针对单目机器人在运动过程中，相邻关键帧之间基线过短造成的相机相对位姿无法恢复的问题，提出了一种利用已知特定参照物迭代优化本质矩阵，从而快速分解出机器人相对位姿的方法。该方法首先利用已知特定参照信息获取本质矩阵估计值，然后通过相邻关键帧之间匹配的特征点组和本质矩阵估计值对本质矩阵迭代优化，最后通过本质矩阵的约束关系求得平移向量估计值，利用基于李群的姿态表示得到旋转矩阵估计值，并进一步优化得到唯一解。该方法避免了由于基线过短而无法恢复相机相对位姿的弊端。实验结果表明，该方法可快速解算并优化机器人相对位姿关系，且相对位移估计绝对误差小于0.03 m，优于传统方法。

During the movement of the monocular robot, the camera pose cannot be recovered due to the extremely short base line between the key frames.Aiming at this problem, an iterative optimization algorithm of essential matrix was proposed to quickly decompose the relative pose relationship of robot by using the known specific reference objects. In this method,firstly, the estimated value of essential matrix is obtained using a specific reference with clear three-dimensional information. Then, the essential matrix is iteratively optimized with the matched feature point groups of adjacent key frames and the estimated value of essential matrix. Finally, the estimation of translation vector is obtained by the constraint relation of essential matrix, and the estimation of rotation matrix is obtained based on Lie group. The unique solution is obtained by further optimization. The problem that the relative pose of the camera cannot be restored due to the too short base line is avoided in this method. The experimental results show that the relative pose relationship of the robot can be quickly solved and optimized in this new method, and the absolute error of relative displacement estimation is less than 0.03 m, which is superior to the traditional method.