A new Robust Method of Line Detection in a Structured Light System

摘要

This paper presents an integrated 3D face scanning system using the structured light technique. A new pattern consisting of horizontal colored strips using the De_Bruijn sequence is designed in a manner that can ensure a minimal distance between any two strips of the same color. After illuminating the scene with this pattern, a first image is taken and used to obtain 3D information. To detect the strip centers in the captured image we use a smoothing Gaussian filter with a large kernel applied to the V component in the HSV color space. The size of this kernel is computed separately for each column. Then, a connection algorithm is used to connect the isolated points in the detected strips. The next step is to determine the colors of these detected strips. For this purpose we exploit the fact that the resulting strips are connected. Firstly, we use the H component of the HSV color space to determine the color of the easy sets of pixels in these strips. Then, we apply a region growing algorithm to assign the colors to the remaining pixels. Finally, each connected and colored part of strip is matched to a strip in the projected pattern and triangulated with its corresponding one in the projected pattern to generate the 3D model. Using the concept of connected strips we exploit the information from several columns to take the decision in each column, which enhances the robustness of the method used here versus the existed ones. A second image of the scene without illumination is obtained and used to add the texture to the reconstructed 3D model. Experiment results show an accurate 3D resolution with this technique.