Research And Implementation Of Parking License Plate Recognition And Management System Based On Improved Faster R-CNN

With the continuous development of urban transportation and the increasing number of vehicles,road traffic control has become increasingly important.License plate recognition technology can provide efficient monitoring and management means for urban traffic,supporting the identification of vehicle violations.Additionally,it can be used for smart parking lot management,enabling functions such as vehicle entry/exit and parking billing.While extensive research has been conducted on license plate recognition both domestically and internationally,there are still some technical challenges.Existing algorithms have limited effectiveness in complex and variable environments,such as tilted angles,rain/snow,and strong light.Therefore,the development of more advanced algorithms and techniques is necessary to address these issues.In recent years,deep learning algorithms have been widely applied in license plate recognition.They extract feature information from license plate images through model training to achieve accurate detection and recognition of license plates.Based on this,this paper proposes an improvement to the Faster R-CNN network for license plate localization.Firstly,the VGG16 network in Faster R-CNN is replaced with a deeper Res Net-50 for feature extraction.Then,the Res Net-50 residual blocks are enhanced,and the CBAM attention mechanism is introduced.Due to the use of a deeper network,the loading time is longer.Therefore,this paper integrates the improved Faster R-CNN network with traditional algorithms and designs a new algorithmic flow for license plate localization.Experimental results demonstrate that the proposed algorithm significantly improves the efficiency and accuracy of license plate localization in the system.In the license plate character segmentation and recognition module,the intuitive fuzzy knowledge measure model is applied for the first time.Experimental results show that using this theory to process binarized images is beneficial for improving character segmentation and recognition accuracy.Subsequently,an optimized CNN license plate character recognition network structure is designed and trained.Finally,relevant experiments demonstrate that the proposed license plate recognition method achieves higher accuracy and practicality.To sum up and combined with the relevant demand analysis,the parking lot license plate recognition and management system based on improved Faster R-CNN is realized.The system design and development of seven functional modules,mainly parking administrator user management,vehicle in and out of the warehouse,vehicle information inquiry and charging module,user login registration,out of the warehouse payment and personal information management module.The vehicles entering and leaving the warehouse use the improved license plate location algorithm to locate,then use the vertical segmentation method to segment the license plate characters,and finally use CNN model to recognize the characters and output the results.The whole system has been tested repeatedly,and it is proved that the parking lot license plate recognition and management system based on Faster R-CNN can be effectively applied in the parking lot.