Detection of stone obstacles in grass by LIDAR intensity clustering

Abstract

This paper proposes an obstacle detection method using reflective intensity of the Laser Imaging Detection and Ranging (LIDAR) for a mowing robot. In order to safely operate a mowing robot, obstacles around the brush cutter should be avoided. The 3D-LIDAR consisting of the 2D-LIDAR and a tilting-mechanism was used to collect the surface texture information around the brush cutter. However, collecting scanned data using the 3D-LIDAR included some small noises other than vegetation by soil or trash. If the brush cutter's auto operation was controlled using raw scan data, work efficiency is reduced by frequent stop instruction. For smooth control and efficiency suitable for mowing, it is necessary to assess the risk of obstacles. This study evaluated the size of the obstacles by clustering LIDAR intensity data. When LIDAR intensity data clustering using a grid of 10 mm resolution, it could detect about 40 mm stone obstacles within a 1ms processing time.