Automatic Evaluation Method Of Airborne LiDAR Point Cloud Plane Accuracy

Nowadays,with the booming development of computer information network and the arrival of the era of big data,concepts such as digital cities,twin cities,and metaverses have emerged,and "digitization" has become the focus of people’s attention.In the process of digital city construction,airborne Li DAR is widely used due to the direct and efficient acquisition of 3D point cloud data.Using Li DAR point cloud as the data source of 3D city construction needs to evaluate its accuracy.The traditional method based on manual evaluation of point cloud plane accuracy has problems such as low efficiency,high cost and long cycle time.Aiming at the low degree of automation of the airborne Li DAR point cloud plane accuracy evaluation method,this thesis adopts the method of image matching to complete the automatic evaluation of point cloud accuracy.Generate images from the point cloud data,find multiple sets of point at the same position through image matching,and use the spatial position coordinates of the point at the same location to evaluate the plane accuracy of the point cloud.The main research contents and corresponding results of this paper are as follows:(1)Li DAR point cloud plane relative accuracy evaluation is achieved by using geographic coordinates of points with the same position after matching adjacent aerial band point cloud images.Analyze the matching accuracy number of point at the same location and efficiency of different feature matching methods such as SIFT、SURF、ORB、AKAZE and determine the matching algorithm applicable to adjacent strip point cloud images.(2)Li DAR point cloud plane absolute accuracy evaluation by using the geographic coordinates of point at the same position after matching the high-precision DOM image and the point cloud image.DOM and point cloud images belong to heterogenous data.In view of the problem that heterogenous images are difficult to match,this paper uses the Log-Gabor convolution sequence to construct the maximum moment map based on the phase consistency theory,and combines the Laplace transform to complete the feature point extraction.Primary Maximum Index Map(PMIM)is proposed,and the feature descriptor is constructed with reference to the SIFT algorithm idea;in the matching stage,the Euclidean distance between the feature vectors is used to measure the similarity of each feature point,and the matching accuracy is improved through two-way matching.The Random Sample Consensus(RANSAC)algorithm is used to eliminate the gross errors of matching points.(3)Based on manual checkpoints and ground control points,the accuracy of the evaluation method proposed in this paper is analyzed,the influencing factors of the evaluation method accuracy are discussed,and the influence of different point cloud sampling intervals on the evaluation method accuracy is analyzed.The experimental results show that the method based on image matching proposed in this paper can achieve reliable and automatic evaluation of the plane accuracy of airborne Li DAR point cloud.Compared with the traditional method,this method obtains a large number of points in the same position,which can enrich the accuracy statistical samples,which makes the evaluation results more objective.The heterologous image matching method proposed in this paper realizes the matching of optical images and point cloud images,and can also complete some other multi-modal remote sensing image matching tasks.