The detection and recognition of arrow markings recognition based on monocular vision

Road information understanding is a necessary task for both intelligent vehicles and driving assistance systems. Previous research mostly focused on the detection of lane position. Other information provided by arrow markings was scarcely mentioned. In this paper, Arrow extraction is carried out by projection histogram on Inverse perspective image and an arrow markings recognition algorithm is presented based on multi-class support vector machines. An improved Haar wavelet feature extraction approach is utilized to describe the feature of arrow markings. In order to guarantee generalization performance, the F-score method is used for feature reduction. The results show that the algorithm can detect and recognize arrow markings effectively, and it's robust to occlusion by other vehicles or poor visibility.