多视角深度图配准算法下破损古建筑三维重建

3D Reconstruction of Damaged Ancient Buildings Based on Multi View Depth Map Registration Algorithm

为解决因古建筑结构复杂且序列杂乱导致的三维重建耗时过长、几何细节效果较差等问题,利用多视角深度图配准算法设计了新的破损古建筑三维重建方法。在标定摄像机坐标的基础上,确定摄像机与真实三维场景间的位置关系,然后采用棋盘角点信息标定法标定转轴的具体坐标,从而确定局部空间坐标。随后分别在相机坐标和世界坐标系统内提取建筑破损部分的所有点,融合两个点集即可实现三维重建。在仿真环境中,以图像工具库Open GL为处理平台,对破损大佛进行三维重建。实验结果表明:所提方法最终的呈现效果较好且抗干扰能力更强,能够满足实际操作要求。

In order to solve the problems of 3 D reconstruction time-consuming and poor geometric details caused by complex structure and disordered sequence of ancient buildings, this paper designed a new three-dimensional reconstruction method of damaged ancient buildings based on the image registration algorithm of view depth. Firstly, the position relationship between the camera and the real three-dimensional scene was determined via the camera coordinate calibration. Secondly, the chessboard corner information calibration method was used to calibrate the specific coordinates of the rotation axis, determining the coordinates of the local space. Then, all the points of the damaged part of the building in the camera coordinate system and the world coordinate system were extracted respectively. Finally, through the fusion of the two-point sets, the three-dimensional reconstruction was realized. In the simulation environment, Open GL was utilized as a processing platform to reconstruct the damaged Buddha. The results show that the method has excellent reconstruction effect and strong anti-interference ability, meeting the requirements of practical operation.