Abstract: The face verification is the branch of face recognition, and it is the research hotspot in criminal investigation. According to the characteristics of face verification, the Scale-Invariant Feature Transform（SIFT） algorithm is used, at the same time the image blocking method is combined to divide the SIFT features into number features and position features, thus the goal of face verification can be achieved. Firstly, the image registration is applied, then SIFT algorithm is used to find the matching feature points. Secondly, the images are blocked into several blocks, and the feature points are calculated in each block to achieve the feature vector. Lastly, the number of feature points is computed. If it reaches the preset threshold, the similarity of the matching vector is calculated. Otherwise, if any of the conditions is not satisfied, the images are defined as a non-verification pair. CAS and FERET databases are used to test the algorithm, and the false alarm rate is 19%, while the missing alarm rate is 0.3%. The experimental results show that the proposed algorithm is a new effective way in face verification. This method can be used in face verification after optimization.

Key words: face verification, Scale-Invariant Feature Transform（SIFT）, number features, position features

摘要： 人脸验证是人脸识别领域的一个分支，是安防领域的研究热点。根据人脸验证的特殊性，使用尺度不变特征（SIFT）算法，并利用图像分块方法，将特征点划分为数量特征以及位置特征，达到人脸验证的目的。所需验证的一对图像配准后，使用SIFT算法寻找出匹配特征点，对待匹配的2幅图像进行分块，统计各个分块中的特征点数量，获得匹配向量。判断两幅图像的特征点数量是否达到阈值，若达到则计算两幅图像的匹配向量相似度，若相似度达到标准，则认为图像对匹配，若有任何一个条件没有满足，则认为不匹配。利用CAS和FERET数据库进行测试，虚警率达到19%，漏警率达到0.3%，验证了算法的有效性以及安全性。该算法经优化后，可用于人脸验证。

关键词: 人脸验证, 尺度不变特征（SIFT）, 数量特征, 位置特征