摘要
如果单纯地采用静态人脸识别算法对视频图像进行检测、识别,忽略了视频中重要的前后帧相关的特性,会造成识别结果的不稳定。为了解决识别结果不稳定的问题,对视频监控中的人脸识别算法进行了改进,根据视频前后帧相关性对人脸运动进行估计并将其应用在视频监控的软件产品中。实验证明,该算法提高了传统单帧人脸识别算法的稳定性,具有较好的应用价值。
Most of the traditional face detecting algorithms recognize face only in a static image and ignore the important relationship between video frames and that will lead to unstable results.This paper improves the face recognition algorithm according to the relationship between video frames and applies it to a video monitor product.Experiments show that the improved algorithm enhances the stability of recognition result and has a preferable application value.
出处
《工程图学学报》
CSCD
北大核心
2011年第6期53-56,共4页
Journal of Engineering Graphics
基金
合肥工业大学青年教师基金资助项目(081002F)
关键词
模式识别
人脸检测
运动估计
人脸识别
视频监控
pattern recognition
face detection
motion estimation
face recognition
video monitor
