Plane Segmentation of Point Cloud Data Using Split and Merge Based Method

Abstract

In indoor environments, segmentation of planar surfaces such as wall, floor, and door can contribute the efficiency of robots in performing tasks. In this study, a split and merge based method is developed to segment planar surfaces via point cloud data in indoor environments. Apart from the previous split and merge studies, fixed-size regions are used instead of octree data structure. In this way, the segmentation time can be decreased as low as possible. In the split phase, the fixed-size regions are assigned to one of the three categories, the outer edge, the inner edge, and the non-edge. In the merge phase, each of these categories is processed separately. Thus, the segmentation success can be increased. The proposed method is tested with point cloud data captured in ESOGU Electrical Engineering Laboratory building modelled in Gazebo simulation environment. In addition, RANSAC and region growing methods are implemented for comparison. Experiments are conducted to analyze performance of the proposed method in terms of segmentation time and segmentation success.