基于遗传算法的正面人脸定位

GA Based Front View Human Face Localization

利用人脸的椭圆轮廓特征，使用参数化可变形椭圆模板和遗传算法（ＧＡ）对正面人脸进行定位．首先取得人脸的边缘图像，并定义边缘图像的能量场，然后定义参数化的可变形椭圆模板，对椭圆模板的参数进行编码，并定义合适的适应值函数，使用ＧＡ以椭圆模板在人脸边缘能量场图像中对人脸轮廓进行最优化匹配．实验结果表明，该方法在中等复杂背景环境下的人脸定位应用中取得了较好的性能，定位更加准确。

According to the feature of hum an face contour, this paper introduced a novelm ethod to find front view hum an face in cluster background using param eterized deform able ellipse tem plate and genetic algorithm s. First, extractthe edge offace im age and define the energy field ofedge im age. Then, define the param eterized deform able ellipse tem plate, code the param eters ofellipse tem plate and give a fitness function. Last, the m atch betw een ellipse tem plate and face contour w as optim ized by GA in the edge energy field offace im age. The experim entalresultshow s thatthism ethod offace localization has a good perform ance in m oderately com plex background scene and is m uch m ore accurate and robust.