Deep learning for air pollution detection: Analyzing scots pine needles with SEM/EDS
The goal of the activity described in the paper is to develop a smart air pollution identification system. The use of artificial intelligence, using the analysis of surface images of selected air pollution indicators to build a machine learning algorithm, allowed to develop an inexpensive and effect...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2025-04-01
|
| Series: | Desalination and Water Treatment |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1944398625001651 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The goal of the activity described in the paper is to develop a smart air pollution identification system. The use of artificial intelligence, using the analysis of surface images of selected air pollution indicators to build a machine learning algorithm, allowed to develop an inexpensive and effective method of identifying hazardous substances. What was used to build the model, were scanning electron microscopy images of two-year-old needles of Pinus sylvestris L. Scots pine from representative research catchments of the national network of Integrated Monitoring of the Natural Environment and from selected locations with specific urban, industrial and transport pressure. The system enhances pollutant classification by incorporating SEM imagery combined with EDS for chemical characterization. The microscopy images were processed in a graphics programme, so that the particles classified based on their size, shape and chemical composition had the same attribute (colour). The layers (masks) used were an appropriate element to develop a machine learning algorithm identifying pollutants divided into previously defined categories. The use of neural networks to build a self-learning algorithm allowed to optimise the analysis of deposited pollutants imaged on the surface of pine needles. The developed system for identifying natural and anthropogenic particles in the form of categorised layers provides a high-level prediction efficiency. This study aims to bridge the gap between traditional bioindication techniques and AI-powered environmental assessment, thus contributing to academic research on machine learning applications in air pollution monitoring. |
|---|---|
| ISSN: | 1944-3986 |