为了更好地利用人脸图像中所蕴含的生物信息来提升人脸年龄估算的精确度，提出了一种基于主动形状与局部二值模式的年龄估算方法(An Age Estimation Method Based on Active Shape Model and Local Binary Pattern，AEM-ASLB)。首先，提取人脸图像的特征集合，该特征集合由脸部几何特征点和纹理特征信息组成。利用主动形状模型来定位脸部关键位置的特征点，面部皮肤的纹理信息则采用局部二值模式进行抽样提取;然后，利用高斯过程回归模型对提取到的几何特征信息进行训练，得到特征的直方图，并进一步归一化为纹理特征向量。经过实验证明，该方案能够利用人脸图像对年龄进行精准地估算，实现上也可以减少分类器的使用，整个方法思路清晰且简单易实现。
Age Estimation Method Based on Fusion of Active Shape and Local Binary Pattern
To make better use of the biological information contained in facial images,which will improve the accuracy of age estimation,an age estimation method based on active shape model and local binary pattern(AEM-ASLB)is proposed.First,the image feature set consists of the facial geometric feature points and texture feature information,and the facial feature points of the key position are carried out by active shape model,while the texture information of the facial skin is sampled by local binary model.Then,the geometric features and texture information extracted are trained by Gaussian process regression model to get the straight square of the features,and further normalized to texture feature vector.The experiment shows that the proposed method can precise-ly estimate the age by facial image,and reduce the use of classifiers,the whole method is clear and easy to realize.
facial imageage estimationgeometric proportiontexture featuresGaussian process regression
国家科技期刊平台
NSTL
万方数据
