面向移动设备人机交互的眼动跟踪系统

发布时间：2018-03-20 01:11

本文选题：眼动跟踪　切入点：移动设备　出处：《浙江工业大学》2016年硕士论文　论文类型：学位论文

【摘要】：眼动跟踪技术可分析、记录用户的注视点坐标,在心理学、广告评测和人机交互等领域已得到广泛的应用。随着机器视觉、图像处理技术的快速发展和硬件水平的飞速提升,眼动跟踪技术也逐渐在移动设备上得到应用。然而由于受到移动设备硬件水平的限制,例如自带摄像头像素较低,处理器的性能远落后于个人电脑等问题,面向移动设备的眼动跟踪技术依然存在计算效率低下、跟踪误差较大等问题。为了实现面向移动设备人机交互的眼动跟踪系统,本文提出改进的眼动跟踪技术,并针对相关的人机交互应用进行了研究,主要内容如下:(1)面向移动设备眼动跟踪的图像处理技术研究。由于移动设备硬件资源匮乏,其眼动跟踪的计算效率较低,对眼动跟踪效果影响较大。为此,通过逐层缩小图像处理范围的方法,提高图像处理计算效率。首先使用改进的局部二值特征级联分类器对移动设备前置摄像头捕获的图像进行人脸检测,判断并提取人脸区域;然后通过计算人脸区域的相对位置得到人眼预选取区域,并针对该区域使用Haar特征级联分类器进行人眼检测,判断并提取人眼区域。(2)面向移动设备眼动跟踪的注视点计算方法研究。由于人眼区域检测存在偏差等情况,提取人眼区域后将该区域的各边进行适当的扩展得到精确人眼预选区域(包含完整的人眼轮廓,但相较人眼预选取区域更为精确排除头发、眉毛等干扰物),随后使用二值化和轮廓提取的方法得到更为精准的人眼图像外接矩形和虹膜与上眼睑连通区域轮廓,并计算得到人眼图像外接矩形中心(eye center,EC),以及虹膜与上眼睑连通区域轮廓的重心(centre of gravity,CG)。进而建立人眼图像外接矩形中心指向虹膜与上眼睑连通区域轮廓重心的向量(EC-CG向量)。最后通过标定方法,拟合EC-CG向量与注视点的映射关系,实现注视点的准确计算。(3)基于移动设备眼动跟踪的人机交互应用研究。设计与开发了基于智能手机的眼动跟踪原型系统,该系统可以准确、高效的获取用户眼动数据。在此基础上,对眼动数据进行可视化处理与分析,绘制眼动数据热区图、感兴趣区域(Area of Interest,AOI)重点标记图、感兴趣区域之间的视线转换关系图。最后以智能手机上的图片广告为例,应用上述眼动数据可视化方法,实现多用户之间眼动数据可视化结果的分享。本文针对眼动跟踪精度和应用原型系统进行了用户测试。眼动跟踪精度测试结果显示,本文的眼动跟踪精度为2.80°~3.56°,眼动跟踪平均处理速度为6.01帧/s,可见本文系统具有较好的实时性和较高的精度,对眼动跟踪在移动设备的应用和推广有积极的推动作用。应用原形系统用户测试表明,通过眼动数据可视化结果的分享,有效提高了新手用户对图片广告上信息的查看效率和理解水平,进一步验证了本文所提眼动跟踪技术以及人机交互应用的实用性和有效性。
[Abstract]:Eye tracking has been widely used in psychology, advertising evaluation, human-computer interaction and so on. With the rapid development of machine vision, image processing technology and the rapid improvement of hardware level, eye tracking technology can be used to analyze, record the coordinate of the user's fixation points, and have been widely used in the fields of psychology, advertising evaluation and human-computer interaction. Eye tracking technology is also gradually being used in mobile devices. However, due to the limitations of the hardware level of mobile devices, such as the low pixel of the camera itself, the performance of the processor lags far behind that of the personal computer. Eye tracking technology for mobile devices still has some problems, such as low computing efficiency and large tracking error. In order to realize eye tracking system for man-machine interaction of mobile devices, an improved eye tracking technology is proposed in this paper. The main contents are as follows: 1) Image processing technology for eye tracking in mobile devices. Due to the lack of hardware resources of mobile devices, the computing efficiency of eye tracking is low. It has a great effect on eye movement tracking. Therefore, by reducing the range of image processing layer by layer, Firstly, the improved local binary feature cascade classifier is used to detect the face of the image captured by the front camera of the mobile device, to judge and extract the face region. Then the human eye pre-selected region is obtained by calculating the relative position of the human face region, and the human eye detection is carried out by using the Haar feature cascade classifier for the region. To judge and extract the human eye region. 2) Research on the calculation method of fixed point for eye movement tracking of mobile devices. Because of the deviation of human eye region detection, After extracting the human eye region and extending each side of the region properly, the precise human eye pre-selected area is obtained (including the complete human eye contour, but the hair is excluded more accurately than the human eye pre-selected area. The obliterated objects such as eyebrows were then used to obtain the contour of the outer rectangle and iris connected to the upper eyelid by using the method of binarization and contour extraction. The center of gravity of the contour of the connected area between the iris and the upper eyelid and the center of gravity of the contour of the connected area between the iris and the upper eyelid are calculated. Furthermore, the vector of the center of gravity of the external rectangular center of the eye image pointing to the center of gravity of the contour of the iris and the upper eyelid is established. Finally, through the calibration method, Fitting the mapping relationship between EC-CG vector and fixation point, realizing the accurate calculation of fixation point.) the research of human-computer interaction application based on mobile device eye movement tracking is carried out. A prototype system of eye movement tracking based on smart phone is designed and developed, and the system can be accurate. On the basis of this, visualize and analyze the eye movement data, draw the hot zone map of eye movement data, and draw the key mark map of area of interest. Finally, take the picture advertisement on the smartphone as an example, apply the above method to visualize the eye movement data. In this paper, the user test of eye movement tracking accuracy and application prototype system is carried out, and the result of eye movement tracking precision test shows that, The accuracy of eye tracking is 2.80 掳/ 3.56 掳and the average processing speed of eye tracking is 6.01 frame / s. It has a positive effect on the application and popularization of eye movement tracking in mobile devices. The user test of the prototype system shows that through the sharing of visual results of eye movement data, It can effectively improve the efficiency and understanding level of the new users to view the information on the image advertisement, and further verify the practicability and effectiveness of the eye movement tracking technology and the human-computer interaction application mentioned in this paper.
【学位授予单位】：浙江工业大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】