Estimated outdoor PM2.5 concentration data by using mobile phone images in Bangkok, Thailand

Description

Estimating PM2.5 concentrations from images using two sets of images collected at a location. Whereas, PM2.5 concentrations from a nearby monitor stations in Bangkok which collected at the same time were used as ground truth. Hourly PM2.5 concentration data was collected from an air pollution monitoring station which operated by the Pollution Control Department (PCD) of Thailand. The PM2.5 concentrations in the air were measured in μg/m3 from the begin of January 2020 until the end of March 2020 were used as Ground truth data. Sky images were taken with the mobile phone camera at the top of the G Tower building at a fix location with manually marked reference regions. The image data consists of 230 photos taken over 3 months at the same period of the PM2.5 concentration data in Bangkok. The images were captured by ourselves using a smartphone model (iPhone 6 Plus) at the same time everyday (7:00 AM and 6:00 PM local time) for three months. Most of the weather conditions in photos are sunny because during the winter season in Thailand. In order to evaluate our approach, the PM2.5 concentration and the sky image data were randomly divided into two sets: 70% for training dataset and 30% for test dataset. The training dataset is used for training a Convolutional Neural Networks (CCN) by using the Python software program, while the test dataset is used for evaluation.

Steps to reproduce

- Hourly PM2.5 concentration data was collected from an air pollution monitoring station which operated by the Pollution Control Department (PCD) of Thailand. The PM2.5 concentrations in the air were measured in μg/m3 from the begin of January 2020 until the end of March 2020 were used as Ground truth data. - Sky images were taken with the mobile phone camera (iPhone 6 Plus) at the top of the G Tower building at a fix location with manually marked reference regions. The image data consists of 271 photos taken over 3 months at the same period of the PM2.5 concentration data in Bangkok. - The PM2.5 concentration and the sky image data were randomly divided into two sets: 70% for training dataset and 30% for test dataset. - Convolutional Neural Networks (CCN) with RestNet-50 model by a Python software program was used to train and test those to sets of data recpectively. - Root Mean Square Error (RMSE) and R-squared were used to identify the estimation error.

Institutions

Categories

Funding

Licence