一种单幅前视图的人脸识别算法

A face recognition algorithm based on the single front view

本文针对依据单幅正视人脸照片在监控视频中进行目标人物识别的问题,提出了一种通过单应变换获得空间多姿态的虚拟人脸的方法,对得到的人脸提取Gabor特征,并采用主成分分析法(PCA)降维之后,通过支持向量机(SVM)对其训练分类器,最后,采用半监督学习的SVM进行人脸识别。该方法在全国研究生智慧城市视频挑战赛的数据集上得到测试并验证了其有效性。

Based on the single front view of human face images, this paper aims at the solution of the problem by target face recognition in the surveillance video with a method proposed to obtain virtual and multi-pose face samples by homography; it extracts the Gabor feature; using principal component analysis （PCA） to reduce the data dimension; after its use for training SVM classifier, employing the semi-supervised SVM for face recognition. The method is used to verify the validity of the data set of National Graduate contest on smart-city technology and creative design.