Research On Vehicle License Plate Recognition Under The Foggy Condition

Vehicle license plate recognition(VLPR) technique is a significant component of the modern intelligent transportation system.It has been widely used in highway toll management, vehicle access management,law enforcement and auditing of the traffic and so on nowadays. Vehicle license plate recognition systems are mostly used to deal with vehicles in good weather condition.However,due to the effect on scattering and absorption of the atmospheric particles in foggy weather conditions,the collected images are blurred,and it brings a lot of difficulty to identify work.Considering this problem,it focuses the key on the study of vehicle license plate recognition under the foggy condition.In the thesis,the fog of those foggy license plates was removed at first,and then identify the image.It recovers the foggy image effective and improves the recognition rate.The main works are as follows:(1) In the aspect of image haze removal,the two algorithms were introduced using dark channel prior based on the physical model and the Retinex algorithm based on the nonphysical model,dark channel prior algorithm was chosen as the foggy removal method according to the experimental effect.(2) In the link of license plate location,first of all,start to image pretreatment such as gray,gray stretch,median filtering, edge detection,and the mathematical morphology and projection is used to realize the accurate license plate location.(3) For the character segmentation,the images were processed after location by OTSU binarization algorithm,and the improved Radon transform is used to accomplish the horizontal and vertical correction.Finally,Using projection and prior knowledge of the license plate to realize the character segmentation.(4) In the aspect of character recognition,the principle of BP neural network was studied deeply,and optimize the neural network classifier to make it more suitable for the recognition to license plate characters.At last,using the algorithm of this thesis,it has designed the fog license plate recognition system under the condition of laboratory based on Microsoft Visual Studio 2008 and OpenCV,which could complete the license plate recognition of fog and non-fog weather accurately also has timeliness and robustness at the same time.