The Research On Automatic Registration Technology Of Satellite Remote Sensing Images Based On Hybrid Methods

Along with the rapid development of satellite remote sensing,the types of image sensors are increasing,image resolutions are gradually improved,and the available remote sensing data rapidly growth.How to integrated the merits of different sensors,and make comprehensive use of the huge amounts of multi-source,multi-resolution information accurately and efficiently has always been a research hot spot.As a key step of the remote sensing image application,the result of registration will directly affect values of applications,so how to improve the efficiency and precision of remote sensing image registration is of great significance to applications of remote sensing images.This paper proposed a set of methods for automatic image registration based on the study of exist algorithms.In this article,remote sensing image automatic registration work is mainly divided into coarse registration and fine image registration.In coarse image registration works,and subdivided into three steps that includes image preprocessing,feature points extraction and solve image transformation parameters.In the pre-processing step,expounds in detail the pre-processing method firstly,and analyzes problems existed in the remote sensing images registration,using a rolling guidance filter method is proposed for remote sensing image preprocessing.The use of the method in theory can not only remove the noise of the image and also be filtered image texture dimensions,which can improve the efficiency of satellite remote sensing image registration.In the related works of feature points extraction,this paper studied typical algorithms of the field of image registration,combined with features of satellite remote sensing images registration which is large amount of data and more resolutions,and analysis advantages and disadvantages of all kinds of methods to solve remote sensing images registration,and proposed the corresponding improvement direction.On this basis,we chose the efficient fast robust feature(SURF)method for remote sensing image of coarse registration and make a detailed introduction.At the same time,in the related works of solving coarse registration image transformation parameters,this paper puts forward an improved iterative closest point(ICP)method,joined the scale parameter calculation,this method makes the ICP method can calculates transformation parameters of different scales of image data.In the fine satellite remote sensing image registration part,this paper introduces the algorithm of images registration based on Lucas-Kanade optical flow,and introduces a kind of improvement of the algorithm in detail,at the same time of theoretically in this paper,the improvement work,combined with the current existing small displacement error in the coarse registration results which is result from feature points error matching,through necessary optimization to the improved algorithm,eventually making the registration results achieved higher precision,and complete the entire process of the satellite remote sensing image registration from coarse to fine.By simulated experiment to various parts,improved the value that rolling guidance filter can improve the efficiency of the remote sensing image registration.And improved ICP method has high efficiency when calculating transform parameters.The improved Lucas-Kanade optical flow method can better solve the problems existing in the coarse registration results,which further improve overall precision of the algorithm.