Research On Point Cloud Filtering Algorithm Based On UAV Photogrammetry

Low-altitude photogrammetry technology through UAV can quickly and efficiently acquire a large number of measurement points,DSM,DOM and other data through a small number of ground control points.Its continuous development and improvement in the field of surveying and mapping is gradually replacing traditional measurement methods.Become the backbone of the field of surveying and mapping.With the development of tilt photogrammetry technology,the accuracy of point cloud data acquired by UAV has been continuously improved,and the cost of LIDAR point clouds has been greatly reduced.However,the amount of point cloud data acquired by the UAV is large and the density is not uniform.In order to obtain high-precision DEM,a large number of object points need to be removed.The current filtering algorithm has more or less problems.This paper analyzes the airborne.After the similarities and differences between the LIDAR point cloud and the UAV image matching point cloud,the classic LIDAR point cloud filtering algorithm is improved to make it suitable for the UAV aerial survey point cloud.The main results are as follows:The point cloud outlier detection is mainly divided into extreme outlier elimination and general outlier elimination.The extreme outlier is eliminated by the elevation statistics method.For the general outlier,the current mainstream outlier elimination methods are discussed: outlier detection based on virtual grid elevation information,and outlier detection based on k-tree proximity index;it is found that these two methods are difficult to meet the edge of the building.An aerial survey point cloud with many points and uneven cloud density.In this paper,the LOF algorithm is used to carry out the outlier elimination of the aerial survey point cloud.The LOF algorithm is based on the local density outlier detection method,which is suitable for the outlier identification in the data with different density.The experiment shows that the parameter k value is reasonable.In the case of the LOF-based outlier detection method,fewer ground points can be eliminated.In point cloud filtering,this paper is based on the point cloud filtering algorithm of progressive TIN densification.In the seed point selection,the original algorithm divides the study area into a regular grid by the largest building size in the study area.This paper refers to the filtering method of the moving surface to change the regular grid into a moving virtual grid.This method can extract More ground seed points greatly increase the efficiency of subsequent iterations.The OTSU algorithm isintroduced in the selection of the threshold.The distance threshold is calculated by calculating the distance from the pending point to the TIN in each initial TIN,and the slope threshold is calculated by the ground seed points.