RPG Seminar – ResNet-LSTM for Real-time PM2.5 and PM10 Estimation Using Sequential Smartphone Images*
Date:
11 Dec, 2020 (Fri)
Time:
5:00 pm
Webinar Link:
https://hku.zoom.us/j/96758054436?pwd=L3NQVnBvMkQwZkZJTUwrbklVb1l4dz09
Meeting ID: 967 5805 4436
Password: 176535

Add to your calendar: iCal, Google Calendar

HKU Campus Map


Speaker

Mr. Shiguang SONG
Department of Electrical and Electronic Engineering
The University of Hong Kong

Abstract

Attempts have been made to estimate PM2.5 and PM10 values from smartphone images, given that deploying highly accurate air pollution monitors throughout a city is a highly expensive undertaking. Departing from previous machine learning studies which primarily focus on pollutant estimation based on single day-time images, our proposed deep learning model integrates Residual Network (ResNet) with Long Short-Term Memory (LSTM), extracting spatial-temporal features of sequential images taken from smartphones instead for estimating PM2.5 and PM10 values of a particular location at a particular time. Our methodology is as follows: First, we calibrated small portable air quality sensors using reference instruments placed in the official air quality monitoring station, located at Central, Hong Kong (HK). Second, we verified experimentally that any PM2.5 and PM10 values obtained via our calibrated sensors remain constant within a radius of 500 meters. Third, 3024 outdoor day-time and night-time images of the same building were taken and labelled with corresponding PM2.5 and PM10 ground truth values obtained via the calibrated sensors. Fourth, the proposed ResNet-LSTM was constructed and extended by incorporating meteorological information and one short path. Results have shown that, as compared to the best baselines, ResNet-LSTM has achieved 6.56% and 6.74% reduction in MAE and SMAPE for PM2.5 estimation, and 13.25% and 11.03% reduction in MAE and SMAPE for PM10 estimation, respectively. Further, after incorporating domain-specific meteorological features and one short path, Met-ResNet-LSTM-SP has achieved the best performance, with 24.25% and 20.17% reduction in MAE and SMAPE for PM2.5 estimation, and 28.06% and 24.57% reduction in MAE and SMAPE for PM10 estimation, respectively. In future, our deep-learning image-based air pollution estimation study will incorporate sequential images obtained from 24-hr traffic surveillance cameras available in all parts of the city in HK, to provide full-day and more fine-grained image-based air pollution estimation for HK.

 

Biography of the speaker:
Shiguang SONG received the B.Eng. (Hons.) degree in Electrical Engineering from the University of Liverpool and M.Sc. degree in Future Power Networks from Imperial College London, UK, in 2015 and 2016, respectively. He is currently working toward the Ph.D. degree in the Department of Electrical & Electronic Engineering, The University of Hong Kong (HKU). His research interests include deep learning, artificial intelligence, environmental technologies, and image processing.

* This is a joint work with Dr. Jacqueline CK Lam, Mr. Yang Han, and Prof. Victor OK Li.


Organizer

Prof. Victor O.K. Li and Dr. Jacqueline C.K. Lam

Most seminars are open to the general public, free of charge, unless otherwise stated. Registration is not required. Arrangement for car parking facilities on campus please contact us for details.

For enquiries, please contact:
Department of Electrical and Electronic Engineering,
Room 601, Chow Yei Ching Building,
Pokfulam Road, Hong Kong
Tel: (852) 3917 7093
Email: seminar@eee.hku.hk