A novel non-payment vehicle searching method for multilane-free-flow electronic-toll-collection systems

Since the heaviest traffic congestion on toll highways occurs near toll gates where vehicles make a short stop to pay the toll, an electronic toll collection (ETC) system is usually built to eliminate the traffic jams. In order to find out the non-payment vehicles, the violation enforcement usually includes cameras to capture images of license plates, and a license plate reader system to recode photographs and license plate numbers of all vehicles. Thus, automatic license plate recognition (ALPR) technology is often used in violation enforcement. However, the identification precision of ALPR is not always reliable. Human review and correction will be needed to improve the accuracy and therefore will result in extra manual operation cost. In this paper, we formulate the non-payment vehicle searching problem into a bipartite graph matching problem and propose a Photograph-to-Transaction matching algorithm (PT algorithm) without recognizing all license plate images for multilane-free-flow ETC systems. The PT algorithm not only can reduce the human loading to review and correct the image recognition results but also can accurately identify all non-payment vehicles. The performance of the PT algorithm was evaluated in ns-2 simulator and three different traffic scenarios: congested traffic, normal traffic and sparse traffic. Besides, we also propose two methods to simplify the bipartite graph, one based on the location of transaction data relative to enforcement line, and another based on the lane location where images are captured by cameras. The simulation results show that our algorithm greatly reduce the number of plate recognitions, and is more feasible and reliable for ETC enforcement. This will activate some consequent activities against the violation vehicles.