@article{Zhang2026, 
author = {Yan Zhang and Haoran Ma and Mei-Po Kwan},
title = {How drivers’ depth perception of environmental features influences traffic speed},
year = {2026},
journal = {Geo-Spatial Information Science},
volume = {29},
number = {4},
pages = {2990-3006},
keywords = {Driving speed, traffic environments, depth estimate, street view image, explainable machine learning, GeoAI},
url = {https://www.sciopen.com/article/10.1080/10095020.2025.2584862},
doi = {10.1080/10095020.2025.2584862},
abstract = {Drivers’ perception of the traffic environment, particularly their depth perception of surrounding elements, has a significant influence on their driving speed. While street view images (SVIs) capture environmental features from a driver’s perspective (such as trees, buildings, and pedestrians), the perceived distance of these elements from the driver (their depth value) plays a critical role in speed regulation that has been understudied. The same features at different distances can create varying levels of psychological pressure on drivers, with closer objects potentially inducing greater caution and lower speeds. To address this research gap, we used a monocular depth estimation model for simulating drivers’ perceptions and explored their impact on driving behavior. Specifically, we characterized the traffic conditions across 5458 road segments in Wuhan based on 5 million taxi GPS records. Next, we calculated the depth value from 14,115 panoramic street view images. Using the interpretable machine learning method, we map the spatial distribution of perceived environmental features associated with both low and high driving speeds. Our approach could explain 38.1% of driving speed variations. The results reveal a threshold effect between perceived environmental depth and driving speed, clarifying the non-linear relationship between drivers’ subjective perception of spatial distances and their speed choices.}
}