文档介绍：III摘要人脸检测技术在当今社会生活中扮演着相当重要的角色,如视频监控、人机交互、人脸识别、人脸数据库分类管理等业务,均用到了人脸检测这个关键技术。人脸的检测和识别技术是模式识别、计算机视觉和图像理解系统的研究热点之一,它综合了图像理解、计算机视觉、数据库和人工智能等多方面多学科的知识,应用前景广阔,受到了越来越广泛的重视,并得到了迅速的发展,并取得了很多有价值的研究成果。但是,目前的人脸检测算法一般对光照情况和背景的复杂程度的变化适应性较差,特别是在面对恶劣光照环境时候,人脸的检测率非常低。本文主要研究了基于肤色模型的人脸检测算法的基本理论和关键技术,总结分析了基于肤色模型的人脸检测算法的优势与弱点,重点讨论了由于恶劣光照条件带来的对人脸检测结果的不利影响的问题。本文采用针对不同光照环境的光线补偿算法,克服人脸图像在复杂光照环境下的检测困难,提出了一种从粗到细、从整体到局部的基于不同颜色空间的人脸光照补偿算法。该算法不用估计环境光的方向、强度和光源的性质(如:是否点光源、是否平行光、单光源还是多光源等)。针对不同的光照环境情况(主要为光照不足或过亮、有色光源照射、由于偏光造成的局部阴影),采用不同的光线补偿策略,并配合基于混合高斯模型对肤色色度信息进行描述,并采用根据各像素点距肤色中心马式距离来计算与肤色相似度的算法,对很多图像处理都得到了比较好的结果,使得皮肤与非皮肤区域有效的分开。理论和实验证明,本方法能够有效地解决人脸检测和识别中不同复杂光照环境影响的问题,提高肤色区域与非肤色区域的分割准确性,使在不同光照条件下的人脸检测性能得到了很大的提高。而且本方法还具有运算效率高、鲁棒性好的特点。关键词:人脸检测;肤色模型;光线补偿;色彩空间; IVAbstract Human face detection plays an important roll in applications such as video surveillance, puter interface, face recognition, and face image database management. Face detection and recognition technology is one of the hottest spots in pattern recognition, computer vision and image understanding system; it has integrated multi-disciplinary knowledge such as image understanding, computer vision, databases and artificial intelligence, and with the rapid development it get more and more attention and has made many valuable research results. But now the face detection system can not reach our expectation, especially with the bad environment. A face detection algorithm for color images in the presence of varying lighting conditions as well plex backgroundsis proposed by us. Our pensation method which is based on different color space from-wide-to-slender and from-whole-to-part instead of estimating the surround light conditions is brought out. The method does not have to estimate the direction, strength and the nature of light source (such as: whether it is point light source, parallel light, single-source or multi-source, etc.) The method with different pensation strategy adapts different light circumstances the description of skin