Research On Feature Points Extraction And Registration Technology Of 3D Point Cloud

In recent years,with the continuous progress of 3D point cloud technology,the processing technology of 3D data has been applied in more and more fields.Besides,it has also been developing rapidly in the digital city,reverse engineering and other technologies.With the rapid development of point cloud processing technology,the demand for technology is also higher and higher.On the basis of studying the current point cloud data processing,some improvements have been made to the existing methods and technologies in this thesis,which achieved a better processing effect.Firstly,as for the extraction of 3D point cloud data feature point,this thesis studies the feature point extraction effect of point cloud data based on the geometric features of point cloud,such as normal vector,curvature and so on.At the same time,this thesis further analyzes the experimental results.On the basis of the existing methods,a feature point extraction method based on neighborhood radius constraints is developed.The experiment proves that the method can realize the feature information of the original cloud with the least possible data points,and can achieve faster running speed.Secondly,as for the registration of point cloud data,this thesis mainly studies the principle of point cloud initial registration and fine registration.Based on the analysis of the traditional ICP(Iterative Closest Point)algorithm,a position relationship between the foot and the triangle is developed on the basis of the two point cloud data points to complete the search of the corresponding point pairs.The improved ICP registration algorithm is added to eliminate the error point pair by the super line distance constraint method.Experiments show that the new corresponding point search method reduces the registration error effectively,and the super line distance constraint method reduces the probability of mismatch.The example verifies the effectiveness of the development algorithm,and achieves good registration effect.Finally,on the basis of theoretical research,a simple 3D point cloud data processing software system is designed,which mainly realizes the basic operation such as point cloud data visualization,point cloud data preservation and format conversion,point cloud filtering,point cloud registration and point cloud surface reconstruction.