Research On Incremental Ortho-rectification And Stitching Algorithm Of UAV Images

Due to frequent occurrence of natural disasters,rapid acquisition of post-disaster global images is a key step in disaster estimation and emergency rescue.Based on this,fast and high-precision remote sensing has been gradually applied in major emergency disaster relief work.With its unique high mobility,high efficiency and low energy consumption,the UAV low-altitude remote sensing platform can urgently obtain images of disaster areas,and is widely used in disaster emergency response and aerial monitoring of dangerous areas.Due to the poor stability of the UAV’s flight attitude and the fact that the imaging process is easily affected by terrain fluctuations and the curvature of the earth,there is a phenomenon of image point displacement in aerial images.Therefore,the orthorectification of the original image is the core content of the rapid processing of the entire UAV image.The traditional orthophoto correction method generates the DOM of the survey area through photogrammetry post-processing software,space three encryption,overall adjustment,etc.,and the timeliness does not meet the emergency needs.In this paper,UAV aerial photogrammetry is used to quickly obtain images of disaster areas,and appropriate and efficient processing methods are used to correct and stitch the UAV aerial images with small image size and large overlap,so as to achieve rapid production of orthophotos in the disaster area.This paper mainly studies from the following aspects:(1)Research on incremental orthorectification algorithm for single image.The traditional aerial triangulation-based acquisition of orthophotos has the problems of cumbersome processing,long time-consuming,and inability to meet the timeliness requirements of emergency disaster relief.This paper proposes a fast incremental orthorectification algorithm to optimize POS parameters and eliminate occlusion.Integrate classical Kalman filtering,forward filtering,and backward smoothing three filtering algorithms to process POS system data,eliminate pose data transmission errors,and meet the requirements for the accuracy of external orientation elements of images;The DSM data in the geographic information database is extracted,and then the occlusion detection algorithm eliminates the height difference displacement of objects in the image,and uses the inverse solution differential correction to realize the incremental orthorectification of a single image one by one.(2)Research on collaborative optimization algorithm of orthophoto incremental stitching.In view of the low efficiency and local optimality of traditional image stitching algorithms,it is impossible to avoid obvious features with elevations much higher than the ground,so it is difficult to optimize the stitching line to obtain largearea orthophoto images.In this paper,a new energy function and an improved dynamic programming algorithm for global orthophoto stitching collaborative optimization algorithm are proposed.Integrate the three cost functions of gray level difference,gradient difference and texture difference to construct a new energy function,calculate the energy value in the image,generate an image energy map,and quickly identify objects in the image whose elevation is significantly higher than the ground;based on the energy map,the improved dynamic programming algorithm searches and avoids objects whose elevation is significantly higher than the ground,and does not divide the splicing lines of lower objects,so as to achieve seamless splicing of adjacent images.This study can effectively avoid buildings and other objects whose elevation is significantly higher than the ground,and achieve rapid splicing and global optimization of orthophoto images.(3)Applicability of single image correction and large-scale image stitching algorithms.Taking a survey area in Dengfeng City,Henan Province as the research object,simulating the current situation of emergency disaster relief,conducting research such as orthophoto correction of single image by UAV,global image stitching,etc.,to quickly obtain the global orthophoto map of the disaster area.The results show that the incremental orthophoto correction algorithm meets the timeliness requirements of emergency disaster relief;the collaborative optimization algorithm searches for the optimal splicing line and incrementally splices the global orthophoto image,which is suitable for splicing the large-area current situation map required by the disaster area.There are 35 figures,13 tables,and 100 references in the paper.