Research On Airborne Lidar Point Cloud Segmentation Algorithm Based On Improved RANSAC

Aviation LiDAR point cloud segmentation,as one of the key research directions for point cloud data processing is constantly playing an active role in the current social and scientific research fields.But at the same time,it will also be accompanied by many problems.How to improve the accuracy of segmentation,reduce data unfitting rate,and improve algorithm performance in point cloud processing is also the main research direction of point cloud segmentation algorithms.Because the random sampling consistent(RANSAC)method has a strong randomness for point cloud segmentation,and the constraints of important steps such as selecting seed points are not perfect,it is easy to create overfitting or under fitting.Therefore,based on the hyperparameter RANSAC point cloud segmentation algorithm,this paper provides a complete adaptive hyperparameter RANSAC point cloud segmentation algorithm to make more complete functional improvement and performance optimization of the original algorithm.The main contents of this paper are as follows:First,this paper adds a Downsampling process to the point cloud data in the hyperparameter RANSAC algorithm.This process can reduce the number of seed points in point cloud data and further optimize time and processing while preserving the mathematical features of point cloud data.First,voxelize the point cloud through the octree,and use the octree depth as the termination condition of recursion until the point cloud data is completely segmented.After the point clouds are distributed in each model grid,the point cloud data features are weighted and averaged,the residual value is calculated to obtain the weighted surface distance average,and Downsampling replaces the feature points in the original grid to improve the Downsampling average of airborne radar point cloud data.This step is the first step to improve the hyperparameter RANSAC algorithm.Secondly,this paper improves the RANSAC segmentation process of point cloud data after Downsampling,and optimizes the conditions such as selecting seed points.Since the hyperparameter RANSAC algorithm has no obvious limitation in selecting seed points,in order to ensure the feasibility of subsequent operations,the farthest point sampling idea is used in combination with the seed point selection process.In addition,dynamic processing is performed on the height threshold of the RANSAC fitting process,the abandon the manual selection process to reduce the unfitting rate of point cloud data segmentation and obtain more accurate segmentation results.While retaining the main part of RANSAC,the initial segmentation results of point cloud data are obtained through iterative iteration based on the calculated threshold k,and patch merging operations are added to this foundation.This operation is an improvement on the hyperparameter RANSAC algorithm and is the main part of the segmentation algorithm.Finally,this article optimizes the processing of unfitted points,reduces the unfitted rate of the algorithm,and cut down the processing results after improving RANSAC segmentation.By traversing the set of unfitted points,the Euclidean distance between the nearest plane of the point within the threshold range is calculated,and the Euclidean distance results between the 24 poles in the nearest two planes and the point are obtained.The unfitted points are included in the nearest plane,which reduces the algorithm’s unfitted rate as much as possible and reduces the occurrence of undersegmentation and other problems.The improved hyperparameter RANSAC segmentation algorithm proposed in this paper is used to segment airborne radar point cloud data.The improved segmentation algorithm can effectively improve the segmentation accuracy of point cloud data and reduce the non-fitting rate.Under the same parameter conditions,comparative experiments were conducted between the initial algorithm and the improved algorithm.The experimental results show that the algorithm proposed in this paper effectively improves the segmentation accuracy and fitting rate of aviation radar point clouds.