Research Of Key Technologies On Motion Blurred License Plate Recognition System

With the progress of the science and technology,the development of the transportation is developed.It makes the intelligent transportation system come into being.As the key technology of intelligent transportation system,license plate recognition is of great significance for promoting the process of intelligent transportation.In the real life,the license plate images are often blurred by the motion because of the high-speed motion.Motion blurred images bring great difficulties to the license plate recognition.Motion blurred vehicle license plate recognition system can deal with motion blurred images effectively.With the increase of traffic pressure,a motion blurred vehicle license recognition system with high recognition rate is urgently needed.In order to solve the above problem,the key technologies of motion blurred vehicle license plate recognition system are studied in this paper.The main research work in this paper is as follows:First of all,we study the restoration method of motion blurred license plate image.We use Wiener filtering to restore images.The key to restore image by Wiener filtering is to get K value and the parameter of the point spread function.In this paper,we use a combination algorithm based on frequency domain feature algorithm and airspace feature algorithm to obtain the parameters of the degradation function.First,we detect the motion blurred angle by Hough transform algorithm,and then we estimate the motion blur length by using the differential autocorrelation function curve.We proposes an algorithm to automatically estimate K value in this paper.Secondly,we use PSO-BP neural network to recognize the restored license plate image.The link of license plate recognition includes license plate location,license plate segmentation and license plate recognition.In the link of the license plate location,we use a combination algorithm based on pixel projection algorithm,edge detection algorithm and mathematical morphology algorithm to locate license plate.After the license plate location,we use Radon transform to achieve tilt correction of the license plate image.In the link of license plate segmentation,we use the projection algorithm to remove the frame and the segmentation characters of the license plate.Next,we propose a combination algorithm based on region connectivity analysis algorithm,vertical projection algorithm and template matching algorithm to solve the problems of character connectivity and character adhesion in the license plate segmentation.In the link of license plate recognition,we propose to use PSO-BP neural network to recognize vehicle license plates.This method solves the problem of slow convergence rate when we train the BP neural network.It improves the recognition rate of the system.Finally,we simulate the key technology of motion blurred license plate recognition system and we analyze and explain the improved algorithm in detail.The simulation results show that the proposed algorithm has an effect in restoring motion blurred license plate images and the license plate recognition rate of the whole system is improved at the same time.The algorithm proposed in this paper has a certain social application value.