【作者】 俞海珍；

【导师】 冯浩； 曹文明；

【作者基本信息】 浙江工业大学 ， 控制理论与控制工程， 2004， 硕士

【摘要】 人脸识别研究包括人脸检测、人脸跟踪、人脸识别和表情分析四个主要领域，其中人脸检测与跟踪是人脸信息处理领域中的一项关键技术，在自动人脸识别系统、基于内容的检索、视觉监测、新一代人机交互技术等领域具有广阔的应用前景。 本文的主要研究工作和成果包括以下几个主要方面的内容： 首先研究了图像分割技术，总结分析了已知背景下图像分割技术，提出了自适应阈值来分割运动区域，利用聚类分析定位头发区域，再根据人脸的几何特征来定位人脸，在下一步的人脸跟踪中用作模板图像。其后研究了人脸跟踪的四种常见思路，结合遗传算法在优化过程中得到了广泛的应用并取得良好的效果的特点，所以本文在定位出人脸的基础上引入遗传算法进行跟踪。已有遗传算法的运算量均较大，本文提出一种新的选择初始种群方法来减少搜索代数，从而提高实时性。最后，用宁波名钟公司研制的摄像头进行实时采集、定位及跟踪，该摄像头通过USB总线直接连在计算机上。 本文实验所用的图像均是由M6215摄像头实时采集的可见光图像，文中的算法由VC++语言在Windows2000平台下编程实现。更多还原

【Abstract】 Research efforts in face processing include face detection, face tracking, face recognition and expression analysis. The face detection and tracking is key technology of the field efface information processing. It’s widely applied in the fields of automated face recognition system, based on content index, video monitoring and human-machine alternative technology.In this paper several research works are given as follows: First the methods of image segmentation are summarized and analyzed. An algorithm of image segmentation is developed based on self-adaptive threshold value. With the difference algorithm of image sequence, I can eliminate the background and extrude the moving object in images. Then hair clustering analysis, face geometry character are applied to verify very candidate region to get real face regions. The detected face is used as matching template in face tracking. Second thefour general ideas of face tracking are pondered upon and analyzed and generate algorithm has a good application in optimization process. An algorithm of face tracking with generate algorithms is developed. In order to reduce the computation, a new method of selection initial population has been presented. Finally face real-time capture, face detection and tracking have been realized by video camera made in Ningbo mingjong company. The video camera is connected to computer by USB.In this paper, PCCamera M6215 grabbed the tracked images and all algorithms were accomplished by VC++ in Windows2000 environment.更多还原