Road safety has become a critical concern due to the rapid increase in two-wheeler usage and frequent violations of helmet-wearing regulations. Manual traffic monitoring is often inefficient and error-prone, especially in high-density traffic environments. This study aims to tackle these difficulties presents an automated detecting helmets and license plates system utilizing deep learning techniques. The proposed approach employs a YOLO-based object detection model to identify riders, helmets, and the number of the vehicle plates from pictures, movies, and live surveillance feeds in real time with high accuracy. Helmet violations are detected by associating riders with helmet presence, and the corresponding number plates are isolated for further processing. The model learns from a custom annotated set of data and optimized to perform reliably under varying lighting conditions, camera angles, and traffic scenarios. The results of the experiments show that the detection is very accurate, low latency, and strong robustness, making the system work in real life time traffic surveillance and smart city applications. This study shows how well deep learning works-based computer vision systems in enhancing road safety and supporting automated traffic law enforcement.