Research On License Plate Character Recognition Based On GABP Neural Network

The vehicle license plate is a certificate for permitting the car to go on the road normally.It is the main basis for the road traffic management department and the owner to identify and find the corresponding vehicle.As an important part of the intelligent integrated transportation service system,license plate character recognition technology is an important link in the realization of traffic management intelligence in the field of pattern recognition and computer vision technology.An important method in the field of machine learning,BP neural network has become the most widely used artificial neural network with its powerful nonlinear function mapping ability and good classification recognition ability.This paper combines license plate character image processing and machine learning technology.A license plate character recognition model based on integrated GABP neural network is proposed for vehicle license plate recognition.The main workload of this paper includes:(1)Image processing of license plate characters: mainly includes the positioning and segmentation of license plate character images.Through comprehensive analysis of similar license plate characters,the edge operator for edge detection with good vertical and horizontal edge detection is selected for edge detection;In the image segmentation of license plate characters,the license plate character image is firstly subjected to gray threshold segmentation,and then the threshold value is selected for binarization processing to obtain a binarized digital matrix.Finally,the license plate character image is segmented by calculating the pixel point gray value accumulation.(2)License plate character recognition: the binarized matrix corresponds to the input neurons of the neural network,the target identifier in the database is used as the output neuron of the neural network,and the hidden layer neurons are determined according to the number of input and output neurons,thus constructing the simulation experiment(Neural network structure model).Because BP neural network initial weight and threshold are given randomly,the neural network converges slowly and easily falls into local minimum.In this paper,the genetic algorithm is used to optimize the initial weight and threshold of the neural network(GABP neural network model)to accelerate the convergence speed and prediction accuracy of the neural network.(3)Reduce the license plate character recognition error by introducing the integrated learning concept: the license plate characters divided by the license plate character image are divided into 7 independent character images,and then the 7 character images are respectively entered into each sub-base classifier,each base The classifier is a GABP neural network model.The training is independently identified in the model.Finally,the recognition results of each base classifier are output together on the integration side as the final recognition result of the license plate character recognition.