AI BASED SURVEILLANCE SYSTEM FOR DETECTION OF VEHICLES WITHOUT HELMETUSING YOLO TECHNOLOGY

Abstract

The project focuses on developing a machine learning-based surveillance system designed to enhance traffic safety by detecting bike riders without helmets and identifying triple riders in realtime.Leveragingadvancedcomputervisiontechniques,thesystemutilizesConvolutional Neural Networks (CNNs) and state-of-the-art detection algorithms such asYOLO (You Only Look Once) and Faster R-CNN to analyse live video feeds from traffic cameras. Withhighaccuracy(96.66%)andalowfalsealarmrateof0.5%,thesystemreliablyidentifies violations,evenincomplextrafficscenarios.Itprocessesvideofeedsframebyframe,detecting helmet-less riders and triple riding through automated classification and object detection. Integration with pre-trained models like ResNet and EfficientNet further enhances its performance. The adaptive learning mechanism enables continuous improvement through training on new datasets, ensuring the system remains robust across diverse environments. Designed for seamless integration into existing traffic management systems, this solution offers real-time monitoring, automated detection, and scalable deployment, significantly contributing to safer roadways and improved traffic regulation. Inadditiontodetectinghelmetviolationsandtripleriding,thesystemcanbefurtherenhanced by integrating license plate recognition to automate fine generation. By leveraging Optical CharacterRecognition (OCR)techniques, the system can extract vehicle registration numbers from detected violators, streamlining law enforcement and reducing the need for manual intervention.Thisfeatureenablesauthoritiestoissueautomatedpenaltynotifications,ensuring a more efficient and transparent traffic monitoring process. Moreover, the system's scalability allowsittobedeployedacrossmultipleurbanlocationswithvaryingtrafficdensities,adapting to different lighting conditions and environmental factors for consistent performance.