The Simplified Algorithm For Airborne LiDAR Ground Point Cloud Based On Improved Hierarchical Clustering Algorithm

Airborne LiDAR(Light Detection And Ranging,LiDAR)technology has significant advantages in obtaining large-area,high-precision three-dimensional surface data,and has become one of the most important data acquisition methods for building Digital Elevation Model(DEM).However,the amount of point cloud data obtained through LiDAR is huge and complex,which affects the efficiency of subsequent data processing.In some engineering applications,a large number of point clouds does not significantly contribute to improving the elevation accuracy of DEM.Therefore,under the premise of preserving the basic topographic features of the original point cloud,simplifying the massive LiDAR ground point cloud is of great practical significance.Hierarchical clustering algorithm is an unsupervised machine learning method.It clusters points with similar features in the space into a cluster,and can identify clusters of arbitrary spatial distribution.At the same time,the definition of its similarity measurement criterion is relatively simple,so it is used in point cloud simplification and achieves better results.However,when simplifying the LiDAR ground point cloud,the algorithm still has the following shortcomings:(1)The hierarchical clustering algorithm needs to subjectively set the cluster termination threshold.But for regions without prior knowledge,inappropriate termination thresholds will cause inaccurate clustering.(2)Hierarchical clustering algorithms usually use single features such as distance,normal vector,and curvature as similarity measurement criteria.However,a single feature can only describe a certain aspect of the topographic features of the LiDAR ground point cloud,and cannot accurately measure the similarity between ground points,thereby affecting the accuracy of the clustering results.In view of the above problems,the main research contents of this article are as follows:(1)Aiming at the problem of inaccurate clustering caused by subjective setting of the hierarchical clustering termination threshold,a hierarchical clustering algorithm based on spatial autocorrelation is designed.The semivariogram is introduced to study the spatial autocorrelation range of terrain features to determine the termination threshold of hierarchical clustering.The purpose is to ensure that the data in the cluster after clustering has spatial autocorrelation,and the points that are closer in the space are clustered into a cluster.(2)Aiming at the problem of inaccurate clustering caused by using a single feature to measure the similarity between LiDAR ground points,a hierarchical clustering algorithm based on terrain features is designed.Based on multiple terrain feature factors,using cosine similarity and Gaussian distance weights,a weighted cosine similarity model is constructed as the similarity measurement criterion for hierarchical clustering to measure the similarity between ground points to improve the similarity of the topographic features of the data in the cluster after clustering.(3)Aiming at the inaccurate clustering problem of traditional hierarchical clustering algorithm in the simplification of LiDAR ground point cloud,an airborne LiDAR ground point cloud simplification algorithm based on improved hierarchical clustering algorithm is designed.Use the hierarchical clustering algorithm based on spatial autocorrelation and the slope-based LiDAR point cloud sampling rules to cluster,sample and calculate the medium error of the ground point cloud.The medium error specified by the DEM of a specific scale is used as the basis for whether to perform hierarchical clustering and sampling based on topographic features on the hierarchical clustering results based on spatial autocorrelation.The purpose is to improve the accuracy of LiDAR ground point cloud clustering and construct a specific scale DEM with fewer point clouds.Research shows that: compared with the K-Means++ algorithm and slope-based LiDAR ground point cloud simplification algorithm,the airborne LiDAR ground point cloud simplification algorithm based on the improved hierarchical clustering algorithm designed in this paper has a good simplification effect,and can obtain a simplified result with a reasonable distribution without any prior knowledge.The algorithm in this paper realizes the construction of DEM with a specific scale with fewer point clouds,which is of great significance for improving the processing and analysis speed of DEM in practical engineering.