Research On Point Cloud Data Processing Based On Lidar

Lidar technology has applied widely used to all business,and the research on point cloud data processing system through Lidar technology has also made great progress.Since the scanning data of lidar data obtains a large amount of discrete three-dimensional data information,these data recorded in coordinates are called point cloud data.A point cloud is a set of statistical disordered numbers representing three-dimensional information of an actual object,including the three-dimensional coordinates of objects value,color value and gray value of the object.When using the information acquisition and processing system of lidar to collect data,due to the negative influence of external interference factors(such as light reflection,shadow environment),scanning accuracy,and mechanical vibration of items,these influences will cause the collected point cloud data to exist.Lots of noise and holes.Due to the large difference in spatial density of these point clouds,the analysis results cannot directly describe the model of the actual object.For the above-mentioned loopholes,this paper designs a processing algorithm for point cloud data based on lidar.lncluding point cloud data processing simplification,filtering,retrieval,normal vector estimation,surface reconstruction,This paper not only performs a series of processing on the point cloud data scanned by the radar,but also designs an adaptive Euclidean clustering algorithm for data segmentation according to the idea of clustering fusion;Ground point cloud data studied in this paper mainly includes:In terms of point cloud data simplification,we use bounding box voxel block simplification and surface variation to achieve point cloud data simplification.The Surface variation is based on the boundary curvature division criterion.The boundary curvature uses an unbalanced mesh model to simplify the boundary points and neighboring points,and the recognition effect is remarkable.The voxel block simplification essentially produces a cubic mesh,and the way to simplify the point cloud is to generate only one point in the voxel block,so the larger the diameter of each side in the block,the lower the resolution of the point cloud.For the simplified point cloud data,the idea of bounding box is quoted,and the recognized and processed objects are wrapped by a cube box.In terms of point cloud data retrieval,the PCL?kd-tree module is used to search for point cloud information in the neighborhood,and organize a large number of scattered(noisy point cloud data)point cloud data.Establish topological associations with data points.Using kd-tree to search neignborhood relations can achieve fast on Point cloud data.In the aspect of surface reconstruction of point cloud data,the data is first retrieved by neighbors,and then the normal vector is estimated,and the normal of the surface tangent plane is fitted,and the tangent plane is obtained by the least squares method.The surface reconstruction uses the PCL?surface model to complete the fitting of the constructed plane between the points and the surface data.Surface reconstruction of point cloud data using greedy projection triangulation algorithm.For point cloud data segmentation,RANSAC plane segmentation and Euclidean clustering are used.With the help of RANSAC algorithm.Outliers points are removed by random sampling and introverted points are retained.After repeated iterations,the best segmentation effect is finally achieved.The Euclidean aggregation method refers to the collection of point clouds with the highest distance density,and the Point cloud data is contained in cluster as much as possible.It makes the features of the segmented objects more obvious and clearer.Based on the kd-tree data structure,the time complexity of o(log(n)).The retrieval efficiency is fast.In this paper,under the Windows platform,combined with the PCL point cloud library,a visual3 D reconstruction system is implemented.The feasibility of point cloud data processing is proved through the analysis of the software operation results and experiments.