A Real-Time Vehicle License Plate Recognition (LPR) System

INTRODUCTION

            License plate recognition (LPR) is an image-processing technology used to identify vehicles by their license plates. This technology is gaining popularity in security and traffic installation this thesis presents a license plate recognition system as an application of computer vision. Computer vision is a process of using a computer to extract high level information from a digital image.

            There is a need for intelligent traffic management systems in order to cope with the constantly increasing traffic on today’s roads. Video based traffic surveillance is an important part of such installations. Information about current situations can be automatically extracted by image processing algorithms. Beside vehicle detection and tracking, identification via license plate recognition is important for a variety of applications. These include, e.g. automatic congestion charge systems, access control, tracing of stolen cars, or identification of dangerous drivers. Deploying smart cameras for the purpose of video based traffic surveillance has the advantage of allowing direct on site image processing tasks. Most license plate recognition systems in operation today use special hardware like high resolution cameras or infrared sensors to improve the input quality and they operate in controlled settings. A different solution, as proposed in this work, is the continuous analysis and consideration of subsequent frames. However, this implicates that enough frames are captured by the capturing device and are processed by the processing engine.

            We state that for the tasks of car detection and subsequent license plate recognition real-time is a flexible term. As about 20 fps (frames per second) might be enough real-time for this tasks when cars are driving at residential speeds, this is insufficient for country roads and highway traffic. In our terminology, real-time operation stands for fast enough operation in order to not miss a single object that moves through the scene, irrespective of the object speed.

            The major advantages of system over all others are its real-time capabilities in city scenarios and its ability to operate under daytime conditions with sufficient daylight or artificial light from street lamps. The usage of active infrared light is popular because the light is reflected by the license plates only. By using a camera and special filters, the detection of the plates and subsequent character segmentation is relatively easy. The major reason focused on cars driving at residential speeds is the lack of a digital camera with a high speed shutter. The drawback of using standard TV cameras is that motion blur is a critical issue, so that reasonable character recognition becomes almost impossible when vehicles are moving at high speeds. Note that this is rather a restriction of the image acquisition device than a deficit of the system. Our system runs fully autonomous and embedded on a smart camera, and thus makes integration into an existing system, e.g. for access control of parking garages, possible.