Research On Skeleton Extraction Of 3D Tree Models

Most of the 3D models should be preliminarily processed in different ways so that they can be applied in scientific fields like CAD, medical imaging, computer graphics, scientific visualization, computational fluid dynamics and remote sensing, etc. One of the most important processing methods is skeleton extraction. The most difficult problem of skeleton-extracting process to be solved is the vague definitions of the curve-skeleton. Because of this, researchers have done a lot of studies and experiments on algorithms. Different demands and algorithms have different definitions of skeletons, parameters and settings of thresholds, so we can only demonstrate an algorithm in a limited collection of 3D objects. Because of the large volume and redundant property of 3D data, it is of great importance in extracting the skeleton of trees’ 3D models in plant and forestry applications.This paper conducts skeleton extraction algorithm research based on trees 3D triangular mesh models. The main research contents include the following aspects:1. The skeleton extraction methods based on feature points and iteration segmentation of 3D models of trees. Elaborate on geodesic distance and feature points of tree models’ surface, understand constraint condition of extracting feature points: distant mapping function and local extrema. Extracting feature points of trees 3D model is first utilizing the farthest geodesic distance of vertices pair. Then calculate the geodesic distance between mesh vertex and one of the vertices pair, getting the set of local extrema of vertices pair respectively. Then we use a threshold to help the intersection of the two set of extrema to get the set of feature points. Utilize the set of feature points to segment the tree models iteratively, connecting the result fragments, finally, the skeleton which extract by this method have nice Property.2. Skeleton extraction methods based on trees 3D triangular mesh contraction. Elaborate on Laplace operator of the mesh and Laplace mesh optimization framework and realize that mesh contraction is based on this framework. Mesh contraction method by cotangent-Laplace operator, and at the same time, adding traction constraints to constrain vertex normal, contracting the tree models, to generate a none-volume mesh. Utilize connectivity of mesh to remove its connectivity after it contracts. Utilizing a cost function to simplify edge-collapse of mesh and generating a one-dimensional linear skeleton. Finally, refine skeleton by building mapping function of mesh-skeleton.3. Analyzing and comparing the two methods of skeleton extraction results. Analyzing the two methods of skeleton extraction results respectively, and summarizing merit and demerit and range of application by comparing skeletal centrality and noise to sensitivity.