【作者】 王彬；

【导师】 耿卫东； 钱徽；

【作者基本信息】 浙江大学 ， 计算机应用技术， 2011， 硕士

【摘要】 如今,在计算机视觉领域,三维重建技术已经得到了迅速的发展,基于图像的三维重建技术因为其硬件要求简单、可实施性较高而得到了更多的关注。基于图像三维重建技术的基础是对拍摄图像的摄像机完成定标,因此摄像机的定标技术作为三维重建的基础也获得了越来越多的关注。在基于图像的三维重建技术环节中最关键、也是决定三维重建结果的就是图像特征点之间的匹配。为了进行图像之间特征点的稠密匹配,可以先对图像进行矫正变换,使得匹配特征点的搜索从二维降到一维。对图像进行正确矫正变换就需要对摄像机进行正确定标,传统的摄像机定标只是考虑了空间三维点的投影误差最小化,而没有考虑图像匹配特征点之间矫正误差的最小化。本文侧重于对新的定标方法的研究,提出并实现了结合矫正误差的定标优化方法,用该方法求解出来的摄像机参数,在保证空间中部分三维点集投影误差尽量小的前提下,同时也可以保证图像匹配特征点之间的矫正误差尽可能小,从而接下来可以完成图像的矫正。论文中对双目摄像机和多目摄像机结合矫正误差的定标方法进行了较详细的算法描述,并在后面给出了采用新方法求解出的摄像机参数对图像进行矫正,然后进行图像特征点稠密匹配、三维重建后的效果对比图。本文最后将多目摄像机结合矫正误差的定标方法实现,并设计成一个定标系统,该系统作为我们实验室后续三维重建系统的前提系统,为三维重建系统提供摄像机的参数。更多还原

【Abstract】 Nowadays,3D reconstruction technology has been developed so rapidly in the domain of computer vision. Image-based reconstruction technology gains a lot of attention on account of its brief hardware need and high feasibility. The premise of image-based reconstruction is that the cameras which are used to take pictures must be calibrated. So as the base of the 3D reconstruction, camera calibration technology has also gained much focus. The point-matching technology is so important that it will decide whether the result of the 3D reconstruction will be good enough. We can rectify the picture first in order to do dense matching between two pictures, because this can decrease the search of match points from two dimensions to one dimension. Only if the cameras have been calibrated correctly, can the pictures be rectified correctly, only the minimizing of the reprojection error of the 3D points in the 3D space is considered in the traditional camera calibration technology, the minimizing of the rectification error between matching points on the pictures is not considered.This paper paid attention to the new method of camera calibration, presented and realized the method of camera calibration combining rectification error. The camera parameters solved through this method, can not only make sure that the reprojection error of a portion of 3D points in the 3D space is minimized, but can also guarantee that the rectification error of the matching points on the pictures is minimized too. The algorithm about calibration method of binocular and multi-cameras combining rectification error has been described in detail in the paper, and also the result of dense-matching and 3D reconstruction between two pictures using the cameras’ parameters solved by the new method is showed next.At the end of the paper, the new method of camera calibration combining rectification error is realized and designed into a calibration system, which provides camera parameters to the 3D-reconstruction system in our lab.更多还原