Traffic Counting Using Existing Video Detection Cameras : Final Report

2016Final report 566PDFTech Reporthttp://www.ltrc.lsu.edu/pdf/2016/FR_566.pdfFHWA/LA.15/566LTRC Project Number: 12-1SS SIO Number: 30000760Data qualityLogistic regression analysisSignalized intersectionsT testTraffic countingVideo camerasHighwaysAutoscope (video vehicle detection system)operations and traffic managementvideo detection systemNaztecsolo terra autoscopesadvanced traffic management centerCentracsResearch HubUnited StatesLouisianaMississippi State University. National Center for Intermodal Transportation for Economic CompetitivenessIshak, Sherif S.Codjoe, JuliusMousa, SalehJenkins, SyndneyBonnette, JenniferLouisiana State University (Baton Rouge, La.). Dept. of Civil and Environmental EngineeringLouisiana. Dept. of Transportation and DevelopmentMississippi State University. National Center for Intermodal Transportation for Economic CompetitivenessUnited States. Federal Highway AdministrationUS Transportation CollectionThe purpose of this study is to evaluate the video detection technologies currently adopted by the city of Baton Rouge and the Louisiana Department of Transportation and Development. The main objective is to review the performance of Econolite Autoscope cameras in terms of their ability to detect data, ease of use, accessibility to data, security issues and cost. The final goal of this project is to investigate the effectiveness of this video detection technology in traffic data collection at signalized intersections in Baton Rouge and to judge the reliability of integrating the traffic count data from the Autoscopes into a database that could be used to supplement traffic count information at any time. In order to accomplish these tasks, a sample of intersections was selected for analysis from an inventory detailing each site\u2019s traffic volume, lighting conditions, turning movements, camera mounting type, technology used, and geometric characteristics. Volume counts from the video detection technology (camera counts) were statistically compared against ground truth data (manual counts) by means of Multiple Logistic Regression and t-tests. Using these data, the capabilities of the existing video detection system was assessed to determine the quality of the data collected under various settings. The results of this research indicate that the performance of the Solo Terra Autoscopes was not consistent across the sample. Of the 20 intersections sampled, eight locations (40%) proved to show significant statistical differences between the camera and manual counts. The results of the regression analysis showed only lane configuration, time of day, and actual traffic volumes were statistically affecting the performance of the Autoscopes. According to supplemental t-test analysis on the time of day, the least accurate counts were recorded during the morning and afternoon peak hours and late at night. When testing based on traffic volume, the camera performance worsened as the traffic volume increased; when considering lane configuration, there were statistical differences for the through lanes, right lanes, and shared right/through lanes. Due to the fact that 60% of the sampled intersections (the remaining 12 out of the 20) provided reliable performance under high traffic volumes and during the same study period and weather conditions, the research team attributed the poor performance of some of the cameras to poor calibration and maintenance of the system. It was concluded that the recalibration of the Econolite Autoscopes can significantly enhance the performance of the video detection system, and it can therefore be considered a reliable means for traffic counting