融合人脸特征和相关向量机的多姿态人脸检测

Combining Facial Features and Relevance Vector Machine for Multi-pose Face Detection

多姿态人脸检测是人脸检测研究领域中的难点和热点之一,针对这一实际应用中亟待解决的难题,提出融合人脸特征和相关向量机的检测算法。算法首先利用肤色特征快速排除大部分背景,在肤色区域中搜索眼睛和嘴巴区域。根据眼睛和嘴巴区域的几何特征所确定的人脸方向,分割出大致正向的人脸候选区域。最后选用分类性能比支持向量机更优的相关向量机对候选区域进行分类。对比实验表明,算法提高了多姿态人脸的检测率,对光照、表情和遮挡有较强的鲁棒性。

Multi-pose face detection has been one of difficult and hot issues in face detection research. For this practical application problem urgently need to be solved, a multi-pose face detection algorithm based on facial features and relevance vector machine algorithm is introduced. Making full use of skin color information firstly, the most background regions can be quickly excluded. After detecting eyes and mouth in the skin color regions, according to face orientation decided by the geometric features of the eyes and mouth region, the approximate frontal face candidates are segmented. At last, the face candidates are classified by relevance vector machine algorithm, which its classification performance is better than support vector machine. The experimental results demonstrate that the algorithm can further improve the multi-pose face detection accuracy and is highly robust to lighting condition, facial expression and occlusion.