基于TOF-3D相机单视角物体表面重建研究

Research on object surface reconstruction under the single view based on TOF-3D camera

物体的三维重建是计算机视觉、虚拟现实、人工智能等领域研究的重点和难点,为了能够更好的重建物体三维表面,文中提出了一种改进的Delaunay三角剖分算法。此算法与原有的逐点插入Delaunay三角剖分算法的不同在于增加了对三角形边长的约束条件,去除了重建过程中不符合要求的三角形,使其重建效果更加真实可靠。并且基于TOF3D相机拍摄得到的目标物体单视角下的点云数据,结合KNN邻域滤波算法验证了改进的Delaunay算法的可靠性和稳定性。本论文中对单视角下物体表面重建的研究为多视角下物体完整表面3维重建的研究与实现奠定了基础。

In order to perfectly reconstruct the object surface,in this paper we firstly obtain the point cloud data of the object surface under single view by the TOF-3 D camera and use median filter method and KNN neighborhood filter method to denoise,and then by the method of modified delaunay triangulation,we realize the surface reconstruction. Finally,we analysis the difference between the surface reconstruction result of the modified delaunay triangulation and not modified,and find that the modified triangulation method have better performance on the reconstruction at the edge and defect. The works in this thesis is helpful for the object complete surface 3 D reconstruction research under multiple view.