Large Field Test of New Stereo Detection System for the Pedestrian Signal Phase for the Visually Impaired

2012Gibson, DavidPDFTech ReportFHWA-HRT-12-074DTRT5706C10030Pedestrian actuated controllersPedestrian movementVisually impaired personsPedestrian crosswalks--United States--Safety measuresTraffic signs and signals--United States--Safety measuresPeople with visual disabilities--United States--Safety measuresUnited StatesUnited States. Federal Highway AdministrationLing, BoGibson, DavidBentzen, Billie LouiseBurton, PaulBoudreau, NeilBobinsky, MikeHoben, JimUnited States. Federal Highway AdministrationMigma Systems, Inc.United States. Department of Transportation. Research and Special Programs AdministrationTurner-Fairbank Highway Research CenterFederal Highway AdministrationIn March 2010, a large field test study was conducted on a stereo pedestrian detection system in four cities in four States under the Federal Highway Administration's Small Business Innovative Research program. The objective was to prove that the stereo vision would reliably detect pedestrians in real-world environments under all weather conditions (e.g., sunny, cloudy, foggy, rainy, and snowy) and extreme temperatures (e.g., very high temperatures in the southern part of the country and very low temperature in the northern part of the country) and determine that the system effectively actuates pedestrian calls and locator tones on accessible pedestrian signals (APSs). The study was conducted in Tucson, AZ; Somerville, MA; Portland, ME; and Manchester, NH. The stereo pedestrian detection systems were installed to evaluate their performance in complex real-world pedestrian/traffic environments under all weather conditions and extreme temperatures. A total of 17 systems, consisting of one computer and two cameras per crossing, were installed at 9 geographically disparate test sites. These were installed at both midblock crossings and at intersections. When pedestrians were detected at crosswalk ramps, the system automatically actuated both regular and APS pushbuttons to make the service request on behalf of the pedestrians. System performances were evaluated using images saved in external hard disks. Results from the testing showed that the overall positive detection rate for the automated APS actuation zones in all four cities was close to 98 percent, with the average number of false calls per day less than three. This excellent performance was confirmed through the real operations of systems installed at the test sites. Currently, all of the systems are operational 24 h a day, 7 days a week. Comments from the general public, including the blind and visually impaired community, were all positive. This system provides automated actuation of pushbuttons based on the presence of pedestrians. This provides the protection of pedestrian phases, thereby reducing the possibility of collisions between vehicles and pedestrians, preventing unnecessary fatalities, and automatically extending the walk time for seniors and wheelchair pedestrians without slowing down the traffic flow.80