Multi-source Image Automatic And Fast Registration Method

Multi-source image registration refers to the process of using an effective transformation strategy to align two or more multi-source image data obtained from different imaging times,multi-source sensors,and different viewing angles in the same coordinate system.At present,multi-source image registration has become an important research part in the field of image processing.While the requirements for refined image processing are increasing,the real-time requirements are also getting higher and higher.However,it is often contradictory to meet the requirements of refined image processing and real-time requirements at the same time,which requires balanced consideration.Therefore,it has become the focus of research to find ways to improve the real-time of registration while ensuring a certain image registration accuracy.In response to the need for multi-source image registration under spatio-temporal dislocation,a fast and automatic multi-source image registration method based on dual-lens cameras(optical and infrared)with different focal length transformation parameters is proposed under the condition of obtaining pre-transformation matrix.This method not only realizes the rapid and automatic registration of optical and infrared images,but also the registration accuracy can reach sub-pixel level.When the pre-transformation matrix acquisition conditions are not available,a multi-source image registration method based on non-closed multi-dimensional contour feature sequences is proposed.This method realizes rapid registration of multi-source images with obvious contour features.This technology is of great significance to the development of my country's multi-source moving target detection,terrain matching navigation,anti-UAV detection,disaster prevention and mitigation,and power line detection.The specific work of this paper is as follows:Aiming at the problem that the traditional multi-source image registration method has slow registration speed and cannot realize full automation,the research on the rapid automatic registration method of dual-lens camera has been carried out.Using the dual-lens camera to quantify the transformation parameters within the focal length step is approximated as an invariant feature,and a model library of the transformation matrix parameters of the dual-lens camera with different focal lengths is established,in which the transformation matrix parameters reach 100 groups.Based on the matrix parameters,a multi-source rapid automatic registration method based on the different focal length transformation parameters of the dual-lens camera is proposed.The registration process is mainly to first obtain the focal length of the dual-lens camera corresponding to the image to be registered,then use the corresponding focal length transformation matrix parameters to quickly eliminate mismatched corners and irrelevant corners,then use distance information and position information to constrain optimization,and finally use random sampling The consensus algorithm(RANSAC)realizes the robust estimation of geometric transformation parameters.Under the premise that the registration accuracy also reaches the sub-pixel level,the registration time of this method is reduced from the original 1 minute to about 20 seconds compared with the multi-source image registration method based on multi-feature joint constraints proposed by the laboratory.The method can realize automatic registration.In view of the problem that multi-source images cannot obtain sufficient number of point features or complete edge contour features due to different shooting environments,different imaging mechanisms,imaging platform movement and mechanical jitter,etc.,based on non-closed multi-dimensional contours The multi-source image registration method of feature sequence is proposed by us.This method can obtain more obvious edge curve characteristics by denoising the multi-source image,and then use an improved method to estimate the concave and convex points on the curve,and use the method of polyline segment fitting to describe the edge curve characteristics of the multi-source image,And then perform the initial registration of the images,and finally use the iterative optimization method to search for the optimal transformation model parameters to achieve rapid automatic registration of multi-source images.