An analysis of triangulation reconstruction based on 3D point cloud with geometric features

The reconstruction of 3D point cloud is the core of reverse engineering and widely applied in industrial field. Focused on the problem of data redundancy and calculation, the reconstruction procedure is realized and the influence of various triangulation methods on geometric features is analyzed. The point cloud data pre-processing is implemented first based on C++ Point Cloud Library (PCL) in the paper, including filtering and smoothing, outlier removal, valid points extraction, simplification, and hole filling. Then the Greedy Projection Triangulation and the Poisson Reconstruction methods are applied separately to reconstruct the mesh models. The spherical center distance and diameter of calibration board are selected as the geometric characteristic parameters to assess the reconstruction quality. The relative error is calculated according to the true value and the average of multiple measurements on the parameters. For the distance feature, the results show that the two methods have similar accuracy. For the diameter feature, the Greedy Projection Triangulation is further suitable than the Poisson reconstruction, and the relative error of which is less than 0.18%.