Road-feature extraction using point cloud and 3D LiDAR sensor for vehicle localization

This paper proposes a novel road-feature extraction method from a point cloud and a LiDAR (Light Detection and Ranging) sensor for vehicle localization. For a robust localization of a vehicle, it is important to extract unchanging features. Road markings such as lines, arrows, and crosses are effective features for localization in a road environment. Thus, this paper describes a method to extract road marking features from intensity information of both point cloud and a LiDAR sensor. The proposed method is effective for an uncalibrated LiDAR sensor. In addition, it is compatible for real-time localization by extracting only essential features from the road. The proposed method is demonstrated by an autonomous vehicle test conducted in a 2.8 km loop near One-north, Singapore.