Research And Design Of License Plate Monitoring System Based On Deep Learning

With the improvement of social and economic level,automobile has become the main carrier of transportation.The explosive growth of the total number of cars has also put great pressure on traffic.Therefore,smart transportation is very important in practical life.As the core link of intelligent transportation,license plate recognition should be paid attention to.Since traditional license plate recognition technology is prone to the influence of license plate factors and abnormal surroundings,a license plate recognition method based on deep learning is proposed and a set of license plate recognition system is designed and implemented,effectively solving the problem of license plate recognition effect in complex scenes.The research content of this paper is as follows:(1)Improve the license plate detection model of YOLOv5 s.In order to adapt to license plate detection in complex environment and meet the requirements of real-time and accuracy,the improved license plate detection model based on YOLOv5 s adopts lightweight network to improve detection speed and accuracy,and adds attention mechanism to further strengthen feature extraction.The experimental results show that the improved vehicle license plate detection model is more suitable for vehicle license plate detection in complex scenes.(2)This paper uses a CRNN-based LSTM+CTC model for reference to realize license plate character recognition,which is an end-to-end license plate recognition algorithm without character segmentation preprocessing.First,the license plate image is first input,and the feature vector is extracted by convolutional neural network and input into LSTM for recognition,so as to train the LSTM model to mark sequence features.Finally,the characters are decoded by CTC to produce recognition results.The experimental results show that the LSTM+CTC network model,which omits the process of character segmentation and then recognition,has good robustness and accuracy.(3)Design and implementation of license plate recognition system.By loading the improved YOLOv5 s license plate detection module and the end-to-end LSTM+CTC license plate recognition module respectively,the system realizes the visual recognition of license plates.The display part includes the license plate confidence,license plate position coordinates and license plate recognition results.Experimental results show that the license plate recognition system can achieve good real-time performance and robustness in both normal and complex scenes.