3D object mapping by integrating stereo SLAM and object segmentation using edge points

Abstract

This paper presents a method of 3D object mapping using a binocular stereo camera. The method employs edge points as map element to represent detailed shape and applies a variant of ICP algorithm to 3D mapping. A SIFT descriptor is attached to each edge point for object recognition and segmentation. The 3D map is segmented into individual objects using training images of target objects with different backgrounds based on SIFT-based 2D and 3D matching. In experiments, a detailed 3D map was built by the stereo SLAM and a 3D object map was created through the segmentation of the 3D map into 36 object instances. © 2009 Springer-Verlag.