Abstract
The demand that a robot manipulates deformable objects. It is important for a robot to recognize these forms by using image information. Form estimation algorithm of deformable object based on image information is proposed in previous research. In this algorithm, first, paired images are captured by CCD camera with rotating around a target object. Secondly, corresponding points are obtained between paired images by application of optical flow. Finally whole scries of points of a target object in three dimension are reconstructed based on stereoscopic vision. It is important to discriminate whether the captured object is planar object or linear object from points group, so that a robot can calculate a manipulation planning of the object. In this research, algorithm to distinguish between a planar deformable object and a linear one from projection image on vertical section of points group is proposed.
