Research On Algorithms Of Color License Plate Location And Character Segmentation

License plate recognition system (LRP) is an important part of the intelligent transportation system (ITS), it has been widely used in various transportation fields. License plate recognition system is mainly composed of three parts:plate localization, character segmentation and character recognition. The latest domestic and international researches about license plate location and character segmentation in recent years are reviewed in this paper. Based on these researches, color license plate location, binarization and character segmentation are deeply discussed and studied in this thesis. The main work and achievements are as follows:1) Several typical license plate location, binarization and character segmentation algorithms are described and analyzed, and also the comparison and evaluation of different algorithms are made.2) A method of color license plate location based on edge and color information of HIS color space is proposed in this paper. First, candidate regions of license plate are extracted from,the gray image, and then the license plate location is chose in these candidate regions based on edge-color pairs of HIS color space. The results show that:the method can greatly reduce the number of test samples to improve the processing speed; moreover it can fully use the characteristics of color information of license plates to eliminate the affect of similar rectangle plate and similar color regions of the car.3) An improved FCM algorithm based on H and I components of HIS color space is proposed for binarization of color plates. The algorithm can reduce the number of samples by combining the samples in the plate image with the same (H, I) value to improve clustering speed. The results show that:for different types of color plates, this algorithm can reduce the affect of blur and various lighting conditions, and it can get good results of color license plate binarization in short time.4) In the process of character segmentation, a method based on vertical projection and the license character features is proposed for character segmentation. The results show that this method has good robustness, it can not only get good segmentation of ideal plate images, but also be able to handle images with adhesion characters, broken characters and the special character "1" well.