An uncertainty-aware Road Anomalies Detection Based on Dempster–Shafer Theory

Sensors on smartphones have become a ubiquitous component of Intelligent Transportation Systems (ITS), particularly in detecting cracks, bumps, and potholes on roads. In this context, and based on the certainty classification approach, many solutions using smartphone sensors have been proposed. Such an approach to certainty ignores the uncertainty associated with sensors, which is its main flaw. In fact, the sensed information may be imprecise and inaccurate, prone to error, and subject to incompleteness, ambiguity, and sometimes conflict. Moreover, in the ITS field, many factors like the sensor's quality, sensor lifetime, and the position of the sensor in the vehicle have a severe impact on the detection process. Towards addressing this issue, and for more accurate detection of road anomalies, this paper investigates an uncertainty classification approach based on the Dempster-Shafer theory (DST). To evaluate the proposed uncertainty method, we compared it with some existing ones in the literature using a publicly available dataset. As a result of the comparative evaluation, the proposed method outperforms the existing methods.