抄録
For a mobile robot exploring in an unknown static environment, self localization and recognition of target objects are essential problems. In this study, we describe a method to recognize a 3D grid structure using a fixed camera and a RGB-D camera system. Tracking algorithms using a fixed camera are tested along with the Scale Invariant Feature Transform (SIFT) for feature extraction and setting Region of Interest (ROI) with a color marker to improve the efficiency and accuracy of the tracking algorithm. The algorithm of point clouds registration without using feature points matching is also introduced in this work. Finally, the results of recognition of a 3D grid structure with relatively low computational cost is shown experimentally.
