改进的主动形状模型方法在人脸特征点定位中的应用

摘要

Active shape model (ASM)is a statistical parametric model,and is mainly used for extracting the features of image.In original ASM model,the principal component analysis (PCA)approach is used to extract the shape eigenvectors and shape parameters from the train-ing set data,and then they are used to build linear statistical parametric shape model according to the transition matrix.In the paper we pro-pose a new method of locating the facial features with the improved active shape model based on analysing the deficiency of traditional method. This method uses incremental PCA,which can effectively overcome the problems caused by the factors of model matching failure and the in-fluence of the testing image,and can update the texture model of the training set.Experimental results show that the improved method can ef-fectively improve the model matching accuracy,and speeds up the features locating time of the model at the same time.%主动形状模型 ASM（Active shape model）是一种统计参数化模型，主要应用于图像中的特征提取。传统的 ASM方法对训练集数据采用 PCA 方法获得形状特征向量和形状参数，然后根据转换矩阵建立线性的统计参数化的形状模型。在分析传统方法不足的基础上，提出一种改进的主动形状模型定位人脸特征的新方法。该方法采用增量 PCA，可以有效解决模型匹配失败和受测试图像影响等因素，同时可对训练集进行纹理模型更新。实验结果表明，改进的方法可以有效提高模型的匹配精度，同时加快了模型定位特征点的时间。