DISTINGUISING BETWEEN SIMILAR OBJECTS BASED ON GEOMETRICAL FEATURES IN 3D PERCEPTION

In this study, we propose a meta-classifier to identify and distinguish between objects with similar geometrical shapes, colors or sizes. Because objects can be recognized based on different parameters such as texture, geometrical boundaries, shape or surface, a classifier which gathers all this features for an unknown point cloud cluster is presented. Objects can be initially classified based on its VFH features but the method here proposed is able to segment similar surfaces avoiding the scale independence and
therefore recognize successfully objects with same 3D shape but different scales or colors. Furthermore, the classifier metric distance is discussed to find out the most suitable for each feature. We present several experiments to validate the model and corroborate the meta-classifier here proposed.