3D Reconstruction of Chinese Hickory Trees for Mechanical Harvest

Abstract

Vibratory harvest is the most appropriate and efficient way to harvest Chinese hickory (Carya cathayensis Sarg.) nuts because of the hilly growth environment and scattered shooting architecture. To develop a mechanical harvester with tree dynamics parameter identification, a three dimensional (3D) machine vision system to map Chinese hickory trees for mechanical harvest is required. This paper proposes an approach for 3D reconstruction of Chinese hickory trees based on two images with complex background. In the proposed system, tree branches are roughly segmented from image by stripe programming, and refined by the active contour. Then, a graph representing 2D tree structure is established based on the extracted skeletons. To improve the estimation accuracy, branch-to-branch correspondence instead of point-to-point correspondence is employed to recover the 3D skeleton of tree branches. Only few graphical user interfaces are used to generate the seeds of the stripes, and refine map of branches. An example is presented, and the result demonstrates that the reconstructed 3D Chinese Hickory tree is similar to the actual one.