Dual Detection of Licence Plates and Helmets Using an Optimized Yolo and Neural Networks

Computer vision technologies are advancing quickly and are being used more often to improve safety monitoring and surveillance in industrial and transportation settings. Traditional monitoring methods rely heavily on manual supervision. This approach can be inefficient, time-consuming, and prone to errors. Deep learning-based object detection algorithms, like YOLO, have greatly improved the ability to detect and classify objects in real time. Among the existing object detection models, YOLOv8 is widely used for helmet detection and related safety applications because of its speed and accuracy. However, earlier models often face practical challenges. These include reduced accuracy in complex environments, difficulties in detecting small objects, and limits when processing multiple objects at once. This project proposes a new computer vision framework that uses the YOLOv10 architecture for the dual detection of safety helmets and vehicle license plates in real time. The proposed system improves accuracy by using data augmentation techniques, coordinate attention mechanisms, and layers for detecting small targets. These elements help handle complex backgrounds and partially hidden objects more effectively. Additionally, the system is trained on annotated datasets that include various helmet types and vehicle plates under different lighting and environmental conditions to enhance robustness. Experimental results show improved accuracy in object detection, faster inference speed, and reliable performance in real-time situations. The findings suggest that the YOLOv10-based framework offers an efficient and scalable solution for improving safety compliance in workplaces and for automated vehicle monitoring in modern surveillance systems.