Authors

Abstract

The soybean-maize strip intercropping mode has been widely popularized, yet existing plant protection machinery fails to dynamically adjust pesticide application according to crop zones, suffers from poor pesticide application accuracy, and is prone to causing crop damage. Aiming at the above problems, this study proposed a precise strip-zoning pesticide application method for soybean and maize and established an automatic strip-zoning pesticide application control system based on machine vision. Perspective transformation was adopted to correct image viewing angles; after subsequent processing via excess green binarization and opening-closing operations, a dynamic ROI (Region of Interest) positioning and recognition algorithm for crop rows based on the peak value method was applied. The least square method was used to extract crop row lines, the offset was calculated according to the geometric relationship of the centerlines of soybean and maize zones, and an automatic strip-zoning offset compensation model was constructed to achieve precise strip-zoning. Field experiment results show that: under conditions with little shadow and stable light, the recognition algorithm achieved the optimal performance with an accuracy of up to 92%. At operating speeds of 2–5 km/h, the automatic strip-zoning pesticide application control system had a mean strip-zoning deviation of ≤3.16 cm, with the proportion of deviations within 5 cm reaching 84% (up to 94%). Under wind speeds of 1.4–2.9 m/s, the maximum droplet drift deposition amount of the closed anti-drift device was 9.41 particles per cm², delivering favorable strip-zoning pesticide application performance, which provides data support and a reference for field precision plant protection operations under the strip intercropping mode.

Abstract in Chinese

大豆玉米带状复合种植模式广泛推广，但现有植保机械无法随作物带进行动态调整施药，且施药精度差，易造成伤害。针对上述问题，本文基于机器视觉，提出一种大豆玉米精准分带施药方法并搭建自动分带施药控制系统。采用透视变换校正图像视角，再经超绿二值化与开闭运算处理后，基于峰值法的作物行动态ROI（Region of interesting）定位识别算法，采用最小二乘法提取作物行线，依据大豆玉米带中心线的几何关系计算偏移量并构建自动分带偏移补偿模型以实现精准分带。田间试验结果表明：在阴影少、光照稳定条件下，识别算法效果最佳，准确率可达92%。自动分带施药控制系统在2-5km/h的速度下分带偏差均值≤3.16cm，5cm内偏差占比达84%（最高94%）。1.4m/s~2.9m/s风速条件下，封闭式防飘移装置的雾滴飘移沉积量最高为9.41个/cm2，具有较好的分带效果，为复合种植模式下的田间精准植保作业提供数据支撑与参考。