Crosswalk Marking Changes, Traffic Volume, and Pedestrian and Bicyclist Safety: An Exposure-Normalized Modeling Study
ASCE International Conference on Transportation and Development, ICTD 2026, Michigan, Amerika Birleşik Devletleri, 28 Haziran - 01 Temmuz 2026, ss.49-62, (Tam Metin Bildiri)
- Yayın Türü: Bildiri / Tam Metin Bildiri
- Doi Numarası: 10.1061/9780784487020.005
- Basıldığı Şehir: Michigan
- Basıldığı Ülke: Amerika Birleşik Devletleri
- Sayfa Sayıları: ss.49-62
- Yıldız Teknik Üniversitesi Adresli: Evet
Özet
This study links AI-derived crosswalk marking quality from aerial orthomosaic imagery (2020 and 2023) with pedestrian crashes (2020-2023) and traffic volume (AADT). Intersections are classified by the minimum-direction change in marking quality as Faded, No Change, and Improved. Methodologically, 2023 crash frequency is estimated using Poisson Pseudo-Maximum Likelihood (PPML) with a log(AADT10k) offset (i.e., modeling crash rates), with a GLM Negative Binomial as a robustness check; injury severity is analyzed via an ordered logit, and KSI risk via a binary logit, with predicted probabilities reported at AADT quantiles. Results show that, after controlling for traffic volume, prior crashes, and geometry, intersections with improved crosswalk markings are associated with lower crash frequency and a more favorable severity distribution, whereas faded crosswalk locations do not differ significantly from No Change; traffic volume is positively related to more severe outcomes but with modest magnitude, and these directions are consistent across the alternative count specification. Overall, imagery/AI-based marking-quality indicators effectively support network-level safety assessment and investment prioritization, suggesting that prioritizing crosswalk maintenance/refresh at high-volume intersections is a practical near-term safety action.