基于辐射能量的红外人脸识别  

Radiation-energy Based Infrared Face Recognition

作　　者：伍世虔[1] 韦礼珍[1] 方志军[1] 李润午[1] 王正友[1] 

机构地区：[1]江西财经大学信息管理学院,南昌330013

出　　处：《中国图象图形学报》2007年第10期1845-1848,共4页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60665001;60662003;60462003);江西省自然科学基金项目(0611025)

摘　　要：人脸识别至今还是计算机视觉中的一个难题,用红外人脸图像来进行识别是近几年才发展的一个新领域,虽然其独立于光源,不易受外部光线的影响,但研究发现,人脸温谱图因受到一系列内在(生理、心理等)和外在(环境、成像等)条件的影响而使识别率下降。针对这一问题,根据Stefan-Boltzmann定律,提出把易受环境温度影响的面部温度数据转换为较为鲁棒的辐射能量数据来进行识别,以提高红外图像人脸识别的性能。理论分析显示,当人体处于一个稳定的状态时,辐射能量图对周围环境温度的敏感度较低。实验结果表明,基于辐射能量图的识别率高于基于温谱图的识别率。Developing a reliable face recognition system has been an open problem. Infrared （IR） imagery appears to be a good alternative as it is independent of external illumination. However, IR images are also affected by a series of factors ranging from imaging, ambient condition to physiological and psychological conditions. The variations of temperature in IR images have analogous effects of illumination in visual images. In this paper, a method based on Stefan-Boltzmann Law has been applied to convert the facial temperature data into more consistent radiation energy data. The experiments conducted illustrate that the radiation energy data are less sensitive to ambient temperature if the human bodies are in steady state, and significantly outperforms temperature data in terms of recognition rate.

关 键 词：人脸识别 红外图像 辐射能量 

分 类 号：TP319.41[自动化与计算机技术—计算机软件与理论]