Design And Implementation Of License Plate Recognition System In Parking Lot

Since the reform and opening up,with the rapid development of the country's economic level,the living standards of Chinese residents have continued to improve,and automobiles have gradually changed from luxury goods to daily mobility tools.In recent years,the country's car ownership has reached the second in the world.Due to the characteristics of urban development in China,the demand for parking lots in urban transportation is increasing.Most cities have problems of difficult parking and high vehicle management costs.As an important member of the highway transportation system,the development speed of the parking lot is far behind the development speed of the car,and it cannot meet people's demand for daily parking.Against this background,the disadvantages of traditional IC / ID card fetching for parking lot management have gradually become apparent.Adopting the license plate recognition technology can not only reduce the management cost of the parking lot,but also improve the bad time-consuming and labor-intensive experience of users entering and leaving the parking lot in the parking lot managed in the traditional way.Therefore,the use of license plate recognition technology in parking lots has great research value and market space.This paper analyzes and simulates the image preprocessing,license plate positioning,license plate character segmentation,license plate character recognition and other related algorithms in license plate recognition.In the license plate positioning part,the simulation of the license plate localization method based on Sobel edge detection and the license plate localization method based on color localization were carried out respectively.Finally,the coarse positioning was performed by Sobel edge detection,and the license plate positioning was performed after the color determination.method.In the segmentation of license plate characters,horizontal projection analysis is used to remove the horizontal border of the license plate,and then vertical projection analysis is used to combine the prior knowledge of the license plate character encoding to perform single character segmentation on the license plate in the vertical direction.In the part of character recognition,the recognition methods based on SVM and ResNet18 were analyzed and compared respectively.When the number of samples is low,the traditional SVM classification recognition method has higher recognition accuracy.Finally,theSVM classification and recognition method is used to construct a SVM classifier based on HOG features for Chinese characters,English letters and numbers,respectively,to complete the recognition of license plate characters.Comprehensively considering the system's functional requirements,performance requirements,and processor cost,power consumption,and computing power,this paper completes the design and implementation of a license plate recognition system on an Raspberry Pi 4B based on the ARM-Linux operating system.According to the various data requirements involved in the system operation,a logical analysis was performed and the specific implementation of the data table was completed in the mysql database.According to the function and performance requirements of the parking lot license plate recognition system,the module of the parking lot license plate recognition software was decomposed,and each sub-module was designed and implemented separately.Finally,the entire license plate recognition system was debugged and simulated in a residential parking lot.Finally,the running effect after the transplantation was actually tested.Randomly selected vehicles in the residential parking lot for license plate recognition,the recognition success rate reached 95.45%,and the recognition time remained within100 ms,which could effectively meet the performance requirements of the license plate recognition system.