An improved point cloud registration three-stage (3S) method using RGB-D camera

Abstract

An improved point cloud registration Three-Stage (3S) method using RGB-D camera has been proposed in this paper. The proposed 3S registration method consists of three steps starts by motion transforms the source point cloud, initially aligns the motion transformed source point cloud with Sample Consensus Pre-rejective (SACP) algorithm and lastly refines the registration by applying Iterative Closest Point With Normals (ICP-WN) algorithm. The proposed point cloud 3S registration method has been implemented on a P3-AT wheeled mobile robot. Results from the conducted experiments show improvement in point cloud registration quality (reduction in visual odometry and loop closure problem) when compared with three other registration methods.