ALPR - Extension to Traditional Plate Recognition Methods

Automatic license plate recognition (ALPR) methods and software are used in toll collection, traffic monitoring and other areas of road transport industry. Majority of ALPR methods and almost all in industrial use, try to recognize a license plate identifier from a single image. However, in a sequence of images, recognition of a license plate in any frame can be improved by considering the information from preceding and succeeding frames, using video object tracking. A new approach is presented, for combining a video tracking and a single frame ALPR method to improve the recognition rate. Unlike earlier techniques which are tied to specific object tracking and identifier recognition methods, the new method can be used with almost any tracking and single frame ALPR methods. Its key part is a method for clustering and alignment of candidate license plate identifiers in a video track. The results from five video sequences taken from a surveillance camera under various weather and light conditions demonstrate the recognition rate improvements.