基于视频序列的双目视觉立体匹配算法研究

发布时间：2018-04-20 06:49

本文选题：双目视觉 + 立体匹配　；参考：《天津大学》2016年硕士论文

【摘要】：视觉是人类获取外界信息的主要途径和方式,而研究如何模拟人类视觉并应用到实际中能够解决诸多难题,在医疗、航天、娱乐等领域均有着重要的意义。而计算机的迅猛发展,为视觉的模拟提供了有力的技术支撑,形成了一门热门学科——计算机视觉。计算机视觉将计算机和摄像机有机的结合起来,通过摄像机对外界进行成像并利用计算机对数字图像进行感知、识别和分析等,最终实现对现实场景空间的认知理解。双目立体视觉是计算机视觉中举足轻重并充满挑战的课题之一,而立体匹配是其中的核心所在。立体匹配获取的视差结果在很大程度上能够反映三维空间中目标的距离,这使得立体视觉能够获取平面视觉所不能提供的深度信息。因此立体匹配在立体视觉领域起到了至关重要的作用,对于立体匹配进行深入研究探讨具有极大的意义。另外,虽然研究者对立体匹配进行了长期的研究并提出了众多立体匹配算法,也取得了不错的效果,但主要都是针对静态图像的,很少提出针对视频序列的立体匹配算法。然而,对于动态场景比如视频序列而言,充分有效的利用视频序列中获得的动态信息能够很好的提升立体匹配的视差提取效果。因此本文基于立体匹配的重要性和针对视频序列进行立体匹配的不足,研究并提出了一种基于视频序列的立体匹配算法。和现有方法相比,主要提出了一个新的计算模型,对自适应权值部分进行推导和建模,利用视频序列中的光流信息来优化权值的计算,提高其准确度和精确度,最终提取准确视差值。最后通过大量的实验对比和分析工作,验证了本文算法的有效性和可行性,取得了更理想的效果。
[Abstract]:Vision is the main way and way for human to obtain external information. Therefore, it is of great significance to study how to simulate human vision and apply it to solve many difficult problems in medical, aerospace, entertainment and other fields. The rapid development of computer provides powerful technical support for visual simulation and forms a hot subject-computer vision. The computer vision combines the computer and the camera organically, imaging the outside world by the camera and using the computer to perceive, recognize and analyze the digital image, and finally realizes the cognitive understanding of the real scene space. Binocular stereo vision is one of the most important and challenging topics in computer vision, and stereo matching is the core of it. The parallax results obtained by stereo matching can to a large extent reflect the distance of objects in three-dimensional space, which enables stereo vision to obtain depth information that plane vision cannot provide. Therefore, stereo matching plays an important role in stereo vision field, and it is of great significance for the further study of stereo matching. In addition, although many stereo matching algorithms have been studied and many stereo matching algorithms have been put forward for a long time, they are mainly aimed at static images, and few stereo matching algorithms are proposed for video sequences. However, for dynamic scenes such as video sequences, the full and effective use of the dynamic information obtained from the video sequences can improve the parallax extraction effect of stereo matching. Therefore, based on the importance of stereo matching and the deficiency of stereo matching for video sequences, a stereo matching algorithm based on video sequences is studied and proposed in this paper. Compared with the existing methods, a new computing model is proposed, which deduces and models the adaptive weight, and optimizes the weight calculation by using the optical flow information in the video sequence to improve its accuracy and accuracy. Finally, the accurate visual difference is extracted. Finally, the effectiveness and feasibility of the proposed algorithm are verified by a large number of experiments and analysis, and more ideal results are obtained.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】