Research On Image Fusion Of GF-2 Satellite And Mine Development Land Occupation Information

With the development of remote sensing technology and the research and use of various new satellite sensors,remote sensing images have been widely used in different industries.However,in the mineral resources development environment remote sensing monitoring project,most of the extraction of mine development land occupation information by production units still uses the method of visual interpretation of remote sensing images.In response to this problem,this paper studies the image fusion methods of images of the GF-2 satellite and its application in mine development land occupation information extraction,in order to get a more convenient way to extract information on mine development land occupation.(1)This paper selects the Ukulqi coal mine and its surrounding area in the Qapqal Xibe Autonomous County of Xinjiang as the study area,and performs orthorectification,image registration and image enhancement on the panchromatic and multi-spectral images of the GF-2 satellite.This basically eliminates the effects of terrain,noise,and geometric distortion.(2)On the basis of image preprocessing,this paper systematically studies the methods of fusion experiments,consisting of the HIS transform,PCA transform,Wavelet transform,Contourlet transform,HIS+Wavelet transform,PCA+Wavelet transform,HIS+PCA+Wavelet transform,HIS+Contourlet transform,PCA+Contourlet transform,HIS+PCA +Contourlet transform.Then it selects eight objective evaluation indexes including standard deviation,information entropy,average gradient,edge intensity,correlation coefficient,peak signal-to-noise ratio,structural similarity and global fusion quality index to comprehensively evaluate the fusion results combining the subjective evaluation.Finally,the quality of the HIS+PCA+ wavelet transform fusion image is better.(3)In this paper,the multi-spectral image and the fusion image obtained by HIS+PCA+ wavelet transform method are used to further extract the land occupationinformation of mine development.The maximum likelihood classification method and the support vector machine are selected to extract mine development land occupation information in the study area.There are the results.For the fused image,the overall classification accuracy of the maximum likelihood method is 83.32%,the Kappa coefficient is 0.8088,the overall classification accuracy of the support vector machine is 86.76%,and the Kappa coefficient is 0.8480.For multi-spectral images,the overall classification accuracy of the maximum likelihood method is 71.15%,the Kappa coefficient is 0.6669,the overall classification accuracy of the support vector machine is 73.63%,and the Kappa coefficient is 0.6941.The fusion image is effectively classified and extracted,and the extraction precision is improved compared with the multi-spectral image.The classification accuracy of the support vector machine is higher than the maximum likelihood method,which further verifies the choice of the fusion method.