Building Roof Segmentation and Reconstruction from Lidar Point Clouds Using Clustering Techniques

This paper presents an approach to creating a polyhedral model of building roof from LiDAR point clouds using clustering techniques. A building point cloud is first separated into planar and breakline sections using the eigenvalues of the covariance matrix in a small neighbourhood. The planar components from the point cloud are then grouped into small patches containing 6-8 points and their normal vector parameters are determined. The normal vectors are then clustered together to determine the principal directions of the roof planes. Directly using a clustering algorithm on normal vectors presents difficulties due to a lack of a-priori information on approximate roof directions. Therefore, a potential based approach is used iteratively with the k-means algorithm. This generates the necessary planar parameters, and segments the LiDAR roof points. For reconstruction, a plane adjacency matrix is created for the roof using the segmented roof points. Planes that intersect each other are identified and breaklines and roof vertices are generated by solving the intersecting planar equations. A vector polyhedral model of the roof is created.