2D grapevine winter pruning location detection method based on thinning algorithm and Lightweight Convolutional Neural Network

In viticulture, there is an increasing demand for automatic winter grapevine 1 pruning devices, for which detection of pruning location in vineyard images is a 2 necessary task, susceptible of being automated through the using of computer vision 3 methods. In this paper, we present a novel 2D grapevine winter pruning location 4 detection method for automatic winter pruning with Y-shaped cultivation system. The 5 method can be divided into following four steps. First, the vineyard image was 6 segmented by threshold in two times Red minus Green minus Blue (2R-G-B) channel 7 and S channel. Second, extract the grapevine skeleton by Improved Enhanced Parallel 8 Thinning Algorithm (IEPTA). Third, find the structure of each grapevine by judging 9 the angle and distance relationship between branches. Fourth, obtain the bounding 10 boxes from these grapevines, then pre-trained MobileNetV3_small×0.75 was utilized 11 to classify each bounding box and finally find the pruning location. According to the 12 detection experiment result, our method achieved a precision of 0.988 and a recall of 13 0.923 for bud detection, an accuracy of 0.834 for pruning location detection and a 14 total time of 0.423s. Therefore, the results indicated that the 2D pruning location 15 detection method we proposed had decent robustness as well as high precision that 16 could guide automatic device to winter prune efficiently. 17