According to the high resolution of UAV aerial images, the simplified Scale Invariant Feature Transform (SIFT) algorithm is used to extract the feature points and complete coarse matching.Considering the special access of aerial images, a new method for feature point purification is proposed based on course constraint.Experimental results show that this method can purify the matching points effectively, and the purification rate runs up to 25%.Compared with Random Sample Consensus (RANSAC) algorithm, the efficiency is nearly doubled while maintaining the purification rate.