range of parameters (8.5~12.5mm) through geometric calculations. Drawing on the test results of mass, destructive force, and destructive energy of single seed cotton lumps under a moisture regain range of 5.5%~11.5%, the elastic collision process between single seed cotton lumps and spikes during the cleaning stage was analyzed, and the testing range for the spike roller’s rotational speed was defined as 560~760r/min. Building on the definition of the aforementioned testing parameter ranges, a quadratic regression orthogonal rotational combination experimental design was implemented—with spike roller rotational speed, seed cotton feeding rate, gap between adjacent grids, and seed cotton moisture regained as experimental factors, and the machine’s impurity removal efficiency and seed cotton short fiber rate as evaluation indices. From this design, regression equations describing the relationships between the evaluation indices and individual experimental factors were derived. Furthermore, a mapping relationship between seed cotton moisture regain and processing parameters was established by using the non-dominated sorting genetic algorithm with an elitist strategy (here after referred to as NSGA-Ⅱ). Verification tests under multiple operating conditions demonstrated that in comparison with the pre-optimization state, the machine’s impurity removal efficiency was enhanced by a maximum of 14.14 percentage points, while the seed cotton short fiber rate was decreased by a maximum of 11.22 percentage points. For different parameter combinations, the maximum relative error associated with the machine’s impurity removal efficiency was less than 4.59%, and that for the seed cotton short fiber rate was less than 2.80%, thereby validating the reliability of the regression model and the optimized parameters.

Abstract:In response to the challenges of quantitative sampling under negative pressure, low real-time detection accuracy of moisture content, and poor stability during the cotton processing, an online moisture content detection device for lint cotton was developed based on the resistance method. A novel online detection method, combining “dynamic quantitative sampling—weight constant pressure measuremen—multi-parameter compensation” was proposed. Key components, including a flip-type sampling mechanism and weight detection system, were designed to achieve stable sampling and continuous detection of lint cotton in a high-speed flow state. Based on the electrical resistance characteristics of lint cotton and environmental temperature and humidity data, a multi-parameter moisture content prediction model was constructed. After comparing various algorithms, the random forest model was identified as the most accurate predictor, with a coefficient of determination (R2) of 0.98 and a root mean square error (RMSE) of 0.22%. Performance validation tests showed that the average deviation between the proposed online detection device and the standard measurement instrument was 0.1%, with an average detection time of 23.4 s per sample. The impact of sample weight on the moisture content measurement results was weak (Pearson correlation coefficient was 0.47). The experimental results demonstrated that the device offered high detection accuracy and good stability, meeting the real-time and accuracy requirements for moisture content detection in cotton processing. This device can provide reliable technical support for intelligent regulation and quality traceability in the cotton processing industry.

Abstract:Aiming at the problems such as low traction efficiency, short driving range and high energy consumption caused by frequent discharge of the power system due to the intense load fluctuations during the plowing operation of non-road mobile electric tractors, an adaptive energy management strategy for hybrid system combining fuzzy logic control and operation mode was proposed. The non-road mobile operation mode of the hybrid electric tractor was established. On this basis, the energy flow model of the lithium battery and supercapacitor hybrid system under different operation modes was established and the overall power system was modelled. Based on the analysis of the power requirements in the operation process, a fuzzy adaptive dynamic threshold adjustment strategy for the output voltage of the power system was proposed, and its weights were optimized by genetic algorithm with the goal of minimizing the energy-saving rate. Through simulation experiments, the power allocation, the state of charge and energy consumption under four energy control strategies were compared respectively to verify the power allocation effect of the hybrid system in different operation modes under different power allocation control strategies. The simulation experiments showed that compared with the proportional power allocation control strategy, the fuzzy adaptive control method control strategy could relatively increase the total braking energy recovery rate by 61.5%, the relative reduction rate of total energy consumption could reach 19.96%, the peak current of lithium batteries could be reduced by 39.96%, and the driving range of electric tractors could be increased by 19.53%. The results showed that the proposed control method had the performance of improving the traction efficiency of electric tractors, extending the driving range, reducing energy consumption and prolonging the battery service life.

Abstract:Rapid and non-destructive method for online determination of soluble solids content (SSC) in Shine Muscat grapes was developed by using visible-near infrared (Vis-NIR) spectroscopy. The influences of spectral acquisition speed, sampling mode, and preprocessing methods on model performance were systematically investigated, and model simplification was achieved while maintaining prediction accuracy. During spectral acquisition, each grape bunch was divided into three equal segments (front, middle, and rear), and the averaged spectra were used to represent the overall optical response of the whole bunch. Partial least squares regression (PLSR) and support vector regression (SVR) models were constructed to evaluate the effects of different preprocessing approaches. Among them, the combination of Savitzky-Golay smoothing and standard normal variate transformation (SG+SNV) yielded the best results by effectively correcting baseline drift and scattering noise, thereby enhancing spectral quality and model precision. Whole-bunch detection achieved superior performance compared with segmented detection, with SVR producing root mean square error of prediction of 0.49°Brix and correlation coefficient of 0.91, and PLSR yielding root mean square error of prediction of 0.45°Brix and correlation coefficient of 0.94. As the acquisition speed was increased from 0.15m/s to 0.6m/s, the prediction accuracy was gradually declined. For wavelength selection, uninformative variable elimination (UVE) demonstrated the best performance, extracting 314 informative wavelengths from 980 while maintaining high accuracy (PLSR: root mean square error of prediction was 0.41°Brix, correlation coefficient was 0.90) and reducing model complexity. Overall, the proposed Vis-NIR-based approach enabled accurate and stable online prediction of grape SSC at the whole-bunch level, offering a practical theoretical foundation for the commercial development of intelligent online sorting systems for bunch-type fruits.

Abstract:Aiming to address the challenges of dynamic end-effector payload variation, insufficient environmental interaction compliance, and limited trajectory tracking accuracy in strawberry harvesting robots, a fully decoupled parallel wrist joint based on an RPU-RPR-UPR configuration was designed, and its inverse kinematic model was established. To improve the positioning accuracy of the manipulator’s end-effector, a complete system dynamics model was developed, and an adaptive sliding mode control (ASMC) algorithm was implemented. Model uncertainties were estimated and compensated in real time by the ASMC, effectively alleviating the chattering phenomenon associated with conventional sliding mode control. As a result, the average tracking errors of the wrist joint’s α, β, and h degrees of freedom were reduced to 0.0758°, 0.0771°, and 0.0414mm, respectively, significantly outperforming both PID control and traditional sliding mode control (SMC) in terms of trajectory tracking accuracy and system robustness. To enhance harvesting smoothness, a fuzzy inference-based variable admittance controller was developed. Admittance parameters were dynamically adjusted according to interaction forces and system state errors, enabling autonomous modulation of the robot’s stiffness-compliance characteristics and thereby improving its disturbance rejection capability and environmental adaptability. Simulation results showed that, under external disturbances of 10N and 20N, the proposed fuzzy variable admittance control reduced overshoot by 12.5% and 17.8%, and shortened settling time by 0.14s and 0.25s, respectively, compared with fixed-parameter admittance control-demonstrating superior dynamic performance. The designed parallel wrist joint was integrated into a mobile strawberry harvesting platform. Acceleration response tests revealed that, during end-effector load variations, compliant control reduced the average acceleration fluctuation amplitude by 45.5% compared with position control, indicating excellent dynamic compliance. In harvesting trials, a success rate of 93% was achieved with an average cycle time of 15s per fruit, satisfying practical requirements for harvesting efficiency and reliability.

Abstract:Aiming to address the challenges posed by the natural growth of asparagus branches and leaves obstructing pathways and the limited working space in raised beds, which result in significant map construction noise, large localization errors, and low mapping accuracy during the autonomous navigation of asparagus harvesting robots, an autonomous navigation system was presented based on 3D LiDAR SLAM for greenhouse asparagus harvesting robots. Initially, 3D point cloud data from the greenhouse environment were acquired by using a Velodyne 16-line 3D LiDAR sensor combined with an N100 inertial measurement unit (IMU). An adaptive point cloud filtering method was employed to preprocess the point cloud data, removing noise caused by the asparagus branches and leaves, thereby reducing the computational burden on the navigation system. Subsequently, a global re-localization process was performed by using the Cartographer pure localization algorithm based on extended Kalman filtering (EKF). For path planning, the Dijkstra algorithm was utilized for global path planning, while the dynamic window approach (DWA) was applied for local path planning. Experimental results demonstrated that the optimal parameter combination for the adaptive point cloud filtering method was k1=6.912, s1=0.334, and s2=0.918. The integration of adaptive point cloud filtering with the Cartographer algorithm enabled high-precision mapping in the greenhouse environment, with a maximum absolute error of 0.056m, a maximum relative error of 9.3%, and a root mean square error of 0.035m. The improved localization algorithm achieved lateral deviation no greater than 0.196m and longitudinal deviation no greater than 0.082m in the greenhouse environment. During autonomous navigation at speeds of 0.10m/s, 0.20m/s, and 0.30m/s, the lateral, longitudinal, and heading mean deviations were no greater than 0.082m, 0.091m, and 7.562°, respectively, while their corresponding standard deviations did not exceed 0.078m, 0.092m, and 6.561°. The proposed navigation framework satisfied the high-precision mapping, localization, and navigation requirements for autonomous systems in greenhouse environments, providing a theoretical and technical foundation for the deployment of harvesting robots in agricultural settings.

Abstract:Aiming to address the challenges of greenhouse mobile robot localization and scene recognition in highly dynamic environments with visually similar scenes, a novel scene recognition model was proposed based on local feature selection and aggregation. The model employed a pre-trained vision Transformer (DINOv2) as its backbone network to extract local image features and introduced a learnable query-based feature selection and aggregation strategy to generate discriminative global descriptors. By leveraging cross-attention mechanisms, the model selectively aggregated the most informative local features into compact global representations. Furthermore, a hybrid loss function combining contrastive learning and triplet learning was applied to optimize the recognition model. A comprehensive greenhouse scene dataset containing 2100 scenes and 25000 images was constructed, covering multiple challenging factors such as illumination variations, viewpoint changes, distance scaling, and temporal crop growth. Experimental results demonstrated that the proposed model achieved Top-1 recall rates (R@1) of 88.79%, 96.49% (R@5), and 97.96% (R@10) on the collected dataset, outperforming state-of-the-art scene recognition benchmarks, including NetVLAD, GeM, CosPlace, EigenPlaces, MixVPR, and SALAD by 23.70, 19.24, 10.64, 3.30, 3.90, and 0.44 percentage points in R@1, respectively. The model exhibited strong robustness under varying illumination conditions (R@1 fluctuation <5 percentage points), moderate viewpoint changes (93.12% accuracy within 15° deviation), and scaling variations (63.94% R@1 at 2×distance). However, performance declined under extreme viewpoint/distance changes and long-term crop growth variations (61.14% R@1 after 5 days). Real-world validation on a greenhouse mobile robot confirmed the model’s practicality, achieving an average recognition rate of 85.88%. The proposed learnable query-based feature aggregation mechanism, combined with the carefully selected feature extraction backbone, significantly improved recognition accuracy in greenhouse environments. This framework can provide a viable technical solution for vision systems in agricultural mobile robotics.

Abstract:Aiming to address the seedling leakage issue caused by missing seedlings in fully automatic pepper tray seedling transplanters, the seedling separation process of the transplanter was analyzed and a dual-seedling cylinder independent drive system with empty-cell replenishment for seedling separation was proposed, aiming to minimize missed planting. An electric drive double group seedling cup common rail type separating seedling system was designed. The two sets of seedling cylinders operated on identical trajectories. While one set remained stationary to receive seedlings, the other set can continue performing seedling-dropping operations or execute empty-cell replenishment tasks, ensuring a continuous supply of seedlings to the planting device’s duckbill mechanism. An optical fiber identification sensor was selected for seedling absence detection, and a PLC-based automatic control system was established to acquire the sensor output signals in real time. Based on the determination of the empty-cell state derived from a binary threshold algorithm, the system triggered the actuator to initiate accelerated motion for empty-cell replenishment operations. With planting frequency, air supply pressure, and the status of the detection and compensation function (on/off) as experimental factors. The trial evaluated transplant failure rates and duckbill seedling reception rates of the planting device to identify optimal parameter combinations, followed by field validation. Results showed that under the combination of 120 plants per minute planting frequency and 0.4MPa air supply pressure with detection and compensation function was disabled, the average duckbill seedling reception rate was 86.46%. When the detection and compensation function was activated, the rate was increased to 98.27%, demonstrating a significant enhancement in seedling reception efficiency. This research can provide valuable insights for the development of seedling separation devices for chili pepper tray seedlings.

Abstract:Aiming at the problem of easy wear of the slideway in the link-slide type seedling picking mechanism, a Watt-Ⅱ six-bar seedling picking mechanism that met the trajectory requirements for automatic transplanting of vegetable plug seedlings was proposed. It adopted a fully hinged linkage structure to avoid slideway wear. Firstly, the trajectory, attitude and working requirements of the seedling picking mechanism in the three processes of seedling picking, seedling conveying and seedling releasing were analyzed, an open trajectory was designed where the forward operational path coincided with the return path. Furthermore, five key poses along this trajectory were identified and defined. Then the Watt-Ⅱ six-bar mechanism was simplified into a double-rocker mechanism and a crank-rocker mechanism. The relative displacement matrix equation of the planar 2R open-chain mechanism was established, and the kinematic synthesis design equations were constructed according to the joint constraints. Through the homotopy algorithm, two sets of mechanism parameters meeting the design requirements were obtained to form the double-rocker mechanism. Combined with the quick-return ratio K, the crank-rocker mechanism was designed, and the two were combined to form the Watt-Ⅱ six-bar mechanism. Finally, the structural design, virtual simulation and prototype test of the seedling picking mechanism were carried out. The results showed that the theoretical trajectory, simulation trajectory and the motion trajectory of the bench test were basically consistent, and the success rate of plug seedling picking was no less than 91.4%, which met the design requirements of the automatic seedling picking mechanism for leafy vegetable plug seedlings. This verified the rationality of the design of the Watt-Ⅱ six-bar seedling picking mechanism, which had a good application prospect.

Abstract:Aiming at the problems of large machine and low cutting quality of potato seed potato cutting machine in China, a multi-functional small roller type potato seed potato grading and cutting machine was designed. The structural design of the whole machine and key components was carried out. Through theoretical analysis and numerical calculation, the main factors affecting the grading effect were determined to be the inclination angle of the roller group, the speed of the first row of roller group and the feeding amount. The main factors affecting the cutting effect were determined to be the vertical center distance of the clamping roller, the angle between the center line of the clamping roller and the horizontal direction and the speed of the clamping roller. Based on EDEM-RecurDyn coupling simulation and theoretical analysis, the range of rotation speed and feeding amount of a row of rollers was determined. The inclination angle, the rotation speed and feeding amount of rollers were taken as experimental factors, and the classification efficiency and classification accuracy were taken as evaluation indexes., Box-Behnken experimental design method was used to determine the optimal parameter combination of the feeding grading device. The inclination angle of the roller group was 10.53°, the rotation speed of the first row of roller groups was 88.7r/min, and the feeding amount was 52.22kg/min. At this time, the grading accuracy was 92.65%, and the grading efficiency was 52.04kg/min. The prototype test showed that under the optimal parameter combination, the grading accuracy was 91.67%, the grading efficiency was 49.44kg/min, the qualified rate of the cutting block was 95.75%, the cutting efficiency was 48.24kg/min, and the uniformity of the potato block was good, which met the agronomic requirements of the seed potato cutting block.

Abstract:Aiming to address the current issues with garlic combined harvesters’ stem-cutting devices, including suboptimal cutting performance, high cutting resistance, delayed cutting causing bulb damage, and clogging of gripping and conveying mechanisms, an SPH-FEM coupled algorithm was employed to design a dual-disc garlic stem-cutting device. Dynamic analysis and parameter optimization were subsequently conducted. Firstly, the material model for garlic stems was established based on their structural characteristics, physical parameters, and mechanical properties. A simulation model of the garlic cutting process was constructed by using ANSYS/LS-DYNA. Finite element simulation results determined the optimal blade disc parameters: disc thickness of 2mm, blade angle of 15°, and blade overlap of 15mm. Single-factor simulation tests were conducted by using the model to establish the operational ranges for the garlic cutting device: feed rate of 1.5~2.5km/h, disc rotational speed of 400~600r/min, and disc spacing of 1~3mm. Finally, a Box-Behnken design, a three-factor, three-level orthogonal combination test plan was implemented. Bench tests determined the maximum cutting resistance of the disc under each factor level. Design-Expert 13 was employed to conduct variance analysis and response surface analysis on the test results, yielding the optimal operating parameters for the garlic stem-cutting device: feed rate of 2.1km/h, disc rotational speed of 560r/min, and disc spacing of 1mm. Bench test results indicated that under optimal operating parameters, the maximum cutting resistance was 7.33N with an error margin of 7.5%. This dual-disc cutting device exhibited low cutting resistance, stable performance, and produced relatively flat stem cross-sections. Crucially, no bulb damage occurred during testing, fulfilling the stem-cutting requirements for garlic harvesting. It provided valuable reference for designing combined garlic harvesting machinery.

Abstract:Aiming to address the issues of low efficiency in manual harvesting and high damage rate in mechanical brushing harvesting of marigolds, an automated marigold harvesting platform with a clamp-feed type design was proposed. The structural design and testing of the harvesting end were carried out to meet the harvesting requirements. Through field tests of the handheld harvesting end mechanism, factors affecting the damage rate of marigolds were identified, including the center distance of the synchronous pulley, the rotation speed of the synchronous pulley, and the harvesting angle. A three-factor, three-level central composite design and response surface analysis were used to study the interaction effects of these factors on the harvesting success rate. A quadratic regression model was established with harvesting damage rate as the response variable, and the significance of each factor on the harvesting success rate was ranked. By optimizing the factors with the damage rate as the objective, the optimized parameters for the center distance of the synchronous pulley, the rotation speed of the synchronous pulley, and the harvesting angle were determined to be 56mm, 200r/min, and 30°, respectively, with a predicted damage rate of 6.59%. Three sets of validation tests were conducted with the optimized parameters, and the results showed that the end harvesting device could effectively complete the marigold harvesting task, with damage rates of 8%, 4%, and 8%, respectively. The relative error between the test values and predicted values was less than 4%. A marigold harvesting device with a mechanical arm module was designed based on the actual working environment, its design closely focused on the growth characteristics of marigold and the core requirements of precision and efficiency for harvesting operations. It can accurately and quickly move the end effector and locate the target marigold position to complete the picking action with stable support and drive. On this basis, a test platform was built, and harvesting tests were conducted. The success rate of the harvesting tests was 93.01%, confirming the feasibility of the clamp-feed type marigold harvesting device.

Abstract:To address the rollover risk of crawler-type sugarcane harvesters operating on hilly terrain-caused by a high center of gravity and narrow track width—this study establishes a real-time vibration measurement and prediction framework. Vibration acceleration signals of the harvester frame are collected during tilt tests on a dedicated experimental platform. The signals are processed in the frequency domain to extract key vibration features that characterize different inclination states. A hybrid BiLSTM-Transformer model is proposed to predict potential rollover conditions.In the proposed method, vibration acceleration data are first preprocessed and decomposed using empirical mode decomposition (EMD) to obtain denoised and reconstructed time-frequency components. The BiLSTM network effectively captures long-term temporal dependencies in the vibration sequences, while the Transformer module focuses on extracting local temporal and attention-based contextual features. The complementary strengths of these two networks enhance both learning efficiency and predictive stability.Experimental results demonstrate that the proposed hybrid model achieves a prediction accuracy of 95.39% with an average response time of 11.87ms, meeting real-time monitoring requirements. To further validate model effectiveness, t-SNE dimensionality reduction visualization and confusion matrix analysis are performed, confirming the model’s discriminative capability across different tilt states.This research provides a reliable theoretical and technical foundation for the development of real-time rollover warning and automatic leveling control systems for crawler-type sugarcane harvesters in complex hilly environments.

Abstract:The problems of high manual labor intensity as well as high harvesting cost become more serious for Camellia oleifera industry, which promote the research of Camellia oleifera harvesting technologies with high efficiency. In order to effectively improve the energy utilization rate of mechanized harvester of Camellia oleifera fruit, the cable-driven excitation approach was employed with negligible cable mass to replace the traditional rigid vibration. Based on a bifurcated basic unit of tree, cable-tree reduced system was constructed with considering two constrained cables. The energy transfer mechanism in the reduced system was investigated to deduce the calculation formula of natural frequency. An ideal truncated cone model with variable circular cross sections for branches was utilized, and the measurement approach of measuring elasticity modulus of branches with variable cross-sections was presented. Taking Camellia oleifera varieties in Hunan Province as an application example, the basic parameters of the tree and the cable were measured, the natural frequency of the cable-tree reduced system was obtained. The theoretical values of the natural frequency for the cable-tree reduced system were compared with simulation and the experimental results. It showed that the relative errors of simulation results and theoretical results were less than 6.4%, the relative errors of the experimental results and the theoretical results were less than 8.6%, which verified the efficiency of the proposed calculation formula of natural frequency for Camellia oleifera with constrained cables. By comparing the resulting modulus of elasticity from the presented measurement approach and the conventional three-point bending test, the relative measuring error was within 7% and thus the feasibility of the presented measurement approach was validated. The research results can provide a theoretical basis for the design of a flexible cable-driven branch-vibrating harvester for Camellia oleifera fruits.

Abstract:In conventional drum-type maize threshing, the tightly arranged kernels on the maize ear lead to a high inter-kernel support force during the initial stage of threshing, resulting in strong impacts from the threshing elements and severe kernel damage. To address this issue, a pre-threshing principle for maize ears was proposed. This approach involved pretreating the ear before threshing by inducing internal splitting of the cob from the inside out into fragments, thereby disrupting the original compact kernel arrangement, loosening the kernels, and significantly reducing the inter-kernel support force. Based on this principle, a novel internal cob-splitting pre-threshing device for maize ears was developed. The device consisted primarily of an actuating cylinder, a prismatic splitting wedge, and an ear-holding fixture. Through analysis of the ear-splitting process, key structural parameters were identified, including the number of prism edges, operating air pressure, and fixture spacing. Single-factor experiments and Box-Behnken response surface tests were conducted to investigate the influence of these factors. The results showed that the significance of each factor affecting the proportion of small fragments and kernel breakage rate followed the descending order as follows: number of prism edges, fixture spacing and air pressure. The optimal combination of parameters was determined to be 8 prism edges, 2cm fixture spacing, and 0.55MPa air pressure. Under this configuration, five validation tests were conducted, yielding an average small fragment proportion of 74.38% and a kernel breakage rate of 1.07%, closely aligning with the predicted values. The results confirmed that the device can effectively split the maize ear into fragments with a high proportion of small cob pieces, while maintaining low kernel damage and significantly reducing the inter-kernel support force. Comparative threshing tests were carried out between a pre-threshing unit with a reciprocating flexible threshing unit and a 5TY-45-150 thresher, and a 5TY-45-150 thresher fed directly with the whole cob. The results showed that the threshing method of “pre-threshing + 5TY-45-150 thresher” and the threshing method of “pre-threshing + reciprocating flexible threshing device” were comparable to the threshing method of feeding the whole cob directly into the 5TY-45-150 thresher. The seed breakage rate of the threshing method was reduced by 2.73 percentage points and 2.97 percentage points, respectively. The reductions were as high as 55.94% and 60.86%, respectively. It proved that the pre-threshing device can significantly improve the operation quality of the whole threshing process. The research result can provide an approach and technical support for improving the quality of mechanized maize threshing.

Abstract:Aiming to address the issues of uneven crushing and straw drop during the cotton stalk fragmentation and returning process in Xinjiang, the design of a secondary wind-assisted cotton stalk crushing and returning device was optimized. The device employed a dual-helix crushing blade set and a wind-assisted conveying blade set for collaborative operation, integrating a diversion air duct to achieve primary crushing and secondary conveying of stalks. The structural configuration and parameters of key components were determined through theoretical analysis. To investigate the influence of operational parameters on the internal flow field during the crushing and conveying process, CFD simulations were conducted to analyze the effects of crushing blade shaft speed, conveying blade shaft speed, and the number of wind-assisted blades. A three-factor three-level orthogonal experiment was performed with crushing blade shaft speed, ground clearance, and conveying blade shaft speed as experimental variables, and stalk drop rate and crushing qualification rate as evaluation metrics. Parameter optimization yielded the optimal combination: crushing blade shaft speed of 2490r/min, ground clearance of 77mm, and conveying blade shaft speed of 2870r/min. Field validation tests under these conditions resulted in a stalk drop rate of 3.41% and a crushing qualification rate of 97.85%, with relative errors between experimental and theoretical values below 5%, meeting industry standards. This research can provide technical support for enhancing the efficiency of cotton stalk returning equipment.

Abstract:In order to solve the problems of low impurity removal efficiency and high cleaning cost in the process of cleaning after machine harvesting of membrane impurities, a membrane impurity separation method was proposed, which was first turbulent diffusion and then air gravity separation, and a perturbative membrane impurity separation device was designed. Through the theoretical analysis of the process of residual membrane turbulent diffusion and membrane miscellaneous air separation, it was determined that the key factors affecting the separation performance were the inlet air velocity, turbulence angle, diffusion chamber outlet height and conveying speed. Based on the discrete element model of residual film, cotton straw and soil miscellaneous materials, the EDEM-Fluent coupling simulation was used to numerically simulate the separation process of membrane impurities, which revealed the movement law and distribution of membrane miscellaneous materials after crushing in the separation device, and provided a basis for the structural parameters and working parameters of the separation device. Experimental tests with the prototype device, using impurity removal rate and winnowing loss rate as evaluation indices, determined the influence of different factors on performance. Optimal parameters were found to be: a turbulence angle of 48°, inlet air velocity of 5.8m/s, and diffusion chamber outlet height of 178.5mm, achieving an 80.26% impurity removal rate and a 12.35% film leakage rate. These results can meet subsequent resource utilization requirements and offer valuable guidance for advancing winnowing impurity removal technology in mulch film separation.

Abstract:Aiming to tackle the issues of high labor intensity and low efficiency associated with manual removal of residual plastic film from straw feed in Xinjiang, a needle-belt type electrostatic separation device for residual plastic film in straw feed was designed. Based on the principle of electrostatic adsorption, the device consisted of core components, including a high-voltage electrostatic generator, electret discharge rod, and metal conveyor belt. Through theoretical analysis, the critical conditions for separating residual plastic film from straw feed were deduced, verifying the feasibility of applying the needle-belt type electrostatic separation technology to residual plastic film removal. Using COMSOL simulation software, the effects of geometric parameters and operating parameters of corona needles on electric field distribution and negative ion density were studied, and the optimal parameters were determined as follows: needle tip cone angle was 12°, needle length was 14mm, and needle spacing was 20mm. Single-factor experiments showed that the electrode spacing (40~60mm), voltage (15~25kV), conveyor belt speed (21~28m/min), and material feeding rate (8~12kg/min) had significant impacts on the separation efficiency. A quadratic regression model was established by using the Box-Behnken experimental design, and the optimal parameter combination of the device was determined as follows: electrode spacing was 51.3mm, voltage was 22.9kV, belt speed was 24.5m/min, and feeding rate was 9.5kg/min. Finally, the separation efficiency measured in the verification experiment was 91%, which was in good agreement with the predicted value of 92.6%, meeting the technical requirements for removing residual plastic film from straw feed.

Abstract:Accurate remote sensing image segmentation of cultivated land (CLRSIS) is crucial for yield prediction, agricultural management, and national food security. However, it remains challenging due to the high-resolution, large size and various remote sensing farmland images with irregularly boundaries and complex background. Convolutional neural networks(CNNs) and Transformers have been widely applied to RSI segmentation, but both of them have limited ability to handle long-range dependencies because of inherent locality or computational complexity. Aiming at the limitation of CNNs and Transformers, and the technical difficulties in CLRSIS, a multi-scale attention visual Mamba U-Net (MSAVM-UNet) model for CLRSIS was proposed. This model achieved performance breakthroughs through three innovative modules: firstly, modified visual state space module (MVSS) adopted a bidirectional selective scanning mechanism, enabling long-range dependency modeling while maintaining linear computational complexity. Secondly, channel-aware attention visual state-space (CAAVSS) effectively enhanced the discrimination between cultivated land and background features through dynamic spectral-spatial feature recalibration. Finally, multi-scale feature aggregation module (MSAA) built a cross-level feature pyramid to achieve multi-granularity information fusion. Experiments on public cultivated land datasets showed that this method was significantly superior to existing methods in terms of segmentation accuracy and computational efficiency, with the average segmentation precision accuracy and DSC achieving 85.60% and 84.46%, respectively. The research result can provide reliable technical support for the precise monitoring of cultivated land in smart agriculture.

Abstract:Accurate and timely estimation of spatiotemporal dynamics of farmland soil respiration rate is crucial for revealing carbon emission patterns during agricultural production. Traditional methods can simulate soil respiration dynamics at small scales, but they remain insufficient in characterizing spatial heterogeneity at the farmland scale. To achieve accurate estimation of soil respiration at high spatial resolution, it was focused on summer maize in a typical region of central Inner Mongolia. The experiment included one control treatment (Tr1: irrigation at 100% ET, where ET represents evapotranspiration) and three deficit irrigation treatments (Tr2, Tr3, Tr4). Soil respiration rates were monitored at different growth stages by using the static chamber method, while vegetation indices and soil surface temperature (TUAV) were retrieved from UAV-based multispectral and thermal infrared data. The TUAV, together with the simple pigment ratio index (SRPI), the green-blue normalized difference vegetation index (NDVIg-b), and the normalized pigment chlorophyll index (NPCI), were incorporated into the Lloyd-Taylor model to develop an improved vegetation-heat index (VHI) model for soil respiration rate estimation. The performance of this model was further compared with that of a back propagation neural network (BPNN) model. The results showed that under Tr1~Tr4 treatments, temperature of the soil surface (TSF) was significantly correlated with seasonal soil respiration rate, with correlation coefficients of 0.946, 0.886, 0.898 and 0.766, respectively. Among the nine vegetation indices indicative of crop photosynthetic capacity, SRPI exhibited the strongest correlation with seasonal soil respiration rate. The VHI model based on SRPI and TUAV achieved the best fitting performance (R2=0.73), which was comparable to the BPNN model (R2=0.81). Overall, it was demonstrated that integrating UAV multispectral and thermal infrared data with the VHI model enabled high-resolution characterization and mapping of soil respiration rate heterogeneity at the farmland scale, thereby improving estimation accuracy.

Abstract:Aiming to address the problems of small strawberry individuals and serious inter-individual occlusion, a strawberry ripeness detection method was proposed based on MSCS-YOLO in an unstructured environment. The method introduced the multi-scale dilated attention (MSDA) mechanism in the Neck part of the YOLO v8n model, which enlarged the sensory field of the model and solved the problem of small strawberry fruits and easy to ignore features. Meanwhile, the improved C2f-Triplet attention structure was utilized to replace the C2f structure in the Neck part to capture the information of the strawberry image more comprehensively from the three dimensions, which enhanced the model’s target recognition ability in the case of fruit occlusion. Embedding the improved SAHead detection head into the YOLO v8n model enhanced the model’s recognition accuracy for strawberries with different ripeness levels in unstructured environments. The experimental results showed that the MSCS-YOLO model achieved an average accuracy of 94.22% in the task of recognizing three types of strawberries: ripe, moderately ripe and unripe, which was 1.38 percentage points and 5.42 percentage points higher than that of the YOLO v8n and RTDETR-L models, respectively;among them, the accuracy of recognizing ripe and moderately ripe strawberries achieved 96.35% and 92.00%, which was 0.82 percentage points and 3.66 percentage points higher than that of the YOLO v8n model, respectively. The MSCS-YOLO model demonstrated better recognition performance and higher accuracy regardless of evening, sunny day, direct sunlight or light irradiation conditions. In addition, the model size of the improved model was 6.42 MB, which was 45.22% and 86.93% smaller than that of the YOLO v7-tiny and YOLO v9c models, respectively, and achieved the synergistic optimization of accuracy and efficiency while maintaining a similar model size with YOLO v8n. Therefore, the MSCS-YOLO model was more advantageous for deployment and application in resource-limited environments, and it can provide reliable technical support for later practical applications on strawberry maturity.

Abstract:Accurately and efficiently detecting straw coverage is crucial for soil protection and sustainable agriculture, as straw coverage not only affects soil fertility and moisture retention but also plays a key role in controlling soil erosion and improving the ecological environment. However, existing straw coverage detection models are often susceptible to interference from natural environmental factors such as lighting and shadows in practical applications. When the similarity between the straw and the soil in terms of color and texture is high, the accuracy of these models significantly decreases, leading to inaccurate coverage assessments and ultimately affecting the efficiency and reliability of farmland management decisions. Aiming to address the challenges posed by the diverse morphology of straw in images captured by vehicle-mounted cameras, including issues of image reflection and shadows, a novel semantic segmentation method called UMU-KAN for detecting straw of varying scales in natural environments was proposed. The replacement of conventional dilated convolutions in the atrous spatial pyramid pooling module with depth-wise dilated separable convolutions was proposed to enhance the extraction of fine-grained straw-related detail information. Additionally, a strip pooling branch captured features of widely spaced straw more effectively, integrating feature information from various branches through skip connections to reduce information loss. This series of improvements constructed a mixed pooling dilated spatial pyramid module, applied to the top semantic layer of the backbone network, thereby obtaining multi-scale information for sparsely distributed straw. Furthermore, a unified attention fusion module appeared during the decoding phase to effectively restore detailed edge information of straw segmentation, enabling the model to better learn features from different levels. Experimental results demonstrated that UMU-KAN achieved a mean intersection over union (mIoU) of 85.36% and a mean pixel accuracy (mPA) of 91.71% on the constructed straw dataset. Compared with the Unet, Swin-Unet, and DeepLabv3+ models, UMU-KAN improved mIoU by 4.20, 3.26, and 1.25 percentage points, respectively, and mPA by 3.58, 2.39, and 0.77 percentage points, respectively. Additionally, the parameter count of UMU-KAN was significantly lower than that of Swin-Unet and DeepLabv3+. UMU-KAN successfully achieved accurate detection of straw in images captured by agricultural machinery cameras, ensuring high detection efficiency even under dynamic and uncontrolled outdoor conditions. This not only highlighted the model’s adaptability and precision but also further demonstrated the significant developmental potential of the KAN architecture in the field of precision agriculture, contributing to the promotion of sustainable agricultural practices and enhancing the efficiency of agricultural management.

Abstract:Aiming to achieve real-time and accurate detection of multi-scale rice pests in complex backgrounds, a dataset containing images of various rice pests was constructed and a pest detection model called YOLO v8-FDI was proposed. The model was based on the YOLO v8n architecture, utilizing the more efficient FasterNet as its backbone network. This design optimized the network structure while maintaining sensitivity to pest features. Dynamic Head technology was incorporated, allowing the model to dynamically adjust the detection heads in the output layer. This improved the model’s accuracy and generalization for pests of different types and sizes. Furthermore, the Inner-IoU loss function was adopted to automatically adjust scaling factors during loss calculation process. This accelerated training convergence and further improved model performance. Experimental results showed that the YOLO v8-FDI model processed a single pest image in an average time of 12.43 m, achieving a processing speed of 80 frames per second (FPS), meeting real-time requirements for practical applications. On the test set, the model’s detection precision, mean average precision mAP@0.5:0.95 and F1 score were 97.7%, 94.0%, and 97.2%, respectively. Compared with YOLO v3-tiny, YOLO v5n, YOLO v7-tiny, YOLO v8n,YOLO v9t,and YOLO v10n, precision was improved by 5.2, 2.7, 6.7, 3.4, 2.2, and 3.2 percentage points, mAP@0.5:0.95 was increased by 10.8, 5.4, 18.1, 2.3, 1.0, and 6.4 percentage points, and F1 score was raised by 2.6, 2.0, 4.9, 1.2, 1.3, and 2.9 percentage points. The novelty lied in the improvements made to the YOLO v8n architecture by integrating FasterNet, Dynamic Head, and the Inner-IoU loss function. These enhancements significantly improved the model’s accuracy and generalization, offering strong technical support for real-time and accurate pest monitoring in complex backgrounds.

Abstract:In the field of precision breeding of cattle, weight is a key indicator for measuring their health and production performance. Traditional weighing methods are not only inefficient but also costly, while existing dynamic weighing algorithms are limited by insufficient robustness and stability. In response to this issue, the hidden information and behavioral information of cattle dynamic weighing signals were quantitatively analyzed, and existing dynamic weighing algorithms were improved by proposing a cattle dynamic weighing algorithm based on classification of time-frequency domain motion state and compensation for error prediction. Firstly, preliminary weight prediction values were obtained through modal decomposition algorithm, and the reference error with static weighing parameters was calculated. Secondly, weights of the window function were optimized to establish an adaptive window function, obtain reliable signal time-frequency domain feature parameters, and explore their relationship with motion labels and corresponding reference errors in the state. Finally, a motion state classification model and two types of error compensation models were established, and the slime mold algorithm (SMA) was used to perform hyperparameter optimization on the latter. Based on this, a complete dynamic weighing model for cattle was established. The experimental results indicated that the dynamic weighing prediction model for cattle performed well. The accuracy of the motion classification model was 98.4%. In low and high activity states, the root mean square error (RMSE) of the final weight prediction values were 4.03kg and 8.96kg, respectively, and the mean percentage error (MAPE) were 0.53% and 0.87%, respectively. This algorithm had good robustness and generalization ability, which can provide reference for intelligent weight monitoring in practical breeding scenarios, and it had certain significance for promoting the development of precision breeding.

Abstract:Using ear temperature instead of body temperature to predict and analyze the estrus of sows. To address the issues of high cost and limited scalability in the automated detection of sow estrus, a prediction model (denoted as GA-LightGBM) was proposed based on the improved genetic algorithm (GA) optimizing the light gradient boosting machine (LightGBM). Compared with infrared cameras and infrared imaging equipment, low-cost and low-power electronic ear tags were used to collect real-time ear temperature data from sows, and an hourly average temperature resampling strategy was innovatively proposed. Comparative experiments showed that this strategy significantly reduced the risk of overfitting. To overcome the problems of slow convergence speed and falling into local optima in traditional genetic algorithms, a forced offspring iteration (FOI) mechanism was designed to improve the algorithm, enhancing the convergence efficiency while maintaining the global search capability. In experiments, a LightGBM model optimized by the particle swarm optimization (PSO) (denoted as PSO-LightGBM) was introduced for comparison. After verification using a dataset containing 311 sows provided by Zhejiang Huateng Animal Husbandry Co., Ltd., the improved GA-optimized LightGBM model (denoted as FOI-GA-LightGBM) achieved an accuracy of 83.91% and an AUC of 0.8390 on the test set, significantly outperforming the GA-LightGBM and PSO-LightGBM models. At the same time, FOI-GA-LightGBM was also compared with random forest (RF) and support vector machine (SVM) in terms of performance. The overall performance of FOI-GA-LightGBM was superior to RF and SVM.

Abstract:Timely and accurate acquisition of soil moisture information is essential for comprehensively understanding soil water content, improving water use efficiency, and conserving valuable water resources. At present, remote collection of soil moisture data typically requires both soil moisture sensors and wireless transmission modules. However, these approaches are limited to discrete point measurements and cannot achieve comprehensive regional assessment. Moreover, the high cost of soil moisture sensors restricts large-scale deployment. A low-cost method for predicting soil water content over an area was proposed, utilizing the working characteristics of underground wireless signal propagation which was strongly influenced by soil moisture, and its advantages in wireless communication. Based on LoRa technology and a fast machine learning approach with strong generalization capability, a LoRa-based underground wireless sensor network was designed and developed for underground environments to remotely collect data such as received signal strength indicator (RSSI), signal-to-noise ratio (SNR), and soil temperature. Using the collected data, a kernel extreme learning machine (KELM) model optimized by particle swarm optimization (PSO) was proposed to estimate soil water content, addressing the challenges of strong nonlinearity, poor fitting, and low convergence in large-scale data modeling. Experimental results showed that the proposed PSO-optimized KELM model achieved higher prediction accuracy than the extreme learning machine (ELM), support vector regression (SVR), and long short-term memory (LSTM) models. In the training dataset, the KELM model achieved a mean absolute error (MAE) of 0.76% and a root mean square error (RMSE) of 1.02%, while in the testing dataset, the MAE and RMSE were 0.72% and 1.07%, respectively. The proposed method can achieve effective prediction and remote acquisition of soil moisture content without relying on the dense deployment of soil moisture sensors, providing a low-cost solution for large-area soil moisture detection.

Abstract:Aiming to address the issues of low efficiency and insufficient automation in traditional soil available nitrogen, phosphorus and potassium (NPK) content detection, as well as the limited path planning and tracking performance of vehicle-mounted detection equipment in complex farmland environments, an autonomous detection system for soil available NPK in field empowered by an unmanned operation platform was established. It integrated Beidou positioning and a four-wheel drive steering platform, optimized sensor detection models, designed a hierarchical-adaptive traversal path planning method for fertilizer-measuring points and a path tracking control algorithm based on genetic algorithm (GA)-optimized linear quadratic regulator (LQR), and built an autonomous fertilizer detection system with cloud-edge architecture. For in-situ, rapid detection of soil fertility indicators, random forest (RF) prediction mode was selected as the optimal algorithm after comparing four machine learning models. Its relative prediction errors for available N, P and K were below 16.44%, 19.26% and 13.91%, respectively, with an average 49.46% accuracy improvement over the direct measurement accuracy of sensors without prior modeling. To accurately characterize the spatial distribution of field-scale fertility indicators, the optimal detection point spacing was determined by quantifying the Christiansen uniformity coefficient (CUC) across different sampling densities;three continuous traversal path planning schemes—comb path (CP), comb cross-row path (CCP), and comb fifth-order Bézier path (CFBP)—were designed for flat obstacle-free, large-obstacle, and small-obstacle farmland environments, respectively. Simulation tests showed that the GA-LQR controller improved the platform’s path tracking accuracy by an average of 24.32%, with maximum absolute lateral deviation and heading angle of 1.71cm and 1.69°, respectively. Field tests demonstrated that under straight path tracking, the GA-LQR algorithm reduced the maximum absolute lateral deviation, maximum absolute heading angle, average absolute parking error at detection points, and total detection time by 15.62%, 19.59%, 20.79%, and 13.43%, respectively, compared with the conventional LQR algorithm;under curved path tracking, the corresponding indicators were decreased by 19.28%, 27.34%, 22.94%, and 15.76%. Additionally, the optimized algorithm shortened the detection time per hectare by 927s, reduced the single-point detection process by 6.28s, and improved the positioning and detection accuracy by an average of 14.26%. The research result can provide a reference for the autonomy, informatization, and intelligentization of soil available NPK content information acquisition.

Abstract:Accurately estimating the spatiotemporal dynamics of global gross primary productivity (GPP) and its underlying mechanisms is crucial for understanding the global carbon cycle and climate change. Satellite remote sensing enables continuous observation of large-scale vegetation dynamics, providing valuable opportunities to study the spatial and temporal variations of GPP on a global scale. However, different GPP products often exhibit significant discrepancies in global GPP estimates, and a comprehensive validation and comparison of these products at the global level has not yet been conducted. Therefore, the spatiotemporal consistency and interannual trends of eight GPP products (EC-LUE, GLASS, GOSIF, MOD17A2H, MuSyQ, PMLv2, EC-LUE, and VPM) during 2003—2014 were evaluated by using observations from 147 global flux towers. Spatiotemporal analysis revealed exceptionally strong temporal correlations among products (R2>0.960). Spatially, all products exhibited high comparability (R2≥0.702) except GOSIF, which showed weaker consistency with others (R2≤0.573). Annual GPP estimates ranged from 678.3g/(m2·a) (MOD17A2H) to 1223.0g/(m2·a) (GOSIF). All products except GLASS displayed increasing trends, with the northern hemisphere dominating the GPP increase. The proportion of ascending areas varied substantially across products, peaking in VPM (72.4%) and reaching a minimum in GOSIF (45.2%). Validation against flux tower data identified PMLv2 as the best-performing product (R2=0.664), while EC-LUE (R2=0.547) and GLASS (R2=0.572) showed relatively lower accuracy. Systematic overestimations were observed in GOSIF, GLASS, and PMLv2 across most sites. Furthermore, the products demonstrated higher accuracy in America, high-latitude regions, and wetlands (WET) and evergreen needleleaf forests (ENF). The research results were significant for improving the ability of ecosystem process models to simulate different regions, and they contributed to a deep understanding of regional ecosystem carbon dynamics and the global carbon cycle.

Abstract:Ecological environment is closely related to the development of society and economy. Analyzing the process of ecological environment evolution is of great significance for guiding regional ecological environmental protection. However, analysis of regional ecological environment evolution based on a single indicator such as surface water area, groundwater level, normalized difference vegetation index (NDVI), etc., has limitation in a single perspective. Taking the Minqin Basin in the lower reaches of the Shiyang River as the study area. Firstly, spatial and temporal changes of four indicators, including surface water area and storage of Qingtu Lake, the terminal lake of the Shiyang River, groundwater level and storage anomaly, land use, and NDVI in Minqin Basin were analyzed. Then a comprehensive analysis of ecological environment evolution was conducted for the Minqin Basin from multiple perspectives, including surface and groundwater resource variations, land use type changes, and NDVI dynamics. The results shown that the water surface area and storage of Qingtu Lake generally followed a trend of drying up, increasing and shrinking. The groundwater water level and storage anomaly in the Minqin Basin generally showed a downward trend, with a significant slowdown in the decline from 2007 to 2013. The changes in land use types in the Minqin Basin were mainly characterized by the conversion between cropland, grassland, and barren, and since 2017, desertification in the Minqin Basin intensified. The NDVI in the Minqin Basin showed a non-significant increasing trend with a rate of 0.0005a-1. By integrating the characteristics of the changes of surface water and groundwater resources conditions, land use types, and NDVI changes, the process of the ecological environment evolution in the Minqin Basin from 2003 to 2022 was divided into three stages of rapid degradation (2003—2006), recovery (2007—2013), and re-degradation (2014—2022) based on multi-perspective. Changes of water resources conditions induced by human activities constituted the primary factor influencing ecological evolution in the Minqin Basin. Establishing a regional economic and social development model that was coordinated with and compatible to the water resource carrying capacity of the Minqin Basin should be prioritized as a key direction for ecological conservation in the region.

Abstract:Aiming to identify impact factors of unbalances of water supplies and water demand (referred to WSD) of Yellow River Basin under extreme dry years to support water resources management, the WSD was measured by ranges and depth of WSD. Besides, responses of WSD on metro-hydrological elements, economic and society, reservoir regulation, ecology and environments were explored by the principal component analysis method (PCA). Moreover, uncertainties of surface water supplies and discharges of reservoirs were measured by stochastic functions that were verified by Kolmogorov-Smirnov test (K-S) and augmented dickey fuller (A-D) approaches. The influences of single and interactive parameters on WSD were studied by three-level factors method. The results showed that WSD of Yellow River Basin ranged from medium, relatively high and high degree. Surface water supplies and discharges of reservoirs had the most impacts on WSD and explanation rate reached about 55.055%. Xiaolangdi reservoir had the biggest impact on WSD, followed by surface water supplies and Wanjiazhai reservoir. These results could identify key impact factors of WSD under interactions among hydrometeorology, socioeconomics, reservoir operation and ecol ogical environment. Besides, it could explore impacts of uncertainties on WSD, supporting water resources management of Huanghe River.

Abstract:The direction of soil water flux is a key parameter in saturated soil flow fields. Soil texture significantly affects pore connectivity, which introduces randomness into the direction of water flow. Therefore, measuring water flux direction requires consideration of an appropriate spatial scale. The direction of saturated soil water flux can be determined by combining the ratio method with the principle of vector composition. Based on the above requirements, a penta–needle heat pulse probe (PHPP) was designed to measure water flux magnitude in any two mutually perpendicular directions within a plane and to determine flux direction through vector composition. Experiments were conducted in saturated sand, sandy loam, and silt loam, with each soil type repacked three times. The experimental results showed that the accuracy of this method in measuring soil water flux direction was significantly influenced by soil texture. For fluxes greater than 4 cm/h, the mean absolute percentage errors (MAPE) of angle measurements in sand, sandy loam, and silt loam were 4.96%, 6.18%, and 15.06%, respectively. This indicated that the accuracy of water flux direction measurements was decreased with finer soil texture. Compared with fluxes below 4 cm/h, the standard deviations of angle measurements in sand, sandy loam, and silt loam were decreased by 10.40°, 6.65° and 6.71°, respectively, for fluxes above 4 cm/h. This indicated that the accuracy of water flux direction measurements was improved with the increase of flux. Stable water flux angle measurements, with absolute errors below 7.5°, were achieved in sand at fluxes above 6 cm/h and in sandy loam above 3 cm/h, but not in silt loam. These findings suggested that pore connectivity in packed soils varied with texture under different fluxes and hydraulic gradients, thereby affecting measurement precision. Additionally, the geometric relationship between soil particle size and probe spacing affected the measurement accuracy of the ratio method. Optimizing the probe spacing of the PHPP based on soil particle size distribution may improve the reliability of water flux angle measurements by using the vector composition method. These findings can contribute to the development and practical application of heat pulse technology.

Abstract:In order to investigate the effect mechanism of protein phosphorylation on apoptosis and myofibrillar protein degradation during maturation, psoas major (PM) muscles injected with protein kinase A (PKA) and alkaline phosphatase (AP) ware used as the experiment subjects. Mitochondrial dysfunction, apoptosis, myofiber type and the degradation of myofibrillar protein were measured and analyzed. The results showed that the AP group had higher mitochondrial membrane permeability, cytochrome c (Cyt-c) oxidation level, and caspase activity at 12~72h postmortem compared with both the PKA and control groups （P<0.05), showing more apoptosis. At 2~48h postmortem, the number of type Ⅰ muscle fiber was significantly increased in both the PKA and AP groups （P<0.05). At 2~72h postmortem, the levels of desmin degradation, troponin-T degradation, and calpain autolysis in the AP group were significantly higher than those in both the PKA and control groups （P<0.05), indicating greater myofiber damage. In addition, the peak activity of caspase-9 occurred earlier in the AP group (at 6h postmortem) than in both the control and PKA groups (at 12h postmortem), and the peak activity of caspase-3 in the PKA group occurred later (at 48h postmortem) than in the AP and control groups (both at 12h postmortem). In conclusion, dephosphorylation induced by AP treatment during postmortem increased mitochondrial functional damage, promoted caspase-mediated apoptosis, and facilitated myofibrillar protein degradation, ultimately exacerbating myofiber damage and improving the tenderness of postmortem meat.

Abstract:Agricultural equipment in the field operation process continues to produce mechanical vibration, capture its vibration energy and realize agricultural machinery self-power supply is of practical significance. For the application of vibration energy harvesting device can’t match the external excitation or mechanical structure vibration frequency, energy harvesting efficiency is low, a resonance frequency adjustable piezoelectric-electromagnetic coupling vibration energy harvesting device was designed, combining the use of bicrystalline piezoelectric cantilever beam structure, and the combination of electromagnetic induction and piezoelectric effect, to achieve the resonance frequency adjustable from 14.8Hz to 31.0Hz, and to broaden the bandwidth of the vibration energy harvesting. Based on Hamilton’s principle, an electromechanical coupling mathematical model was established, which revealed the regulation mechanism of the resonant frequency of the system. Through indoor experiments, the effects of different load resistances, excitation acceleration, excitation frequency, and magnetic distance on the power generation performance were investigated, and the optimal load resistance configurations and resonant frequencies under different magnetic distances were determined. Field tests were carried out by using EH2604 high power tractor to study the power generation performance of the vibration energy recovery device under the soil surface of farmland. The results showed that when the optimal load resistances for the cantilever piezoelectric patch, electromagnetic coil, and bending piezoelectric transducer were 850kΩ, 990Ω, and 300kΩ, respectively, and the optimal magnetic distance was 53mm, the vibration energy output power peaked. The device can effectively collect vibration energy under different magnetic distances and vehicle speeds, and the output power tended to increase with the increase of vehicle speed, and the magnetic distance was 38mm, and the maximum output power was 25.97μW at a vehicle speed of 12km/h. The research result can provide technical solutions and ideas for the development of self-supply energy system of agricultural machinery for reference.