The Research Of Remote Sensing Image Fusion Method Optimization Based On The GF-2 Satellite In The Mountain Area

With the rapid development of science and technology, image fusion technology has been widely applied in different fields. At the same time, various kinds of new satellite sensors also made the data volume of remote sensing images increase rapidly. GF-2 satellite successfully launched into orbit on August 19th,2014, and the first image data released on September 29th,2014, which is a major breakthrough in the development of Chinese remote sensing industry. The high spatial resolution panchromatic image’s spatial resolution reached 0.81m. Even the multispectral image’s reached 3.24m. It is also show that the spatial resolution of our country’s independently developed remote sensing satellite had reached the international advanced level of submeter level. However, in view of the widespread and shadow effects obvious mountain landform in China, the research of image fusion method based on the GF-2 data and its application effect test were rarely reported. Then, on the foundation of the summarizing and analyzing for the previous literatures and results, the mountain area which is less studied was selected as the study area, and the GF-2 remote satellite data was chosen as the research object. This paper comprehensive explored and established the GF-2 remote sensing satellite data based on pixel level fusion and optimize method in the mountain area. Which the scientific and social significance are obviously.This paper’s main achievements are shown as follows:(1) Shangyong and Mohan which are the border regions of Yunnan province was chosen as the study area. In the mountain area, the shadow area turns up because the elevation is remote, the slope is steep and the brightness value of the image is different, which is unfavorable to obtain the better image fusion effect. So terrain correction which is based on the slope matching model applied was carried out in the fusion image pretreatment of the GF-2 satellite data in the mountain area. After terrain correction, the mean values in the brightness and shady of the image were smaller. It is show that the terrain correction pretreatment before image fusion had a better result, and could basically eliminated the impact of mountain shadow. It had laid a solid foundation for the image fusion of the GF-2 satellite data in the mountain area.(2) On the basis of the terrain correction and generated a new green band through the green wave band and the near infrared band, used the GF-2 satellite data, this study systematically carried out various kinds of method, for example IHS transform, PCA transform, wavelet transform, IHS+ wavelet transform, IHS+PCA transform, PCA+ wavelet transform, IHS+PCA+wavelet transform and so on. Eight objective evaluation indexes which are mean value, spectral correlation coefficient, spatial correlation coefficient, entropy, average gradient, space frequency, gray variance index and peak signal to noise ratio were selected to comprehensively evaluate the optimization of fusion method combined with subjective evaluation. Finally, IHS+ PCA+wavelet transform method was selected as the optimized fusion method in the mountain area.(3) Demonstration and field verification of object-oriented land use information extraction was carried out based on the optimized fusion image by IHS+PCA+wavelet transform method of the GF-2 satellite data in the mountain area. It is also proved that the IHS+PCA+wavelet transform of the GF-2 satellite data in the mountain area is the best method of image fusion in this paper. Based on the object-oriented classification method, the above optimized fusion method image results of IHS+PCA+wavelet transform was segmented by multi-scale segmentation algorithm. The land use information was extracted through making full use of image object spectral, texture, and other related class information. Combining with the field survey and visual of high resolution image select samples, and using nearest neighbor classification method. Which were on the support of eCognition software. Through error matrix based on samples accuracy evaluation. It is show that the fusion image which is processed by IHS +PCA+wavelet transform is further suitable for land use classification. This also testified from the aspect of application effect that the IHS+PCA+wavelet transform was the optimal processing method for the GF-2 satellite data fusion in the mountain area.