Algorithm Improvement Of Feature Extraction Of Point Cloud Data

Point cloud feature extraction is very important in the field of face recognition,computer vision, digital entertainment, topography and so on. The birth of thethree-dimensional laser scanning brings a new life to point cloud data.However, how toanatomical the characteristics of point cloud data is urgent question at present.Traditional feature extraction technology can not meet the technical requirement, how toimprove the algorithm to get the perfect point cloud features is the main contents of thispaper.In view of the scattered data obtained by laser scanning, on the one hand,this paperhas introduced the traditional point, line, surface feature extraction method.On the otherhand, the paper studied how to quickly and precise extract feature. The main contents asfollows:1) Describe the principle of Bezier,B-spline and NURBS in the extraction of curesand surfaces and analyze the advantages and disadvantages of the three methods ofextraction line, surface by Matlab；2) Describe the method of the curvature extreme value to improve base thesemethod and analyze and compare that through data.；3) Describe growth method of discount and set the threshold to reduce the candidatefeature points to improve the feature line extraction.Then give the conclusion by palmdata；4) Analyze and advance the characteristics of surface extraction algorithm based onthe least squares method,and use the elevation for the point cloud data block. Then fitwith a least squares average value for the block data adjacent boundary, last smooth it.According to those ways,fit16control points and analyze the result by Matlab.Point,line,surface and feature have been extracted by by two different types of dataand verified. Conclusion can be drawn from the experimental illustration:the algorithm of point and line has more advantage in extraction of complex target,the extraction ofsurface is to be further simplified of block,but the overall is coherence.