【摘要】 基于肤色和器官的人脸检测是视觉监控领域中广泛应用的经典方法,但是辨别每个肤色像素和提取候选区域非常耗时且对噪声敏感,很多时候不能满足实时人脸检测的需要。通过引入肤色单元的概念,提高了该方法的快速性和鲁棒性,最终将其应用于实时视频序列中。首先,采用单元化的方法进行肤色分割,提取出人的肤色部分;接着,根据人脸长宽比例的范围,确定出候选人脸;然后,再对候选人脸区域分别进行眼睛和嘴巴的定位;对眼睛和嘴巴定位之后,我们可以利用眼睛和嘴巴呈倒三角关系的几何特征反过来进行人脸的精确定位。实验结果表明,该方法的识别率较高,并能满足实时视频人脸检测的快速性要求。更多还原

【Abstract】 Face detection based on complexion and organs location is one of the classical methods widely used in the field of visual monitoring,but discriminating each complexion pixel and extracting candidate areas are very time consuming and noise sensitive,which often cannot meet the requirements of real-time face detection.The concept of skin cells is introduced to improve the speed and robustness of the processing and applied to real-time video sequence.Firstly,complexion-like regions are extracted by classifying color skin units.Secondly,candidate face region are filtered with aspect ratio of the regions. Finally,real face regions are well located by utilizing the geometrical features of eyes and mouth and their inverted triangle relation.The experimental results show that this method has high detection accuracy and satisfies the requirements of real-time face detection.更多还原