Research On Robust Weighted Total Least Squares Registration Algorithm Based On Point Cloud Correlation

As an emerging technology in the field of measurement,3D laser scanner has been widely used in various fields of engineering measurement,and the point cloud matching criterion is a key step in 3D laser point cloud data processing.However,Most point cloud registration algorithms do not consider the correlation effect of random errors,and there is no corresponding method for estimating the robustness.Therefore,the research of the robust weighted point cloud registration algorithm considering the point cloud correlation has important theoretical significance and application value.Aiming at the core problem of coordinate transformation in point cloud registration,this paper studies the physical properties and geometric characteristics of point cloud data and the corresponding robust estimation point cloud registration algorithm based on the Gauss-Helmert model.And achieved certain results.The main research contents and conclusions of this paper include:(1)The Iterative Closest Point algorithm of Delaunay point set search is studied.According to the classical Least Squares(LS)theory,the linear model of point cloud registration(Bursa-Wolf,BW)and nonlinear model(Gauss-Markov,GM)are derived;Based on the Total Least Squares(TLS),the Gauss-Helmert(GH)model of point cloud registration is derived.The point cloud registration effects of Delaunay-ICP algorithm,LS-GM algorithm and TLS-GH algorithm are compared and analyzed by simulation experiments.The experimental results show that the registration accuracy of TLS-GH algorithm is the highest when the registration point pair is determined,and the registration accuracy of LS-GM algorithm and TLS-GH algorithm is higher than Delaunay-ICP algorithm.(2)On the basis of the equal-point cloud registration Gauss-Helmert model algorithm,considering the physical and geometric characteristics of the scanned point cloud data,the ranging angle,spot area,angle of incidence and correlation are given.And the construction method of the co-factor matrix,the Weighted Total Least Squares(WTLS)point cloud registration algorithm is realized,and some beneficial results are obtained through relevant experiments.The experimental results show that the precision of ranging angle measurement,spot area weighting,incident angle weighting and correlation weighting are equivalent The WTLS algorithm with various weights has certain improvement on point cloud registration accuracy.(3)Based on the ranging angle measurement,spot area,incident angle and correlation weighting algorithm,the Gauss-Helmert model point cloud registration robust robust estimation algorithm is studied,combined with the idea of standardized residual structure weight function function.The anti-difference function for Huber,IGG and IGG? is constructed,and the corresponding iterative calculation method and accuracy evaluation formula are derived,and the relevant experiments are designed to verify the algorithm.The experimental results show that the robust weighted total least square(RWTLS)algorithm has significantly improved the accuracy of the WTLS point cloud registration algorithm,and the IGG III weight function has better robustness and stability.Figure[23];Table[17];References[100]...