【作者】 陈国龙；

【导师】 黄子强；

【作者基本信息】 电子科技大学 ， 光学工程， 2014， 硕士

【摘要】 随着显示技术的快速发展,为了获得更舒适的观看体验,裸眼式自由立体显示技术方案被提出。其中光栅式自由立体显示因其独特的优势被各大公司和学者研究,但传统的光栅式自由立体显示技术只在空间中形成有限的最佳视点,当用户头部移动到最佳视点之外时,双眼会看到串扰的立体图像,影响了立体视觉体验。针对这种问题,在跟踪式裸眼3D显示系统中通过实时定位人眼的空间坐标,由人眼坐标对图像像素进行重新排布改变最佳视点的区域,可有效地减小人眼移动过程中的图像串扰。因而本文以人眼的检测识别为主要研究目的,在2D图像领域研究实现了人眼的检测跟踪,而随着光学、计算机学的发展,基于结构光投影的三维轮廓术逐渐成为研究热点,本文创新地研究了三维图像信息在人眼检测领域的应用,并同样将人眼检测应用到了裸眼3D显示下。主要内容为:1.本文基于人眼的双目视差原理,分析了立体显示器实现立体显示的必要条件,并依据狭缝式光栅立体显示原理制备了原型机;2.针对2D图像处理领域的人眼检测进行分析研究,通过分析Haar特征值和积分图计算原理,以及对Kalman滤波器原理的研究实现了一种AdaBoost级联分类器和Kalman滤波相结合的人眼跟踪识别算法,并通过实验验证了该改进算法的性能;3.对三维轮廓术领域内基于位相原理的四步相移法进行了Matlab软件仿真,并对相位解包裹算法进行了介绍和实现。提出了一种基于二值光栅投影的人眼中心位置检测方法,提出并实现了光栅条纹边界的编解码算法,利用像素的偏移量来获取该位置点的高度轮廓分布,并利用计算机进行模拟仿真和实际实验进行方案可行性验证。同时研究了涉及到的图像二值化和噪声滤除技术;4.在人眼检测跟踪研究的基础上,整合了跟踪式狭缝光栅裸眼3D显示系统的人眼跟踪模块和子像素移动的硬件模块,实现了当人头部移动产生串扰时,相应地通过移动子像素来减小串扰。更多还原

【Abstract】 With the quick advances in the display technology, the naked-eye stereoscopic display solutions have been proposed in order to achieve a more comfortable viewing experience. Among these technologies, the grating auto-stereoscopic display is widely researched by major display companies and academic institutes because of its unique advantages. However, traditional raster stereoscopic display technologies can only form limited best viewpoints in space, which means when the user’s head moves outside of the best viewpoints, inevitable crosstalk would seriously affects the three-dimensional visual experience. In response to this problem, it is believed that an eye-tracking stereoscopic display system can effectively reduce the image crosstalk by tracking the position of the human eyes and moving the RGB sub-pixels to change the location of the best viewpoints in real-time.Therefore, the human-eye detection is identified as the primary research purpose of this project. Furthermore, because of the successful research of a two-dimensional human eye tracking and the development of optics and computer science, this project proposed and researched the three-dimensional image application in the field of human-eye detection. Moreover, the human-eye detection is applied to the naked-eye 3D display as well. The milestones of this project are:First, based on the principle of binocular vision of the human eye, this project analyzes the necessary conditions to achieve a three-dimensional stereoscopic display. Furthermore, a prototype is built based on the stereoscopic display principle; Second, based on the analysis of the Haar eigenvalue and the integral image computing principles, as well as the research on the Kalman filter, an improved recognition algorithm is proposed by combining the AdaBoost cascade classifier and Kalman filtering. And experiments have verified this new improved algorithm. Third, using Matlab do the experimental simulation of four-step phase shifting which is a three-dimensional profilemetry based on the principle of phase, and phase unwrapping algorithms are introduced and implemented. The human eye center position detection method, based on a binary raster projection,is proposed and the coding and encoding algorithm of grating boundaries is proposed and implemented, using pixel offsets to get the points’ height profile of the distribution, and both of the software simulation and actual experiments are conducted to verify the feasibility of the program. Moreover, the image binarization and noise filtering technology are involved during the implementation. Fourth, on the basis of the human eye detection and tracking studies, a human eye tracking module of a tracking slit grating naked eye 3D display system and hardware module of sub-pixels rearrangement are integrated to achieve a lower crosstalk effect when head movements occur by rearranging the corresponding sub-pixels.更多还原