Pavement performance monitoring and anomaly recognition based on crowdsourcing spatiotemporal data

Pavement performance is a critical factor toward riding comfort experience and drastically affect traffic management and the safety of road users. Since road quality declines over time and current periodic inspection on a vast road network is laborious and costly to the authority. This paper proposes a participatory system to conduct pavement performance monitoring of a country-wide road network based on crowdsourcing spatiotemporal data. By conducting cloud computing of a statistical grading mechanism with respect to the vertical and lateral acceleration behavior, the perception of riding comfort, which has a high correlation with pavement quality, can be reflected faithfully based on the spatiotemporal data acquired from a smartphone-driven progressive web application. Moreover, a deep learning technique is leveraged to identify road anomalies from the on-site images for a cross-check mechanism, which ensures the reliability of the monitoring pavement conditions and facilitates the automation level of road anomaly labeling and documenting. The proposed pavement performance monitoring was validated by the road network of Taipei city, Taiwan, which rendered promising results with an accuracy up to 98% and a false positive rate smaller than 1.3% showing the practicality and adaptability in a complex road network.