Motion Blur Image Recognition Algorithm Based On Evolutionary Algorithm Research

In recent years, with the improvement of people’s living standard and the increase in China’s automobile, the government is also on the construction of infrastructure so fast. At present, with the increase of overpasses, highways and packing lots, but it lacks reasonable and effective vehicle management measures. In order to reinforce the traffic management and improve the efficiency in the use of highway, have produced all kinds of intelligent transportation system. License plate recognition system is an important part in intelligent transportation system. Now license plate recognition technology has been widely applied, but mostly in the good quality vehicle image. In practical applications, the collected vehicle images at high-speed are blurry. So it is useful to recognize high-speed blurry license plate.This paper mainly discusses the motion blur image restoration, vehicle license plate location, the tilt correction, character segment and character recognition method. At the conclusion of the previous research, the new algorithms were put forward on the basis of the existing ones. The main contents and innovative points as follows:(1) In moving vehicle image restoration, first the model of motion blurred image degradation was introduced. And then the evolutionary algorithm based on moving vehicles image recovery was put forward. According to the motion blurred image degradation model to determine fitness function, and then choose, crossover and mutation operation of evolution algorithm. The simulation results show that, compared with the blind recovery method, evolutionary algorithm can improve recovery capacity of the motion blurred image.(2) In the license plate localization, the license plate localization method was put forward based on chaotic evolutionary algorithm. This method used chaos to generate the initial population. In the evolutionary process, the chose individuals were optimized by chaos, and license plates were localized according to the global optimization of chaotic evolutionary. The experimental results show that this method is better than others.(3) In character division, the binary algorithm based on evolutionary algorithm was put forward. The fitness function was determined by Otsu method. Using this method can quickly and accurately get two threshold values, and it is better to realize the binary of vehicle images.(4) In character recognition, the license plate character recognition method based on evolutionary algorithm and the BP neural network were put forward in this paper. First it was global optimized by evolutionary algorithm, and then the initialized weights and offset were gained. Finally it was accurately optimized by BP algorithm. The experiments show that combining the two methods can get the higher character recognition probability.