Automated framework for extracting sidewalk dimensions from images using deep learning

Ayman Halabya, Khaled El-Rayes

Research output: Contribution to journalArticlepeer-review

Abstract

State and local governments are required by federal and state laws to provide and maintain accessibility on their sidewalks and pedestrian facilities. They need to conduct and frequently update self-evaluation to assess the compliance of their sidewalks and pedestrian facilities with accessibility requirements and identify any barriers that limit or deny access for people with disabilities to public programs, services, or activities. This paper presents the development of an automated framework that is capable of (1) providing a cost-effective and practical methodology for conducting self-evaluations using sidewalk images, (2) creating 3D models of existing sidewalks that can be used in analyzing their conditions, and (3) automatically extracting sidewalk dimensions and geometry from sidewalk input images. A case study of a small pedestrian network that includes 830 m of sidewalks was analyzed to test the framework performance and demonstrate its novel and practical capabilities.

Original languageEnglish (US)
Pages (from-to)1049-1058
Number of pages10
JournalCanadian journal of civil engineering
Volume49
Issue number6
DOIs
StatePublished - 2022
Externally publishedYes

Keywords

  • Americans with disabilities Act (ADA)
  • deep learning
  • machine learning
  • neural network
  • self-evaluations
  • sidewalks

ASJC Scopus subject areas

  • Civil and Structural Engineering
  • General Environmental Science

Fingerprint

Dive into the research topics of 'Automated framework for extracting sidewalk dimensions from images using deep learning'. Together they form a unique fingerprint.

Cite this