论文部分内容阅读
为了提高移动机器人的自主导航能力,提出一种基于视觉的用于城市环境中移动机器人姿态估计的优化方法。该方法利用单视图中的竖直消失点以及两视图中匹配的水平消失点完成估计。为了准确地提取消失点,设计了一种误差识别的消失点估计方法以获得具有最小方差的估计结果。通过分析消失点估计过程中的误差传递,将消失点的最小方差估计问题转化为一个凸优化问题。然后,利用垂直判据检查所有提取的候选水平消失点,并对两视图中的水平消失点进行匹配。物理实验表明,该算法能够快速、准确地估计出移动机器人的姿态。
In order to improve the autonomous navigation ability of mobile robot, a vision-based optimization method for attitude estimation of mobile robot in urban environment is proposed. This method makes use of the vertical vanishing point in a single view and the matching horizontal vanishing point in both views. In order to extract the vanishing point accurately, a vanishing point estimation method of error identification is designed to get the estimation result with the smallest variance. By analyzing the error transfer in the vanishing point estimation process, the problem of minimum variance estimation of vanishing point is transformed into a convex optimization problem. Then, using the vertical criterion, all extracted candidate horizontal vanishing points are examined, and the horizontal vanishing points in the two views are matched. Physical experiments show that the algorithm can quickly and accurately estimate the attitude of the mobile robot.