Segmentation of point cloud from a 3D LIDAR using range difference between neighbouring beams

Apart from detecting traversable region, an autonomous mobile robot needs to robustly identify different objects in its surroundings. This information it can then use to localize itself, as well as to plan its path. Multi beam laser scanners like Velodyne are now-a-days commonly used to obtain range data of the environment in the form of point clouds. The geometry of laser scanning, particularly the range difference between neighbouring beams has been used by various researchers for the detection of obstacles. In this paper, we exploit the same basic concept in a different form and introduce the notion of an unevenness field that enables easy identification of obstacles in contrast to the ground. Then, we go ahead and use the same unevenness field with modified limits for segmenting point cloud into different objects. Finally, we demonstrate results of segmentation in various environments in our research centre. The conceptual and computational simplicity of this approach, and its effectiveness makes it an attractive method for range image segmentation.