A study of automated detection technologies was undertaken as part of the Massachusetts Highway Department (MassHighway) Research Program. The objective of this research was to identify and evaluate existing technologies that may accurately and efficiently detect, count, and classify non-motorized modes of transportation (i.e., pedestrians and bicyclists). In addition to accuracy and efficiency, other criteria considered included: applicability to both on-road and off-road locations; flexibility in detecting and classifying non-motorized activity under multiple conditions; portability; and cost effectiveness. The research process began by identifying detection technologies currently used in the transportation industry. Microwave, ultrasonic, acoustic, video image processing, piezoelectric, passive infrared, active infrared, magnetic, and traditional (inductive loops and pneumatic traffic classifiers) were considered. The research team selected active infrared for further analysis. An Autosense II Active Infrared Imaging sensor was purchased and evaluated. The experiment was conducted during the summer and fall of 2001. The Autosense II device was frame-mounted 18 feet above the selected observation location and connected to a desktop computer. Data consisted of the manually collected trail user volumes, separated into pedestrian and bicyclist volumes, and the detection and classification data from Autosense II. The data obtained from Autosense II were compared with the manual counts to evaluate the performance of the device in detection and classification. The results showed that Autosense II was very effective in both detection and classification of bicyclists and the detection of pedestrians. Ninety-seven percent of the bicyclists observed were accurately detected. Classification of bicyclists was less accurate as on 77% of the bicyclists detected were classified (as motorcycles). Ninety-two percent of pedestrians observed were successfully detected; however, no pedestrians were classified correctly since the algorithms were not designed for this function. Nevertheless, nearly all observations classified as "unknown" were pedestrians. The results of this research indicate that none of the market-available intelligent transportation systems (ITS) devices are effective at both pedestrian and bicyclist detection and classification. Nevertheless, active infrared is a technology with the capability of pedestrian and bicycle detection and classification with some modifications.
Abstract