Bundle Adjustment is an important component in structure from motion,in order to overcome the drawbacks existing in the robustness,accuracy and space time efficiency of the existing algorithms,a fast and robust bundle adjustment (FRBA) algorithm has been proposed.Firstly,in order to avoid the effects of outliers,we use the Cauchy loss function to potentially down-weight outliers for improving accuracy of the proposed method.Secondly,a large-scale bundle adjustment problem can be divided into some small ones by making use of the sparsity between 3D points and the cameras for reducing the requirements of memory.Finally,according to the inherent property of the matrix after its sparse decomposition,we use a fast matrix factorization algorithm to solve the normal equation.The proposed FRBA algorithm was compared with some state-of-the-art methods on the synthetic dataset,BAL dataset and real image datasets respectively.Experimental results show that the proposed FRBA algorithm outperforms the state-of-the-arts on both time efficiency,space efficiency as well as precision.%集束调整是运动推断结构的核心,针对现有算法在大规模场景下易受外点影响,空间占用率过高和效率较低问题,提出一种快速鲁棒的集束调整(fast and robust bundle adjustment,FRBA)算法.首先,为了避免外点(outliers)的影响,采用Cauchy损失降低外点的权重,提高算法精度.其次,充分利用运动推断结构中三维点与摄像机之间的稀疏性对大规模集束调整进行稀疏分解,降低内存空间的使用.最后,根据稀疏分解后矩阵的固有特性,采用快速矩阵分解法求解正态方程的解.在合成数据集、BAL数据集和真实图像数据集上对FRBA算法进行测试,并与现有经典算法进行比较.实验结果表明无论在时间效率还是精度上,FRBA算法均处于领先位置.
展开▼
