Research On Point Cloud Data Processing Of Lidar

In the fields of environmental mapping and unmanned driving,the point cloud data collected by the lidar equipment often contains a lot of noise due to the intricate road topography,vast landforms and the difference in the reflectance of the scanned object surface.Moreover,in the process of data collection,the point cloud data overlap rate between frames is relatively low due to the occlusion of buildings and obstacles.These problems directly affect the subsequent 3D scene reconstruction and the final decision of the unmanned driving system.Therefore,it is very important to study the point cloud data processing of lidar.Starting from the point cloud data processing technology,this paper deeply studies the two core problems of point cloud filtering and point cloud registration in lidar point cloud data processing technology.Firstly,according to the characteristics of noise source and noise model of lidar scan data,this paper studies in detail the 3D guided filtering algorithm with better filtering effect in point cloud filtering algorithm.The point cloud model processed by the 3D guided filtering algorithm has the defects of local noise residue.The secondary guidance filtering method in 2D image processing is extended and introduced into 3D point cloud data filtering,so as to effectively filter useless noise points while maintaining the local contour of the model.In this way,the local contour of the model is well maintained while useless noise points are effectively filtered out.Then,this paper studies in detail the implementation process of the traditional point cloud registration algorithm,aiming at the disadvantage of the current ICP point cloud registration algorithm,when the overlap space between the original point cloud and the target point cloud is relatively small,the registration effect of the ICP point cloud is poor.An improved algorithm based on normal vector clustering and point representative election is adopted to improve the effect of point cloud registration significantly.Finally,this paper studies the objective evaluation method of point cloud filtering and point cloud registration,and compares the experimental results processed by this method with those processed by classical point cloud filtering method and classical point cloud registration method from subjective and objective perspectives.Experiments show that this method is superior to other classical algorithms.