Abstract: A study is made on the method used for the normalization of human face’s Point Distribution Model（PDM） in Active Shape Model（ASM），an improved method for the normalization of human face’s PDM is proposed based on a geometric transformation.This approach makes full use of the geometric characteristics of human faces，so it can eliminate the“non-shape” factors in PDM better.Further more，this method normalizes all the models one-time without iteration，so this approach saves computing time than the old method.At last，it maintains the accuracy of the original method.These statements are verified by the experiments conducted on the ORL face database.

Key words: face recognition, Active Shape Models（ASM）, feature extraction, Procrustes method

摘要： 对主动形状模型（ASM）中用于人脸点分布模型归一化的方法进行了研究，以几何变换为基础，提出了一种改进的人脸点分布模型的归一化方法。这种方法充分利用了人脸的几何特征，因此能够更好地消除人脸点分布模型的“非形状”因素，而且该方法无需迭代，可以一次性将所有模型归一化，因此比原方法节省了运算时间，而且能基本保持原有的精度。通过在ORL人脸数据库上的实验，很好地验证了上述论断的正确性。

关键词: 人脸识别, 主动形状模型, 特征提取, Procrustes方法