Fine-Grained Infer PM_2.5 Using Images from Crowdsourcing

Shuai Li, Teng Xi, Xirong Que, Wendong Wang

Abstract: Among all air pollutants, PM $$_{2.5}$$ , which can be inhaled into lungs, is most harmful for peoples’ health. However, the number of fixed air quality measurement stations is insufficient. In order to make people be more aware of the air quality around them, this paper have proposed a method to infer fine-grained PM $$_{2.5}$$ concentration. We leverage different type of collected by crowdsourcing for data mining. Then, features which have strong correlation with PM $$_{2.5}$$ concentration are extracted. Furthermore, we train the proposed model using integrated radial basis function (rbf) kernel based ridge regression. The performance of the proposed method is evaluated thoroughly by real dataset collected by crowdsourcing. The results show that, our method can accurate infer the PM $$_{2.5}$$ concentration.