头部多角度人脸快速跟踪算法DSP实现

针对传统驾驶员疲劳检测人脸跟踪算法复杂,DSP实现时实时性不强,不能有效地实现多角度人脸跟踪的问题,提出了一种快速人脸跟踪算法.该算法通过对YCbCr肤色模型进行图像预处理、肤色检测,提取人脸区域,通过对亮度信号Y进行统计运算,判断人脸边界,再进行相似度判断,从而实现人脸区域的跟踪.实验结果表明,该方法简单、鲁棒性强,能够快速地实现彩色图像人脸多角度跟踪.

Fast face tracking algorithm of head multi-position based on DSP

In the field of real-time human face tracking for driver fatigue detection,the classic algorithms are so complex that the DSP system can not track the face in multi-angle state quickly and exactly,so a new face tracking algorithm is presented.In YCbCr human face color model,the image is preprocessed first,then the face region is extracted through the face color detection.Through calculating the brightness signal Y,the face edge can be detected.Then a symmetry similarity measure is used to check the factuality of the face tracking.In this way,the face region can be tracked.Experimental results indicate that this algorithm is so simple and can realize the tracking of face in multi-angle of color image.