A deep neural framework for real-time vehicular accident detection based on motion temporal templates

Area of Science:

• Computer Vision
• Artificial Intelligence
• Intelligent Transportation Systems

Background:

• Vehicular accident prediction and detection are crucial for Intelligent Transportation Systems (ITS) and smart cities.
• Existing methods often lack real-time capabilities and high accuracy.

Purpose of the Study:

• To present a novel vision-based framework for real-time vehicular accident prediction and detection.
• To improve the safety and efficiency of transportation networks.

Main Methods:

• Utilizes motion temporal templates and fuzzy time-slicing for object detection and tracking.
• Employs deep neural networks (DNN) trained on extracted local features to identify abnormal vehicle behavior.

Main Results:

• Achieved a high hit rate of 98.5% with a low false alarm rate of 4.2%.
• Demonstrated effectiveness on real-world vehicular accident videos.
• Ensured delay guarantees for real-time traffic monitoring.

Conclusions:

• The proposed framework offers a promising solution for real-time vehicular accident prediction and detection.
• It significantly enhances the capabilities of ITS and contributes to safer smart cities.