基于双平行线特征的位姿估计解析算法

提出了一种基于特征直线的单目单帧计算机视觉相对位姿估计算法.用几何的方法代替常规代数方法,回避了特征点算法中传统的通过求解旋转矩阵和平移向量来获取位置和姿态参数的过程.以特征直线的平行关系和四直线围成的面积作为输入,推导出了几何意义下的位置和姿态参数解析表达式,求解过程简单、直接.理论推导和实验验证表明了该算法的正确有效性.

Analytic algorithm of pose estimation based on two pairs of parallel lines

P4P problem under the condition that the four points come from the comers of a parallelogram was studied and a novel analytic solution to the problem based on feature lines is proposed. The definitions of object coordinate system and camera coordinate system are given first. Based on the pinhole model of camera and the theory of perspective projection, a geometric method instead of algebraic one is used to calculate the object 3D pose parameters relative to the camera coordinate system, which avoids computing the rotation matrix and translation vector. The parallel correlation of the lines and the area of the quadrangle formed by the four lines are used as inputs to obtain geometric values of pose parameters, and the process of the solution is an analytic one. Mathematical deduction and experimental results verify the validity and effectiveness of the algorithm.