随着计算机技术的发展，三维重构的研究逐渐成为一个热点，而基于序列图像的三 维坐标计算方法是三维重构技术非常重要的研究内容之一。在对图像进行建模时，只需 要拍摄物体不同角度的照片，而不需要与物体直接接触，这样就不会对物体造成任何损 坏，所以，它可应用于文物保护、航空航天等相关领域的三维建模方面的研究。因此， 本课题的基于序列图像的标记点三维坐标计算方法研究具有一定的现实意义。 首先，本文讨论了三维坐标计算方法的理论基础，包括摄像机成像模型涉及的坐标 系、针孔成像模型和立体视觉原理等内容。针对本课题的基于序列图像的标记点三维坐 标计算方法研究的特点，对现阶段的摄像机定标方法进行了分析与比较，并在分析与比 较后，选用DLT 法求解摄像机内、外参数及标记点三维坐标的初值。其次，为了提高 摄像机内、外参数及标记点三维坐标值的精度， 本文采用光束法平差与 Levenberg-Marquardt 方法相结合进行优化计算，光束法平差具有理论严密、精度高等优 点，Levenberg-Marquardt 方法具有较好的迭代速度，且对参数初值不敏感、能够收敛等 优点。最后，利用摄影测量学中的误差理论与测量平差精度评定方法对计算结果进行评 估，经过评估与分析，表明本课题的标记点三维坐标计算方法是可行有效的。 本文利用Visual C++编程工具编程既实现了DLT 法求解摄像机内、外参数及标记 点三维坐标初值，又实现了光束法平差与Levenberg-Marquardt 方法相结合对标记点三 维坐标优化计算。

With the development of computer technology,the research of 3D reconstruction has been becoming a Hot Spot, nevertheless, the 3D coordinates calculated method based on image sequences is one of the uppermost research contents in the 3D reconstruction field. Modeling technique based on images,it only wants to take photos of objects from different positions,but not to be in touch with objects directly,so it will not take a toll on objects anymore,and it can be applied in the research of 3D modeling that are liked Historic Preservation 、Aeronautics and Astronautics etc. In a word,the subject which is research on calculating method of 3D coordinates of marked points based on image sequence has certain practical significance. In this paper,first,we discussed the theoretical basis of calculating method of 3D coordinates,including the content that the coordinates involve in camera imaging model、 pinhole imaging model and the principle of stereoscopic vision. In accordance with this subject that is research on calculating method of 3D coordinates of marked points based on image sequence characteristic,we has analyzed and compared the method of camera calibrate at this stage,after the analyzing and comparing, we chose the DLT method to solve the initial values that are intrinsic parameters and external parameters and the 3D coordinates of marked points,then to improve the precision of the values of them,in this paper,we optimized them by combining bundle adjustment with Levenberg-Marquardt method,bundle adjustment has some disadvantages like rigarous theories and high precision etc, Levenberg-Marquardt method has some disadvantages like good iterative speed and is not sensitive to the initial values of parameters and can convergent etc. Finally,we assess the result by using the theory of error and the method of survey adjustment and accuracy assess in photogrammetr,after asssessing and analyzing,it notes that the calculational methods of the 3D coordinates of marked points are feasible and effectual. In this paper,we programmed the initial values of the initial value of the intrinsic and external parameteres and the 3D coordinates of marked points solved by the DLT method and optimizing the values of the 3D coordinates of marked points combined bundle adjustment with Levenberg-Marquardt method by using visual c plus plus programmable tools.