3D Tree Models Reconstruction from a Single Image

Abstract

Due to their geometric complexity, reconstructing 3D tree models is a challenging problem. In this paper, we propose an approach to reconstruct 3D tree models using only a single image of trees without leaves, which is segmented from the background. This reconstruction process requires less user interaction. Firstly, the undirected graph topology of the plant is extracted from the given tree image. This is performed after image thinning, pixels classification, fake-branches deleting, line-segments fitting and width estimation. Secondly, the tree topology of the plant is obtained in terms of the plant growth regularity by determining the direction, the parent, and the first child of each edge in the undirected graph. Finally, the 3D tree models are reconstructed by rotating the branch and creating the 3D geometry of each branch based on the tree topology of the plant