A leaf-wall-to-spray-device distance and leaf-wall-density-based automatic route-planning spray algorithm for vineyards

In this study, to achieve precision spraying of pesticide in vineyards and reduce pesticide waste and pollution, algorithms were proposed for estimating parameters such as the average distance between the leaf wall and the spray device (LW-SD distance) and the leaf-wall density by integrating colour and depth images acquired by Microsoft Kinect. First, the colour video images were segmented using a morphological image segmentation technique to accurately separate the leaf walls from the remaining images. Then, based on the depth images, algorithms for estimating the average LW-SD distance and the leaf wall density were established to estimate the spray control parameters. A spline area-based algorithm for estimating the average LW-SD distance was proposed to accurately estimate the LW-SD distance. Finally, an algorithm for estimating the route deviation, as well as an algorithm for correcting and planning the spray route, was established to guide and maintain the spray device on the optimal path. The experimental results demonstrate that the spray parameter estimation algorithms produce relatively small errors in the estimation of the average LW-SD distance and the leaf wall density. In addition, the differences between the spray distance and route deviation estimated using the route planning algorithm and the measurements were also relatively small, demonstrating the accuracy of the algorithm.