To address challenges in hybrid rice seed production—specifically labor dependence, low uniformity of pollen distribution, and low operational efficiency—which collectively drive up large-scale production costs, technological innovations are critical. However, despite the demonstrated potential of UAV-assisted pollination, the quantitative relationships between its operational parameters (altitude, speed, flight patterns) and pollen dispersal dynamics remain poorly understood, impeding standardization efforts. In this study, guided by agronomic pollination requirements, we developed an integrated analytical framework linking “pollen density-yield” dynamics to elucidate the governing mechanisms of flight parameters on pollination quality. A DJI T50 UAV was used to carry out the assisted pollination test on two varieties of hybrid rice, Changtian You 405 and Wanxiang You 377, to explore the effects of different flight speeds, altitudes, and trajectories of the UAV on pollination quality and to evaluate the cost-effectiveness ratio, taking the yield and its composition as the evaluation indexes. The experimental results showed that the UAV flight operation parameters had a significant effect on the pollination quality, and the best pollination quality was obtained when the flight altitude was 4 m and the speed was 3 m/s, achieving yields of 2.64 and 3.15 t/hm²; the average yields of the UAV-assisted pollination were 2.10 and 2.61 t/hm², and the filled grain percentages were 15.76% and 34.2%, respectively. These increased the yields by 21.4% and 11.06%, respectively, and the filled grain percentages by 8.69% and 3.95%, compared with artificial pollination. The results also showed that the cost-effectiveness ratio of UAV-assisted pollination was 28.11% lower than that of artificial operation. The results indicate that UAVs have great application prospects in hybrid rice pollination.

Water-sensitive paper (WSP) has been widely used to assess the quality of pesticide sprays. However, fog droplets tend to overlap on WSP. In order to accurately measure the droplet size and grasp the droplet distribution pattern, this study proposes a method based on the optimized XGBoost classification model combined with improved concave-point matching to achieve multi-level overlapping-droplet segmentation. For different types of overlapping droplets, the corresponding improved segmentation algorithm is used to improve the segmentation accuracy. For parallel overlapping droplets, the centre-of-mass segmentation method is used; for non-parallel overlapping droplets, the minimum-distance segmentation method is used; and for strong overlapping of a single concave point, the vertical-linkage segmentation method is used. Complex overlapping droplets were gradually segmented by loop iteration until a single droplet was obtained or no further segmentation was possible, and then ellipse fitting was used to obtain the final single-droplet profile. Up to 105 WSPs were obtained in an orchard field through drone spraying experiments, and were used to validate the effectiveness of the method. The experimental results show that the classification model proposed in this paper achieves an average accuracy of 98% in identifying overlapping-droplet types, which effectively meets the needs of subsequent segmentation. The overall segmentation accuracy of the method is 91.35%, which is significantly better than the contour-solidity and watershed-based algorithm (76.19%) and the improved-concave-point-segmentation algorithm (68.82%). In conclusion, the method proposed in this paper provides an efficient and accurate new approach for pesticide spraying quality assessment.

The upright state of pot seedlings in the process of rice mechanized throwing operations has an important influence on the growth rate and yield of rice, and pot seedling uprightness is affected by the influence of soil backfilling during trenching. Due to the complexity of the furrow opener–soil–pot seedling interaction mechanism in the rice pot seedling planting process, the soil backfilling process is difficult to observe. In order to improve the uprightness of pot seedling planting, this paper constructs a soil model and a soil–pot seedling model step by step, based on the discrete element method (DEM), as well as a coupled model of the pot seedling planting system to study the process of furrow opener–soil–pot seedling planting, the reliability of which is then verified. The results showed that the simulation results of the constructed soil model and soil–pot seedling model deviated from the actual calibration results by <6%, and the model could accurately simulate the pot seedling throwing process. The simulation analysis of the trenching process revealed that the soil backfilling process during trenching showed a three-stage evolution pattern of “backfilling-covering-stabilizing”; in addition, the forward speed of the machine was 0.8 m/s, and the falling speed of the seedling discharge cylinder was 3.5 m/s, which made it possible for the model to simulate the pot seedling throwing process accurately. In addition, when the pot seedling with a forward speed of 0.8 m/s and a drop speed of 3.5 m/s fell into the trench after 0.15 s of trenching, its lateral and longitudinal uprightness were 67.0 ± 1.2° and 65.2 ± 1.5°, respectively. After optimization of the structure of the trenchers, the width, depth, and length of the main body were 40 mm, 37.87 mm, and 32.32 mm, respectively, and the lateral and longitudinal uprightness of the pot seedlings increased to 70.0 ± 1.0° and 69.4 ± 0.8, respectively. The coupled model bench validation test showed that its reliability error was <5%. The coupled model provides technical support for the design and parameter optimization of rice planting equipment.

UAV pollination holds significant promise for enhancing hybrid rice seed production, yet the mechanisms of pollen diffusion under UAV downwash and the lack of theoretical guidance for operational parameter optimization remain critical challenges. To address this, this study employed a coupled Computational Fluid Dynamics–Discrete Phase Model (CFD–DPM) numerical simulation to systematically investigate the interaction between the UAV-induced wind field and pollen particles. A validated CFD model was first developed to characterize the UAV wind-field distribution, demonstrating good agreement with field measurements. Building upon this, a coupled wind field–pollen CFD–DPM model was established, enabling a detailed visualization and analysis of airflow patterns and pollen transport dynamics under varying flight parameters (speed and height). Using the pollen disturbance area and effective settling range as key evaluation metrics, the optimal pollination parameters were identified as a flight speed of 3 m/s and a height of 4 m. Field validation trials confirmed that UAV-assisted pollination using these optimized parameters significantly increased the seed yield by 21.4% compared to traditional manual methods, aligning closely with simulation predictions. This study establishes a robust three-tier validation framework (“numerical simulation—wind-field verification—field validation”) that provides both theoretical insights and practical guidance for optimizing UAV pollination operations. The framework demonstrates strong generalizability for improving the efficiency and mechanization level of hybrid rice seed production.

A multi-rotor UAV-based oscillating-tube broadcasting device incorporating replaceable material-specific screw metering units was developed for low-rate seed and granular-fertilizer application. The key structural parameters of the oscillating tube were screened through theoretical analysis and single-factor experiments using pre-germinated conventional Huanghuazhan rice. An L16(4³) orthogonal experiment was then conducted solely for exploratory screening of the main effects of flight height, forward speed, and broadcasting material on the single-pass broadcasting-uniformity coefficient of variation (${\mathrm{C}\mathrm{V}}_{\mathrm{b}}$) and effective spreading width (We). Forward speed significantly affected ${\mathrm{C}\mathrm{V}}_{\mathrm{b}}$, whereas all three factors significantly affected We, with flight height showing the strongest effect. Separate two-factor central composite design experiments were subsequently conducted for conventional Huanghuazhan rice, hybrid Yongyou 12 rice, Yangguang 131 rapeseed, and Stanley compound fertilizer. Material-specific quadratic models were established, yielding model-recommended flight height and forward speed combinations of 3.04 m and 2.35 m s⁻¹, 2.79 m and 1.70 m s⁻¹, 3.00 m and 1.80 m s⁻¹, and 3.20 m and 2.25 m s⁻¹, respectively. These combinations represent constrained model-based recommendations rather than global or independently verified optima. In a limited consecutive-pass field demonstration, the field-sample coefficient of variation (${\mathrm{C}\mathrm{V}}_{\mathrm{f}}$) ranged from 13.76% to 20.34%, and the mean absolute pointwise deviation from the target count ranged from 15.11% to 19.05%. The results demonstrate the preliminary mechanical and operational feasibility of the prototype under the tested materials, operating settings, site, and environmental conditions.

Aiming at the path tracking oscillation problem caused by sideslip of navigation agricultural machinery in complex farmland environments, this paper takes the Yanmar YR-70D transplanter as the research object and proposes an adaptive control method based on deep belief network (DBN) sideslip identification. Based on the preview tracking model, this method dynamically adjusts the preview distance by real-time identifying the sideslip status via DBN. In this research, the Optuna framework is utilized to optimize the DBN model, and the SHAP framework is introduced to quantify the feature contribution. The results show that, considering both real-time performance and accuracy, when the combined variable of “lateral deviation + heading deviation” with a data length of 20 is adopted, the prediction accuracy of the model reaches 88.04%, among which heading deviation (HD) has the highest contribution with a SHAP value of 0.957. Comparative experiments indicate that the accuracy of DBN is comparable to that of mainstream models such as LSTM and Transformer, but its forward propagation speed (0.0487 ms) is better than other models’. Upland field experiments verify that under different test factors of vehicle speed and slope angle, the line-on time (1.83–5.60 s), line-on distance (1.46–5.07 m), and maximum overshoot (0.32–15.75 cm) of the adaptive control group are all superior to those of the fixed-parameter control group. Paddy field experiments further confirm that under the working condition of 1.0 m/s vehicle speed and 15° slope, the above three indicators of the experimental group are 0.85–3.1 s, 0.75–3.14 m and 0–7.44 cm, respectively, all of which are better than those of the control group. The proposed method demonstrated superior path tracking performance under the operating conditions of the tested rice transplanting machines, indicating its potential application in lateral slip perception adaptive navigation control. However, further verification is still needed with different types of transplanting machines, field environments, and operating conditions to assess its general applicability.