A Comparative Study of Pedestrian Crossing Behavior and Safety in Baltimore and Washington, D.C., Using Video Surveillance

Abstract

There is a national effort to reduce pedestrian deaths. To achieve this goal, a better understanding of pedestrian travel behavior and interactions with vehicles is needed. Using video surveillance of locations in Baltimore, Maryland, and Washington, D.C., this study compares pedestrian-related travel behavior in the two neighboring cities. A computer vision pipeline approach will be used to identify pedestrians and vehicles from video surveillance footage in order to extract key metrics such as walking speed, gap acceptance, and type of unsafe maneuvers, to characterize pedestrian crossing behavior and associated traffic patterns. Statistical analyses of these metrics will determine which factors - such as land use, infrastructure, and sociodemographic characteristics - contribute to pedestrian travel behavior decisions and safety.

Universities and Sponsoring Organizations Involved

Morgan State University

University of Maryland

U.S. DOT Office of the Secretary/Research

Principal Investigators

Dr. Celeste Chavis, MSU, celeste.chavis@morgan.edu

Kofi Nyarko (MSU), kofi.nyarko@morgan.edu

Cinzia Cirillo (UMD), ccirillo@umd.edu

Funding Sources and Amounts

USDOT: 120,000, Morgan: $40,000 (Match), UMD: $20,000 (Match)

Start Date

9/1/2020

Expected Completion Date

8/31/2020

Expected Research Outcomes

The following outcomes will be generated from this work:
• A video-based tracking of pedestrians and vehicles using a computer vision pipeline approach
• A detailed database of pedestrian movement along a variety of facilities in two major cities: Baltimore, Maryland, and Washington, D.C.
• A pedestrian behavior and safety analysis at each location with provided recommendations

Expected Equity Impacts and Benefits of Implementation

Pedestrians are an important component of urban mobility. Though there is an understanding of the variability of pedestrians by age, gender, and mobility, there is a lack of understanding of the geographic and temporal variations in pedestrian-related travel behavior. The results of this study can inform policy and infrastructure decisions to improve the safety of pedestrians, roadways' most vulnerable user group. The pedestrian and vehicle tracking model has many future real-time applications. Furthermore, the results of this research can improve microsimulation calibration and inform dynamic traffic signalization.

Subject Areas

Safety, travel behavior, pedestrians, computer visioning