Fast Keypoint Features From Laser Scanner for Robot Localization and Mapping

Abstract

Detecting features in sensor measurements and distinguishing among them is an important capability for robot localization and navigation. Despite the wide diffusion of range finders, there are few works on keypoint features for 2-D LIDAR and there is potential for improvement over the existing methods. This letter proposes two novel keypoint detectors for the stable detection of interest points in laser measurements and two descriptors for robust associations. The features defined by combining keypoints and descriptors allow stable and efficient place recognition. Experiments with standard benchmark datasets assess the performance of the detectors and descriptors investigated. One of the proposed features, termed FALKO-BSC, achieves higher repeatability score and similar descriptor performance compared with the FLIRT state-of-the-art feature. FALKO-BSC is also shown to enable effective localization.