License Plate Character Recognition Based On Quantum Neural Network

The intelligent transportation is the future direction of the traffic development, the license plate recognition system as an important part of the intelligent transportation system, plays an important role in intelligent transportation, and has important application in the actual production of life. The license plate recognition system is based on digital image processing, pattern recognition, computer vision technology, and automatic extraction of vehicle license plate information. In this paper, based on the background of the vehicle license plate recognition, understand the main process of recognition of license plate recognition system. The license plate image preprocessing, license plate location, character segmentation and recognition algorithm of the license plate recognition system is further researched. The specific contents are as follows:(1) Based on the license plate image processing algorithms and common license plate location algorithm. The detection effect of the image in the vertical direction is enhanced by using the improved Sobel operator, combined with mathematical morphology processing method, a new license plate location method based on the improved Sobel operator and mathematical morphology is proposed. Then using Hough transform algorithm to determine the inclined plate, and the plate tilt correction.(2) Through the research of character segmentation algorithm, understanding the key factor impact the license plate character segmentation accuracy and realize License plate character segmentation. In this paper, based on the projection character segmentation algorithm and with license plate its own shape and size characteristics. The license plate character segmentation algorithm based on projection and prior knowledge is proposed. This algorithm has better segmentation for adhesion and fuzzy character. Because of the divided plate characters have different sizes, we use bilinear interpolation algorithm for character normalization.(3) Research on quantum neural network model, using quantum neural network’ parallelism and ambiguity, build a quantum neural network license plate character classifier and put forward the quantum neural network classifier based on multi-level transfer function, and introduced momentum factor on the basis of gradient descent,improve the existing neural network convergence rate, reduce the vibration of the network training process. Then use the coarse network features extraction method extract the character feature and train the network.At last, writing a test interface by VC++ and MATLAB software system in this paper. Through experimental analysis, this paper proposes the license plate locating algorithm, character segmentation algorithm and character recognition algorithm has better positioning, segmentation and recognition effect.