3D Tree Skeleton Reconstruction Based on Enhanced PyrLK Optical Flow Algorithm

The reconstruction of tree skeleton suffers from the problems of low accuracy and heavy user interaction. In this paper, we propose a new reconstruction method based on an enhanced Pyr LK optical flow algorithm to alleviate such problems. First, affine transformation of feature points and interactive tracking are incorporated into traditional Pyr LK to improve the robustness of feature matching. Second, we also propose to use 3D voxel-based flooding and least square to extract the tree skeleton after point cloud reconstruction with higher efficiency and accuracy. Experiments show that our reconstruction method is effective and efficient in 3D tree skeletons.