Seeing Green: Measuring Pedestrian-Level Street Greenery with Mapillary Imagery
Introduction: Why Street Greenery Matters
Urban green infrastructure, such as street trees, parks, and green corridors, plays a vital role in promoting environmental sustainability, public health, and urban aesthetics. However, how pedestrians actually perceive greenery at the street level remains an understudied question. Many existing studies rely on top-down spatial analysis or car-based imagery, which may not accurately capture the pedestrian experience.
In our recently published study, Measuring pedestrian-level street greenery visibility through space syntax and crowdsourced imagery: A case study in London, UK, we address this gap by integrating Visibility Graph Analysis (VGA) from space syntax and the novel Pedestrian Green View Index (PGVI), calculated using crowdsourced street imagery from Mapillary. This innovative approach allows us to assess street greenery from a dual perspective: geometric visibility and human-scale perception.
A map of the study area, the city of London. Roads are symbolized in shades of green--the darker the green, the higher the percentage of street-side greenery.
The Challenge: Measuring Green Visibility Accurately
Most urban greenery assessments use car-based street view imagery (e.g., Google Street View) to calculate the Green View Index (GVI). However, these images are captured from a driver’s perspective, often missing crucial sidewalk greenery or distorting its visibility due to factors like road width, parked vehicles, or seasonal variations.
Our study aimed to move beyond vehicle-based assessments and develop a method that better reflects what pedestrians actually see. We combined:
- Space Syntax’s Visibility Graph Analysis (VGA): A computational model that evaluates urban visibility based on spatial configurations.
- Pedestrian Green View Index (PGVI): A greenery visibility metric computed using crowdsourced street-level images from Mapillary, ensuring a more realistic representation of pedestrian perspectives.
- Volunteer Surveys: We engaged 183 participants to validate our findings and compare computational models with actual human perception.
Overview of methodological framework integrating street-level imagery, Space Syntax, and perception surveys
Using Mapillary to Capture the Pedestrian Experience
Unlike conventional car-based imagery, Mapillary provides pedestrian-level images, making it an invaluable data source for our research. We utilized Mapillary to:
- ✅ Collect thousands of street-level images across diverse London streetscapes.
- ✅ Capture sidewalk greenery, which is often invisible in car-based images.
- ✅ Improve data granularity by including cycle paths, pedestrian-only streets, and narrow alleys.
- ✅ Compare vehicle-based GVI with pedestrian-based PGVI to identify discrepancies in green visibility.
Through deep learning-based image segmentation, we extracted green pixels from Mapillary images to calculate PGVI scores, offering a more accurate measurement of how much greenery pedestrians truly see.
Comparison of GVI and PGVI segmentation results across street types. PGVI, using pedestrian-eye-level images, captures greenery more accurately than GVI—highlighted by the stark contrast in the T5 examples, where PGVI reflects a much greener pedestrian experience than vehicle-based views suggest.
Key Findings: How Pedestrians Perceive Greenery
Our study revealed several critical insights into urban greenery visibility:
- 📌 PGVI (Mapillary-based) correlates more strongly with human perception than traditional GVI. Car-based GVI often underestimates pedestrian greenery due to obstructed views.
- 📌 VGA provides useful structural insights but lacks human-centric accuracy. While space syntax helps understand potential visibility zones, it cannot replace real-world pedestrian experience data.
- 📌 Safety and visual transparency are key factors in how pedestrians perceive green spaces. Streets with higher greenery visibility tend to be rated as more inviting and comfortable for walking.
Illustration of nine street greenery types categorized in this study, based on and adapted from existing street typologies to fit London’s urban context. These types range from no greenery (T1, T2) to complex combinations of grass, shrubs, and trees (T9), providing a basis for systematic comparison of greenery visibility.
Implications for Urban Planning and Design
Urban designers and planners can leverage these insights to:
- 🏙️ Improve walkability by prioritizing pedestrian-scale greenery, rather than just roadside trees seen from cars.
- 🌱 Use Mapillary imagery to conduct large-scale greenery visibility audits, ensuring a data-driven approach to greening urban streets.
- 🚶♂️ Ensure both aesthetic and functional benefits of greenery by strategically placing shrubs and trees to maximize visibility.
- 📊 Combine computational modeling (VGA) with real-world perception data (PGVI) to create more inclusive, human-centric green infrastructure.
Visibility Graph Analysis (VGA) results for the City of London, showing how much pedestrians can see from different points along streets. Red and orange areas are highly visible, while blue and green areas are more visually enclosed. The detailed views highlight how buildings and trees shape pedestrian visibility.
Next Steps and Future Research
While this study has provided valuable insights, there are still challenges to address:
- 🔹 How does greenery visibility change with seasons? Future research could incorporate seasonal Mapillary imagery to track year-round changes in green perception.
- 🔹 How does urban morphology influence green visibility? Examining different street layouts and urban forms could refine our models further.
- 🔹 How do people interact with visible greenery? Integrating behavioral studies (e.g., pedestrian movement tracking) could enhance our understanding of how greenery impacts urban life.
Conclusion: A More Pedestrian-Friendly Future
Our research demonstrates that street greenery should be measured and designed from a pedestrian’s perspective, not just from above or from cars. By leveraging crowdsourced street imagery from Mapillary, we can create more accurate, scalable, and inclusive methods to evaluate and enhance urban greenery.
As cities worldwide strive for sustainability and livability, data-driven urban planning powered by crowdsourced geospatial data will be crucial. With Mapillary’s rich street-level imagery, urban designers now have an unparalleled resource to ensure greener, healthier, and more pedestrian-friendly cities.
Additional Resources
📄 Read the full paper here: https://doi.org/10.1016/j.ufug.2025.128725
/Mingze Chen
Ph.D. Candidate, Urban Nature Design Research Lab, University of British Columbia; Founder of Nature AI
/Yuxuan Liu
Master of Architecture, University College London