Abstract: To get the entire three-dimensional(3D) image of the object scanned separately by cone beam computed tomography(CBCT), it needed to process the reconstructed image of each region by 3D image mosaicing. As an important step of the mosaicing approach based on feature point, feature point matching buildt the one-to-one relationships between the points detected in the overlap regions. Aiming at the mismatch problem that caused by similar features in the feature matching process of SIFT, a 3D feature point matching method was presented based on spatial relations called Distance Feature Set Intersection(DFSI). This method firstly used easy-calculating 3D distance features to form descriptors, which avoided the large computation cost by expanding the statistical range. Then, distance feature set intersection was devised as the similarity measure, which solved the problem of feature vector elements not corresponding in previous method based on spatial relations. The experimental results show that the proposed approach improves the matching accuracy when images have multiple similar regions.