Walnut shell–kernel separation is an essential step in the deep processing of walnut. It is a crucial factor that prevents the increase in the added value and industrial development of walnuts. This study proposes a walnut shell–kernel detection method based on YOLOX deep learning using machine vision and deep-learning technology to address common issues, such as incomplete shell–kernel separation in the current airflow screening, high costs and the low efficiency of manually assisted screening. A dataset was produced using Labelme by acquiring walnut shell and kernel images following shellshock. This dataset was transformed into the COCO dataset format. Next, 110 epochs of training were performed on the network. When the intersection over the union threshold was 0.5, the average precision (AP), the average recall rate (AR), the model size, and floating point operations per second were 96.3%, 84.7%, 99 MB, and 351.9, respectively. Compared with YOLOv3, Faster Region-based Convolutional Neural Network (Faster R-CNN), and Single Shot MultiBox Detector algorithms (SSD), the AP value of the proposed algorithm was increased by 2.1%, 1.3%, and 3.4%, respectively. Similarly, the AR was increased by 10%, 2.3%, and 9%, respectively. Meanwhile, walnut shell–kernel detection was performed under different situations, such as distinct species, supplementary lighting, or shielding conditions. This model exhibits high recognition and positioning precision under different walnut species, supplementary lighting, and shielding conditions. It has high robustness. Moreover, the small size of this model is beneficial for migration applications. This study’s results can provide some technological references to develop faster walnut shell–kernel separation methods.

In order to explore the equivalent static condition of the impact damage of the Korla fragrant pear under free fall, the Korla fragrant pear was chosen as the research object in this paper, and the height of free fall and rebound were recorded by a high-speed camera to obtain the impact damage energy of the fragrant pear. In addition, a mechanical compression test was conducted on the fragrant pear to obtain its static deformation energy, and the equivalent static condition of the impact damage was calculated based on the mechanical energy conservation formula using the damage volume of the fragrant pear as the evaluation standard. The test verification was carried out on the damaged fragrant pear using a scanning electron microscope test. This study showed that the falling and rebound height of the fragrant pear within the range of the test parameters and under the free fall test conditions conform to linear fitting. The coefficient of determination (R²) was 0.9951, and the impact damage occurred when the corresponding energy reached 205.89 × 10⁻³ J. When the damage energy was within 203.4 × 10⁻³ J and 498.23 × 10⁻³ J, the static damage volume of the fragrant pear was linearly correlated to the impact damage volume, R² was 0.9944, and the maximum deviation between the predicted damage volume and the actual damage volume of the fragrant pear in the verification test was 4.4%. These data can provide a reference for research on the equivalent method of impact damage of the Korla fragrant pear, and provide theoretical guidance to the fruit industry to reduce damage and increase efficiency.

A multi-duct orchard sprayer is designed in this paper to address the problem of low droplet coverage during plant protection in walnut orchards in South Xinjiang, China. This spray comprises a base frame, a pesticide tank, an air delivery system, a copying base frame and a drive system. Effects of the sprayer structure on the flow field were simulated using computational fluid dynamics. Operation parameters were optimised using the response surface analysis and validated using a field test. The results demonstrated that five jet tubes in the simulation design could fulfil the design requirements of GB/T 32250.3-2022. The upper, middle and lower droplet coverages were 74.92%, 90.01% and 69.9%, respectively, when the sprayer’s advancing speed, the angle of jet tubes and the nozzle diameter were 0.53 m/s, 70.27° and 0.81 mm, respectively. The machine can effectively enhance the spraying efficiency and droplet coverage of each canopy during plant protection operations in the new walnut orchards of South Xinjiang, China, and provide references for designing and optimising fruit tree spraying machinery.

Effective post-emergence weed control in maize must suppress existing weeds quickly while also limiting inter-row vegetation after application without sacrificing field efficiency. To address this requirement, a 4.5 m-wide machine-vision-guided air-curtain sprayer was designed to coordinate weed-targeted foliar treatment with low-dose residual application between crop rows. Real-time perception-to-actuation control was implemented through three-camera visual sensing, a lightweight YOLO11n detector, grid-intersection triggering, and spatiotemporal delay compensation. On a held-out test set from the same site-season dataset, the detection module achieved a precision of 0.96, recall of 0.88, mAP@0.5 of 0.94, and mAP@0.5:0.95 of 0.66, with a recognition-to-control-output latency of 120−180 ms. Ground tests showed a clear speed effect: target hit rate remained at 93%−94% at 2.0−3.0 km/h but declined to about 70% at 6.0 km/h, indicating 4−5 km/h as a practical operating range under the tested ground conditions. A single-site, single-season field experiment compared uniform spraying (UNI), targeted spraying (TS), targeted plus low-dose residual spraying (TSM), and an untreated control (CK). Multi-temporal UAV and hyperspectral observations indicated that, within the monitored period after application, TSM preserved a clearer row–inter-row vegetation pattern than TS. Relative to UNI, TSM reduced total formulation volume applied by 47.3% and was associated with the highest mean single-plant grain yield at maturity (180.2 g plant−1), compared with 139.0 g plant−1 under CK. Taken together, the results support coordinated targeted foliar spraying and low-dose residual inter-row application as a more input-efficient option for post-emergence weed management in maize.

To address the issues that conventional least squares fitting of main navigation lines tends to deviate from seedling populations and visual row detection suffers from low reliability during the sparse and discretely distributed maize seedling stage, this paper proposes a GNSS–vision fusion navigation method based on point-by-point connection driven by the prior spatial distribution of seedling points. In the global stage, seedling points extracted from UAV orthophoto imagery serve as the prior input, connected point-by-point according to directional consistency and adaptive distance constraints to form chains, from which a directional histogram extracts the main trend and eliminates outliers to generate a main navigation line that follows the natural orientation of the seedling population. In the local stage, the front-mounted camera detects near-field discrete seedling distribution and converts it into a look-ahead deviation as PID feedforward compensation to correct GNSS tracking commands in real time, forming a collaborative framework of "global prior line determination + local feedforward seedling avoidance." Field experiments using lateral deviation MAE as the metric compared three modes: GM, VM, and FM. The front-wheel MAE values were 6.68 cm, 4.30 cm, and 3.80 cm, respectively, and the rear-wheel values were 4.70 cm, 3.48 cm, and 2.68 cm, respectively, with the proposed method achieving the best performance in both cases. Results demonstrate that the point-by-point connection strategy effectively extracts the main navigation trend under missing-seedling and row-gap conditions, while the feedforward compensation mechanism improves trajectory smoothness.