A novel method for extracting license plates in complex background

This paper presents a new method of extracting license plates in complex background from rough search to accurate location based on different features and relative position. The whole process can be divided into three parts: in the first part, cars´ front or rear wheel is to be located, which is used to locate the rough position of license plates according to the relative position, in the second part, we will search the stronger density regions in the form of blocks and form final connective components. After that, a accurate search will be handled in final bounding rectangle of connective components, because the final bounding rectangle are relatively small region with a little noise or interference, in which we can use the fastest way to extract license plates candidates. Finally the license plates candidates will be verified by different ways. The experiment shows that this proposed method has less constraint, and have a fast process of extracting, plus a satisfied accuracy.