多摄像机环境下人脸最优视角选择算法

A Human Face Best Viewpoint Selection Algorithm in Multiple Cameras Environment

多摄像机环境中的人脸最优视角选择在多通道人机交互,视频会议,序列图像中的人脸识别等领域有着广泛的应用。论文给出了一种基于肤色与几何信息的人脸最优视角选择算法。该算法能利用肤色的聚类特性对人脸进行粗检测,并根据几何信息对头部进行定位,然后给出一种最优视角的准则,并在该准则下通过分析检测结果中肤色在头部区域中所占比例来间接求出人脸相对于摄像机的偏转程度,从而达到最优视角选择的目的。实验结果表明,该算法在多摄像机环境下能根据人脸正面的程度自动地进行视角切换,得到最优的视角。并且具有运算简单,实时性高,对环境的变化不敏感,多摄像机可以任意放置,无需校准的特点。

Best viewpoint selection in multiple cameras plays an important role in many applications such as human-machine interaction,videoconference,and face recognition in sequence images.In this paper,a method of face best viewpoint selection based on information of skin color and geometry is given.First,this method makes use of the clustering character of skin to look for human face and locates head according to information of geometry,Second,a rule of best viewpoint selection is proposed,then we analyze the proportion of skin in the area of head to compute the degree between human face and cameras.According to the rule,we can get the best viewpoint in multiple cameras environment.The result of experiment shows the method is real-time,robust,what's more,the multiple cameras can be located randomly and not be calibrated.