Disabled Parking CV: Scalable Methods to Analyze Disability Parking Using Computer Vision and High-Resolution Aerial and Streetscape Images

Term Start:

June 1, 2024

Term End:

May 31, 2025

Budget:

$200,000

Keywords:

Accessibility, Computer Vision, Disability Parking, Parking, Public Transportation

Thrust Area(s):

Data Modeling and Analytic Tools, Equity and Understanding User Needs

University Lead:

University of Washington

Researcher(s):

Jon E. Froehlich

People with disabilities disproportionately rely on public transportation to access employment, education, and healthcare services; however, public transit is not always available or equally distributed, which excludes social and community participation (Bascom & Christensen, 2017). Car transit is thus the only viable alternative. Since the Americans with Disability Act (ADA) of 1990, 4-8% of public parking spaces need to be reserved for drivers/passengers with disabilities (Myron et al., 2010), providing wide, accessible spaces close to destinations. And yet, there has been no systematic, large-scale study of the allocation and sizes of disability parking spaces across the US. The limited prior work that does exist has used questionnaire methods to survey disabled drivers (Lu, 2019).

In this project, we propose building and evaluating state-of-the-art computer vision (CV) methods applied to emerging high resolution aerial photography—such as the open 0.08 meter/pixel orthoimagery of Washington DC (DC Orthophoto, 2021)—to semi-automatically:

  • Track the allocation of disability parking in public and commercial lots
  • Examine characteristics of said parking (e.g., size, access area, % of allotment to normal parking) as well as public transportation ridership usage
  • Create new analytic metrics enabled by the approach such as the proximity to POIs (e.g., the distance to an entrance).

The overarching goal of this work is to create open datasets and analytics for ADA-accessible parking as well as to infuse this information into modern mapping tools (e.g., OpenStreetMaps). The proposal builds on the extensive prior work in developing novel crowdsourcing+AI techniques to infer accessibility characteristics of the built environment such as the award-winning Project Sidewalk (Saha et al., 2019) and BusStopCV (Kulkarni et al., 2023).

To validate the proposed techniques, we will select 3-5 pilot cities and neighborhoods based on characteristics such as population density, zoning, surrounding building age, and socio-economic data (subject to image availability). The research team will manually label parking lots for training and testing the CV model. Using a 70/20/10 train/validate/test set, we will report precision, recall, and f1 score. Drawing on the existing partnerships with WorldEnabled.org, Disability Rights Washington, and the center for Research and Education on Accessibility Technology and Experiences, we will share the results and interview participants who rely on disability parking about the findings. As a stretch goal, we also plan to examine on-street accessibility parking allocations via CV methods trained on streetside parking signs.

Project scope

Scroll to Top