High resolution images can be obtained by using UAV technology. In order to obtain high precision homographic matrix and to improve image matching efficiency, this paper improves RANSAC algorithm, eliminates match gross error through by adding image gray level information. Then the root mean square error is used to evaluate quality of image matching. In order to verify the reliability of the algorithm, a series of buildings images were taken to test the algorithm. The results show that rate of gross error match points elimination was increased by 7.45^%, and the new algorithm had a small root mean square error, which increased the accuracy significantly.