基于二维几何特征与深度特征的人脸识别技术研究
Study on Face Recognition Technology Based on Two Dimensional Geometric Features and Depth Features

人脸识别一般可分为识别及验证等两种应用场景。早期的人脸识别输入数据，大多是通过二维灰图像实现，同时，一些方法也引入了立体视觉系统来获取三维信息进行识别。由于受到使用环境上的诸多限制，如二维脸部数据较易受到环境光源、脸部方位、及化妆等影响而造成准确性降低，而三维人脸识别信息虽可克服上述缺点，但仍有高运算量及易受脸部表情影响等限制。因此，结合二维与三维的优点，本文实现一个实用且稳定的人脸识别技术，本文通过对目前该领域的研究，作一详细的介绍与比较，相信能对建基于多维信息的现代人脸识别系统的发展有所帮助。
The application of face recognition technology generally focuses on two scenarios including identification and validation. Conventional approaches use two-dimensional intensity or color images, and some approaches introduce stereovision system to acquire three-dimensional information for the recognition process. In general, two dimensional based approaches suffer from severe problems related to environmental conditions, such as variation of ambient light, occlusion due to face orientation and costume. On the other hand, three dimensional based ones have to overcome the challenge of computational cost and expression variation, though some of the previous two-dimensional problems can be avoided. As the depth sensor technology is being improved, more and more depth measurement equipments are utilized to generate three-dimensional data, particularly the depth of the facial object, for numerous systems. The integration and combination of these two spectrums is thus now a very active research interest, which aims to construct a reliable and efficient face recognition system. The purpose of this paper is to provide a comprehensive survey of the proposed methods in this area and a comparison among them, in order to yield fundamental basis for future developments of a practical face recognition system.