基于ToF相机的三维重建技术

3D RECONSTRUCTION BASED ON TOF CAMERA

针对利用ToF相机实现物体的重建提出一种新的三维重建算法,通过分析物体重建过程的特点,对KinectFusion的重建算法进行改进。点云匹配过程包括粗匹配和精匹配两个过程,最后进行全局优化,提高相机位姿估计的精度,从而得到更精确的三维重建结果。利用泊松重建算法重建物体表面,相比于TSDF算法,能够实现完整表面的重建。提出结合强度图对深度图进行多边滤波的算法以及一种新的补空洞法则,增强深度图像。多边滤波算法在PSNR和SSIM的评估中都优于双边滤波结果,提出的三维重建算法与KinectFusion三维重建结果对比,表面更完整,重建结果更优。

This paper mainly proposes a new 3D reconstruction algorithm for object reconstruction using ToF camera.By analyzing the characteristics of the object reconstruction process,we improve the reconstruction algorithm of KinectFusion.The process of point cloud matching includes two processes:rough matching and fine matching,and finally we perform global optimization for better result.We could obtain more accurate 3D reconstruction results by improving the estimation accuracy of the camera poses.Compared with TSDF algorithm,the whole surface was reconstructed by using poisson reconstruction algorithm,which could realize the reconstruction of the complete surface.This paper proposes an algorithm for multilateral filtering of depth maps by combining intensity maps and a new filling holes rule to enhance the depth image.The multilateral filtering algorithm is superior to the bilateral filtering results in the evaluation of PSNR and SSIM.Compared with the KinectFusion 3D reconstruction results,our algorithm has more complete surface and better reconstruction results.