Abstract

This present study, proposes a method for reconstructing a human model based on high-resolution feature template mesh from point cloud to establish a point-to-point correspondence between human bodies. First in order to get an ideal feature template, human body is segmented, feature contours are extracted and B-spline curve based on smooth constraint is introduced to fit contour point cloud. Second, an optimization match algorithm is described by fitting point cloud into template mesh. The objective function is a weighted combination of three measures: approximation to point cloud, smooth constraint and coplanar constraint. The proposed method has high match accuracy and speed. It's robust to data with noisy or ghosting.

Paper Details