Water Detecting And Navigation On Unstructured Road

Unstructured road navigation technology is one of the difficulties in autonomous navigation of mobile robots, while it is usually difficult to detect water on the road and it might cause large damage to robots. In this paper, we consider avoiding all harmful obstacles as much as possible, propose an unstructured navigation system framework based on detecting water, and design a related navigation and detection algorithm, making robots real-time accurate to identify the travelable area of unstructured roads, which provides a guarantee for safe autonomous navigation. This paper is consisted by three main parts:First, aiming at image segmentation of unstructured navigation based on detecting water, considering the real-time demands of navigation outdoors, we propose a method for image partitioning combining with laser radar data, which accomplishes quick and accurate partition of unstructured roads. It is proven that this algorithm can partition images accurately under the circumstances of fulfilling the real-time demands, which satisfies the needs of unstructured road navigation.Water detecting on the assumption that the close area in front is road, combining with the results of image segmentation above, water area is detected through comparing the color and textural features between known road area and unknown area.At last, we use an incremental support vector machine based on a near-to-far study method to detect the travelable area, make it detect long-distance travelable area only in need of very little environment information, on the basis of without human intervene, which makes it convenient to further route planning.