Content area
Full text
Abstract: Traffic data of multiple vehicle types are important for pavement design, traffic operations and traffic control. A new video-based traffic data collection system for multiple vehicle types is developed. By tracking and classifying every passing vehicle under mixed traffic conditions, the type and speed of every passing vehicle are recognised. Finally, the flows and mean speeds of multiple vehicle types are output. A colour image-based adaptive background subtraction is proposed to obtain more accurate vehicle objects, and a series of processes like shadow removal and setting road detection region are used to improve the system robustness. In order to improve the accuracy of vehicle counting, the cross-lane vehicles are detected and repeated counting for one vehicle is avoided. In order to reduce the classification errors, the space ratio of the blob and data fusion are used to reduce the classification errors caused by vehicle occlusions. This system was tested under four different weather conditions. The accuracy of vehicle counting was 97.4% and the error of vehicle classification was 8.3%. The correlation coefficient of speeds detected by this system and radar gun was 0.898 and the mean absolute error of speed detection by this system was only 2.3 km/h.
(ProQuest: ... denotes formulae omitted.)
1 Introduction
The proportion of large vehicles (LVs) affects the interactions among vehicles and the efficiency of roadway. Highway capacity manual [1] requires LV data for the analysis of traffic conditions, traffic safety and pavement design. Therefore traffic data of multiple vehicle types are important for the research of traffic flow.
Owing to the complexity of the mixed traffic conditions, induction loops are not useful to collect such data. Most of existing video-based detectors can only provide several macroscopic traffic parameters such as flow and mean speed without vehicle classification. Many of these detectors are sensitive to road reflections, illumination changes, shadow interferences etc. These detectors always ignore cross-lane vehicles and may repeatedly count one vehicle, which influences their accuracy of vehicle counting. Their accuracy of classification is obviously affected by vehicle occlusions. Therefore no effective tool is available to obtain traffic data of multiple vehicle types.
A new video-based traf fi c data collection system for multiple vehicle types is developed in this paper. This system tracks every...