3D shape measurement of a large cloth extremely close to a fisheye stereo camera

Existing 3D stereo shape measurement techniques do not consider the case of a large item very close to the stereo camera (i.e., covering almost the entire field of view). This case is crucial for humanoid robots with a head-mounted stereo camera performing support tasks because this is a natural configuration when handling large objects (e.g., adult clothes, bath towels, and linens). To address this, this paper describes the development of a 3D shape measurement method using a fisheye stereo setup. The method is successfully applied to measure the 3D shape of a large coat hung over the arms of a humanoid robot. However, as the object approaches the camera, new 3D shape measurement challenges arise. The causes of these difficulties are clarified, and suitable improvements to the algorithms are proposed. Experiments validate the effectiveness of the proposed improvements.