Road surface conditions affect transport safety and driving comfort, particularly in snowy areas. This paper proposes a new method for detecting road surface conditions based on crowdsourced mobile sensing technology. The method can efficiently detect road surface conditions using motion sensors embedded in smartphones mounted on vehicles. Detecting road conditions using such sensors, which are usually loosely placed in the vehicle, nonetheless poses a challenge. Our approach comprises two modules: a clustering module and classification module. The clustering module uses a K-means algorithm to distribute the vehicle travel mode into a proper cluster. The classification module uses the random forest classifier, which is assigned to each group of vehicle travel mode, to classify the road surface conditions. We defined new road surface conditions as the estimation target, considering both the substance that covers the road surface and the shape of the road surface itself. The results show that our approach can detect road surface conditions with accuracy as high as 90%.
展开▼
