改进SIFT融合五官特征的旋转人脸检测算法

Face Detection Rotation Based on Improved SIFT Integration of facial features

摘要:

为了有效解决由于采集设备位置的不理想或被检测者不知情等情况下,采集到的人脸图像不是正面的,进而造成人脸的误检和错检等问题；首先,将主成分分析(PCA)与尺度不变特征变换(SIFT)方法结合,分别利用PCA方法的降维和SIFT算法的旋转、平移、缩放及部分仿射不变性快速完成旋转人脸的初检测；然后,利用人脸五官特征对人脸进行矫正与标定,提升检测准确率；最后,通过改进AdaBoost方法训练人脸分类器并计算关键点匹配率,完成旋转人脸准确检测；实验结果表明:与传统方法相比,该方法保证了高检测率,同时错检率也明显下降。

Abstract:

In order to effectively solve the location of acquisition device is not ideal,collected face image is not positive,and cause false detection and error detection. First,the principal component analysis (PCA) and the Scale Invariant Feature Transform (SIFT) algorithm combined,respectively rotary dimensionality reduction and SIFT algorithm PCA algorithm,pan,zoom,and quickly rotating part affine invariant face early detection. Then,using the facial features of a human face human face correction and calibration,improve face detection accuracy. Finally,the improved algorithm AdaBoost face classifier training and match rate calculation of key points,complete and accurate detection of the rotation face. The results showed that:compared with the traditional method false detection rate was significantly reduced, at the same time ensure a high detection rate.