基于多分类器融合算法的3D人脸年龄识别

A 3D Facial Age Recognition Algorithm Based on Multi-Classifier Fusion

为了提高人脸识别中待测人脸图像年龄估计的正确率,提出了一种基于多分类器融合的3D人脸年龄识别算法。首先,利用人脸的纹理信息将二维图像映射到标准三维模型上,并以贝叶斯决策理论为基础,对K ittler提出的多分类器融合算法理论框架及其组合规则进行了详细的研究、讨论和改进,然后应用改进后的多分类器组合规则将多个单独识别分类器加以融合以达到分类未知年龄目标人脸的目的,并估计人脸年龄。实验结果表明,算法可有效估计目标人脸年龄,并减小估计误差。

In orde to improve the accuracy of age estimation in face recognition, a modified 3D facial age recognition algorithm based on multi -classifier fusion is presented in this paper. First, 2D face images are mapped to 3D standard models with their texture information. Then on the basis of Bayesian decision theory, the multi - classifier fusion algorithm and fusion rules proposed by Kittler are systemically discussed and improved. With the improved fusion rule, several individual classifiers are combined so as to classify the target faces with unknown ages, then ages can be estimated further. Experiments show that the algorithm effectively enhances the recognition rates of different facial ages.