Multi-Feature Airborne Lidar Strip Adjustment Method Combined with Tensor Voting Algorithm

In the airborne LiDAR system data acquisition task, it is necessary to use the calibration data for calibration, but there may be calibration failures or lack of calibration data. On the other hand, even after the corrected data, the features of the same name between the strip will have a large deviation. This paper proposes a multi-feature airborne LiDAR strip adjustment method combined with tensor voting algorithm for these two problems. First, use the tensor voting method (TVM) to calculate the plane feature intensity value of each point, set an appropriate threshold according to the plane feature intensity value to remove non-plane points, and use the CSF algorithm to remove ground points. Second, use the RANSAC algorithm to extract the building plane. Then, according to the Euclidean distance between the centroids of each plane point cloud, the adjacency of the planes is judged, and the intersection line of the adjacent planes is calculated. After that, the minimum Hausdorff distance (MHD) is used to determine the line pair. Then, rough registration is performed based on the average L1 distance on the set of minimized matching lines. Finally, the points on the plane are used to adjust the strip using the point-to-plane ICP algorithm. Proved by real data, the method proposed in this paper can have better results in data with large deviations between strip caused by correction failure, and the accuracy is slightly better than TerraMatch software. At the same time, it does not require any manual operation and auxiliary data or original data conversion.