中文摘要：

针对主动形状模型中局部灰度模型结构简单、初始搜索位置偏移较大的问题,提出一种基于ASM的改进型人脸特征点定位方法。利用左右瞳孔和嘴巴中心的坐标,调整模型的初始搜索位置,在以特征点为中心的一定矩形区域,建立二维局部加权灰度模型和局部纹理模型,并采用新的模型匹配方法定位目标特征点。实验结果表明,基于ASM的人脸特征点定位改进方法相比传统ASM方法,定位精度提高了26.15%。

英文摘要：

In order to solve those problems that the structure of local gray model in active shape model algorithm is too simple,and initial model position has a misregistration,an improved facial feature points localization method based on ASM is proposed.Using the coordinates of pupils and mouth to adjust the initial search location,and then establishing a 2Dlocal weighted gray model and local texture model in an expanded rectangular area around the center of the feature point,finally the feature points can be located with the new matching algorithm.Experimental results show that our method improves26.15% on the positioning accuracy than the traditional ASM algorithm.