Automatic Road Anomaly Detection Using Smart Mobile Device

Maintaining the quality of roadways is a major challenge for governments around the world. In particular, poor road surfaces pose a significant safety threat to motorists, especially when motorbikes make up a significant portion of roadway traffic. According to the statistics of the Ministry of Justice in Taiwan, there were 220 claims for state compensation caused by road quality problems between 2005 to 2007, and the government paid a total of 113 million NTD in compensation. This research explores utilizing a mobile phone with a tri-axial accelerometer to collect acceleration data while riding a motorcycle. The data is analyzed to detect road anomalies and to evaluate road quality. Motorcycle-based acceleration data is collected on twelve stretches of road, with a data log spanning approximately three hours, and a total road length of about 60 kilometers. Both supervised and unsupervised machine learning methods are used to recognize road conditions. SVM learning is used to detect road anomalies and to identify their corresponding positions from labeled acceleration data. This method of road anomaly detection achieves a precision of 78.5%. Furthermore, to construct a model of smooth roads, unsupervised learning is used to learn anomaly thresholds by clustering data collected from the accelerometer. The results are used to rank the quality of the road segments in the experiment. We compare the ranked list from the learned evaluator with the ranked list from human evaluators who rode along the same roadways during the test phase. Based on the Kendall tau rank correlation coefficient, the automatically ranked result exhibited excellent performance. Keywords-mobile device; machine learning; accelerometer; road surface anomaly; pothole;