CrackSense: A CrowdSourcing Based Urban Road Crack Detection System

As a common road surface distress, cracks pose a serious threat to road infrastructure and traffic safety in cities today. Consequently, road crack detection is considered as an essential step for effective road maintenance and road structure sustainability. However, due to the high cost incurred by dedicated devices and professional operators, it is impossible for existing systems to achieve universal spatiotemporal coverage across citywide road networks. To fill this gap, in this paper, we present the CrackSense, a mobile crowdsourcing based system to detect urban road crack and estimate its damage degree. Specifically, for the heterogeneous crack data, we put forward a crowdsourcing data quality evaluation and selection mechanism. And then, by utilizing the multi-source sensing data aggregation, we propose tow algorithms, namely RCTR and RCDE, to recognize road crack types, i.e., horizontal crack, vertical crack, and net crack, and estimate the crack damage degree, respectively. We implement the system and develop a smartphone APP for mobile users. By conducting intensive experiments and field study, the results demonstrate the accuracy and effectiveness of our proposed approaches.