Extracting Digital Elevation Model Of Mountain Forest Area From Satellite Remote Sensing Image And Photon Radar Data

For a long time,the digital elevation model extraction from satellite optical remote sensing images has been faced with problems such as low efficiency and insufficient accuracy.In recent years,the new generation of satellite-borne lidar technology represented by ICESAT-2 /ATLAS photon counting detection radar has made great progress,providing favorable conditions for solving this problem.Based on the optimization of satellite remote sensing image data processing and photon point cloud data processing methods respectively,a joint processing algorithm based on linear point cloud and planar point cloud registration is proposed to improve the uncontrolled positioning accuracy of digital elevation model in mountain forest region.The main work and innovations of this thesis are as follows:(1)In the data processing of satellite remote sensing image,Fourier polynomial instead of general polynomial is proposed as the systematic error compensation term,and the adjustment solution of RFM model is carried out to improve its positioning accuracy.The experimental results of ZY-3 and GF-7 images show that the proposed adjustment model can effectively improve the positioning accuracy of optical images.The plane and elevation accuracy of ZY-3 images are 2.55 m and 0.72 m,respectively,and the plane and elevation accuracy of GF-7 images are 0.22 m and 0.28 m,respectively.(2)In the process of photonic point cloud data,a photonic point cloud filtering algorithm based on grid continuity constraint is proposed.The experimental results of ICESat-2 /ATL03 data show that the proposed algorithm can effectively filter noise points and extract surface elevation points,the filtering success rate is higher than 94%,the elevation extraction accuracy is better than 5.4m,and it can achieve stable processing effect for photon point cloud data with different topographic changes,ground object types and laser energy.(3)The digital elevation model is extracted by joint processing of photon point cloud and optical remote sensing image,and a joint processing algorithm based on line-plane point cloud registration is proposed.The experimental results show that with the introduction of ICesat-2/ATL03 photon point cloud,the accuracy of ZY-3 photographic point cloud in X,Y and Z directions reaches 0.46 m,2.43 m and 2.13 m,respectively,and that of GF-7 photographic point cloud in X,Y and Z directions reaches 0.36 m,1.34 m and 0.59 m,respectively.The extraction accuracy of digital elevation model is improved significantly.