Research On Registration Method Of Oblique Image And High Resolution Satellite Image Assisted By Imaging Model

Nowadays,people's understanding of remote sensing information technology is deepening,and remote sensing technology has become one of the important ways for humans to obtain spatial information.The diversification of remote sensing image acquisition methods provides a rich means for earth observation.While massive and multi-source remote sensing images bring convenience to spatial information collection,they also bring the challenges of automatic,real-time and high-precision processing of information data.Among the massive multi-source data,high-resolution satellite imagery and UAV oblique image are important data sources for effectively acquiring ground object information.The full and efficient use of the two sources has become one of the key technologies for remote sensing earth observation,and image registration of the two sources is often the premise of their comprehensive utilization.The registration technology of oblique image and high-resolution satellite imagery is the basic technology in industrial production and machine vision,such as target change monitoring,image fusion,map update and 3D reconstruction.And it is particularly useful to study the automatic registration of oblique image and high-resolution satellite imagery.Therefore,this paper studies an imaging model-assisted automatic and fast registration method for UAV oblique image and high-resolution satellite image from coarse to fine,which provides an idea for automatic,fast and accurate registration of heterogeneous images under large-scale,large-angle,and large-light differences.The research content of this paper mainly includes:(1)An image coarse registration method based on imaging model is proposed.With the central projection model of the oblique image and the RPC model of the satellite image,the three-dimensional coordinates of the object point are used as a bridge to establish the approximate correspondence between the oblique image and the high-resolution satellite imagery points as the matching points for coarse registration.An image extraction strategy was developed to extract high-resolution satellite image required for subsequent registration from large imagery,which reduces the workload of subsequent fine registration.(2)The RPC parameters attached to the satellite imagery usually contain systematic errors.In order to make the coarse registration result more accurate,the refined RPC model is needed to calculate the image point coordinates of the ground points on the satellite imagery.To this end,a refinement method for directly optimizing partial RPC parameters using a small number of ground control points is proposed.(3)The two types of images extracted after coarse registration still have great differences in illumination,gray scale and viewing angle.For the heterogeneous optical remote sensing image registration,an affine invariant region feature registration method based on combination features of Hessian-Affine and MSER is proposed to effectively balance the speed and accuracy of registration.(4)Aiming at the problem of mismatching after using RANSAC algorithm to refine matching points,a global mismatching elimination strategy is proposed to further improve the accuracy of matching and registration.