一种基于目标识别的运动视定位方法

A LOCALIZATION METHOD USING MOTION VISION

针对室内移动机器人定位问题提出了一种基于目标识别的运动视定位方法.该方法确定机器人相对于其接近目标的位姿.选取目标上与相机投影中心相对于地面等高的两个点作为特征点,首先拍摄一幅包含这两个特征点的图像,机器人移动后再拍摄一幅包含这两个特征点的图像,该方法就能根据这两个特征点在这两幅图像中的坐标计算出机器人移动前后的位姿.该方法只需两个特征点和两幅图像.因而比较简单且有较高的实时性.试验结果验证了该方法的有效性.

This paper proposes a motion vision localization method based on objects recognition aiming at the indoor mobile robot localization problem. This method determines the pose of a robot relative to the object to which it moves. Two points with the same height from the floor as the projection center of camera are selected as feature points. Before the robot moves, an image containing the two feature points is taken. Another image containing the two feature points is taken after the robot moves. Then, the pose of the robot before and after motion can be determined according to the image co-ordinations of the two feature points in the consecutive two images. This method only requires two feature points and two images, Therefore, it is fairly simple and is sufficiently real time. Experimental results show that it is effective.