License plate verification method for automatic license plate recognition systems

Automatic license plate recognition (ALPR) is the technology for identification of license plate numbers from a video stream or an image without human interaction. ALPR has many applications such as registration or accounting (toll roads, checkpoints, and parking lots), search and tracking (recovery of stolen vehicles, capture of offenders, and other regulations of traffic laws), etc. There are many approaches used in contemporary ALPR systems that use different image processing techniques and machine learning algorithms for recognition. These systems display promising results of recognition on data with high resolution and quality. However, these systems encounter difficulties of recognition due to partial occlusions, not uniform illumination, dirty plates, and unwanted elements. Moreover, the types of license plates vary by different countries and sometimes even by regions. Therefore, many ALPR systems require specific approaches to attain high recognition rates. ALPR systems must also be able to make all the processing in real time because in most cases recognition time should not have influence on external processes. In this work, we consider different approaches of ALPR and provide our approach of license plate verification method which demonstrates very promising results.