毕业论文：实时瞳孔检测与跟踪.doc

PAGE PAGE I 摘要 实时瞳孔检测与跟踪是计算机视觉领域一个重要的课题，在人脸检测、人 脸识别、司机驾驶行为监控、人机接口等方面具有重要应用。瞳孔检测与跟踪 可分为三个阶段:人脸检测、瞳孔检测、瞳孔跟踪。近年来，科学家们在上述 各个方面进行了很多研究，但这些问题仍未得到很好解决。 本文试图对一些现有的人脸检测、瞳孔检测与跟踪方法进行改进，以得到 一个性能良好的瞳孔检测与跟踪系统。 首先，研究了不同彩色空间下不同的肤色模型。针对己有肤色模型在人脸检 测中存在的不能较准确检测出皮肤、对光照条件敏感等问题，在大量实验数据基 础上，建立了一种基于YCbCr色彩空间的肤色模型。实验结果表明:本文方法在 不同肤色以及较大的光照变化条件下均具有较好的检测率。 其次，针对基于人面部各个特征部位(眉毛、瞳孔、鼻子、嘴巴等)之间的 几何结构的眼睛检测方法只考虑了特征部位相互间的相对位置信息的不足，将人 脸特征部位的对称性引入该方法的评价函数中。实验证明改进后的评价体系瞳孔 检测更准确。 最后，将卡尔曼滤波与Mean-shift算法相结合用于瞳孔跟踪的方法，不仅提 高瞳孔跟踪的准确性，还提高了跟踪速度。 关键字:肤色模型，人脸几何模型，瞳孔检测与跟踪 卡尔曼滤波，Mean-shift算法 Abstract Real-time pupil detection and tracking technology is an important research topic in computer vision area. It has many applications in face detection, face recognition, drivers monitoring, human-computer interface etc. There are mainly three steps in pupil detection and tracking: face detection, pupil detection and pupil tracking. In recent years, many works have been done in this area but it is still a challenging problem. In this paper, some methods to face detection, pupil detection and tracking are modified to better performance. Firstly, different skin models in different color space are studied and a novel skin model in YCbCr color space is proposed, which is insensitively to illumination changes. Experimental results indicate that this skin model performance well under different skin color and various light conditions. Secondly, face symmetry is considered in the evaluation function in the geometrical face model based eyes detection method. Experimental results show that the improved estimation system works better in pupil detection. Finally, Kalman filtering and Mean-shift algorithm is combined in pupil tracking to improve both tracking accuracy and tracking speed. Key words: Skin model, Geometrical face model, Pupil detection and tracking Kalman filtering, Mean-shift algorithm 目录 第一章 绪论, 1. 1研究背景及意义 1.2典型应用二 1. 3研究现状 1.4相关理论介绍 1.5图像采集系统组成， 1. 6主要研究内容和章节安排 第二章 基于新肤色模型的皮肤