三维重建是计算机视觉和计算机图形学领域的一个重要研究方向，在无人驾驶、文物数字化、现代医学、3D动画游戏等多个具体的应用场景中都表现出了极大的价值。目前基于多视角图像的三维重建技术主要有运动恢复结构及同时定位和地图构建两种方法。本文将讨论基于单目相机的三维地图重建，在过程中使用增量式的运动恢复结构算法实现稀疏点云地图的构建，并在此基础上进一步讨论深度图估计的方法，进而完成稠密地图的构建。本文使用SIFT算法对每一帧图像完成特征的提取，利用特征点的匹配关系来估计相机的运动变化，并通过对当前帧与参考帧进行三角测量来计算出像素点对应的空间深度。本文以校园中的建筑为真实数据来源进行地图重建实验，以此分析不同相机参数对三维重建的影响，并总结了算法研究中存在的问题。

3D reconstruction is an important subject in the field of computer vision and computer graphics, which has shown great value in many specific applications, such as unmanned driving, digitization of historical relics, modern medicine, 3D animation games and so on. At present, 3D reconstruction technology based on multi-view images mainly include two methods: Structure from Motion and Simultaneous Localization and Mapping. In this paper, 3D-map reconstruction based on monocular camera will be discussed. During this process, the algorithm of incremental structure from motion is used to realize the construction of sparse point cloud map, from which the method of depth map estimation is further discussed, and the dense map will be constructed. This paper use SIFT algorithm to extract features from every image and estimate camera’s motion by the matching results. And then, the spatial depth corresponding to pixels will be calculated by triangulation of the current frame and reference frame. This paper will carry out experiments by using the real data of campus’ buildings, analyze whether different camera parameters will influence the reconstruction, and summarize the problems existing in the research.