基于视频的人脸跟踪检测和识别系统的研究与实现-计算机技术专业论文.docx

摘要摘要 摘要 摘要 视觉是人类感知外部世界的主要方式，由于人的视觉特性所在，人脸又是人 类视觉中最为感兴趣的区域，人们往往能够轻易地对两个人脸进行区分。随着 计算机技术的飞速发展以及视频监控系统在城市中的广泛使用，人们开始思考 如何让计算机像人一样迅速地发现和识别人脸，从而帮助我们处理大量的视频 信息，解决传统身份验证系统的局限性，提高相关领域的工作效率。 本文致力于构建一套基于视频的人脸跟踪检测和识别系统，并统计视频中共 出现的人的个数，系统主要包括肤色区域的寻找和跟踪，人脸的检测与识别等 部分。本文对肤色检测方法、人脸跟踪方法、人脸检测和识别方法都进行了深 入探讨，并对比了各个方法之间的优缺点，提出了结合单类支持向量机和HSV 色彩模型的肤色检测方法，使肤色检测更加精确；改进了传统的CAMShift算法， 从而实现了跟踪多个对象并提高了跟踪精度；提出了理论上效果更好的结合 Gabor特征和单类支持向量机的人脸检测方法。在对系统所采用的肤色检测、对 象跟踪、人脸检测和识别的方法细致介绍后，又详细说明了本文系统的设计与 实现，并对系统所实现的多种方法作了对比。 本文最后对系统的工作成果及创新之处进行了总结，同时，说明了系统的一 些不足之处，并对未来的发展进行了展望。 关键词：视频图像肤色检测目标跟踪人脸检测人脸识别 AbstractAbstract Abstract Abstract Vision is the main method for people to perceive the external world．Because of the characteristics of human beings’vision，face is the greatest interest of people’S vision．Thus，people can easily recognize two other person’S face．With the rapid development of computer technology and the broad application of video surveillance system in cities，people begin to thiIll(how to make computers find and recognize face as human beings，SO that computers Can help US process numerous video information，eliminate the limitations of traditional identity authentication system and increase work efficiency of related areas． This paper aims to construct a set of face track，detection and recognition system based on video and count the number of people appearing in the video．The system mainly incorporates the search and track of complexion region，the detection and recognition of face and SO on．This paper elaborates on the method of complexion detection，face track，face detection and recognition，and compares different methods． This Paper proposes complexion detection method integrating with One—class Support Vector Machine and HSV color model to make complexion detection more accurate； improves traditional CAMShifl algorithm SO that it Can track several objects and improve track precision；and presents theoretically better face detection method integrating with Gabor features and One—class Support Ve