International Journal of Scientific Research and Engineering Development

Autonomous vehicles require intelligent systems to ensure passenger safety because there is no human driver to handle emergencies. Existing in-cabin monitoring systems mainly focus on drivers and do not support multi-passenger safety, medical emergency detection, or security threat handling. This paper presents an AI-based in-cabin monitoring and emergency alert system using deep learning and computer vision techniques. The proposed system continuously monitors passengers inside the vehicle using a cabin camera. It uses YOLO for real-time object and harmful weapon detection, Faster R-CNN and ResNet for multi-person pose estimation, and a CNN model for facial expression recognition. These models help detect unsafe behaviors, violent actions, and medical emergencies such as unconsciousness or distress. The system combines all detected information through a decision-making module to identify emergency situations accurately. When an emergency is detected, the system automatically sends alerts to hospitals, police, and emergency contacts using V2X communication and SMS services. In medical emergencies, the vehicle can navigate to the nearest hospital. Experimental results show that the system works in real time with high accuracy and low response time. This system improves passenger safety and security in autonomous vehicles by providing fast, reliable, and automated emergency responses.

Mr.G M Anand Reddy, Ms. Nayana C R, Ms.Sushmitha H, Ms.Tejashree D T, Ms.Pallavi M, "Towards Accurate Multi-Person Pose Estimation and Emergency Alert System in Vehicles Using Deep Learning" International Journal of Scientific Research and Engineering Development, V8(6): Page(2601-2607) Nov-Dec 2025. ISSN: 2581-7175. www.ijsred.com. Published by Scientific and Academic Research Publishing.