Abstract: In order to improve visual navigation accuracy of bi-directional automatic guided vehicles, it extracts the centerline of guide path by processing color images. According to measurement accuracy between the target and automated guided vehicles, a method of path model classification based on the average slope difference and cornerity distinguish index estimation is proposed, which divides path into lines, arcs and non-circular turning three kinds of models, and it uses MMSE method for parametric regression to linear model, uses the Levenberg-Marquardt method to fit the arc model, according to the position of cornerity point, it divides non-circular turning model into sections and fits these sections respectively. Experimental results show that this method not only has a good performance on path identification which is based on monocular vision, but also achieves the measurement accuracy requirements of target.

Key words: visual navigation, automatic guided vehicles, average slope difference, cornerity distinguish index, arc fitting

摘要： 为了提高双向型自动导引车的视觉导引精度，通过处理彩色图像提取导引路径中心线；根据测量目标对自动导引车的导引精度，提出一种基于平均斜率差及拐点分辨指数估计的路径模型分类方法，将路径分为直线、圆弧拐弯和非圆弧拐弯3种模型，并采用最小均方差法对直线模型参数进行回归，用Levenberg-Marquardt法对圆弧模型进行拟合，根据拐点位置将非圆弧拐弯进行分段拟合。实验结果表明，该方法对基于单目视觉的路径识别有很好的效果，同时测量精度也达到目标要求。

关键词: 视觉导航, 自动导引车, 平均斜率差, 拐点分辨指数, 圆弧拟合