The Implementation Of License Plate Recognition Arithmetic Based On FPGA

The License Plate Recognition(LPR)is an automatic vehicle license plate recognition technology by images and videos processing or pattern recognition based on Personal Computer(PC),embedded product or Field Programmable Gate Array(FPGA)platform.The LPR is the key technology on Intelligent Transportation System(ITS),and is widely used in traffic flow control,vehicle access control and electronic toll collection and so on.So the LPR has constantly been a hot topic since it is put forward by ITS America in 1990.This paper,as the hardware implementation of LPR arithmetic to the research object,made a thorough study of LPR arithmetic,hardware implementation and verification platform building,and completed the LPR arithmetic based on FPGA.According to the characteristics of the hardware implementation,the license plate location algorithm is designed to be pipeline structure,which improved performance on large-sized image processing speed.And then for the plates in complex environments,a new character segmentation method based on connected component labeling and lookup table has been proposed to improve the accuracy of character segmentation.Finally BP neural network is used for normalized characters recognition.In addition,the hardware construction of LPR arithmetic is designed by top-down approach.And the mutual control between different modules in different levels is completed by the multi-stage state machine nested.This paper verified the theory of LPR algorithm by Matlab software,designed hardware for LPR arithmetic by VerilogHDL language,simulated functions by Modelsim software,and through the FPGA development platform from Terasic Company as the core,implemented the LPR verification system combining with the camera and monitor.After the actual license plate video verification,when working frequency of the system is 100MHz and input videos size is 640 X 480,processing speed can be up to 50 frames/s,which satisfy the real-time processing requirement of PAL and NTSC specifications.LPR accuracy is above 90%.The characteristics and innovation of this subject line in that,according complex environment and the characteristics of hardware implementation,a complete LPR algorithm and hardware implementation are proposed.Compared with those implementations which based on PC or DSP processor,the LPR arithmetic based on FPGA can take advantage of parallel and pipeline processing of hardware structure to greatly improve the recognition speed,and reduces the pressure on calculation of the control system.At the same time,the license plate location and character segmentation method of this paper is more suitable for hardware implementation,promoting the LPR arithmetic into hardware and chip.Paper also proposed a fast connected component labeling algorithm based on FPGA to improve the processing speed of license plate location and character segmentation algorithm.