Related Experiment Video
Updated: Jun 15, 2026
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
9.6K
Advancing Air Pollution Exposure Models with Open-Vocabulary Object Detection and Semantic Segmentation of
Zhendong Yuan1, Jules Kerckhoffs1, Pi-I Debby Lin2
1Institute for Risk Assessment Sciences, Utrecht University, Utrecht 3584 CM, Netherlands.
Environmental Science & Technology
|September 27, 2025
Summary
Street-view images enhance urban air pollution mapping by adding visual data to land use regression (LUR) models. This improves accuracy in predicting nitrogen dioxide (NO2), black carbon (BC), and ultrafine particles (UFP) at a hyperlocal level.
Area of Science:
- Environmental Science
- Urban Planning
- Geospatial Analysis
Background:
- Mobile monitoring and Land Use Regression (LUR) models are effective for urban air pollution mapping.
- Traditional LUR models often lack detailed built environment and emission source data.
- Street-view imagery offers a rich source of visual information for environmental modeling.
Purpose of the Study:
- To develop a framework integrating street-view image features into LUR models for enhanced air pollution mapping.
- To assess the impact of visual features on the accuracy of nitrogen dioxide (NO2), black carbon (BC), and ultrafine particle (UFP) predictions.
- To identify novel object-level and segmentation-level predictors from street-view data.
Main Methods:
- Developed a framework combining object-level and segmentation-level visual features from street-view images.
- Integrated these visual features into stepwise regression and random-forest-based LUR models.
- Utilized 5.7 million mobile air pollution measurements and 0.37 million street-view images in Amsterdam.
Main Results:
- Incorporating street-view images improved LUR model performance, increasing R² by 0.01-0.05 and reducing MAE by 0.7-10.3%.
- Visual features demonstrated stability across different years and seasons.
- Identified previously unrecognized predictors like chimneys, traffic lights, and shops, contributing 8-18% feature importance.
Conclusions:
- Street-view imagery significantly enhances the accuracy of hyperlocal air pollution mapping.
- Visual data provides valuable insights into the built environment and emission sources.
- This approach offers a promising method for improved air quality assessment and exposure analysis.
