基于显著点信息的人脸年龄估计算法
关键词 年龄估计，预处理，年龄特征向量，估计模型
毕业论文设计说明书外文摘要
Title Facial Age Estimation Algorithms Based on Landmark Information
Abstract
Image-based age estimation has become an active research topic in pattern recognition and human-computer interaction field. As aging process is personalized from person to person, finding an appropriate age representation arithmetic and building an appropriate estimation model have always been an open problem in age estimation. In this dissertation, firstly, facial images are pre-processed in three ways. Secondly, we extract shape feature and texture feature based on gray and multi-scale local binary pattern respectively. Finally, several classification models and support vector regression model are respectively built with shape, texture and combination of these two features. Experimental results show that, among classification models with different features, collaborative representation based classifier model with the shape + statistical gray texture feature performs the best. Moreover, the support vector regression model with the statistical gray texture feature achieves a better performance than with other features.
Keywords age estimation, pre-processing, age feature vector, estimation model
目 次
1 绪论 1
1．1 课题研究背景 1
1.1.1 人脸年龄估计的意义 1
1.1.2人脸年龄估计的重要应用 1
1.1.3人脸年龄估计面临的困难 1
1.2研究综述 2
1.2.1年龄特征 2
1.2.2估计模型 3
1.3主要研究内容 4
1.3.1主要工作和成果 4
1.3.2论文的组织结构 5
2 预处理 6
2.1人脸年龄数据库的介绍 6
2.1.1 FG-NET数据库 6
2.1.2 MORPH数据库 7
2.2图像预处理 8
2.2.1图像灰度化 8
2.2.2 图像的旋转 9
2.2.3图像的裁剪 9
2.2.4 尺度归一化 10
2.2.5 直方图均衡化 10
2.3本章小结 11
3 年龄特征的提取 12
3.1 主成分分析 12
3.2形状特征 13
3.3纹理特征 14
3.3.1纹理归一化 14
3.3.2 统计灰度纹理特征 16
3.3.3形状特征+统计灰度纹理特征 17
3.3.4统计MLBP纹理特征 18
3.3.5 形状特征+统计MLBP纹理特征 21
3.4本章小结 22
4 估计模型的建立 23
4.1最小距离分类器 23
4.2 K-近邻分类器 23
4.3基于协同表示的分类器 23
4.4支持向量机 24