Integration application of terrestrial laser scanner point clouds and unmanned aerial vehicle image point clouds

This study was to establish a high-precision 3D point cloud model of bridges, inspect bridge health conditions, and fit and draw the bridge geometry. Firstly, close-range photography, orbiting flights, and grid flights with drones were employed to acquire detailed texture data of a certain large twin-line bridge. Subsequently, the data collected from different flight routes were processed in context capture software for 3D reconstruction, merging the main and detailed images to generate a comprehensive bridge point cloud 1. The Trimble SX12 instrument was used for integrated scanning of the entire bridge, obtaining a complete bridge point cloud 2. An iterative closest Point (ICP) algorithm based on a bi-directional KD-tree optimization was proposed to register and merge the drone-surveyed bridge point cloud 1 with the terrestrial laser-scanned bridge point cloud data 2. The encrypted bridge point cloud was then used to establish a refined 3D realistic model of the operational twin-line bridge for railways. Additionally, a Principal Component Analysis (PCA) algorithm based on KD-tree was introduced to extract the suspension point cloud of the bridge. The least squares method was applied to fit the bridge arch axis alignment, and the RANSAC algorithm was used to fit the bridge deck profile. Validation of the effectiveness of the fusion modeling was conducted through comparative analysis with the accuracy and completeness of single unmanned aerial vehicle (UAV) and terrestrial laser scanning. The results indicate that the horizontal accuracy of the fusion model is 1.71 cm. The vertical accuracy is 1.25 cm. This represents an improvement of 16.59% in the horizontal direction and 20.89% in the vertical direction compared to the accuracy of the single UAV modeling. The completeness of the fusion model is 98.17%, providing a more realistic texture effect. The model can identify bridge pier conditions, such as honeycomb surface and water seepage, as well as arch rib issues like painted rust and cracks. This study can provide valuable insights and references for the application research of 3D point cloud models for bridges, demonstrating promising prospects. © 2024, Central South University Press. All rights reserved.