Research And Application Of Segmentation Methods For Point Cloud In Autobody Surface Reversion

Along with the product digitization and the fast manufacture technology development, the reverse engineering technology obtained extensive research and the application in the auto-body design. The auto-body surface was generally produced by quadratic surface and free-form surfaces through extension, transition and cut out. Fitting complex auto-body surface with a sole surface mathematical model can't guarantee the precision of reconstruction surface. Moreover, the auto-body surface quality specification requests the boundary line shape light between various curved surface, and adjacent region has no obvious distortion. Therefore, for the convenience of surface reconstruction and partial revision, and enhancing the quality of reconstruc- tion surface, the point cloud segmentation is need. The point cloud segmentation is an extremely essential step in the reverse model design. The segmentation results will affect the surface fitting and the CAD model reconstruction quality directly.In view of the question in three dimensional scattered point cloud segmentation, the concept through withdrawing transition surface between the curved surface to segment the point cloud is proposed, and two methods for region segmentation algorithm has been realized and improved based on this concept.First, a method for region segmentation algorithm according to the planarity has been realized and improved, which extracts transition surface between the curved surface through the average squared error of the least squares plane fit of data point and its neighborhood. This method can segment the auto-body surface including incisive boundary and small blend surface effectively.Next, a method for region segmentation algorithm according to the logarithmic value of normal vector has been realized and improved, which determinates data point character through the logarithmic value of normal vector ratio. The curved surface which contain smooth transition region can be segmented effectively. The logarithmic method and the data point weighted average algorithm are used, which reduce the influence on segment precision of grid asymmetry and the scattered data point undulation, and strengthen the algorithm stability and validity.Finally, the improved planarity segmentation algorithm and the improved logarithmic segmentation algorithm were applied in the auto-body surface point cloud segmentation. The algorithms validity and usability have been proven through the auto-body surface segmentation example.The preliminary application of above two methods in the auto-body surface point cloud segmentation indicated that, these two methods can segment the complex auto-body surface including incisive boundary, small blend surface and smooth transition region effectively. These methods are easy to implement, and are capable of providing a fast segmentation for point cloud.The segmentation results can reflect the model construction well, which provides the good foundation for carry on the high quality surface fitting. The preliminary application results show that these methods are suitable for the auto-body reversion model design, and have the nicer applied potential.