LANE-BASED TRAVEL TIME ESTIMATION FOR MULTIPLE VEHICLE TYPES: A VEHICLE REIDENTIFICATION APPROACH

Travel time are crucial information for route guidance and traffic management. On urban roads with many lane-changing movements, different types of vehicles will have different traveling speeds and the traffic conditions of different lanes are not the same. However, little attention has been given to estimate travel times for different types of vehicles traveling on different lanes. This paper proposed a lane-based travel time estimation for multiple vehicle types through the matching of low-resolution vehicle images from surveillance cameras. Vehicle size and high-dimensional deep-learning features are extracted from vehicle images for vehicle classification. Lane-based bipartite graph matching technique, which based on matching probabilities estimated from vehicle features, are adopted to reidentify vehicles. Case study is conducted on a congested four-lane urban road in Hong Kong, and the results shows that the proposed model generally give an accurate estimation of travel times for different types of vehicles traveling on different lanes.