中文摘要：

在人脸识别算法中,尺度不变的SIFT特征是人脸局部特征的重要描述方式之一。在传统SIFT方法中,由于特征点的坐标是根据高斯差分空间的局部极值点来确定的,造成特征点匹配困难。通过基于回归的局部二值特征对人脸进行校准,确定对人脸有意义的特征点位置。用SIFT特征描述子的不变特性描述人脸的局部特征,能够有效地提高识别速度以及识别率。对特征点进行区域加权,能够对人脸的姿态变化以及角度偏转有一定的鲁棒性。

英文摘要：

The scale invariant feature transform（SIFT）in face recognition algorithm is one of the important descriptive approaches of the face local features. In traditional SIFT method,because the coordinates of the feature points are determined according to the local extreme point of Gaussian difference space,which leads to difficult match of the feature point. The face is calibrated by local binary feature based on linear regression to determine the location of the feature point which is meaningful to the face. The face local feature is described by the invariant characteristic of SIFT feature descriptor,which can effectively improve the face recognition speed and recognition rate. The region weighing for feature point has certain robustness to the face posture change and face angle deflection.