Research On Vehicle License Plate Location And Segmentation In Color Images

License Plate Recognition is the core technique for Intelligent Transportation Systems. It takes an important role in information collection, vehicle monitor system, automatic fare collection system and other fields.The thesis aims at researching the following three aspects in the License Plate Recognition System: license plate location, the detection and correction of license plate and character segmentation.Vehicle license plate location is the first step in automotive vehicle license plate recognition system, and the accuracy of locating plays an important role in the system. First, we analyze the characterization of domestic license plates, and discuss recent developments in the state of license plate location based on color information. A major development tendency of pattern recognition is also shown. Then an algorithm of vehicle plate location based on gradient projection together with color area segmentation is put forward. Simulation results demonstrate our algorithm efficiency.The detection and correction of oblique license plate is one of the key technologies in automatic license plate recognition. An algorithm for oblique license plate correction is proposed based on the corner point information and inertial principal axis. Firstly, the corner points of the license plate are found using Harris algorithm; then the inertial principal axis of those corner points is found to determine the oblique angle of the plate; thus the oblique plate can be corrected. The experimental results are given; it shows that the proposed algorithm has high speed and high precision.During the course of character segmentation, we make the best of the apriority knowledge of the normal license plate. A color image segmentation method is presented based on the SRG method. Seeds can be automatically selected in the Lab color space by making use of both color differences and relative Euclidean distances in the new algorithm; the new method of SRG is used to form the initial segmented regions; and the initial regions are hierarchically merged based on the region distance defined by the color spatial and adjacent information. Then, single characters are obtained. The proposed algorithm can overcome the disadvantages of not selecting seeds automatically and leading to the result of over-segmented by the traditional SRG method.