论文部分内容阅读
摄像机的位姿估计在计算机视觉和机器人学等领域得到了广泛的研究与应用。通过已知的3D几何特征和其对应的2D几何特征获得摄像机的旋转和平移参数是实现摄像机位姿估计的一种重要方法。在此基础上,文中针对单目相机位姿估计提出了一种新的基于3D~2D对应几何特征实现摄像机位姿估计的算法。该算法利用单位四元数表示旋转矩阵,通过建立位姿参数约束方程解出旋转和平移参数。相对于Ansar的NLL算法,该算法能更好地克服图像噪声对位姿计算精度的影响,且能保证其计算出的旋转矩阵具有正交性。模拟实验将该算法同NLL算法进行了比较,结果表明:该算法在鲁棒性、计算精度及复杂度等方面具有明显的优势,实验结果验证了该算法的有效性。
The pose estimation of camera has been widely studied and applied in the fields of computer vision and robotics. Obtaining the rotation and translation parameters of the camera by the known 3D geometric features and their corresponding 2D geometric features is an important method to realize camera attitude estimation. On the basis of this, a new algorithm for pose estimation of camera based on 3D ~ 2D corresponding geometric features is proposed for attitude estimation of monocular camera. The algorithm uses the unit quaternion to represent the rotation matrix, and the rotation and translation parameters are solved by establishing the pose and parameter constraint equations. Compared with Ansar’s NLL algorithm, this algorithm can better overcome the influence of image noise on pose accuracy and ensure the orthogonality of its calculated rotation matrix. Simulation results compare the proposed algorithm with the NLL algorithm. The results show that the proposed algorithm has obvious advantages in robustness, computational accuracy and complexity, and the experimental results verify the effectiveness of the proposed algorithm.