Paper Title
Simulation Learning Algorithm Applied to Yolov4 License Plate Image Recognition in Rainfall Scenes

Incorporating add-on algorithm to enhance license plate recognition over rainfall scenes and to discuss the system effectiveness analysis, this proposed approach has been applied the convolutional neural network YOLOv4 object detection system to train license plate data sets in different situations. The training processes are conducted in different states and evaluated on different combined datasets: the Application Oriented License Plate(AOLP) dataset whose real images are used for benchmark tasks, and a generated dataset with synthetic images recreating a variety of lighting and rainfall conditions. For license plate and character training and recognition, we have prepared four kinds of data sets, which are the original license plate image from AOLP, and the simulated rainfall factor interference to the original license plate image for training. The experimental recognition results show that the license plate detection and recognition rate is better in the rainy scene where the overall character recognition reached 84.5%, effectively improving the recognition ability in the rainfall conditions. Keywords - Deep Learning, Convolutional Neural Network, Yolov4, License Plate Recognition