The Utilization of Artificial Intelligence and Deep Learning for Identifying Vehicle Wheels

The technology using artificial intelligence (AI) allows, through the preprocessing of video and image data, their transformation into traffic information, the recognition in abusive conditions of elements, the specification of the different characteristics of the vehicle (type, brand, model, color), object detection and tracking, automatic calibration and positioning. Algorithms for performance evaluation can evaluate recognition quality and detect malfunctions due to external factors, reduced lighting conditions, or incorrect camera installation. This article discusses the application of AI and deep learning techniques to identify vehicle wheels. It highlights the preprocessing of image and video data to extract traffic information, detect vehicle attributes, and perform object tracking. The study focuses on understanding AI and convolutional neural networks (CNNs) for image processing, emphasizing the significance of feature extraction algorithms. The research outlines the methodology for creating a dataset of annotated wheel images, selecting the YOLO detection algorithm, and using the Darknet Github framework for training and testing CNNs. It discusses the training process, including configuration, data augmentation, and evaluation metrics like Mean Average Precision (MAP) and Intersection Over Union (IOU). Finally, the article concludes with insights into the practical applications of the developed wheel detection system, suggesting future improvements and potential integrations with other projects.