Self-adaptive road detection method based on vision and cluster analysis

A self-adaptive road detection method based on vision and cluster analysis is proposed for automatic guided vehicles. Based on the K-means algorithm, an automatic sample selection method is developed - while moving, the vehicle automatically selects new road samples and takes cluster analysis at a specified time interval to get the latest road features. Verified by experiments on campus roads, the proposed method is adaptive to the changes of road conditions. The influence of illumination, shadow, and road texture to the detection results is effectively reduced. Much less manual operations are needed compared to the traditional approaches based on learning algorithms.