| Airborne LiDAR(Light Detection and Ranging,also known as laser detector)filtering process is to separate the ground points and Non-ground points from the cloud data points.It increases the complexity of the filtering process since LiDAR data being short of topological relationships and texture features.Progressive morphology filtering algorithm,irregular triangulation densification filtering algorithm(Progressive TIN Densification for short PTD),moving surface filtering algorithm are all filtering algorithms with good filtering effect in a certain area.The progressive morphological filtering algorithm,PTD filtering algorithm and moving surface filtering algorithm are all filtering non-ground point clouds and retaining ground points.This paper studies the above three types of filtering algorithms,and seeks better filtering algorithms to complete the separation of ground and non-ground points of LiDAR point clouds.In this paper,the defects of the above three classic filtering algorithms are improved:a method of drawing an isoline to determine the critical parameter threshold is proposed for the problem that the threshold of the critical parameter in the progressive morphological filtering algorithm is difficult to determine;there is a problem of gross error and a confidence interval estimation theory is proposed to test the gross difference for the initial seed point in the PTD filtering algorithm;a hierarchical clustering classification algorithm is proposed for the problem that the classification result of the classic mobile surface filtering algorithm is too dependent on the distance between adjacent points,which uses the horizontal distance between adjacent points the distribution law of distance from height difference determines the type of data point.There are 15 sets of sample data provided by the International Society for Remote Sensing and Surveying(ISPRS)in our experiment.This data is used to test the three classical filtering algorithms improved in this paper.The applicable areas of the three algorithms are calculated,and the filtering accuracy of various algorithms is analyzed through multiple sets of data verification. |
