Research On Multimodal Remote Sensing Image Registration Algorithm

Multimodal images provide richer,more comprehensive,and comprehensive spatial information than single-modal images.This makes it possible to compensate the deficiencies of the single image source.Automatic high accuracy registration technology of multimodal remote sensing images is the necessary prerequisite and key technology for point cloud coloring,image fusion,image stitching,target recognition and change detection,and it is also essential for many emerging geospatial environments and engineering applications.Therefore,the registration of very large multimodal,multitemporal,with different spatial resolutions data is,still,an open matter worth study.In this paper,the characteristics of multimodal images are analyzed,and the influence of multimodal remote sensing image differences on registration and the registration difficulties are summarized.We have studied Multimodal remote sensing image registration algorithms systematically and the analyzed shortcomings of the existing algorithms.At present,multimodal remote sensing image registration in various applications relies on human-computer interaction,which is labor-intensive and time-consuming.Powerful feature descriptors do not have the same robustness for multimodal images.At the same time,experiments are carried out on the current mainstream feature-based and transform domain-based registration algorithms.It is found that these methods are only effective for images of specific modalities or scenes,and cannot meet the sub-pixel level registration accuracy.In light of this,we propose a Generic and automatic Markov Random Fields(MRFs)-based registration framework of multimodal image using grayscale and gradient information.The proposed approach performs non-rigid registration and formulates an MRF model while grayscale and gradient statistical information of a multimodal image is employed for the evaluation of similarity while the spatial weighting function is optimized simultaneously.Besides,the value space is discretized to improve the convergence speed.The developed automatic approach was validated both qualitatively and quantitatively,demonstrating its potential for a variety of multimodal remote sensing datasets and scenes.As for the registration accuracy,the average target registration error of the proposed framework is less than 1 pixel,while the maximum displacement error is less than 1 pixel.For calculation efficiency,it takes 7-8 minutes to take time.In the meantime,the proposed approach had the partial applicability for the multimodal image registration of large deformation scenes and multitemporal dataset to locate the change region accurately.It is also proved that the proposed registration framework using grayscale and gradient information outperforms the MRFs-based registration using only grayscale information and only gradient information while the proposed registration framework using Gaussian function as spatial weighting function is superior to that using distance inverse weight method.In summary,based on the latest research results and technology,this paper systematically studies the multimodal remote sensing image registration algorithm,and carries out corresponding experiments and analysis.It can effectively realize highaccuracy and automatic registration of multimodal images,and promote the application of multimodal images in different fields.