The comparision of local region selection strategies in point clouds

Local region growing strategy that is a preliminary step of polynomial fitting in point cloud processing plays an important role. Most common utilized region growing algorithms has been compared by incorporating the normalized eigenvalue analysis method. The results of this study which includes circular region growing method, nearest neighbor region growing method, Euclidean minimum spanning tree based region growing method and hybrid region growing method are compared over mean square error of polynomial fitting and algorithm runtime. Hybrid region growing method gives best results according to the mean square error criteria while circular region growing method takes first place according to the runtime criteria by only about 17% difference. Due to the fact that mean square error is a more important parameter than runtime for polynomial fitting process, hybrid region growing algorithm is a more preferable method.