Abstract: A robust and efficient driver’s face tracking algorithm in video that can detect faces is presented in this paper. The algorithm ensures the localization of mouth , nose and eyes with Horizontal gray-level projection region to construct Feature Triangle including isosceles triangle and right triangle for face tracking. The algorithm creats latency rectangle region with rigid body constraints. Then it constitutes forty-eight optimization filters according to 2D-Gabor kernel functions to acquire forty-eight feature point inside the latency rectangle region.Our algorithm can conquer different sizes,different lighting conditions,varing pose and expression,and noise problems. The algorithm can also detect faces from the side way. Experimental results demonstrate that an approximately 98% success rate is achieved.

摘要： 本文提出一种驾驶员面部跟踪的鲁棒、有效的视频图像人脸跟踪的算法。该算法根据水平投影曲线获取眼睛、鼻孔和嘴等区域构造特征三角形包括等腰三角形和直角三角形，再根据刚体约束生成潜在人脸跟踪矩形区域。再根据二维Gabor核函数构造最优滤波器48个，在特征跟踪矩形区域内获取48个人脸特征点。该算法能够在不同尺寸、不同光照、不同姿态和不同表情甚至不同噪音情况下跟踪人脸有效率达97.81%。