Research On Key Technology In License Plate Recognition

License-plate recognition (LPR) is an important technology in intelligent transportation system and is widely used in electronic police, highway management and parking payment system. The accuracy of LPR system are related with the image quality. So current LPR algorithms are mainly based on large, high-resolution plate and have lower accuracy in distinguishing small, low-resolution plate.This paper first briefly introduced technology in LPR. After that, a robust LPR system with high accuracy for low-resolution plate is presented.The proposed system is composed by three subsystems:plate location, character segmentation and character recognition, and we combine these three subsystems into the one system at last.In plate location section, we introduced both grayscale and color information into mean-shift algorithm, and improved former mean-shift plate location algorithm using region analysis method. Such algorithm also has higher accuracy and fewer thresholds compared with traditional plate location algorithm.In character segmentation section, we modified traditional region-based segmentation algorithm and improved its accuracy identifying low-resolution plate character.In character recognition section, we presented a novel plate character recognition based on shape context. This algorithm is more accurate in distinguishing confusing characters than common statistical learning algorithm.Finally, we introduced the system implementation and did performance for the system.