城市虚拟地理场景非定标三维重建算法仿真

Simulation of Non Calibration 3D Reconstruction Algorithm for Urban Virtual Geographic Scene

为建立城市地理场景,用于辅助城市规划高度逼真、运行可靠的虚拟场景,提出城市虚拟地理场景非定标三维重建算法。使用几何节点、灯光节点、组合节点等构成简单的虚拟城市场景结构,以深度遍历方式获取楼房、灯光等位置信息;确定相机坐标系与世界坐标系,采用消失点相机自定标方法建立方程组,获得相机内参数矩阵,结合Moravec基本思想,得出像元兴趣值,选择最大兴趣点作为特征点,利用最小二乘影像算法实现特征点匹配,计算特征点坐标,经过去噪处理,实现非定标三维重建。实验结果表明,重建方法可以减少定标工作量,降低对计算性能的要求,使重建后的虚拟场景更加逼真。

In order to establish a virtual scene with high fidelity for urban planning,a non-calibration 3 D reconstruction algorithm of virtual geography scene of city was presented.Firstly,geometry nodes,light nodes and combinatorial nodes were used to form a simple structure of virtual city scene.Secondly,the location information of buildings and lights was obtained by depth traversal.Thirdly,camera coordinate system and world coordinate system were determined.After that,the equations were established by camera self-calibration method based on the vanishing point,and the internal parameter matrix was obtained.Combined with Moravec idea,pixel interest value was obtained,and the maximum interest was selected as the feature point.Moreover,the least square image algorithm was used to match the feature points and calculate the coordinates of feature points.Finally,the non-calibration 3 D reconstruction was achieved after denoising.Simulation results show that the reconstruction method can reduce the workload of calibration and the requirement of computational performance,and thus make the reconstructed virtual scene more realistic.