Typical Elements Classification And Reconstructionof Shield Tunnel Point Cloud

The ground 3D laser scanning data can be used for feature extraction of buildings.The processing and classification of point cloud data is an important part of the ground 3D laser scanning system,which can provide accurate and streamlined point cloud data for subsequent point cloud applications.In terms of automatic classification of typical elements of shield tunnel point cloud,the main problems include: 1)low efficiency of data processing;2)element segmentation accuracy is not high;3)When the typical element distance threshold is small,feature extraction errors are prone to occur.In this paper,the optimal scanning spacing and optimal resolution of the shield tunnel are selected,and then the global splicing scheme is optimized based on the target to integrate the traditional single shield tunnel scanning scheme.The author calculated the parameters of the best scan and scanned the experimental area.Finally,the Cyclone was used to preprocess the point cloud data,and the point cloud was spliced and manually deleted.During the construction of the shield tunnel,the environment is complex,the types of tunnel elements are diverse,and the efficiency of massive point cloud data processing is very low.The extraction of tunnel elements is the key to establishing an intelligent three-dimensional management platform.In this paper,the data compression algorithm and the downsampling algorithm are considered,and the noise points and outliers are automatically eliminated.Finally,the point cloud classification algorithm suitable for narrow and cylindrical objects is used to fit the shield tunnel.An automatic classification method for typical elements of shield tunnels considering geometric features in complex environments is proposed.By simplifying the point cloud and filtering the outliers,the traditional RANSAC algorithm is improved in operational efficiency.The experimental results show that the method quickly and accurately segmentes the typical elements of the shield tunnel with small adjacent threshold.The error between the fitted tunnel radius and the design radius is only 3 mm,which meets the design requirements.Compared with the traditional RANSAC algorithm,the improved algorithm runs 17 times faster,greatly improves the efficiency of automatic classification,and achieves the target classification with the adjacent distance threshold up to 1cm.Based on the classification and extraction of typical elements of shield tunnel,in order to achieve efficient reconstruction of the scene,this paper uses the moving least squares algorithm to make the point cloud smoother.In this paper,the threedimensional reconstruction algorithm of shield tunnel based on point cloud is selected to solve the problem of splicing misalignment between adjacent shield tunnel models,and the multi-segment model is automatically spliced.By comparing the Poisson reconstruction algorithm and the greedy projection triangulation algorithm,this paper finds that the greedy projection triangulation reconstruction does not reconstruct the distortion because the model object is not closed,so it is selected as the reconstruction algorithm of the tunnel element.The maximum possible value of each triangle side length generated by the greedy projection triangulation algorithm has the greatest influence on the algorithm.After selecting the optimal side length of 0.25 m,the number of holes is reduced from 4974 to 158,and the hole is repaired by Geomagic Studio.The algorithm hole problem is solved,and a three-dimensional model is provided for establishing a three-dimensional visual shield tunnel construction management platform.