Estimating Road Surface Condition Using Crowdsourcing

Road surface conditions have a significant impact on transport safety and driving comfort, particularly in snowy areas. This paper proposes a new method for estimating road surface conditions by using a motion sensor embedded in a smartphone. The method is based on a mobile sensing framework that can collect sensor data using crowdsourcing. In this study, we have defined new road surface conditions as the estimation target, which takes into account both the substance that covers the road surface, and the shape of the road surface itself. The paper also describes a method of feature selection, comprising two steps: First, an initial feature set is directly calculated using various features published in previous studies, using raw sensor data. Second, three feature selection algorithms are compared: Principal Component Analysis (PCA), Relief-F, and Sequential Forward Floating Search (SFFS), and the most effective of the three chosen. In this study, the SFFS algorithm showed higher accuracy than the others. The road surface condition classification was performed across different speed ranges using the Random Forest Classifier, and results show that the best accuracy, of about 91%, was obtained in the 50 km/h–80 km/h range.