基于未标定图像的三维重建算法

3D Reconstruction Algorithm Based on Uncalibrated Images

提出一种基于多幅未标定图像的三维重建算法。在标记点匹配的基础上进行射影重建,通过施加度量约束将射影重建升级为欧氏重建,即利用未标定的透视图像恢复相机的内、外部参数以及标记点的三维空间坐标,实现场景的三维重建。标记点易于进行点对精确匹配,较手动拾取匹配提高了效率。实验结果表明,利用该算法能够大幅减小再投影误差。

This paper presents a 3D reconstruction algorithm based on uncalibrated images.The projective reconstruction is performed based on the reference points matching between the different uncalibrated images.The projective reconstruction is updated as the Euclidean reconstruction by using metric restriction.The camera intrinsic parameters,Euclidean motion and shape are recovered by uncalibrated images.Compared with previous methods,the operation efficiency is improved since the manual matching between different uncalibrated images is not involved by automatic reference point matching.Experimental results show that the reprojection error is remarkably reduced by the presented algorithm.