Perceiving Metropolitan-scale Pothole Profiles With Crowdsourcing

With the rapid development of economy, auto mobiles have become a indispensable means of transport.However, with the increasing of vehicles number, the road condition becomes more severe.The roads in bad condition have quite bad influence on the experience of drivers and the safety of those drivers and passengers. The road condition is also concerned by the government. Accessing to timely and accurate metropolitanscale road condition information, especially about dangerous potholes is of great importance to the public and the government.In this paper, we propose a novel scheme, which utilizes smartphones to sense and estimate the metropolitan-scale pothole profile. In particular,the pothole profile refers to the depth and length of those potholes.In our method, we utilize the sensors built in smartphones to collect all kinds of the sensory data of vehicles.In particular, our system can actively learn the knowledge of a vehicle and adopt a one degree-of-freedom(DOF) vibration model to infer the depth and length of pothole while the vehicle is moving. Furthermore,we leverage the strong power of crowdsourcing by aggregating these individual estimations derive a globally optimized result. By design, our system is light-weighted and robust to various conditions such as poor light, bad weather and different vehicle types. We have implemented a prototype system to prove the feasibility of our system. Extensive experiment results demonstrate the efficacy of the design of our system. On average,our method can achieve relatively accurate estimation. And the low depth and length estimation error rates of 13% and 16%, respectively.