Development and validation of automated microscale walkability audit method. (January 2022)

• Record Type: Journal Article
• Authors: Koo, Bon Woo; Guhathakurta, Subhrajit; Botchwey, Nisha
• Is Part Of: Health & place. Volume 73(2022)
• Publication Date: 2022-01
• Subjects: Reliability -- Automated audit -- Computer vision -- Google Street View images; Health -- Social aspects -- Periodicals; Health services accessibility -- Periodicals; Public health -- Periodicals; Political planning -- Periodicals; Social medicine -- Periodicals; Epidemiology -- Periodicals; Health Policy -- Periodicals; Health Services Accessibility -- Periodicals; Public Health -- Periodicals; Public Policy -- Periodicals; Sociology, Medical -- Periodicals; Épidémiologie -- Périodiques; Politique sanitaire -- Périodiques; Santé, Services de -- Accessibilité -- Périodiques; Health services accessibility; Health -- Social aspects; Political planning; Public health; Social medicine; Periodicals; 613
• DOI: 10.1016/j.healthplace.2021.102733
• Languages: English
• ISSNs: 1353-8292

Abstract

Abstract: Measuring microscale factors of walkability has been labor-intensive and expensive. To reduce the cost, various efforts have been made including virtual audits (i.e., manual audits using street view images) and the introduction of computer vision techniques. Although studies have shown that virtual audits (i.e., manual audits using street view images) can reliably replicate in-person audits, they are still prohibitively expensive to be applied to a large geographic area. Past studies used computer vision techniques to help automate the audit process, but off-the-shelf models cannot detect some of the important microscale walkability characteristics, falling short of fully capturing the multi-facetted concept of walkability. This study is one of the earliest attempts to use the combination of custom-trained computer vision models, geographic information systems, and street view images to automatically audit a complete set of items of a validated microscale walkability audit tool. This study validates the reliability of the automated audit with virtual audit results. The automated audit results show high reliability, indicating automated audit can be a highly scalable and reliable replacement of virtual audit. Highlights: A complete set of questions of a validated walkability audit tool is automated. Street view images, computer vision, and GIS are used for automation. Validation conducted by comparing automated and manual audit results. Automated audit showed high reliability; future improvements are discussed.