Simplification Of LIDAR Point Based On Spatial Autocorrelation

In recent years,with the improvement of the accuracy of LIDAR cloud acquisition equipment,the collected LIDAR cloud data has grown geometrically to a magnitude of one million.Although the LIDAR point cloud data collected by the precision device is very rich in detail,it is suitable for expressing complex irregular geometry and rich terrain features,but the large LIDAR point cloud data is processed.The existing algorithms and existing data processing software are severely challenged.Huge amounts of point cloud data can make data processing very slow,even impossible.Therefore,on the premise of preserving the details of the point cloud,it is of great practical significance to simplify the data of LIDAR point cloud.In this paper,we discard the morphological parameters of topography features such as curvature,height difference,density,surface area,and dispersion,which are commonly used in the existing simplified algorithm.We propose a LIDAR point cloud simplification method based on spatial autocorrelation,and refer the spatial autocorrelation theory to LIDAR point cloud simplification.The high precision DEM is constructed with a high simplification rate for the large-scale LIDAR ground point cloud.First,two experimental areas with mountainous and original terrain features were selected to preprocess massive LIDAR ground point cloud data and remove non-Surface points.Secondly,we divide the point cloud in the experimental area with the idea of surrounded box partitioning,calculate the spatial autocorrelation index of the point cloud in each surrounded box,qualitatively classify the point cloud surrounded box,and formulate sampling rules based on the spatial autocorrelation characteristics of each type of point cloud in surrounded box.Only the point closest to the center point is reserved for the box with high spatial autocorrelation,and the point closest to the center,the elevation maximum point,and the elevation minimum point are reserved for the box with low spatial autocorrelation.Then,using the cross-validation method to optimize the sampling parameters,using the partition parameters and sampling thresholds of different envelope boxes,a multi-point cloud space autocorrelation sampling scheme was developed to simplify the point cloud in the experimental area,and the degree of simplification and quality of the point cloud were comprehensively evaluated.The results of higher simplification rate and higher precision of DEM were selected,and the optimal simplification parameters were put forward to build the self-correlation optimal sampling method of the point cloud space,guide the simplification of the point cloud in the enveloping box,and merge the enveloping box to achieve the entire experimental area.Simplification of point cloud data.Then,the two commonly used point cloud simplification methods of simplification and uniform sampling simplification are used as comparison objects,and the point cloud simplification rate and DEM accuracy are used as evaluation indicators.The simplification effect of the simplified method in different terrain is evaluated from the aspects of the simplification rate and the error of the simplified post-point cloud reconstruction DEM.Finally,the error spatial distribution of DEM reconstructed by simplified point cloud is analyzed by constructing DEM terrain description error field.Three simplified methods were extracted to reconstruct the skeleton lines of DEM,and compared with the original point cloud to construct the skeleton lines extracted by DEM,the coincidence of skeleton lines under different terrain was analyzed.The method of point cloud simplification is better than the method of mass grid simplification and uniform sampling simplification in both the simplification rate and DEM precision.Using the point cloud simplification method proposed in this paper to simplify the reconstructed DEM,the terrain description error space distribution is uniform,and the DEM error space distribution is not caused by the change of terrain.The terrain feature line extracted by DEM is preserved well.It is almost completely aligned with the original DEM skeleton line.To sum up,the LIDAR point cloud simplification method based on spatial autocorrelation is proposed in this paper,and the large scale ground point cloud data is divided and processed by the encircling box algorithm,and the processing efficiency is improved.Compared with the two commonly used point cloud simplification methods,the experiment on mountainous and plain terrain has great advantages in terms of simplification rate,DEM accuracy,terrain compatibility,and strong terrain adaptability.This paper provides a new idea for LIDAR point cloud simplification,complements the existing LIDAR point cloud simplification method,and implements a high-precision DEM with a small number of point clouds..