A new method of cylinder reconstruction based on unorganized point cloud

As the key objects of interest, cylinders such as pipes or circular pillars abound especially in the point cloud of terrestrial laser scanner(TLS). Therefore, the efficient reconstruction of these circular primitives has been very important to point cloud processing. Based on the study of fitting methods for cylinders abundant in unorganized point cloud, a new method using region growing for cylinder modeling is presented. In this method, using the point normals and distance of point to surface as growing constraints, the cylinder reconstruction is implemented combined with the point connectivity. Meanwhile, for the existing noise in point cloud, the RANSAC algorithm is introduced to improve the robustness and reliability of the method. The experiments of the presented algorithm show its automation and efficiency.