3D Map Merging Based on Overlapping Region

Abstract

Aiming at the problems of low matching efficiency and large computation time in SLAM (Simultaneous Localization and Mapping) 3D map merging of multi-robot, we propose a 3D map merging algorithm based on overlapping regions. This algorithm uses the viewpoint feature histogram and region growth segmentation to search overlapped region. Aiming at the low differentiation degree of traditional similarity function, we build a new similarity measure function by adding weight coefficient. In addition, we propose KP-PDH (Key Point and Point Dense Histogram), a new 3D descriptor, which is composed of the key point in the new coordinate and the point density projected onto the 2D plane by the neighborhood of the key point. KP-PDH descriptor effectively improves the efficiency of 3D map merging. Experimental results show that the algorithm of searching overlap region and KP-PDH descriptor can effectively reduce the matching time and mismatching rate. In this paper, our proposed algorithm increases the speed of 3D map merging, reduces the mismatching, and can be effectively applied to the multi-robot SLAM 3D map merging.