Feature Line Extraction System for Painter Robots

Our laboratory is constructing a painter robot. The first step in painting is to extract desirable feature lines such as silhouette and boundary lines, which represent the object to be drawn. For this purpose, this thesis proposes a system that extracts feature lines from a textured 3D polygonal model. First, the system builds a textured 3D model of the target object from multiple cameras using the technique of multiple-view geometry. Here we employ a graph-cut algorithm to refine the separation between the foreground and the background region in the captured images. As a result, the system can generate a high-quality 3D model. Then, the system extracts and combines two types of edges: geometry edges obtained from polygonal data, and photometry edges obtained from a rendered image. Both of them are observed from a virtual viewpoint. We apply the proposed system to a textured 3D polygonal model of a human and verify that the system is effective to obtain appropriate feature lines for drawings.