Skeleton Abstraction Of 3D Scanned Human Body Model

With the development of 3D scanning technique, skeleton extraction from scanned human body model is one of the most active research areas in virtual human modeling. The method of skeleton extraction based on 3D scanned human body firstly builds realistic human body models with the human body data by 3D scanner, and then extracts skeleton information of the model. Many methods are proposed to deal with skeleton extraction. However, there are still some shortcomings in these methods. Some methods are dependent on postures of models, so the results are not robust, when the posture is changed slightly. Some methods are not efficient, because many markers have to be placed on models by hand. In this paper, we research key technique about end points extraction and skeleton extraction of the human body model. We propose and realize related algorithms. The main contributions are as follows: 1) An accurate and robust end feature point extracting method. We propose an end feature point extracting method based on geodesic distance. The method searches the end feature points by calculating the maximum of geodesic distance of all points. By experiments on scanned human body models, this method is proven better than methods based on the sum of geodesic distance of all points. Time complexity decreases from O(n3) to O(n2). 2) Skeleton extracting method. We propose a method to meet the requirement of skeleton extracting. The method is based on Morse Theory, and an exquisite Morse function is defined as geodesic distance from points to the feature point on top of head, which can be calculated by end feature point extracting method. Experiments prove that our method is pose-independent, efficient, and automatic. To compare with the method based on ideal Morse function, time complexity of our method decreases from O(n3) to O(n2).