Research Based On HD Image Multi-License Plate Recognition System

License Plate Recognition System(LPRS) is an important part of the Intelligent Transportation System(ITS). It has been widely used in the electronic toll collection, traffic control and parking intelligent management, so researchers have carried out many studies on LPRS both at home and abroad. However, most of researches have only located a single vehicle recognition processing in the surveillance area. The paper combines the application of LPRS and the situation of multi-lane in the current monitoring highway and uses HD cameras to capture the video stream on the road, in order to recognize license plates for multi-lane vehicles.The main focuses and innovations of this paper are summarized as follows:1. License plate location module. License plate location is the key of LPRS. First, pre-process a series of vehicle images which HD camera collects, then make use of the cursory scan, the initial location and precise location of the three progressive steps to locate multi-lane license plates. Compared to use single feature, the paper combines the edge scan, jump and color features of license plates. The experimental results show that the method can locate plate clearly and accurately.2. Character segmentation module. First, character segmentation pre-processes. License plate tilt correction is the key of this part. According to the correct criteria to judge if the license plate need correct, and propose a complementary tilt plate correction method to correct it. Most of studies only deal with the tilt license plate under specific situation, while this method can accurately correct the tilted plates in various situations on the road. Expand the correction range and improve the correction quality. Then, the paper combines with the jumping character and the priori knowledge of the characters, and then do separately level segmentation and vertical projection to license plates, which can realize the accurately segmentation for license plate.3. Character recognition module. This module proposes character recognition algorithm and process of characters, letters and numbers. Most of all, character feature extraction is focus, through PCA to take down the dimension of the texture information. Secondary identification is necessary to the confusing characters, in order to improve the overall recognition accuracy.The LPRS has achieved under the software of Visual C++, experimental results show that the method of this paper can locate license plate accurately on the multi-license plates, system identification accuracy is higher. The algorithm of each modules for the LPRS is easy, it can meet the general requirements and the accuracy of the LPRS.