【摘要】 为了提取人脸的全局和局部特征,以实现复杂环境下的人脸特征有效表达,提出一种基于梯度方向直方图(HOG)金字塔人脸识别方法.该方法通过多尺度分析、HOG特征谱提取来构建人脸图像的HOG金字塔,对HOG金字塔各层特征谱划分为非重叠小块来构建统计直方图,并将各层的统计直方图连接起形成特征向量来实现整个人脸的特征表达,计算不同序列特征的相似度并采用最近邻分类器来进行人脸识别.在光照和时间环境等复杂变化的人脸识别技术(FERET)库中进行人脸识别实验,结果表明:该方法具有较强的人脸特征鉴别能力,方法的鲁棒性也较好.更多还原

【Abstract】 In order to extract the local and global facial features for effective face recognition in complex environment,a face recognition method based on histograms of oriented gradients(HOG)pyramid model was.This method constructs the HOG pyramid of face image by multiscale analysis and HOG feature maps extraction.The feature maps in each layer of HOG pyramid are respectively separated into several non-overlapping blocks from which the HOG histograms are built and concatenated into an improved feature vector that can be used as the face descriptor later.Face recognition is performed throughout the similarity calculation among above feature vectors by using nearest neighbor classifier.Face recognition experimental results on the face recognition technology(FERET)database that captured under complicate changes of light and time environments demonstrate that the proposed method is of highly discriminable ability and good robustness in face recognition.更多还原