节点文献
单摄像机视线跟踪技术研究
Study on Eye Gaze Tracking Using One Camera
【作者】 刘瑞安;
【导师】 靳世久;
【作者基本信息】 天津大学 , 精密仪器及机械, 2007, 博士
【摘要】 视线跟踪技术是未来智能人机接口的关键技术之一,在工业控制、机器人学和临床医学上有着广泛的应用前景。基于角膜反射原理和图像处理的视线跟踪技术,具有非侵入的优点,近年来取得了较快进展。但由于视线跟踪技术还没有完全成熟及眼动本身的特点,现有的视线跟踪技术要解决的主要问题是:提高视线跟踪系统的精确度;提高系统的可靠性;简化校准程序;减少对用户的限制。针对上述问题,论文对单摄像机视线跟踪技术进行了深入研究,减少了对用户头部的限制;提高了系统的精确度、可靠性及便利性。论文的主要创新性研究工作如下:1.提出了一种单摄像机视线跟踪方法,利用三个光斑和瞳孔中心的位置关系计算视线方向,校准参数随用户头部位置不同而自动选择;同时改进了特征检测算法。提出的方法减少了系统在使用过程中对用户头部的限制,提高了系统的精确度、可靠性及便利性。2.对视线跟踪技术的光学采集系统进行了研究,设计了一种实时跟踪眼睛与调光、变焦、聚焦相结合的CCD摄像机自动调节系统。在不同外部环境光强及用户不同使用位置情况下,摄像机可由大视场下的头肩图像通过自动调节,实时跟踪眼睛并捕获到小视场下的眼睛清晰图像。避免了人工调节的繁琐并提高了调节的准确性。3.对眼睛图像中瞳孔定位算法进行了研究,提出了一种眼睛红外图像中瞳孔椭圆边缘及中心的亚像素定位算法,能够精确地确定瞳孔中心位置,提高了系统的精确度。4.提出了一种眨眼检测算法和在人脸图像序列中判别眼睛睁、闭状态的方法,提出的算法检测正确率高,可应用于视疲劳监视及眼控鼠标。
【Abstract】 Eye gaze tracking is one of the key techniques of the intelligent interface in the future between human and computerized devices and has widespread prospects of application in industry control, robotics and clinical medicine. The technique of eye-gaze tracking, based on corneal reflection and image processing, has the advantage of unobtrusiveness and has made considerable headway in recent years. However, owing to that it is also a developing technique and the characters of the eye movement, there are still many unresolved problems preventing the use of non-intrusive eye gaze systems in actual applications. The most prevalent of these are: accuracy, restricted head movement, robustness, and ease of calibration. To solve these problems, the technique of single camera remote eye gaze tracking has been studied; the constraints on the user’s head imposed by most existing eye trackers have been significantly relaxed. The accuracy and robustness of the eye gaze tracking system have been improved.The main innovative researches in this paper are as follows:1. A single camera remote eye-gaze tracking technique is proposed to calculate the gaze direction by the relationship between the positions of the pupil center and three corneal reflections, the calibration parameters can be automatically adjusted to allow free head movements. Also, robust feature detection algorithms are developed. The system can work under natural head movements and show high accuracy and very good robustness.2. The image capturing system of the eye-gaze tracking technique is studied, a real-time eye tracking method and an automatic adjusting method for object lens of CCD camera in combining the aperture regulation with zooming and focusing are proposed. This system can track the eye from a head-should image and acquire clear images of the magnified eye in real time by the adaptive regulation of the CCD camera in different external lightings and user position conditions. The drawback of intensive labor is eliminated and the accuracy is improved.3. The algorithms to detect and locate the pupil in eye images are studied. Also a sub-pixel ellipse contour detection algorithm is proposed in order to locate the center of the pupil accurately in IR image of the eye. The proposed algorithm achieved sub-pixel accuracy, which will significantly enhance the performance of the eye-gaze tracking system.4. A method is presented to detect the user’s eye blinks and to classify the eye as either open or closed at each frame of a real-time video stream of the face. A high success rate of detection is obtained using the proposed method, which can be used for fatigue monitoring and eye-controlled mouse.
【Key words】 Eye gaze tracking; Blink detection; Eye-controlled mouse; Automatic focusing; Sub pixel; Ellipse fitting;