关键词: 核桃树, 休眠期, 深度相机, 点云, 配准, 三维重建

Abstract: Aiming at the problems of low precision and high cost of 3D reconstruction of fruit tree in dormant period, a method of reconstruction based on consumer Kinect V3 camera was proposed under field conditions. In this paper, the dormant walnut tree was taken as the research object. Firstly, a visual perception system based on two Kinect V3 cameras was built in view of the effect of light in natural environment, and walnut tree point cloud data on both sides of 0° and 180° was collected. Secondly, the point cloud of walnut tree was coarse registration using the marked points, in which ICP algorithm was used to accurately register, and the complete point cloud was obtained. Finally, Alphashape algorithm was used for surface reconstruction of the complete walnut tree. The experimental results showed that the average registration error was 0.58 cm, and the size error of surface reconstruction was less than 2 mm. The reconstruction accuracy was high, which provided data support for automatic pruning of walnut tree.

Key words: walnut tree, dormant period, depth camera, point cloud, registration, 3D reconstruction