Reconstruction Technology Of Railway Fastener Based On Point Cloud Data

As an important part of rail transit facilities,railway fasteners will directly affect the running safety of trains,so it is of great practical significance to improve the research and development capabilities of railway fasteners.However,due to the complexity of the overall surface structure of railway fasteners,it is difficult to complete the accurate modeling of the fasteners through data measurement.Based on this,the research on the surface reconstruction technology of railway fastener point cloud data is carried out to realize the accurate modeling of the fastener,which provides a better initial model for the optimal design,model analysis and measurement of the shape and size of the railway fastener.Based on the structural characteristics of railway fasteners,it is difficult to obtain complete point cloud data from a single perspective.The PhoXi3D scanner is used to obtain point cloud data from multiple perspectives,and the combination of conditional filtering and statistical filtering is used to remove noise points and outliers,which provides the initial value for the subsequent processing of the point cloud.Aiming at the problem that the incorrect matching point set affects the registration efficiency and accuracy during the initial registration of point cloud,a Sample Consensus Initial Alignment algorithm based on sub-nearest neighbor matching strategy and the optimization of the corresponding point set with double thresholds is proposed.Firstly,the ISS algorithm is used to extract the feature points of the point cloud,and the sub-nearest neighbor matching strategy based on FPFH descriptor is proposed to complete the search of the initial corresponding point set.Secondly,a double threshold constraint based on the Angle between point pair Euclidean distance and normal vector is proposed to eliminate false corresponding point pairs.Experimental tests show that the accuracy and efficiency of the improved fastener point cloud initial registration algorithm are improved.Aiming at the problems of ICP precision registration algorithm,such as low registration efficiency and easy to fall into local optimum,a fastener point cloud ICP precision registration algorithm is proposed based on uniform sampling and adaptive threshold correspondence optimization.Firstly,uniform sampling is used to simplify point cloud data.Secondly,the k-d tree structure is used to accelerate the search speed of corresponding points,and an adaptive threshold constraint is proposed to remove the error corresponding point pairs.The experimental results show that the improved fastener point cloud precision registration algorithm improves the accuracy and efficiency of registration.After multi-view registration,there are many redundant points in the point cloud of fasteners,and the point cloud data is irregular,which will lead to problems such as uneven reconstruction surface.Therefore,a Poisson surface reconstruction algorithm based on voxel down-sampling and MLS optimization is proposed.The redundant points are removed by voxel down-sampling,and the MLS algorithm is used to smooth the fastener point cloud;Meanwhile,the weighted covariance matrix is used to estimate the normal vector,which improves the accuracy and robustness of the normal vector estimation.The experimental results show that the improved surface reconstruction algorithm is more efficient,the reconstructed surface is smoother and the details features are preserved well.