融合RPC类透视相机模型和SuperPoint+SuperGlue模型的定位算法OA

In order to improve the positioning accuracy and efficiency of multi-source high-resolution remote sensing images in disaster areas without control points after earthquakes,this paper researched,designed and proposed a multi-source remote sensing image SFM orientation algorithm that integrates an similar to RPC perspective camera model and a deep learning algorithm.The specific steps are as follows:Firstly,the complex RPC camera model was transformed into a simple and universal RPC perspective camera model,and the camera parameters were solved;then,the fusion algorithm of SuperPoint and SuperGlue was used to carry out feature extraction and matching;after geometric verification,the incremental SFM strategy was finally adopted to achieve SFM orientation,so as to improve the orientation accuracy and efficiency of multi-source remote sensing images.In order to verify the advantages of the algorithm,the paper selected three high-resolution remote sensing image sub-regions with different characteristics for experiments.The results show that compared with the traditional RPC model regional network adjustment method,the method used in the research effectively reduces the complexity of RPC model construction,the directional matching accuracy is better,and the directional matching efficiency is increased by more than 3 times.