Point Cloud Data Based On Support Vector Machine Repair

Reverse engineering is a advanced technology in engineering, it can exploit and produce products through rebuilding scan data model. It has two contents which are data acquisition and data processing respectively. Data repair can shorten operational time for succeeding reconstruction and model building. Therefore, research about data repair in RE in this paper has a high practical value. In the technical background of collecting data within three-dimensional structured light, the defect data repairment is researched in this paper.Defect point cloud includes merely the coordinate value of points, it has no topological relations in points. In order to automatically search out boundary contour which needs repaired, threshold angle is introduced as the standard of judging boundary points. A new method is proposed to project space points to least-squares plane, and then uses the comparison of angles between projection line sections and threshold angle to judge the boundary points. Based on the distance between boundary points, boundary line is generated. Finally, the direction of boundary line, the internal boundary line and region which needs repaired can be found out.The paper proposes a method to repair the defect region of scattered point cloud. The method builds mathematical model with non-linear regression algorithm of support vector machine. The feasibility of method is verified through repair examples. The relationship between the effect of surface patch and the space fall of boundary points is analyzed. At the same time, accurate repairing scope, situation of application as well as the limitations of this method are be discussed.