Development and application of vehicle-license number recognition system using real-time image processing

The authors have proposed a system architecture and a high-speed method to extract characters (a four-digit vehicle number) on a vehicle license plate for a real-time vehicle license number recognition system. This paper describes the results of field tests of a system which is based on the previous work but with some further developments. To identify characters on a vehicle license plate, a template-matching method is used which compares the skeleton pattern of characters and its background pattern. To use the system at night, a continuous illumination with two light sources is employed. The experiments including simulations confirm the usefulness of the proposed method at night. The field tests show that about 75 percent of all passing vehicles were recognized successfully in the daytime and about 55 percent at night. The average processing time is about 150 ms, and the 98 percentile value is about 410 ms. These values are sufficient for practical measurements of a vehicle traveling time (between two points). This paper describes also a system to give traveling information to vehicle users, by employing message boards as an example of application of the proposed system.