High-precision detection method for large and complex steel structures based on global registration algorithm and automatic point cloud generation

Abstract

For extremely large and complicated steel structure buildings, the traditional monitoring method is inefficient and labor-intensive. However, LiDAR and UAV close-range photogrammetry can directly collect data in various large and complex environments, register sites, and fit nodes to monitor and analyze their structures. Taking a long-span steel structure monitoring as an example, this paper introduces the deformation monitoring scheme of steel structure with lidar and unmanned aerial vehicle close-range photogrammetry technology. The method of feature-based global registration is adopted to register the site point cloud. The initial value of feature constraint is taken as the error equation of observation value column, and the overall adjustment is carried out. The spatial transformation parameters and unknown point adjustment values are solved through the bundle adjustment model. Through iterative global registration algorithm, the correction error of observation value is controlled within a certain threshold range through constant constraint weighting and reconciliation of observation value until registration is completed, and the entire grid structure is generated. The eccentricity calculation of the spherical node and column in the grid structure is carried out by using the spherical node multi-link center point algorithm. Then, the dense point cloud generated by the photos taken by UAV close-range photogrammetry and the grid structure generated by LiDAR point cloud are used as mutual references, so that the deformation monitoring of long-span steel structures becomes more detailed and comprehensive.