High-precision Extraction Method Of DIM Point Cloud Based On LiDAR-assisted Non-control Aerial Images

With the rapid development of intelligent and science technology,the demand for 3D models are higher and higher.Recently,two types of technology can directly gain point clouds including DIM(Dense Image Matching)and Li DAR(Light Detection and Ranging).The point clouds which are obtained by DIM are high density,significant edge characteristics and low cost.However,these point clouds easily exist error matching points because of the shade,radiation intensity,spatial texture and so on.In addition,in order to obtain real and measurable models,it is necessary to collect ground control points for absolute orientation.By this way,it will be high consumption of manpower and material resources.Through Li DAR system,the high-precision point clouds can be quickly collected.Li DAR system has an advantage of high speed,high precision,little interference from the weather,strong timeliness,permeability from vegetables and so on.However,the acquisition of Li DAR point clouds is expensive.Besides,the density of Li DAR is lower than DIM.These disadvantages lead to confining Li DAR in some departments.If only DIM or Li DAR is applied,it will be unable to the great point clouds.Li DAR are applied to assist DIM,not only obtaining high-quality DIM point clouds but also obtaining real models to save cost under the condition of non-control points.Recently,there are many methods of combining the DIM and Li DAR.However,these methods are not mature enough and ignores three factors.Firstly,it is ignored that the high-precision PT(point location)information of Li DAR point cloud can be considered as control points to assist the absolute orientation of aerial images.Secondly,it is ignored that there is some problems of the fusion of DIM point clouds and Li DAR point clouds because the principles of DIM technology and Li DAR technology are different.Thirdly,it is ignored the reasons leading to the miss matching error of DIM.Therefore,it has important theoretical and application value for the absolute orientation and optimization of dense image matching point clouds to combine DIM technology and Li DAR technology.How to complete the registration of aerial images and Li DAR point clouds and how to complete the optimization of DIM point clouds are high valuable subjects.This thesis studies those problem from four aspects mainly.(1)Based on the principles of DIM and Li DAR,a modified filtering algorithm based on CSF(cloth simulation filtering)algorithm is proposed to filter multi-source heterogeneous point cloud.It prepares for the registration and fusion of these heterogeneous point clouds.(2)Aiming to the absolute orientation of non-GCP aerial images,the LiDAR-assistant orientation method is proposed.This method considers Li DAR as “soft control” to complete orientation of aerial images,completing the orientation of aerial images.(3)In order to eliminate the mismatch points and optimize the information in shelter areas of DIM,the Li DAR-constraint optimization method is proposed.(4)In order to valid the effectiveness of the method(2)and method(3),an experiment is applied in the study area and the experiment result is analysis qualitatively and quantitatively.The result shows that through introducing Li DAR into DIM technology,not only solve the orientation problem of non-GCP aerial images,but also eliminate the mismatch points and optimize the information in shelter areas.