Measured Point Cloud Completion Method for Bridge Based on Design Information

To tackle the issues of missing details and difficult control point extraction in three-dimensional laser-scanned point clouds of bridges, which result from inadequate point density and occlusion, a method based on the design information is proposed for completing the measured point clouds of bridges, so as to enable high-accuracy coordinate extraction of control points and spatial posture analysis for complex structures during the monitoring of bridge construction. First, a cross-section point cloud segmentation algorithm is developed using a K-dimensional tree algorithm and principal component analysis, allowing an accurate segmentation of cross-section point clouds by adjusting the slice thickness and orientation. Second, a cross-section cloud point completion algorithm based on the CAD design information and coarse-to-fine two-stage registration is put forward to generate high-accuracy cross-section design point clouds, which are further used to complete the scanned sections with the registration of design and measured point clouds. In addition, the effect of different interpolation parameters on the completion accuracy is also analyzed. Finally, the coordinates of control points are extracted from the completed point clouds, and the bridge’s spatial posture is also analyzed. The entire methodology was experimentally validated on an east-west special-shaped steel arch bridge, which was still under construction over the Haihe River in Tianjin Binhai New Area. Results demonstrate that the root mean square error was 0.315 mm at an interpolation spacing of 1 mm, realizing a high-accuracy cross-section cloud point completion. The measured in-plane relative deflection angle between the two bank sections to be constructed was 6.80×10^-4 rad, and the relative central height difference was 35.9 mm. The result of spatial posture analysis indicated southward deflection on the eastern bank and northward deflection on the western bank. The proposed method in this paper can complete incomplete measured point clouds of bridges and provide reference and algorithm support for spatial posture detection, demonstrating its potential for engineering applications.