巡线机器人视觉导航中的摄像机现场标定

Camera calibration in visual navigation of inspection robot on site

为在工作现场对巡线机器人视觉系统中的摄像机进行标定,提出了一种自动化的标定方法。方法首先通过视觉检测地线和线上的运动目标,发现目标后,控制机器人沿地线作低速运动。检测到目标进入处理区域时,记录目标上最前端点的坐标,然后变换2次视角,获取3个视角下的一系列图像。在处理中,提取每个视角下间隔Δn帧的n幅图像中目标上的最前端点,以通过该点的地线的垂线与地线两侧边缘线及轴线的交点形成2×(n-1)的棋盘格,并对格点进行校正。然后利用地线尺寸、机器人速度和被处理帧间的时间差等已知条件,获取图像中的点与其三维空间点间的对应关系,最后利用张氏标定法相同的处理方式,解算出摄像机的内外参数。结果通过实验验证,以张氏标定结果为准,该算法对内参中的焦距误差不超过5%。通过本文算法,可实现摄像机的工作中的自动标定,无需人工拍摄标定参照物,不需人工干预,能较好的实现自动化。

In order to realize calibration of the camera carried by the inspection robot on site and obtain the intrinsic parameters such as the focal length, an automatic calibration method is proposed based on the Zhang＇s algorithm. In the method,the ground wire and moving objects on it are detected by vision method,and then the robot is controlled to move slowly along the wire when the obieet is found. When the object enters processing area, the front point coordinates of the object are recorded, and then two an- gles of view are transformed to obtain a series of images from three viewing angles. In image processing, the front points on the object in n images from each visual angle spaced at An frames are extracted,and then a checkerboard with 2 × （n-1） grids is formed at the intersection point of the vertical line through the points and the axis line and the two sides of the ground wire. Then,by using the known conditions such as the size of the wire,the speed of the robot and the time difference between the processed frames, the relation between the grid points in the image and its three-dimensional space points are obtained. Fi- nally,the intrinsic and extrinsic parameters of the camera are calculated by the same processing method of Zhang＇s calibration method. It is proved by experiment that the error of the focal length is less than 5 % in terms of Zhang＇ s calibration results. The automatic calibration during the work of the camera can be realized in the method, and manual calibration model is not need, so the whole work can be done by machine.