A method of defect extraction for point clouds

In recent years, with the development of 3D scanning technology, it is possible to easily express real objects with CG. However, in the point cloud obtained by the 3D scanner, data loss occurs due to the influence of occlusion and the material of the measurement object. Therefore, it is necessary to interpolate the point cloud data obtained by 3D scanning. In order to automatically interpolate the defect, it is an important theme to detect the defect region. Studies to extract defects have been done since long ago. However, since defects are detected based on the local positional relationship of points, there is a problem that it is greatly affected by the density of point cloud. Specifically, when detecting a defect from a local region based on information on the angle formed by the vector and the distance between the points, it is difficult to judge whether the point is a defect region or a density sparse region. Therefore, according to the previous method, a region having a low density may be detected as a defect in some cases. Therefore, in this paper, we propose a method of extracting defects that can be applied to complex point cloud data by applying Voronoi division to the target point. In this method, defects are detected using information on the area of the Voronoi region and circularity. In addition, noise is eliminated by grouping the detected defect points. Since this method is not a mesh data but a defect extraction method for point cloud data, it is not affected by mesh method.