"Road Collision in Indian Road intersection"

"AbstractThis study investigates the role of machine learning-based object detection models in improving accident prevention and traffic management for Connected Autonomous Vehicles (CAVs) operating on urban and rural roads in Rajasthan. Utilizing a YOLO architecture trained with annotated images of car and truck accidents, the model\u2019s performance is assessed through comprehensive evaluation metrics\u2014including bounding box loss, classification loss, distribution focal loss, precision, recall, and mean Average Precision (mAP). The results demonstrate steady decreases in all loss functions and substantial improvements in precision and recall scores (up to 0.8), indicating strong generalization and robust detection capabilities across diverse road geometries and intersection designs. The research highlights how model insights correlate accident occurrences with geometric road parameters such as alignment, lane width, sight distance, intersection design, and roundabouts, informing CAV system integration and urban planning. Future directions include multi-sensor fusion, edge device deployment, and traffic policy optimization to further enhance safety and support CAV adoption in mixed traffic environments."