Research On Registration Technology Of Infrared And Visible Remote Sensing Images With Different Resolution

Earth observation from space is developing towards high resolution,high spectrum and high sensitivity.Fusion and utilization of remote sensing images in different bands can provide more abundant information,so it has great application prospects.Image registration is the basis of image fusion,and its results directly affect the practical application of remote sensing images.Infrared and visible remote sensing image registration is an important mode in multi-sensor(multi-band)image registration.It can effectively utilize the complementarity of infrared and visible light information to help people better understand the scene and identify the target,so in the environment The fields of monitoring,target tracking,and military intelligence acquisition have been widely used.The main contents of this thesis on infrared and visible remote sensing image registration technology with different resolution include the following:1.The image registration technology has been studied in detail from the aspects of model,composition and specific implementation methods.By comparing the spectral characteristics and imaging characteristics,the differences and similarities between infrared and visible images are analyzed in principle.According to their similarities and differences,the key problems that need to be solved between them are studied from the aspects of feature description,feature reproducibility,computational complexity and common features.2.A symmetrical edge information extraction method is proposed.The edge information extracted by this method is applied to infrared and visible image registration and compared with traditional edges.The results show that,compared with the traditional edge,the symmetry edge has higher registration precision and mutual information peak in image registration based on mutual information measure.3.An image registration algorithm based on regional edge mutual information is proposed.The algorithm achieves registration between infrared and visible images by adjusting image scale,extracting symmetrical edges,selecting feature regions,and registering the feature regions by normalized mutual information.The results show that the proposed method has high registration precision and repeatability in infrared and visible image registration.4.The proposed image registration algorithm is applied to infrared and visible remote sensing image registration with different resolution of Tiangong 2 Wide-band Imaging Spectrometer(WIS).The WIS contains remote sensing images of visible light,short-wave infrared and long-wave infrared,and the spatial resolution of subsatellite points in each spectrum is 100,200 and 400 meters,respectively.Two sets of remote sensing image registration experiments,“short-wave infrared-visible” and “long-wave infrared-visible”,were set up.The results show that the root mean square error(RMSE)of the proposed method in these two sets of experiments reached RMSE < 1 and RMSE < 2,respectively,which can meet the requirements of image registration with pixel-level precision.