BJUT-3D三维人脸数据库及其处理技术  

BJUT-3D Large Scale 3D Face Database and Information Processing

作　　者：尹宝才[1] 孙艳丰[1] 王成章[1] 盖赟[1] 

机构地区：[1]北京工业大学计算机学院多媒体与智能软件技术北京市重点实验室,北京100124

出　　处：《计算机研究与发展》2009年第6期1009-1018,共10页Journal of Computer Research and Development

基　　金：国家自然科学基金项目(60533030;60825203);北京市自然科学基金项目(4061001);国家科技支撑计划基金项目(2007BAH13B01)~~

摘　　要：BJUT-3D是目前国际上最大的中国人的三维人脸数据库,其中包括经过预处理的1200名中国人的三维人脸数据,这一数据资源对于三维人脸识别与建模方面的研究有重要意义.首先介绍了BJUT-3D数据库的数据获取条件、数据形式,并针对数据库建立过程中数据预处理技术进行了讨论.最后作为数据库的直接应用,进行了多姿态人脸识别和人脸姿态估计算法的研究.实验结果证实,该算法具有良好的性能.3D face recognition has become one of the most active research topics in face recognition due to its robustness in the variation on pose and illumination. 3D database is the basis of this work. Design and construction of the face database mainly include acquisition of prototypical 3D face data, preproeessing and standardizing of the data and the structure design. Currently, BJUT-3D database is the largest Chinese 3D face database in the world. It contains 1200 Chinese 3D face images and provides both the texture and shape information of human faces. This data resource plays an important role in 3D face recognition and face model. In this paper, the data description, data collection schema and the post-processing methods are provided to help using the data and future extension. A 3D face data dense correspondence method is introduced. Dense correspondence means that the key facials points are carefully labeled and aligned among different faces, which can be used for a broad range of face analysis tasks. As an application, a pose estimation and face recognition algorithm across different poses is proposed. Eexpremental results show that the proposed algorithm has a good performance.

关 键 词：三维人脸数据库 人脸识别 三维人脸模型 形变模型 网格重采样 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]