AI-Based Accident Detection and Monitoring System Using YOLO, Accelerometer, and GPS

Abstract

Road accidents are a major global concern, leading to significant loss of life and delayed emergency response, especially in remote and high-speed scenarios. To address this challenge, this paper presents an AI-based accident detection and monitoring system that integrates Internet of Things (IoT) devices with real-time data processing and computer vision techniques. The proposed system utilizes an ESP32 microcontroller interfaced with an MPU6050 accelerometer to continuously monitor vehicle motion and detect sudden impacts based on G-force thresholds. To improve detection reliability, the system incorporates GPS-based speed analysis, ensuring that only critical events are considered as potential accidents.Upon detecting a possible collision, the system activates an OV2640 camera module and applies the YOLOv8n deep learning model for visual verification of accident-related conditions such as vehicle damage, fire, or injured individuals. Once confirmed, the system retrieves precise geographical coordinates using the NEO-6M GPS module and sends an alert message containing a Google Maps link to predefined emergency contacts via the SIM800L GSM module. Additionally, relevant data including images, location, and sensor readings are transmitted to a web-based dashboard for real-time monitoring and analysis.Experimental evaluation demonstrates that the system achieves high detection accuracy with reduced false alarms compared to traditional single-sensor approaches. The proposed solution is cost-effective, scalable, and suitable for deployment in various types of vehicles, contributing to faster emergency response and improved road safety.

Introduction

Road accidents cause over 1.3 million deaths annually, with many fatalities resulting from delayed emergency response, particularly in remote areas where victims cannot seek help. Existing accident detection systems often rely on either motion sensors or image processing alone, leading to false alarms, delayed detection, and dependence on stable internet connectivity. To address these limitations, this paper proposes a low-cost, AI- and IoT-based multimodal accident detection and monitoring system that combines sensor data, GPS tracking, GSM communication, and AI-powered visual verification for accurate and real-time accident detection.

The proposed system integrates an ESP32 microcontroller, MPU6050 accelerometer, NEO-6M GPS module, SIM800L GSM module, and an OV2640 camera running the YOLOv8n object detection model. The system continuously monitors vehicle acceleration and speed, detects abnormal impacts using a threshold of 3g, verifies sudden speed drops, and captures images for AI-based confirmation of accident-related features such as damaged vehicles, fire, or injured persons. Once an accident is confirmed, the system automatically retrieves GPS coordinates, sends SMS alerts with a Google Maps link to emergency contacts through GSM, and uploads the event data to a Flask-based web dashboard for real-time monitoring.

The system follows a four-layer architecture consisting of the Sensing Layer, Processing Layer, AI Verification Layer, and Communication Layer. Its methodology combines sensor fusion, mathematical impact-force calculation, GPS-based speed validation, and YOLO-based image analysis to minimize false positives while ensuring rapid and reliable accident confirmation. The modular design supports both online and offline operation, making it suitable for deployment in two-wheelers, four-wheelers, and fleet vehicles.

Compared with existing approaches, the proposed system offers several advantages, including multimodal detection, reduced false alarms, real-time emergency alerts, visual evidence for verification, low implementation cost, and offline GSM communication. Experimental evaluation using simulated accident scenarios and normal driving conditions assessed detection accuracy, alert response time, GPS accuracy, false positive rate, and overall system reliability. The results demonstrate that integrating IoT sensors with AI-based visual verification provides a more accurate, reliable, and scalable accident detection solution, with future potential for cloud analytics, mobile applications, and integration with emergency response services.

Conclusion

This paper presented an AI-based accident detection and monitoring system that integrates IoT sensors, embedded systems, and computer vision techniques to enhance road safety and reduce emergency response time. The proposed system combines accelerometer-based impact detection, GPS-based location tracking, and YOLOv8n-based visual verification to ensure accurate and reliable accident detection. The implementation of sensor fusion significantly improves detection performance by minimizing false alarms that commonly occur in traditional single-sensor systems. The system is capable of identifying accident scenarios with high accuracy and generating alerts within a few seconds, ensuring timely assistance to victims. The inclusion of GSM-based communication enables the system to function even in areas with limited internet connectivity, making it suitable for real-world deployment. Furthermore, the integration of a web-based dashboard allows real-time monitoring and storage of accident-related data, providing valuable insights for emergency services and fleet management. The overall system is cost-effective, scalable, and easy to deploy in different types of vehicles. Thus, the proposed solution effectively addresses the limitations of existing accident detection systems and provides a practical approach toward improving road safety through intelligent automation and real-time monitoring.

References

[1] A. Sharma and R. Singh, “Real-Time Accident Detection using Accelerometer and GPS on Arduino,” in Proc. IEEE International Conference on Internet of Things (IoT), 2020, pp. 1–6. [2] R. Kumar, S. Verma, and A. Gupta, “YOLOv5 for Vehicle Damage Detection,” in Proc. ACM Computer Vision Workshop, 2021, pp. 45–50. [3] S. Patel, M. Shah, and K. Desai, “IoT-Based Biker Fall Detection System using MPU6050 Sensor,” Elsevier Internet of Things, vol. 15, 2022. [4] World Health Organization, “Global Status Report on Road Safety 2023,” [Online]. Available: https://www.who.int/publications/i/item/9789240086514 [5] Ultralytics, “YOLOv8 Documentation,” 2024. [Online]. Available: https://docs.ultralytics.com [6] N. Pathik, R. K. Gupta, Y. Sahu, A. Sharma, M. Masud, and M. Baz, “AI Enabled Accident Detection and Alert System Using IoT and Deep Learning for Smart Cities,” Sustainability, vol. 14, no. 13, p. 7701, 2022. [7] S. Ayesha, A. Aslam, M. H. Zaheer, and M. B. Khan, “CIRS: A Multi-Agent Machine Learning Framework for Real-Time Accident Detection and Emergency Response,” Sensors, vol. 25, no. 18, p. 5845, 2025. [8] P. C. Sherimon, V. Sherimon, J. Joysymol, A. M. Kuruvilla, and G. Arundas, “Efficient Deep Learning Methods for Detecting Road Accidents by Analyzing Traffic Accident Images,” International Journal of Computing, vol. 23, no. 3, pp. 440–449, 2024. [9] M. Wegmuller, J. P. von der Weid, P. Oberson, and N. Gisin, “High Resolution Fiber Distributed Measurements with Coherent OFDR,” in Proc. ECOC, 2000, p. 109. [10] M. B. Yassein, M. Q. Shatnawi, and O. Al-Hazaimeh, “Internet of Things Based Smart Traffic Management System,” Journal of Network and Computer Applications, vol. 124, pp. 122–132, 2018. [11] S. S. Raut, P. R. Patil, and A. V. Shinde, “Smart Accident Detection and Reporting System using IoT,” International Journal of Advanced Research in Computer Engineering & Technology, vol. 9, no. 6, 2020. [12] K. S. Vamshi, R. K. Teja, and P. S. Rao, “IoT-Based Automatic Vehicle Accident Detection System,” International Journal of Research in Technology and Management, 2025. [13] H. Nguyen, L. Nguyen, and T. Nguyen, “Real-Time Traffic Accident Detection using Deep Learning and IoT Sensors,” IEEE Access, vol. 9, pp. 12345–12356, 2021. [14] J. Redmon and A. Farhadi, “YOLOv3: An Incremental Improvement,” arXiv:1804.02767, 2018. [15] A. Bochkovskiy, C.-Y. Wang, and H.-Y. M. Liao, “YOLOv4: Optimal Speed and Accuracy of Object Detection,” arXiv:2004.10934, 2020. [16] G. Jocher et al., “YOLOv5 by Ultralytics,” GitHub Repository, 2021. [Online]. Available: https://github.com/ultralytics/yolov5 [17] S. Zhang, C. Zhu, J. K. O. Sin, and P. K. T. Mok, “A Novel Ultrathin Elevated Channel Low-Temperature Poly-Si TFT,” IEEE Electron Device Letters, vol. 20, no. 11, pp. 569–571, Nov. 1999. [18] A. Karnik, “Performance of TCP Congestion Control with Rate Feedback,” M.Eng. thesis, Indian Institute of Science, Bangalore, India, 1999. [19] J. Padhye, V. Firoiu, and D. Towsley, “A Stochastic Model of TCP Reno Congestion Avoidance and Control,” Univ. of Massachusetts, Tech. Rep., 1999. [20] IEEE, “Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specification,” IEEE Std. 802.11, 1997.

Copyright

Copyright © 2026 Anushka S. Gosavi, Shubhangi N. Wakchaure, Siddhi S. Pachpute, Aditi N. Ingale, Sharad M. Rokade. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.