基于普通数码相机双目立体视觉理论的三维重建方法研究
发布时间：2018-09-16 20:50
【摘要】：双目立体视觉理论提出了一种由二维影像恢复场景三维信息的解决方案，是当前三维重建领域的研究热点。基于此本文以双目立体视觉理论为基础，以日常生活中的数码相机为工具，建立双目立体视觉三维重建系统，并对系统中的立体匹配、相机标定、三维重构等组成部分进行详细的探讨和研究。主要完成以下工作： (1)影像预处理：采用自适应中值滤波和Wallis锐化分别对立体像对进行预处理，目的是消除噪声干扰、增强影像边缘及细节信息。 (2)立体匹配：在SIFT算法的基础上，，结合单应矩阵与极线双重约束条件进行约束匹配，并利用RANSAC算法消除误匹配，实现了一种由粗到精的立体匹配方法，提高了匹配的速度及精度。 (3)相机标定：采用张正友平面标定法，在标定的过程中结合Harris角点检测法与二次曲面拟合法提取棋盘格模板角点的亚像素级坐标，确定相机的内参数。 (4)三维重建：利用立体匹配结果以及相机标定获得的内外参数，依据SFM原理计算出匹配点的三维坐标，并且提出一种坐标比较法验证三维坐标的精度，进一步确定方案的可行性。 各章节的实验结果表明：每个组成部分都能到达预期的效果，整个系统能够快速获取精度较高的空间点的三维信息，有力的证明本文提出方法的可行性，为数码相机拥有率不断提高的普通用户们提供一种基于数码照片就能恢复出空间景物三维信息的新的解决方法，也为推动数字城市建设等需求三维模型领域提供了另一种参考方案！
[Abstract]:Binocular stereo vision theory proposes a solution to restore 3D information from 2D images, which is a hot research topic in the field of 3D reconstruction. Based on the binocular stereo vision theory and the digital camera in daily life, this paper sets up a binocular stereo vision 3D reconstruction system, and calibrates the stereo matching and camera in the system. Three-dimensional reconstruction and other components are discussed and studied in detail. The main works are as follows: (1) Image preprocessing: adaptive median filter and Wallis sharpening are used to preprocess stereo image pair respectively, in order to eliminate noise interference, Enhance image edge and detail information. (2) Stereo matching: on the basis of SIFT algorithm, combining monoclinic matrix with polar line double constraint condition to perform constraint matching, and using RANSAC algorithm to eliminate mismatch, A stereo matching method from coarse to fine is realized, and the matching speed and precision are improved. (3) camera calibration: using Zhang Zhengyou plane calibration method, In the process of calibration, the sub-pixel coordinates of the chessboard template corner are extracted by using Harris corner detection method and Quadric surface fitting method. (4) 3D reconstruction: based on the results of stereo matching and the internal and external parameters of camera calibration, the 3D coordinates of matching points are calculated according to SFM principle. A coordinate comparison method is proposed to verify the accuracy of the three-dimensional coordinate and to determine the feasibility of the scheme. The experimental results of each chapter show that each component can achieve the desired effect, and the whole system can quickly obtain the three-dimensional information of the higher precision spatial points, which proves the feasibility of the method proposed in this paper. To provide a new solution for ordinary users, whose digital camera ownership rate is increasing, which can restore 3D information of spatial scene based on digital photos. Also for the promotion of digital city construction and other requirements of three-dimensional model areas to provide another reference scheme!
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P234
本文编号：2244749
[Abstract]:Binocular stereo vision theory proposes a solution to restore 3D information from 2D images, which is a hot research topic in the field of 3D reconstruction. Based on the binocular stereo vision theory and the digital camera in daily life, this paper sets up a binocular stereo vision 3D reconstruction system, and calibrates the stereo matching and camera in the system. Three-dimensional reconstruction and other components are discussed and studied in detail. The main works are as follows: (1) Image preprocessing: adaptive median filter and Wallis sharpening are used to preprocess stereo image pair respectively, in order to eliminate noise interference, Enhance image edge and detail information. (2) Stereo matching: on the basis of SIFT algorithm, combining monoclinic matrix with polar line double constraint condition to perform constraint matching, and using RANSAC algorithm to eliminate mismatch, A stereo matching method from coarse to fine is realized, and the matching speed and precision are improved. (3) camera calibration: using Zhang Zhengyou plane calibration method, In the process of calibration, the sub-pixel coordinates of the chessboard template corner are extracted by using Harris corner detection method and Quadric surface fitting method. (4) 3D reconstruction: based on the results of stereo matching and the internal and external parameters of camera calibration, the 3D coordinates of matching points are calculated according to SFM principle. A coordinate comparison method is proposed to verify the accuracy of the three-dimensional coordinate and to determine the feasibility of the scheme. The experimental results of each chapter show that each component can achieve the desired effect, and the whole system can quickly obtain the three-dimensional information of the higher precision spatial points, which proves the feasibility of the method proposed in this paper. To provide a new solution for ordinary users, whose digital camera ownership rate is increasing, which can restore 3D information of spatial scene based on digital photos. Also for the promotion of digital city construction and other requirements of three-dimensional model areas to provide another reference scheme!
【学位授予单位】：辽宁工程技术大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P234
【参考文献】
相关期刊论文 前10条
1 管业鹏,顾伟康;基于双目视觉的基准差梯度立体匹配法[J];传感技术学报;2004年01期
2 李金岭;刘鹂;乔书波;郭丽;;关于三维直角坐标七参数转换模型求解的讨论[J];测绘科学;2010年04期
3 张卡;盛业华;叶春;;基于数字视差模型和改进SIFT特征的数字近景立体影像匹配[J];测绘学报;2010年06期
4 潘国荣;周跃寅;;两种坐标系转换计算方法的比较[J];大地测量与地球动力学;2011年03期
5 闫庆庆;王宝山;訾国杰;;基于VC的七参数坐标转换模型研究及实现[J];地理与地理信息科学;2010年05期
6 隋婧,金伟其;双目立体视觉技术的实现及其进展[J];电子技术应用;2004年10期
7 谢萍;邹峥嵘;肖奇;;基于Harris角点和SIFT特征的近景影像匹配[J];测绘科学;2012年04期
8 梁志敏;高洪明;王志江;吴林;;摄像机标定中亚像素级角点检测算法[J];焊接学报;2006年02期
9 王涛;李恒宇;谢少荣;;一种双目视觉传感器的快速自标定方法[J];计算机工程;2012年12期
10 赵文彬;张艳宁;;角点检测技术综述[J];计算机应用研究;2006年10期
本文编号：2244749
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2244749.html
