Robust Image Matching for Camera Pose Estimation Using Oriented Fast and Rotated Brief

Abstract

This paper presents a novel image matching method for camera pose estimation based on point cloud segmentation. The Oriented Fast and Rotated Brief (ORB) is employed to extract the key points, which are then extracted based on matched point cloud planes. The point cloud planes are segmented based on the depth image, and then matched by the distance of the centroid between planes. The putative corresponding key points on the planes are generated based on the distance of their 3-D coordinates and the descriptors of the key points are further matched based on the putative corresponding key points. As an additional constraint, the spatial relative position in 3-D spaces solves the problem that the descriptors of each key point in some scenarios are too similar which may lead to a mismatch. According to the experimental results, the superiority of the proposed approach is illustrated by comparing with the existing matching methods.