Background and Purpose
Over 25% of adults in the US are obese,1 contributing to 300,000 deaths and costing the US healthcare system $147 billion annually.2 Federal governments to local non-profit agencies have instituted policy and built environment (BE) changes in effort to reduce obesity and increase physical activity (PA). A challenge in evaluating the success of policy and BE change is the capacity to capture a priori PA behaviors and the ability to eliminate researcher and respondent bias in assessing post-change environments. A novel collaboration between public health and computer science is presented here with the goal of automatically analyzing existing public data feeds in innovative ways to quantify BE intervention effectiveness.
Objectives
The Archive of Many Outdoor Scenes (AMOS) has collected over 225 million images of outdoor environments from more than 12,000 public webcams since 2006.3 AMOS uses publicly available webcams and a custom web crawler (similar to web search engine or Google) to capture webcam images with a time stamp, capturing one photographic image per camera each half hour.4 Many of these environments have experienced BE improvements or policy change (e.g., complete street policies, bike shares, and walking school bus programs). AMOS provides a unique opportunity to measure BE change and associated behavioral modification.Washington, DC, provides an excellent case study due to the ubiquity of webcams and recent efforts to increase active transportation and PA including the addition of bike lanes, curb bulbs, painted crosswalks, and a bike share program. Available webcams have been spatially matched with specific BE changes. The present objective was to use innovative crowd-sourcing methods to capture the prevalence of pedestrians and cyclists prior to and after the BE changes at two distinct intersections in Washington, DC. Counts of active transportation pre and post-changes allow for the unbiased and retrospective evaluation of the BE changes.
Methods
Two AMOS webcams (to date) have captured example BE change in Washington, DC; at Pennsylvania and 9th St. NW (Intersection One; a new bicycle lane) and 7th St. and Independence Ave SW (Intersection Two; new curb bulbs and painted crosswalk). Using the AMOS dataset, all 120 webcam photographs were captured between 7:00am and 7:00pm during the first work week of June prior to the BE change and the 120 photographs were captured from the same week of June following the BE change. All data is from 2007-2010. The use of this captured webcam data allowed for a pre-BE change and post-BE change travel mode analysis. The Amazon Mechanical Turk (MTurk) website was used to crowd-source the image annotation.5 MTurks are simple tasks that have not yet been automated by computers. MTurk workers were paid US$0.01 in March and July 2012, to mark each pedestrian, cyclist, and vehicle in a photograph. Each image was counted 5 unique times (n=600 per week, per camera), a process completed in under four hours. Counts per transportation mode were downloaded to SPSSv.19 for analysis. The odds of observing each transportation mode in year two compared to year one was examined. Recent research has revealed MTurk workers to be reliable.6, 7
Results
The odds of the webcam at Intersection One (bike lane) capturing a cyclist present in the scene post-BE change increased 3.5 times, compared to pre-change (OR=3.57, p<0.001). The number of cyclists per scene increased four-fold between pre (mean=0.03; SD=0.20) and post (0.14; 0.90; F=36.72, 1198; p=0.002). There was not a significant increase in pedestrians.The odds of the webcam at Intersection Two (curb bulbs and painted crosswalk) capturing a pedestrian in the scene post-BE change increased by 33% (OR=1.33, p<0.001). The number of pedestrians increased by over 50% between pre (mean=2.12; SD=2.59) and post-BE change (3.41; 3.64; F=46.13, 1198, p<0.001) There were not increased odds of seeing a cyclist (presence / absence) between the two years, however there was a significant increase in the number of cyclists captured per scene between the two years (0.06 and 0.14, respectively; F=12.62, 1198; p<0,001).
Conclusions
Findings suggest publicly available web data feeds and crowd-sourcing have great potential for capturing behavioral change associated with BEs. The addition of bike lanes at Pennsylvania and 9th was associated with a significant four-fold increase in the number of cyclists captured per scene. The addition of curb bulbs and a painted crosswalk at 7th and Independence was associated with a 33% increase in pedestrians per scene.This initial, novel research presents an unobtrusive surveillance of the effectiveness of a PA policy and BE intervention. The use of public webcam scenes and MTurks offer an inexpensive (US$12.00/scene) means for public health, urban design, and governments to evaluate effectiveness of BE policy change and interventions. Future research includes testing with additional bike lane and BE change scenes in Washington, DC, and expanding to include parks and safe routes to schools. Current efforts are also focused on the utilization of computer algorithms to automate the counting of transportation modes per scene.
Support / Funding Source
The authors would like to acknowledge the Washington University in St. Louis University Research Strategic Alliance for providing the funding for this cross-university research.
