Unstructured Road Navigation Image Feature Extraction And Classification

In the research of mobile robot, navigation is one of the kernel approaches. Vision-based navigation system should detect the running environment and locate the boundary of the road in real-time. The recognition of non-structured road is the important component in the Vision-based navigation system. In the cross-country environment, the non-structured road holds the very great proportion, so it has vital significance to recognize non-structured.This paper mainly research region extraction and road classification of non-structured road in the Vision-based navigation system. For the characteristic of non-structured road's image, we choose space HSV to carry on color analysis which conforms to human vision sensation, and use gray co-occurrence matrix and Gabor filter method which is more suitable to analysis natural texture to extract texture character's statistics of non-structured road. For outdoor non-structured road's environment, we use the support vector machines' method which mainly contraposes the two classification problem to segment non-structured road's region. Then, for the road region which have been extracted, we use the improved k- nearest neighbor method to classify different roads, and make the robot have better information about non-structured road's environment, and let it travel much more steadily, safely and speedily.