Hilly and mountainous areas serve as important production bases for grain, oil, sugar and distinctive agricultural products in China. However, restricted by complex topography and landforms, the level of agricultural mechanization in these regions is roughly 20% lower than the national average, posing a huge challenge to the promotion of agricultural mechanization. At present, agricultural production in hilly and mountainous areas still relies mainly on manual labor, which is inconsistent with the actual demand due to population aging and the loss of young and middle-aged labor force. As a power platform in hilly and mountainous areas, electric tracked vehicles feature excellent trafficability and environmental friendliness. Nevertheless, the existing motor control theories are difficult to meet the operational requirements of tracked vehicles in such scenarios. Specifically, permanent magnet synchronous motors (PMSM), which is viewed as power core of agricultural tracked vehicle, exhibit unsatisfactory control performance in the zero and low-speed range, making it difficult for agricultural tracked vehicles to achieve precise operation in hilly and mountainous areas. To address this issue, a hybrid control strategy combining the high-frequency signal injection method and a nonlinear observer is proposed based on existing research foundations and the actual operating conditions of hilly and mountainous regions. Through a systematic analysis of the impacts of inverter dead time and nonlinear factors on the motor operating characteristics, a corresponding compensation scheme is designed to effectively attenuate the adverse effects caused by the aforementioned factors. The mathematical model of a PMSM is established, and the three-phase current waveforms of the motor under the non-compensation scheme and the hybrid compensation strategy for dead time and nonlinear factors are compared. The results demonstrate that the dead-time effect and chattering phenomenon in the motor current are significantly suppressed after compensation, and the stability of the current waveform is greatly improved. An experimental platform is built, and motor starting and operating condition switching tests are carried out under set load conditions. The experimental results verify the effectiveness of the proposed hybrid control strategy; the system can maintain stable operation even during the switching process between the high-frequency signal injection method and the flux observer. Data comparison reveals that compared with non-compensation control, the total harmonic distortion (THD) of the motor current is reduced by 6.34 percentage points under no-load conditions and by 5.26 percentage points under load conditions after adopting the hybrid compensation strategy. Meanwhile, a comparative test of super-twisting sliding mode active disturbance rejection control shows that the control strategy can basically eliminate the speed overshoot during motor operation and further enhance the control precision of the system. It should be acknowledged that satisfactory progress has been made in the present research. However, limited by experimental conditions, only simulations and tests under partial operating conditions have been completed. Further optimization of the switching strategy and improvement of the field test scheme are required in future work, so as to fully verify the robustness of the hybrid algorithm.

Maize is one of the most significant cereal crops worldwide. It is often required to realize the seed purity and accurate maize counting for high-quality control in modern production. However, conventional computer vision cannot fully meet the requirements of the large-scale performance during conveyor-belt inspection, including the spatial dense distribution of multiple seed varieties, subtle morphologies among categories, and motion-induced blur under the relative movement between the camera and the seeds. In this study, an improved YOLOv10n architecture was proposed to dynamically detect and then count maize seeds in real time. Specifically, the framework was also optimized for high-performance detection on the moving conveyor belts. The YOLO-corn detection model significantly enhanced the baseline YOLOv10n architecture using four structural improvements. 1) RFAConv (Receptive Field Attention Convolution) was integrated into the redesigned C2fE modules. The feature smearing effect of motion blur was mitigated in the standard convolutions with fixed parameters. Spatial features within the receptive field were adaptively re-weighted to concentrate on discriminative micro-textures, such as the seed embryos and grain contours. 2) Diverse Branch Block (DBB) was incorporated to extract the shallow feature. A multi-branch topology during training was utilized to capture diverse scale-space information after structure re-parameterization. Subsequently, a single kernel of inference was fused to enhance local edge perception with less computational overhead. 3) The lightweight ADown down-sampling module was utilized to avoid the loss typical of conventional pooling. The spatial fidelity was combined with the parallel average pooling and strided convolution. Structural cues were then preserved throughout the hierarchy. 4) A composite FPIoU-v2 loss function was proposed to accelerate convergence. The segmented linear re-weighting of Focal-IoU was coupled with the pixel-level boundary sensitivity of PIoU. The loss function was effectively recalibrated for the hard-to-detect overlapping samples after training. The BoT-SORT algorithm was also implemented for the temporal tracking task. Camera motion compensation was used to filter out mechanical vibrations, while an improved Kalman filter was used for smoother trajectory estimation. A virtual line-crossing logic was further integrated to map trajectories into discrete counts, effectively neutralizing redundant counts by tracking ID instability. Experimental evaluations demonstrated that the YOLO-corn framework substantially outperformed existing benchmarks. According to a self-curated dynamic dataset with five varieties of maize seed, the better performance of the model was achieved in a Precision of 89.2%, a Recall of 88.4%, and an mAP@0.5 of 94.0%. Compared with the original YOLOv10n, the performance was improved by 2.1, 2.1, and 1.6 percentage points, respectively, while the number of parameters increased by only 0.17 M. In terms of throughput, a high inference speed of 110.5 and 49.5 FPS on a high-performance workstation and an NVIDIA Jetson Nano after TensorRT optimization, respectively, indicates its readiness for edge deployment. Ablation studies confirmed that the synergistic interaction between RFAConv and DBB was crucial to reducing the motion-induced blurring. Furthermore, counting experiments revealed that an average accuracy exceeded 89.3% at low-to-medium velocities (0.1-0.3 m/s) and stayed above 81.5% even at the increased speeds (0.4-0.5 m/s), indicating strong operational robustness. The YOLO-corn framework can offer a robust, high-accuracy, and real-time solution for the online monitoring of the moving maize seeds. These findings can provide a technical foundation for the agricultural inspection to balance the architectural complexity with detection fidelity. Physics-based deblurring pre-processing and 3D point-cloud data can be expected to resolve extreme occlusion and stacking in high-speed industrial environments.

Field roads have been the most important transportation ways for commercial grain and agricultural products in hilly and mountainous regions in China. Especially, the youth labor can rapidly migrate from the rural to the urban areas against ever-increasing urbanization. Current manual production cannot fully meet the large-scale and precision agriculture in recent years, owing to the protracted yield cycle and low rate of return. Therefore, it is urgent to implement the mechanization and intelligent agriculture for the national food security in hilly and mountainous regions. Among them, the autonomous and safe operation of intelligent machinery can be critical to navigate or circumvent obstacles on field roads. However, the conventional machine vision cannot accurately and rapidly construct the visual guidance lines in such terrains. In this study, the extraction was proposed to enhance the visual guidance and obstacle avoidance using millimeter-wave radar and vision fusion. The state information was also detected from the target objects on the field road. The specific steps were as follow. Preprocessing techniques were employed to filter the portion of the radar object data. A multi-target tracking was utilized to eliminate the interference data for the continuous tracking of dynamic objects. Accurate radar object data was then obtained for subsequent data fusion. A semantic segmentation network was created using Deeplabv3+, and then leveraged a dataset of the adjacent field roads. The millimeter-wave radar and vision data were synchronized in both time and space via the timestamp alignment and least squares-based coordinate transformation. Extraction approach was then established for the visual guidance lines, particularly for the scenarios where dynamic target state was both available and unavailable. A series of experiments were carried out to validate the extraction of visual guide line. The average errors of detection were ranged from 1.60 to 9.20 pixels at the real road midpoints in the scenes without dynamic targets. Moreover, the visual guidance lines were successfully extracted for the obstacle avoidance proactively when encountering dynamic targets. Evidently, the integrated approach was effectively overcome the constraints of traditional machine vision, thereby enhancing the safety and reliability of machinery operation in hilly and mountainous terrains. Meanwhile, the efficacy was depended mainly on the precise projection of millimeter-wave radar data onto the visual plane. Simultaneously, the variables were markedly introduced the positional discrepancies, such as the posture of platform (including pitch and roll). Furthermore, the driveability of the operational area was effectively evaluated in the forthcoming regions. At the same time, the real-world conditions were more complex in the frequent presence of common elements like livestock, vehicles, and tricycles. The resilience of approach can be expected to improve the radar data processing for the detection of static objects. Moreover, positional deviations can also be mitigated from mechanical movements. As such, the precise visual guidance lines were established in the hilly and mountainous terrains. Therefore, the radar data processing can be focused on the dataset expanding and position deviation that induced by various factors. The intelligent level of agricultural machinery can be improved in hilly and mountainous area. The finding is of great significance to consolidate national food security.