Loading [a11y]/accessibility-menu.js
Modeling the Risk of Truck Rollover Crashes on Highway Ramps Using Drone Video Data and Mask-RCNN | IEEE Conference Publication | IEEE Xplore

Modeling the Risk of Truck Rollover Crashes on Highway Ramps Using Drone Video Data and Mask-RCNN


Abstract:

Automatic vehicle detection is essential in autonomous driving, traffic analysis, and transportation management. Deep neural networks (DNNs) have emerged as a powerful to...Show More

Abstract:

Automatic vehicle detection is essential in autonomous driving, traffic analysis, and transportation management. Deep neural networks (DNNs) have emerged as a powerful tool for object detection in images and videos, and more recently, Oriented Bounding Boxes (OBBs) have gained popularity as they offer more precise bounding boxes. In this paper, we focus on truck rollover crashes, which pose a significant risk on highway ramps, resulting in blockages and severe congestion. Identifying the primary factors contributing to these rollovers is crucial for developing effective preventive measures. We propose a novel pipeline for detecting, tracking, and analyzing vehicles in highway and ramp traffic. It includes stages for vehicle detection, tracking, and identifying high-risk events like slowdowns or stalls on roadways and gore areas. We created an aerial image dataset for highway vehicle analysis, consisting of 3,150 diverse images from nine locations captured using high-resolution cameras mounted on drones. It contains 30,228 annotated instances of vehicles categorized into three classes and two sub-classes, viz. tractors and trailers. This research demonstrates how data from a swift round of drone deployments can be efficiently processed by the proposed pipeline, giving us vehicle behavior and traffic patterns, enabling improved traffic management and incident detection. This trajectory-based safety analysis approach is valuable for understanding driver behavior near highway ramps and conducting road safety audits. Link to GitHub repository: https://github.com/z00bean/HighwayTrafficAnalysis.
Date of Conference: 24-28 September 2023
Date Added to IEEE Xplore: 13 February 2024
ISBN Information:

ISSN Information:

Conference Location: Bilbao, Spain

Funding Agency:


Contact IEEE to Subscribe

References

References is not available for this document.