I-014
Estimating the Positions of Target Objects Based on Object Recognition by Processing 3D Point Cloud Data

◎Qi Li・Jun Ohya・Atsuo Takanishi・Takashi Matsuzawa（Waseda Univ.）・Kenji Hashimo（Meiji Univ.）

Nowadays, robots grasping technic are being developed. Conventional methods to support grasping have some problems such as: not robust against changes in illumination conditions and cannot directly provide the position of the target object. This paper proposes a method for recognition and estimating the position of objects by analyzing 3D point cloud data. The procedure is as follows: (1) Pre-processing. (2) Segmentation. (3) Obtain the points’ features and match them. (4) Obtain coordinates. (5) Obtain the rotation and translation matrix. As the result of experiments, the accuracy of detection is 89 %. The mean error of estimating the distance is 0.835%.