License Plate Location And Recognition Algorithm Based On Real-time Video Stream

Intelligent transportation system is an automatic transportation managing system which is based on the computer science technology, communication technology, and electronic engineering technology. It takes the huge modification about the national transportation management, the using of the transportation resource, the progress of the transportation. Vehicle license plate location and recognition system is the core of the intelligent transportation system, and it is based on the computer image process and analysis and pattern recognition. It provides the most important function, and it implements the auto location and recognition of vehicle license plate and provides the transportation management and service. Research work about the vehicle license plate location and recognition algorithm which is based on the real-time video stream is presented in this paper and its destination is to satisfy the requirement of the real environment.In the license plate location and recognition based on real-time video stream, it is necessary to locate and recognize the plate in real time, but localization is badly impacted by the motion-blurred image, illumination, shadow. Through the research about the video stream, this paper presents a fast vehicle license plate location method. By effective image pre-process, we get the figure structure feature of the character and the horizontal cross-points statistic feature. Based on these features, we locate the vehicle license plate.By setting the video detection window, we accelerate the image process. By image pre-process, and mathematical morphology operation and the change feature of the figure of license plate, we locate the license plate accurately. The presented method can improve the effect of motion-blur and environment and is fast and robust. In the experiment, the license plate localization rate achieves 92%, and the average time of localization is less than 40ms.By the analysis of the projection and connected domain of the vehicle license plate image, we separate the character and normalize the character. Through researching the traditional template match algorithm, we supply a quick template match algorithm based on compressing the image pixel. This algorithm pick-up a new template which is based on pixel compressed according the feature of the image gray and structure. We use the compressed template to match and recognize the data. By effectively picking up the feature, the algorithm reduces the operation times, and accelerates the speed of template matching. In the experiment, the correct recognition rate of the letter and number achieves 96%, but the correct recognition rate of the character only achieves 79%. The average recognition time of a vehicle license plate is only 5 ms.In conclusion, under the environment of real-time video stream, this paper presents a fast vehicle license plate location algorithm based on the horizontal cross-points statistic feature, and presents a new template matching algorithm based on the image pixel compression. The vehicle license plate location and recognition system costs less than 55ms to finish the whole process, and it fulfils the requirement of the real-time environment.