The existing discrete element model of wheat plants lacks the glume, which hinders the simulation of the entire threshing process. To address this issue, this paper takes wheat at the harvest stage as the research object and constructs a complete discrete element model of wheat plants with glumes based on the Hertz-Mindlin with bonding model in the EDEM simulation software. The parameter calibration of wheat glumes discrete element model is studied through collision bounce experiments, slope experiments, and accumulation experiments. The results show that the coefficient of restitution, coefficient of static friction, and coefficient of rolling friction between glume and steel are 0.488, 0.625, and 0.048, respectively, and the coefficient of restitution, coefficient of static friction, and coefficient of rolling friction between glume and glume are 0.232, 0.966, and 0.059, respectively. The relative errors between the simulation results and the measured values are less than 5%, and the calibration parameters are effective. Based on the structural parameters of the self-developed experiment-bed of tangential axial-flow grain threshing device, a three-dimensional model of the wheat threshing device is established to simulate the whole threshing process of the complete wheat plant, and the bench-scale experiments are carried out with the non-threshing rate as the performance index. The results indicate that the model can completely simulate the separation process of glume and grain and the movement law of different grains, and the relative error of non-threshing rate between the simulation experiments and bench-scale experiments is 4.36%. This further demonstrates that the proposed model can provide a reference for the wheat threshing process research and device performance optimization design.

Widespread infestation of pests and pathogens during winter wheat’s heading stage poses significant risks to yield loss. In this study, an assessment model of health degree (HD) of winter wheat under field conditions was established by using unmanned aerial vehicle remote sensing (UAV RS) imagery. Firstly, non-photosynthetic features were identified from the UAV RS imagery based on different machine learning methods, including Minimum Distance (MD), Maximum Likelihood Estimation (MLE), and Support Vector Machine (SVM). Classification results indicated that MD demonstrates the best performance, according to the values of Overall Accuracy (0.898), Kappa Coefficient (0.863), and Precision (0.856). Therefore, the inversion model between the proportion of pixels classified as non-photosynthetic features and the corresponding ground truth of the incidence of non-photosynthetic features was established. Coefficient of determination (R²), RMSE (root mean square error), and RRMSE (Relative RMSE) of the inversion model are 0.73, 4.86%, and 19.81%, respectively, demonstrating strong correlation and high accuracy. Subsequently, an assessment model for HD of the wheat field was generated based on the predicted incidence of the non-photosynthetic features, and the conclusion was reached that HD1 (pre-symptoms of the infestation of pests and pathogens) dominated in the wheat field, with the proportion of area as 56.16%, while HD4 and HD5 (severe infestation of pests and pathogens) were negligible, with proportions of area of 2.29% and 17.75%. Finally, the assessment model of HD was used to simulate the precision OSMP (One-Spray-Multiple-Protection), and the agricultural chemical could be reduced to 69.11% of the conventional OSMP operation, which provides theoretical and methodological support for the reduction of agricultural chemicals in the domain of precision agriculture.

To realize the visual navigation of agricultural robots in the complex environment of orchards, this study proposed a method for fruit tree recognition and navigation based on YOLOv5. The YOLOv5s model was selected and trained to identify the trunks of the left and right rows of fruit trees; the quadratic curve was fitted to the bottom center of the fruit tree recognition box, and the identified fruit trees were divided into left and right columns by using the extreme value point of the quadratic curve to obtain the left and right rows of fruit trees; the straight-line equation of the left and right fruit tree rows was further solved, the median line of the two straight lines was taken as the expected navigation path of the robot, and the path tracing navigation experiment was carried out by using the improved LQR control algorithm. The experimental results show that under the guidance of the machine vision system and guided by the improved LQR control algorithm, the lateral error and heading error can converge quickly to the desired navigation path in the four initial states of [0 m, −0.34 rad], [0.10 m, 0.34 rad], [0.15 m, 0 rad] and [0.20 m, −0.34 rad]. When the initial speed was 0.5 m/s, the average lateral error was 0.059 m and the average heading error was 0.2787 rad for the navigation trials in the four different initial states. Its average driving was 5.3 m into the steady state, the average value of steady state lateral error was 0.0102 m, the average value of steady state heading error was 0.0253 rad, and the average relative error of the robot driving along the desired navigation path was 4.6%. The results indicate that the navigation algorithm proposed in this study has good robustness, meets the operational requirements of robot autonomous navigation in orchard environment, and improves the reliability of robot driving in orchard.

The facility-based production method is an important stage in the development of modern agriculture, lifting natural light and temperature restrictions and helping to improve agricultural production efficiency. To address the problems of difficulty and low accuracy in detecting pests and diseases in the dense production environment of tomato facilities, an online diagnosis platform for tomato plant diseases based on deep learning and cluster fusion was proposed by collecting images of eight major prevalent pests and diseases during the growing period of tomatoes in a facility-based environment. The diagnostic platform consists of three main parts: pest and disease information detection, clustering and decision-making of detection results, and platform diagnostic display. Firstly, based on the You Only Look Once (YOLO) algorithm, the key information of the disease was extracted by adding attention module (CBAM), multi-scale feature fusion was performed using weighted bi-directional feature pyramid network (BiFPN), and the overall construction was designed to be compressed and lightweight; Secondly, the k-means clustering algorithm is used to fuse with the deep learning results to output pest identification decision values to further improve the accuracy of identification applications; Finally, a detection platform was designed and developed using Python, including the front-end, back-end, and database of the system to realize online diagnosis and interaction of tomato plant pests and diseases. The experiment shows that the algorithm detects tomato plant diseases and insect pests with mAP (mean Average Precision) of 92.7%, weights of 12.8 Megabyte (M), inference time of 33.6 ms. Compared with the current mainstream single-stage detection series algorithms, the improved algorithm model has achieved better performance; The accuracy rate of the platform diagnosis output pests and diseases information of 91.2% for images and 95.2% for videos. It is a great significance to tomato pest control research and the development of smart agriculture.