近年来,EPnP算法作为一种相机位姿估计的解析算法,因其较低的计算复杂度而得到广泛的关注,但该算法对图像噪声的稳健性不强。提出了一种基于EPnP算法的迭代算法,即IEPnP算法。IEPnP算法保留了EPnP算法的主要思想,构造了4个虚拟控制点,利用弱透视投影模型获得相机的初始位姿,计算出虚拟控制点在相机坐标系下的坐标,然后通过高斯-牛顿法对虚拟控制点在相机坐标系下的坐标进行优化求解,最终通过解决绝对定向问题来获得对相机位姿的估计。IEPnP算法简化了EPnP算法的计算过程。在不同的图像噪声水平下进行仿真实验,结果表明,相比于EPnP算法,IEPnP算法不仅保持了较高的计算效率,而且对图像噪声具有更强的稳健性。

As an analytical camera pose estimation algorithm, EPnP algorithm has attracted much attention in recent years for its low computational complexity. However, it is not robust to image noise. Hence, an iterative version of EPnP algorithm, called IEPnP, is proposed. The primary ideas of EPnP are preserved in IEPnP, 4 virtual control points are introduced, and their coordinates in the camera coordinate system are obtained through an initialization process based on weak perspective projection model. Gaussian-Newton algorithm is applied to optimize the coordinates of the virtual control points in the camera coordinate system. Finally, the pose estimation result is acquired by solving an absolute orientation problem. Meanwhile, the computational process is simplified in IEPnP. Simulations under different image noise levels are implemented, and the results show that IEPnP is more robust than EPnP to image noise while maintaining a high computational efficiency.