Algorithm 1: an algorithm to achieve image based prediction of PM2.5

Input: SP denotes the normalized probability that the image belongs to Severe Pollution class; MP denotes the normalized probability of Mild Pollution class, and NP denotes the normalized probability of Nominal Pollution class.

Output: PM2.5 concentrations predicted from the image (PM_Image).

Begin

1 if ((SP > MP) and (SP > NP)) then

2 PM_Image = (150.5 + 275.9* SP) + (150.4 * MP)/2

3 else if ((NP > SP) and (NP > MP)) then

4 PM_Image = (35.4 * (1 – NP)) + ((150.4 * MP)/2)

5 else if ((MP > SP) and (MP > NP)) then

6 if (SP > NP) then

7  PM_Image = (35.5 + 114.9 * MP) + ((500.4 * SP)/2)

8 else if (NP > SP) then

9  PM_Image = (35.5 + 114.9 * MP) + ((35.4 * NP)/2)

10 return PM_Image

End