Research On Vehicle License Plate Localization

The increase of traffic as a result of economy development makes greater demands on traffic control and management. Automatic Vehicle Identification (AVI) technology is used widely in Advanced Traffic Management System (ATMS). Automatic Vehicle License Plate Recognition System (LPRS) is one kind of AVI technology, and License Plate localization is a key step in LPRS.This dissertation focuses on locating the license plate in an image. It divides the procedure into two parts, first find an approximate location of the license plate, then process this part of the image and find a more accurate location in it.It first introduces the fundament concept in the LPRS, the structure of LPRS, the research status in this area, its basic theory and methods, and explains the specifications and characteristics of the license plates currently used in China. After that, based on the analysis of peak-valley searching localization method and line-difference localization method, it proposes a new difference-based localization method. Tests show the accuracy of this new method is higher than the previous two. And it also discusses another locating method base on the texture analysis and proposes this method to be a complement to the first one. Thus we can get an approximate image containing the license plate.Based on the analysis of weak points of traditional binarization method, this dissertation propose a new one by analyzing the histogram of central area in the approximate image. We design an algorithm to correct slanting license plate using edge detection and Radon transform, and an algorithm of locating the top-bottom, right-left of license plate accurately. At last, we test the algorithm using more than 1500 pictures and the results indicate the algorithm having a high localization rate and less computer running times.