| Airborne remote sensing has the ability of repeated observation and large area high resolution remote sensing data acquisition.It has been widely used in the fields of ground reconnaissance,disaster assessment,emergency and disaster relief,topographic mapping,environmental monitoring and so on.However,due to the limitation of the resolution,field size and flight height of aerial camera,the mutual restriction of field of view and resolution exists in single aerial remote sensing image.In order to meet the dual requirements of large scene and high resolution in the application of aerial remote sensing,the image mosaic technology is often used to synthesize multiple aerial images into high-resolution panoramic images,so as to be used for large scenes Emergency monitoring.This paper focuses on how to give full play to the role of aviation remote sensing in disaster emergency,studies the fast geometric rectification,matching and mosaic algorithm of aerial images,improves the traditional image mosaic algorithm and processing process,reduces the resource consumption of mosaic process and improves the efficiency of image mosaic to meet the high timeliness requirements of disaster emergency tasks,and provides a shadow for analyzing the environment and situation of disaster areas Like data security.This paper mainly studies in the following aspects:(1)Fast geometric rough rectification on aerial remote sensing image.In view of the fact that the current aerial image geometric rectification processing method is difficult to meet the high timeliness requirements of post disaster emergency response,this paper studies the realization of rapid geometric Rough rectification on aerial images by using the position and attitude data of positioning and orientation system(POS)and digital elevation model(DEM)data in terrain database without ground control information.According to the camera parameters and image data of aerial camera,DEM data and exterior orientation elements,the collinearity equation is used,and the image regular grid sampling method is used to carry out fast geometric rectification correction positioning for the image,and the image data with geographical coordinates is generated.Experimental results show that for a single 150 megapixel high-resolution aerial image,the geometric rectification accuracy can reach the absolute positioning accuracy of meter level,and the rectification speed is better than 15 s / image.(2)Image registration method based on improved AKAZE and multi-matching strategy fusion.Aiming at the problems of low registration accuracy and multi-image registration error accumulation of AKAZE algorithm in the process of image registration,an image registration method based on improved AKAZE and multi-matching strategy fusion is proposed.Firstly,AKAZE algorithm is used to detect multi-scale features.Then a stable Root SIFT descriptor is used to describe the feature.Secondly,multi-matching strategy fusion method is used to improve the matching accuracy and registration accuracy.Finally,the global registration method based on minimum spanning tree is used to achieve the registration of multiple images,which can effectively alleviate the cumulative error of multiple images registration.(3)Fast seamless mosaic of geometric rough rectification images.In this paper,a method of searching the optimal seamline based on geodesic time is proposed.On this basis,a fast mosaic process of aerial images is designed,which is called "Geometric Rough Rectification-Optimal Seamline Detection".In the process of mosaic,the geometric rough rectification images are not matched and adjusted,and the seamline are directly searched between the geometric rough rectification images,so as to realize the fast and seamless mosaic of geometric rough rectification images.Firstly,the energy cost function is generated by using the gray information difference between adjacent images,and the energy cost function is optimized by combining the gradient and edge information after texture-edge decomposition,so as to avoid the seamless passing through buildings and other obvious objects.Then,a method based on geodesic time is used to improve the efficiency of optimal seamless search,and finally a seamless high-resolution image with good overall visual effect is generated.By selecting different types of image data sets as test data,the effectiveness of the proposed method is verified. |
