The Feature Points Extraction for the Apparel Manufacture from 3D Human Body Scan Data

J. Shin and S. Ozawa (Japan)


3D body scan data, spin-image matching, countour tree al gorithm, automatic recognition


We present a novel method for extracting the 23 feature points on 3D human scans to make clothes not using land mark. In order to extract the feature points, we distribute the feature points into two types which are curvature fea ture points and structure feature points. The curvature fea ture points are determined from curvature distribution of a neighboring area from the feature points which have com mon chracteristics between individual body shapes. The structure feature points can not be determined from cur vature distribution alone because of the great differences between individual body shapes. Therefore, in this paper, we propose the two methods for extracting the curvature feature points based on the shape of a standard model and the structure feature points by using a contour tree algo rithm(which is a graph consisting of the structural relation ships between the features extracted from the shape of an objective model). Finally, we evaluate our algorithm on 11 human body models with different shapes and the results in the experiment shows that our proposal method was effec tive.

Important Links:

Go Back