The invention relates to a feature matching method based on a quadtree in combination with ORB and SIFT. Uniformly distributed feature points are detected by adopting a FAST feature point detection algorithm of a self-adaptive threshold; the feature points are describing by using an SIFT descriptor; the FLANN matcher carries out feature matching, a proportional threshold method is used for carrying out primary screening on matching pairs, then angle rotation consistency is used for carrying out secondary screening on the matching pairs, and finally an RANSAC algorithm is used for carrying outfine matching, so that the matching precision and robustness are improved.