License Plate Recognition System Based on Convolutional Neural Network Multi Model Fusion

In the process of building smart cities, Intelligent Transportation Systems (ITS) technology plays a crucial role. However, traditional motor vehicle license plate recognition technology has disadvantages such as low recognition rate, slow speed, and susceptibility to environmental influences. To address this issue, this paper proposes a multi model fusion algorithm based on convolutional neural networks, using YOLOv5 model to achieve faster and more accurate license plate positioning, using spatial transformation network (STN) to achieve license plate tilt correction, and using LPRNet model to achieve license plate recognition. Based on this, the detection and recognition methods based on neural networks are combined, and the entire network structure is designed to be highly lightweight to achieve automatic license plate recognition.