Deep Learning and IoT-Enabled Accident Detection and Reporting for Smart Cities Domain

• Rawal Javed,
• Tehseen Mazhar,
• Muhammad Aoun,
• Tariq Shahzad,
• Halah A. Al-AlShaikh,
• Yazeed Yasin Ghadi,
• Abdul Khader Jilani Saudagar,
• Muhammad Amir Khan,

Vol. 19, No. 4, April 30, 2025
10.3837/tiis.2025.04.005, Download Paper (Free):

• Deep Learning
• Intelligent transport system
• IoT
• Edge Computing
• Yolo V4

Abstract

Even though numerous ICT-based solutions have been put forth for the detection of accidents and rescue missions, they often suffer from compatibility issues with different vehicles and are accompanied by high costs. This paper introduces a cutting-edge intelligent system designed for accident detection and rescue, mirroring human cognitive functions through integrating the Internet of Things (IoT) and artificial intelligence. A specialized kit for IoT has been created to identify accidents and gather comprehensive incident-related data, such as pressure, position, gravitational force, and speed. This information is then transmitted to the cloud for processing. A deep learning model is implemented in the cloud environment to verify the IoT module's output and activate the rescue mechanism when an accident is detected. The DL module, employing ensemble transfer learning with dynamic weights, aims to minimize false detection rates. Because there isn't an appropriate dataset available, a personalized dataset is generated from diverse online videos. To validate the proposed method, a comparative analysis is conducted on Yolo V4. Experimental results indicate that Yolo V4 outperforms, achieving training, validation, and test accuracies of 98.2% each. To assess the system's real-world applicability, it undergoes validation using a robot car model.

Statistics

Cite this article