基于贝塞尔轨迹的视觉导引AGV路径跟踪研究

Research on path tracking of visual navigation AGV based on Bessel trajectory

为提高视觉导引自动导引车(automatic guided vehicle,AGV)路径跟踪精度,提出一种基于贝塞尔轨迹的精确路径跟踪算法。该算法首先将采集的多种路径特征图像进行预处理得到形状信息,训练SVM多层路径形状特征分类器;然后根据命令,改变AGV采集到的分支路径图像的权重,迭代计算所选择路径的若干最小内接圆;最后,利用最小二乘规则,将最小内接圆的圆心拟合成贝塞尔轨迹,实现AGV的精确路径跟踪。将该算法应用于视觉引导AGV中,并进行路径特征的在线识别和轨迹跟踪实验。结果表明:路径特征的识别准确率为99.7%以上,识别时间约为22 ms,弯道轨迹跟踪准确度为20 mm和20°;与传统方法相比,该方法显著提高路径特征识别和轨迹跟踪的准确率,更能满足工业现场需求。

To increase path tracking accuracy of visual navigation AGV（automatic guided vehicle）, a precise path tracking algorithm based on Bessel trajectory is proposed. Firstly, the algorithm will pre-process the collected feature images of various paths to obtain shape information, and train the SVM multi-path shape feature classifier, the collected images and iteratively calculate and then change the weight of the branch paths of the minimum inscribed circles of the selected paths according to the order. Finally, based on the least squares rule, the centre of the minimum inscribed circle will be fitted into the Bessel trajectory to realizing the precise path tracking of AGV. The algorithm was applied in visual navigation AGV and on-line recognition and trajectory tracking test of path features were carried out and the results shown that the recognition accuracy of path features is up to 99.7%, and the recognition time is about 22 ms, curve trajectory tracking accuracy is 20 mm and 20°. Comparing with the traditional method, the method can improve the accuracy rate of path recognition and path tracking, which meets industrial field applications.