3D LIDAR-based ground segmentation with l1 regularization

Obtaining a comprehensive and accurate model of the complex ground is not only crucial for autonomous driving in the urban and countryside environments, but also the base of the successive obstacle detection and classification. This paper presents an improved ground segmentation method for 3D LIDAR point clouds. According to the distribution character of the 3D LIDAR data, An individual terrain scan is represented as a circular polar grid map, which is then divided into a number of segments. In order to constraint the complexity of the structure of the ground in each segment, l₁ Regularization is used to extract ground for every segment. Experiments are carried out on our Autonomous Land Vehicle in different outdoor scenes. The results show that our approach can get a promising performance.