Author(s):
D. Hartmann, D.R. Middleton, D. Morris
Publication Date
May 1997
Abstract
The research documented in this report analyzed detection capabilities of a trip-wire video image processing system in a freeway setting. Count and speed accuracy, as well as occlusion, were parameters of interest in field testing at Texas A&M University's Riverside Campus research facility. Testing analyzed three camera heights, 9.1 m (30 ft.), 12.2 m (40 ft.), and 15.1 m (49 ft. - 6 in.), in conjunction with three passenger car speeds, 32 km/h (20 mph), 72 km/h (45 mph) and 88 km/h (55 mph). The video image processing system used in the study was the Autoscope¿ 2004. The camera imaging device was a 12.4 mm (1/2 inch) interline transfer microlens charged coupled device (CCD), utilizing a 6 mm, fl.2 auto iris lens. An analysis of variance (ANOVA) test indicated that both camera height and travel lane location affected the system's ability to accurately detect passenger cars. Generally, higher camera heights and travel lanes farther from the camera produced accurate passenger car detection farther upstream from the camera, based on no traffic in other lanes closer to the camera. Also, passenger cars traveling in adjacent travel lanes did not always influence the video image processing system's ability to accurately detect passenger cars in this highly controlled environment. The paired t-test indicated that speeds determined by the video image processing system were significantly different from speeds obtained by radar. Tests at night revealed errors in counts, and daylight truck occlusion was worse than cars for all camera heights. Based on the cost information from Texas Department of Transportation, life-cycle costs of video detection are similar to the cost of detection by inductive loops where many individual loop detectors are replaced by one camera, as might occur at intersections. Motorist delay may cause a different outcome. Where fewer loops are replaced by one camera, as on freeways, the additional investment for video detection will probably not be cost effective.
Report Number:
1467-4
Keywords:
Costs, Inductive Loop Detection, Machine Vision, Video Image Processing, Wide Area Detection
Link(s):
Document/Product
http://tti.tamu.edu/documents/1467-4.pdf
Publication/Product Request
TTI reports and products are available for download at no charge. If an electronic version is not available and no instructions on how to obtain it are given, contact the TTI Library.