This invention discloses an automatic basis removal method for
structured light point clouds based on frame-by-frame analysis, specifically relating to the field of
structured light 3D scanning
point cloud processing. It addresses the problem of erroneous or insufficient basis removal caused by
confusion between the basis and the bottom boundary of an object during
structured light scanning. The method involves acquiring the current frame
point cloud and generating depth and height maps to determine the
usable area, extracting candidate planes, and combining them with the basis model to determine the basis
reference surface and the basis candidate
point set. A basis
mask for the current frame is constructed and its consistency with the previous frame's basis
mask is evaluated across frames. A fringe consistency weight map is generated, and the normal reliable bandwidth and circumferential coherence angular path are calculated to update the basis stability probability. A stable basis
mask and a mask for undetermined zones are then defined. Based on the stable basis mask, basis points are removed while the undetermined
point set is retained. Further boundary consistency
verification is performed on the initial basis-removed
point cloud. Backfilling or removal is performed on the undetermined
point set, and the final basis-removed point cloud is output, updating the basis model and the previous frame's basis mask.