基于极大似然估计的全景立体视觉机器人自定位方法

Self-localization of a robot with omnidirectional stereoscopic vision based on maximum likelihood estimation algorithm

针对室内环境下移动机器人的自主定位问题,研究了一种基于极大似然估计算法的全景立体视觉机器人自主定位方法。通过分析双曲面折反射式全景成像系统原理,利用全景视觉具有探测范围广、能够获取环境信息丰富的特点,提出了一种同向垂直基线的全景立体深度信息测量算法。在三边定位算法的基础上推广,提出了基于多路标的极大似然定位算法,采用最小二乘估计法,计算机器人定位的几何超定方程组,设计了完备的定位算法,实现机器人的绝对坐标解算。该方法能够应用于大范围环境的全局定位,实验结果验证了该算法的可靠性和准确性。

In order to solve the self-localization of mobile robot in indoor environment,a self-localization method based on the maximum likelihood estimation localization algorithm and used for omnidirectional stereo vision robot was proposed. By analyzing the imaging principle of hyperbolic catadioptric omnidirectional vision system,a method of measuring the panoramic three-dimensional depth information adopting vertical baseline with the same direction was proposed. The omnidirectional vision has the characteristics of wide detection range and rich environmental information. On the basis of the trilateration localization algorithm,a maximum likelihood localization algorithm based on multiple landmarks was proposed,the least squares estimation method was used to calculate the geometric overdetermined equations and a complete positioning algorithm was designed to realize the absolute coordinates solution on the robot. The method can be applied to global localization in a wide range of environments,and experimental results verified the reliability and accuracy of the proposed algorithm.