三维特征点距离特征集合求交匹配算法

Distance feature set intersection for 3D feature matching

对于使用锥束CT分区成像的物体,要得到其完整的三维图像,需要对各分区重建图像进行三维拼接。作为基于特征的三维拼接算法中重要的步骤之一,特征点匹配是要对图像重叠区域中检测到的特征点建立对应关系。针对目前三维SIFT特征匹配算法对于相似特征误匹配率较高的问题,提出基于三维特征点空间关系的三维特征点匹配算法:距离特征集合求交法。该算法使用求取简便的特征点三维距离特征作为特征描述符,避免了扩大特征信息统计范围时巨大的计算消耗问题,然后在匹配过程中设计了距离特征集合求交的相似性度量方法,解决了以往基于空间关系方法中特征矢量各项元素不对应的问题。实验证明:该算法在图像存在大量相似特征的前提下,能够有效提高三维特征点匹配的匹配正确率。

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.