跟随路径导引是自主式移动机器人广泛采用的一种导航方式，其中视觉导航具有其他传感器导航方式所无法比拟的优点，是移动机器人智能导航的主要发展方向。为了提高移动机器人视觉导航的实时性和准确性，提出了一个基于路径识别的视觉导航系统，其基本思想是首先用基于变分辨率的采样二值化和形态学去噪方法从原始场景图像中提取出目标支持点集，然后用一种改进的哈夫变化检测出场景中的路径，最后由路径跟踪模块分直行和转弯两种情况进行导航计算。实验结果表明，该视觉导航系统具有较好的实时性和准确性。

Guidance using path following iswidely applied in the field of autonomousmobile robots. Comparedwiththe navigation system without vision, visual navigation has obvious advantages as rich information, low cost, quietness, innocuity, etc. This paper describes a navigation system which uses the visual information provided by guide lines and color signs. In our approach, the visual navigation is composed of three main modules: image-preprocessing, path-recognition and path-tracking. First, image-pre-processing module formulates color models of all kinds of objects, and establishes each object’s support through adaptive subsampling-based binarization and mathematical morphology. Second, path-recognition module detects the guide lines through an improved Hough transform algorithm, and the detected results including guide lines and color signs integrate the path information. Finally, calling different functions according to the movement of straight-going orturning, path-trackingmodule provides required input parameters tomotor controller and steering controller. The experimental results demonstrate the effectiveness and the robustness of our approach.