Rapid identification and segmentation method of fruit trees based on radar point cloud

Accurate identification of fruit trees in production of orchard automation is an important prerequisite for orchard accurate operation．In view of problem that it is difficult to accurately and quickly identify a single fruit tree in a complex orchard environment, fruit tree recognition was studied based on lidar 3D point cloud technology, and an accurate, fast and highly automated fruit tree identification method and device was proposed．By constructing a multi-dimensional sensing fusion identification device, characteristic point cloud data of fruit trees was scanned, and point cloud dataset of fruit trees was established, and the data was screened by point cloud quality and effective angle filtering．Boundary feature point recognition algorithm was used to quickly extract boundary feature data of a single fruit tree, and effective scanning data of fruit trees was constructed by using boundary area features．It can be applied to rapid identification of contours and characteristics of individual fruit trees．In order to verify accuracy of fruit tree identification method, fruit tree canopy contour measurement accuracy test and fruit tree group identification test were carried out, and results showed that this method could effectively identify fruit trees, and comprehensive measurement accuracy of physical indicators was greater than 97.125%, which could provide an effective basis for accurate operation of orchard．