双目立体成像和显示的FPGA视频处理技术研究

发布时间：2018-05-14 04:25

本文选题：立体视觉 + 双目　；参考：《北京理工大学》2016年硕士论文

【摘要】：伴随着增强现实和虚拟现实等技术的快速升温,立体视觉技术正成为当前的研究热点,在军事、公共安全、工业生产以及游戏娱乐等领域的应用正在快速拓展。通常,计算机立体视觉系统是通过双目立体摄像机获取视频图像的。但是,不同的计算机视觉系统,由于用途的差异,设计的图像采集系统都有差异。就立体显示效果而言,图像采集系统参数中影响最大的基线距离。这是因为视觉中枢会将光学系统的基线距离,默认为人的双眼间距。因此,对于基线距离不等于人双眼间距的图像采集系统,需要对双目显示器的输出图像进行调整,以得到比较理想的立体显示效果。论文首先分析了人眼立体视觉的原理,得到了影响人的立体景深感的最重要因素——双目图像的视差。随后研究了目标物体的距离与其双目视差的关系。之后,论文对双目相机的成像原理进行了阐述。根据上述关系,论文推导出了双目立体显示配准公式。利用该公式,在已知了双目立体相机和双目显示器的参数的基础之上,根据目标物体的距离,可以实现对双目显示图像进行调整,从而是观察者获得比较理想的立体观察效果。根据双目显示的调整公式,实现对双目显示图像的调整,目标物体的距离是一个必不可少的参数。为此,论文进一步研究了基于双目显示的测距系统。实现双目测距的首要前提是获得高精度的双目摄相机的参数,所以论文介绍了的张正友相机标定方法。实现双目测距的另一个要求是,获得同目标在双目摄像机的左右相面上的位置。所以论文接下来讨论了SURF图像特征提取与匹配算法。为了压缩特征匹配时的计算量,提高匹配的精度,论文还引入了极线约束。至此,双目显示与测距系统的理论框架搭建完成。随后,论文描述一套在PC平台完成的验证系统,并分析了该系统的实验结果,以检验该系统方案的可行性。为提高系统的性能和便携性,在论文最后还讨论了在基于FPGA的图像处理系统上完成本套设计方案,并提供了摄像机驱动模块,双目显示器驱动模块以及SUFR算法实现方案等设计思路。
[Abstract]:With the rapid development of augmented reality (AR) and virtual reality (VR) technology, stereo vision technology is becoming a hot research topic, and its applications in military, public safety, industrial production and games and entertainment are expanding rapidly. Usually, a computer stereo vision system acquires video images by binocular stereo cameras. However, different computer vision systems, because of the use of differences, the design of image acquisition systems are different. As far as stereoscopic display effect is concerned, the maximum baseline distance is affected by the parameters of the image acquisition system. This is because the visual center sets the optical system's baseline distance, by default, as human binocular spacing. Therefore, for the image acquisition system whose baseline distance is not equal to the distance between human eyes, it is necessary to adjust the output image of the binocular display to obtain an ideal stereoscopic display effect. Firstly, the principle of human stereoscopic vision is analyzed, and the parallax of binocular image is obtained. Then the relation between the distance of the target object and its binocular parallax is studied. Then, the imaging principle of binocular camera is described. According to the above relation, the registration formula of binocular stereoscopic display is deduced. Based on the known parameters of binocular stereoscopic camera and binocular display, the binocular display image can be adjusted according to the distance of the target object, which is an ideal stereo observation effect for the observer. According to the adjustment formula of binocular display, the distance of the target object is an essential parameter to adjust the binocular display image. Therefore, this paper further studies the ranging system based on binocular display. The first prerequisite to achieve binocular ranging is to obtain the parameters of high precision binocular camera, so the calibration method of Zhang Zhengyou camera is introduced in this paper. Another requirement to achieve binocular ranging is to obtain the position of the same target on the left and right side of the binocular camera. Therefore, this paper discusses the SURF image feature extraction and matching algorithm. In order to compress the computation and improve the accuracy of feature matching, pole constraints are also introduced in this paper. So far, the theoretical framework of binocular display and ranging system is completed. Then, the paper describes a set of verification system completed on PC platform, and analyzes the experimental results of the system to verify the feasibility of the system scheme. In order to improve the performance and portability of the system, in the end of the paper, the design of the system based on FPGA image processing system is discussed, and the camera driving module is provided. The design idea of binocular display driver module and SUFR algorithm.
【学位授予单位】：北京理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】