To solve the problems of high labor costs, a low weeding rate and a high seedling injury rate in the direct seeding of rice fields, this paper presents a reciprocating inter-row weeding machine for strip-seeded rice. The machine uses a combination of weeding wheels and weeding shovels to improve the efficiency of weeding between rice rows. Its reciprocating mechanism was designed and optimized. The simulation model of weeding teeth–paddy soil interaction was established in EDEM. The structural parameters of the weeding teeth were optimized, and the bending angle of the optimized weeding teeth was 55°. A prototype trial production and field tests were carried out. The results showed that the prototype’s inter-row weeding rate was between 80.2% and 85.3% and the seedling injury rate was between 3.5% and 5.1% when the prototype’s working speed was 1~3 km h⁻¹. The faster the speed of the prototype, the lower the inter-row weeding rate and the higher the seedling injury rate.

Improving sowing quality is crucial for ensuring wheat emergence and healthy growth. To address issues of poor wheat sowing quality, such as uneven sowing depth and inadequate soil coverage, in the Yangtze River Delta region of China, this study systematically analyzed the effects of the implement’s structural and operational parameters on sowing quality. Based on this analysis, a double-shaft rotary tillage and double-press seeder was designed. Protrusions on the grooving press roller are used to form seed furrows, rotary tiller blades cover the seeds with soil, and the rear press roller compacts the soil. DEM-MBD (discrete element method–multibody dynamics) coupled simulations, combined with single-factor and central composite design (CCD) experiments, were conducted with seeding depth as the evaluation index and four experimental factors: the protrusion height on the press grooving roller, forward speed, seed mass in the seed box, and straw mulching amount. The optimal protrusion height was 29 mm. The effects of rotary tiller blade working depth, rotational speed, and forward speed on soil-covering mass and its coefficient of variation were evaluated through discrete element method (DEM) simulations. The optimal working depth and rotational speed were found to be 55 mm and 350 r·min⁻¹, respectively, based on single-factor and Box–Behnken Design experiments. Field experiments based on optimized parameters showed results consistent with the simulations. The qualified rate of seeding depth decreased as forward speed increased. The optimal forward speed was 4.5 km·h⁻¹, at which the average seeding depth was 25.7 mm, the qualified seeding depth rate was 90%, the soil-covering mass within a 50 cm² area was 143.2 g, and the coefficient of variation was 13.21%, meeting the requirements for wheat sowing operations.

Accurate ripeness assessment of tomatoes in natural orchard settings is challenged by severe occlusion, dense clustering, and scale variation among fruits. This paper introduces YOLOv8n–TiD, a lightweight detection framework designed to overcome these obstacles. The architecture enhances the YOLOv8n backbone with a novel C2f-iRD module, which integrates re-parameterized dilated convolutions and inverted residual blocks to enlarge the receptive field while retaining fine-grained texture details. For feature fusion, we propose C2f-iRMB in the neck to ensure cross-scale consistency. To address semantic drift during upsampling, we replace standard interpolation with the DySample operator. The detection head is re-engineered as TADAH (Task-Adaptive Dynamic Alignment Head), enabling genuine task decoupling via deformable convolutions and conditionally shared features. Additionally, we introduce F–PIoUv2, a regression loss that emphasizes medium-quality predictions and curbs excessive bounding box expansion. Evaluations on a custom dataset show that YOLOv8n–TiD cuts parameters by 45%, FLOPs by 19%, and model size by 41%, while raising mAP@0.5 by 1.9 points—all in real time. On Android devices, it sustains 30 FPS inference and generalizes effectively to a distinct cherry tomato cultivar. These findings confirm the method’s robustness in discriminating occluded, small, and visually similar maturity stages, providing a practical vision system for robotic harvesting and field-based grading.