Newswise —  Air pollution is a critical global health issue, demanding innovative monitoring solutions. Traditional methods, reliant on ground stations, are expensive and geographically limited, hindering comprehensive coverage. Recent strides in technology have spotlighted the potential of using visual data from surveillance cameras as a cost-effective alternative for air quality assessment.

A new study (doi: https://doi.org/10.1016/j.ese.2023.100319) published in Environmental Science and Ecotechnology (Volume 18, 2024) innovates a hybrid deep learning model that significantly improves outdoor air quality monitoring using surveillance camera images. This approach enhances air quality estimations, including PM2.5 and PM10 concentrations and the Air Quality Index (AQI), irrespective of the time of day.

The research team skillfully combined Convolutional Neural Networks (CNN) with Long Short-Term Memory (LSTM) networks, creating a model that intelligently captures both the spatial details present in individual images and the temporal dynamics across a sequence of images. This innovative approach is particularly adept at overcoming the longstanding challenge of accurately estimating air quality during nighttime, a period when traditional image-based methods typically falter due to low light conditions. By analyzing the visual cues in surveillance footage, such as haze and visibility, the model can predict concentrations of particulate matter (PM2.5 and PM10) and the Air Quality Index (AQI) effectively, both day and night.

Highlights

  • Three time-series image datasets were constructed for air quality assessments.
  • CNN and LSTM are combined to achieve an average estimated R2 > 0.9 throughout the day.
  • Our method enhances nighttime air quality estimation and improves overall accuracy.
  • Our method outperforms existing methods with the differences on R2 being 0.02–0.22.

Dr. Xuejun Liu, lead researcher and corresponding author, emphasizes, "Our model's ability to accurately estimate air quality from images, regardless of day or night, marks a significant step forward in utilizing technology for environmental monitoring. It opens up new avenues for comprehensive air quality assessment in regions lacking infrastructure."

This research signifies a substantial leap forward in environmental monitoring, showcasing the potential to enhance air quality assessments significantly. It opens the door to more dynamic, cost-effective monitoring solutions that could vastly improve our understanding and management of air pollution on a global scale.

###

References

DOI

10.1016/j.ese.2023.100319

Original Source URL

https://doi.org/10.1016/j.ese.2023.100319

Funding information

This work was supported by the National Key Research and Development Program of China [2021YFE0112300]; the National Natural Science Foundation of China (NSFC) [41771420]; the State Scholarship Fund from the China Scholarship Council (CSC) [201906865016]; and the Postgraduate Research & Practice Innovation Program of Jiangsu Province [KYCX21_1341].

About Environmental Science and Ecotechnology

Environmental Science and Ecotechnology (ISSN 2666-4984) is an international, peer-reviewed, and open-access journal published by Elsevier. The journal publishes significant views and research across the full spectrum of ecology and environmental sciences, such as climate change, sustainability, biodiversity conservation, environment & health, green catalysis/processing for pollution control, and AI-driven environmental engineering. The latest impact factor of ESE is 12.6, according to the Journal Citation ReportTM 2022.

Journal Link: Environmental Science and Ecotechnology, October 2023