摘要
针对视频中的彩色序列图像,提出了一种人脸检测算法。该算法是一个由粗到精的检测过程。首先采用运动检测分析方法,根据多帧差分图像中运动物体边缘点的水平投影确定目标的水平位置,并结合肤色检测算法进一步确定人脸位置,然后用训练好的支持向量机进行人脸验证。实验结果表明,针对一般的彩色序列图像任意姿态人脸检测问题,该算法快速有效。
To detect human face in color image sequences quickly and effectively, a detection method from coarse to accurate is proposed. Firstly, the horizontal position of moving object is detected based on the horizontal projection of the edge of the difference images. Then, color information is used to detect face at the moving region. Finally, the human face region is confirmed by using a support vector machine classifier. Experiments show that the approach is fast and reliable.
出处
《数据采集与处理》
CSCD
北大核心
2005年第2期208-212,共5页
Journal of Data Acquisition and Processing
关键词
人脸检测
肤色
检测算法
支持向量机
face detection
skin color
detection algorithm
support vector machine (SVM)