Polyhedral Building Segmentation Using Mean Shift Algorithm Based on LiDAR Point Cloud

In most of the applications that LiDAR point cloud is used, segmentation is a preprocessing stage which is of considerable importance. The accuracy of this process has a great impact on the final result of the desired purpose. One of its main applications is 3D building reconstruction, for which different roof facets must be extracted first. So, for the LiDAR point cloud segmentation, in this paper, Mean Shift algorithm is used. Mean Shift algorithm is a non-parametric process which doesn’t require prior knowledge about data and also about clustering parameters such as the counts and the centers of the clusters. The input data is as an irregular LiDAR point cloud, also it is assumed that the buildings’ roofs are composed of some planar facets. In order to evaluate this method, using 2D map, several different buildings were selected and the results were evaluated both qualitatively and quantitatively. The performance of proposed method in dealing with different types of buildings was significant.