Point cloud simplification method based on two-dimensional section filtering marking

LIU Fei-fei, YANG Duo

(College of Mechanical Engineering, Dalian University, Dalian 116622, China)

Abstract: Aiming at the problem that the local redundant points are difficult to be removed correctly during the point cloud simplification process, the effects of filter denoising, coordinate system conversion and two-dimensional plane point processing were studied, and the largescale noise removal in the current point cloud preprocessing process was studied. The point cloud simplification method was summarized and a point cloud simplification method for two-dimensional cross-section screening markers was proposed. Firstly, the large-scale noise in the original point cloud was removed by the filter. The main body of the point cloud was intercepted by cross section and coordinates were transformed to process the position and color information of the points on the two-dimensional section. The non-model points of outliers and contour edges were screened and marked. Finally, the marker points were semiautomatically removed to complete the pretreatment. The point cloud model reconstructed by image method was tested. The results indicate that the overall simplification efficiency of the method is high, and the semi-automatic removal method can accurately remove redundant points and small-scale noise, maintain the sharp features of the model and prevent the model from shrinking and deformation.

Key words: point cloud simplification; dimensionality reduction; cross-section scanning; reverse engineering