Road Scanner: A Road State Scanning Approach Based on Machine Learning Techniques

The state of roads may sometimes be difficult to perceive due intense climate conditions, absence road signs, or simply human inattention, which harmful both vehicles and drivers. automatic monitoring the states represents a promising solution warn drivers about status in order protect them from injuries accidents. In this paper, we present novel application for data collection regarding states. Our entitled “Road Scanner” allows onboard users tag four types segments roads: smooth, bumps, potholes, others. For each tagged segment records multimodal using embedded sensors smartphone. collected concerns mainly vehicle accelerations, angular rotations, geographical positions recorded by accelerometer, gyroscope, GPS sensor, respectively, user phone. Moreover, medium-size dataset was built machine learning models were applied detect right label segment. Overall, results very since SVM classifier (Support Vector Machines) has an accuracy rate 88.05%.