基于轨迹基的三维非刚体射影重建方法

3D Nonrigid Projective Reconstruction Method Based on Trajectory Basis

为了实现三维非刚体的射影重建,在相机为透视投影模型下,提出了一种基于轨迹基的三维非刚体射影重建方法.该方法基于三维非刚体可由若干个特征点的轨迹组成且由轨迹生成的低维子空间的基底可以事先定义的特点,利用图像矩阵的行向量和列向量生成一个低维子空间的性质,对深度因子进行迭代求解,最终实现三维非刚体的射影重建.由于由轨迹生成的低维子空间的基底可以事先定义,因此求解的未知数大大减少,提高了算法的鲁棒性,并且当求解深度因子时利用了图像矩阵的行向量和列向量的约束,提高了重建精度.模拟和真实实验结果表明,所提方法具有收敛速度快、重建误差小等优点.

To obtain 3D nonrigid projective reconstruction,we propose a 3D nonrigid projective reconstruction method with the camera as the perspective model. The nonrigid can be taken as some feature trajectories,and the basis of the low dimension subspace spanned by the trajectories can be predefined. At the same time,the row and the column vectors from the image matrix span a low dimension subspace. On the basis of the above properties,the depth factors can be obtained,and the projective reconstruction is realized. The basis of the subspace spanned by all the trajectories can be defined in advance. Thus,the number of solved unknowns is greatly reduced,and the method becomes more robust. At the same time,while solving the depth factors,the constraints from the row and the column vectors of the image matrix are used,which improves the reconstruction precision. Simulation and actual results show that the proposed method has the advantages of fast convergence and small errors.