Locating 3D edges by direct methods in motion based vision

Mobile vision leads to more robust methods of recovering the structure than those based on binocular stereo vision. To solve the structure from motion problem using direct methods can avoid both optical flow and correspondence computation. In this paper, we propose to obtain the location of 3D straight edges from a camera placed in the hand of a robot using these methods. The algorithm is based on the image regions supporting the geometric features and can be applied with any known camera motion.<>