Multi-class Traffic Sign Detection and Classification Using Google Street View Images

摘要: Maintaining an up-to-date record of the location and condition high-quantity low-cost roadway assets such as traffic signs, is critical to safety transportation systems. Despite their importance, today’s video-based data collection analysis practices are still costly, prone error, performed intermittently. While, databases Google Street View (GSV) contain street-level panoramic images all signs updated regularly, potential for creating a comprehensive inventory has not been fully explored. The key benefit these that once detected, accurate geographic coordinates detected can be automatically determined visualized within same platform. Nevertheless, detecting classifying from GSV imagery challenging due interclass variability particularly changes in illumination, occlusion, orientation. This paper evaluates application computer vision method multi-class sign detection classification images. extracts using API leverages sliding window mechanism detect candidates signs. For each candidate, Histogram Oriented Gradients (HOG) formed concatenated with Color Histogram. HOG+Color descriptors then fed into multiple one-vs.-all Support Vector Machine classifiers classify them specific categories. experimental results average accuracy 95.5% demonstrate leveraging viable solution inventories