【摘要】 人脸特征点自动定位方法在人脸识别、三维人脸模型重建等方面都有重要作用.三维人脸模型重建对下巴特征点精度要求很高.采用一种结合遗传算法和活动外表模型(AAM)的人脸特征点定位方法(GA-AAM),对AAM算法在下巴轮廓提取中的不能精确收敛问题作了改进.对于用实时AAM算法做特征点粗定位得到的结果,在AAM的代价函数中引入代表特征点处的边缘信息,进一步采用遗传算法作优化.实验结果表明该方法对下巴特征点的精确收敛十分有效.更多还原

【Abstract】 Automatic extraction of facial feature points plays an important part in the fields like face recognition or 3D face model reconstruction.In the applications of 3D face model reconstruction,the precision of facial feature points,especially on chin contours,affects the reconstructed result directly.A method(GA-AAM) for searching facial feature points combining AAM with genetic algorithm(GA) is proposed,which can improve the precision of searched chin contours.After coarse searching of real-time AAM,edge information is invited in its cost function for further optimizing.Experimental results show that the proposed method can converge chin contours more precisely and efficiently.更多还原