Research On Curve And Surface Fitting Method Based On Dense Point Cloud

Curve and surface fitting is a challenging subject which is often encountered in theoretical research and practical application.It is also a problem which must be solved in reverse engineering.In the fields of computer image processing,pattern recognition and computer vision,the fitting of point cloud data is also one of the key steps.The fitting problem is closely related to CAD and CAGD,and the fitting effect of point cloud data is directly related to the next work.The main content of this paper is based on the fitting of point cloud data,and the main work is as follows:(1)The background of the project,the significance of the research and the research status of curve and surface fitting of point cloud data are introduced in detail.This paper summarizes the contributions made by Chinese and foreign scholars in the field of curve and surface fitting.At the same time,on this basis,it lists specific application examples and points out the necessity of studying curve and surface fitting.The basic concepts,formulas,and properties of three common fitting models,Bezier,B-spline and NURBS,are introduced in detail.Combined with the relationship between the three,the application of different fitting functions is analyzed.Then the concept of curve and surface smoothness is introduced to analyze the judgment conditions for achieving smoothness and the reasons for the uneven smoothness of curves and surfaces.After that,the curve fitting methods for ordered random points and disordered random points are classified and discussed.(2)Point cloud data preprocessing is a particularly important step in curve and surface fitting.This paper introduces several preprocessing steps of point cloud data.In the aspect of point cloud denoising,it mainly introduces the observation method,curve checking method and chord high deviation filtering method.In this paper,the point cloud data reduction method based on curvature is adopted.The point cloud reduction method also includes the center of gravity compression method based on the bounding box and the simplification method based on cluster analysis.At the same time,the common extraction methods of point cloud feature points are introduced.(3)The expression and properties of cubic B-spline curve are introduced.In this paper,a multi-step feature point selection strategy is proposed,which optimizes the traditional feature point selection algorithm based on previous studies.Through the hierarchical processing of feature points,a better extraction effect is achieved.At the same time,a fitting method for cubic B-spline curve is proposed.In the experimental part,the ordered random points and disordered random points are fitted respectively,and the fitting results are compared,which shows the scientificity and feasibility of the algorithm in this paper.(4)Introduced the solving process of three fitting algorithms: traditional least square method,moving least square method and total least square method,and completed the visualization of 3D point cloud data.The experiment completes the fitting of plane and sphere(standard surface)based on least square algorithm and global least square algorithm.The effectiveness of the proposed algorithm is verified by comparison.Finally,the T-Scan scanner was used to scan the irregular surface,and the point cloud data obtained from the scan was preprocessed and the surface reconstructed.The experiment not only obtained the expected results within the accuracy range,but also made the optimization and reconstruction process of the surface more convenient,which also showed that the algorithm in this paper was correct and had certain advantages.