Few-Shot Lightweight License Plate Recognition in Complex Environments for Intelligent Transportation Systems

Effective license plate identification in intricate contexts is essential for intelligent transportation systems, particularly in situations with constrained data and processing resources. Under these situations, traditional image recognition and deep learning models struggle, resulting in reduced accuracy and flexibility that could compromise traffic safety and hinder the development of intelligent transportation systems. An essential part of intelligent traffic management, the automatic license plate recognition (LPR) algorithm is widely used in parking payment systems, residential vehicle access control, and traffic infraction enforcement. This paper presents a few-shot lightweight license plate recognition technique utilizing advanced image processing and convolutional neural networks to tackle these issues. The methodology commences with an analysis of the mean grayscale values of target photos and employs image enhancement techniques, such as the ACE algorithm and dark channel before dehazing, to preprocess license plate information from intricate contexts. A new method integrating color key elements is presented to attain precise recognition of license plate characters. Experimental findings indicate that the suggested approach markedly enhances recognition performance in intricate situations. It attains a license plate region localization accuracy of 85.95%, a recall rate of 83.24%, and an F1 score of 85.14%. In comparison to conventional image processing methods, the F1 score exhibits an enhancement of 48.26%, and it also exceeds SSD, YOLO, and YOLOv3 by 25.62%, 27.16%, and 18.20%, respectively. The suggested method is efficient, exhibiting minimal temporal complexity, and successfully diminishes noise while ensuring strong recognition of license plate characters.