三维模型的空间匹配与拼接

为获得敦煌莫高窟第172窟的全景三维模型,实现文化遗产数字化保护,提出一种三维模型的空间匹配与拼接方法。首先通过基于四目立体视觉系统的复杂场景三维重建方法,获取单一场景的三维模型;然后以场景真实三维点云数据的对应图像坐标为约束条件,进行高精度特征点匹配;最后以两对同名点在相邻两场景中的空间距离应该一致为条件,保留匹配精度高的点,用于计算旋转矩阵和平移矩阵,进行三维模型间的空间拼接。实验结果表明:与传统拼接方法相比,本文的模型拼接处无明显错位现象,视觉效果更好,传统方法与本方法的匹配误差分别为0.008 2和0.001 6,拼接精度得到显著改善。

Spatial matching and stitching for three-dimensional models

In order to obtain the panoramic 3D model for the 172nd cave of Mogao Grottoes in Dunhuang and realize the cultural heritage digital protection, we propose a spatial matching and stitching method for three-dimensional models. First, we acquire a 3D model of a single scene by a 3D reconstruction method based on the four stereo vision system. Then, we take the corresponding image coordinates of 3D point cloud data in the scene as the constraint condition to obtain high precision feature points matching. Finally, high matching precision points according to the condition that the spatial distance between the two pairs of corresponding points should be consistent in two adjacent scenes are remained. We employ them to calculate the rotation matrix and translation matrix, and achieve space stitching between 3D models. Experimental results show that, compared with the traditional method, the stitching location of the model by the proposal has no obvious dislocation, and the visual effect is better. The matching errors of the traditional method and the proposed method are 0.0082 and 0.0016, respectively. The stitching precision has been improved significantly.