Air pollution has become one of the most pressing environmental issues in many countries, including China. Finegrained PM2:5 particulate data can prevent people from long time exposure and advance scientific research. However, existing monitoring systems with PM2:5 stationary sensors are expensive, which can only provide pollution data at sparse locations. In this paper we demonstrate for the first time that camera on smartphones can be used for low-cost and fine-grained PM2:5 monitoring in participatory sensing.
We propose a Learning-Based method to extract air quality related features from images taken by smartphones. These image features will be used to build the haze model that can estimate PM2:5 concentration depending on the reference sensors. We conducted extensive experiments over six months with two datasets to demonstrate the performance of the proposed solution using different models of smartphones. We believe that our findings will give profound impact in many research fields, including mobile sensing, activity scheduling, haze data collection and analysis.