A Real-Time Road Anamoly Detection Using Smart Phone Sensors

The quality of the conventional transport system has been significantly impacted recently by the swift advancement of emerging technologies and the rapid growth of the economy. ITS which seeks to enhance the transportation network, is gaining popularity. Additionally, one of the main goals of ITS is to enhance traffic safety, and potholes seriously compromise the safety of vehicles. Therefore, by establishing a real-time pothole-detecting system and disseminating pothole information, drivers’ safety may be enhanced. For assessing road conditions, we propose to analyze various supervised machine learning approaches employing mobile sensors and GPS data obtained from cellphones. Our research focuses on classifying three primary class labels: deep transverse cracks, potholes, and smooth roads. We test our hypothesis that employing information from the three sensors’ axes yields more accurate results than using data from just one. We also study how well deep neural networks do in classifying different types of traffic situations. Our findings demonstrate that smartphone sensor axes outperform models using only one axis. This demonstrates that vibration signals for road abnormalities contain information throughout all three axes. We also note that substantially more accurate data classification is obtained when using neural networks. The techniques discussed here can be applied more broadly to monitor road conditions that could endanger people’s safety. Commuters in addition to giving pertinent authorities maintenance information.