Abstract: Abstract: A matching method for high resolution satellite images based on RFM is presented. Firstly the RFM parameters are used to predict the initial parallax of corresponding points. Secondly the approximate epipolar equation is constructed based on projection tracking and its accuracy is analyzed. Thirdly approximate 1D image matching is executed on pyramid images and least square matching on base images. At last RANSAC is imbedded to eliminate mis-matching points and matching results are obtained. Test results verified the method robust and with high matching rate, compared to 2D gray correlation method and SIFT matching method, and solved the question of high resolution satellite image matching with different time and large rotation images.