Corridor-Wide Surveillance Using Unmanned Aircraft Systems Phase III: Exploration of the Implementation of Using Unmanned Aircraft Systems for Freeway Incident Detection and Management: Part B

Details

• Creators:
  Zhang, Yu ; Kourtellis, Achilleas ; Post, Joseph ; Li, Bai ; Porter, Brian ; Nodehi, Mahdi Monemi
• Corporate Creators:
  University of South Florida
• Corporate Contributors:
  University of South Florida. National Institute for Congestion Reduction ; United States. Department of Transportation. University Transportation Centers (UTC) Program ; United States. Department of Transportation. Office of the Assistant Secretary for Research and Technology
• Subject/TRT Terms:
  Automatic Incident Detection Background Subtraction Algorithm Data Collection Drones Freeways Incident Management Thermal Imagery Traffic Monitoring Traffic Surveillance Unmanned Aerial Vehicles (UAV) Unmanned Aircraft Systems Vehicle Detection

  More +
• DOI:
  https://doi.org/10.5038/CUTR-NICR-Y3-4-7.2
• Resource Type:
  Tech Report
• Geographical Coverage:
  United States
• Edition:
  Final Report [August 2022 – April 2025
• Corporate Publisher:
  University of South Florida. National Institute for Congestion Reduction
• Abstract:
  In the third phase of the project, the research team continued utilizing drone systems equipped with thermal cameras to enable real-time detection of traffic incidents and their resulting non-recurrent congestion on freeways, while distinguishing them from recurrent congestion. A comprehensive literature review on existing traffic incident detection methods was conducted. Building on insights from the literature and prior accomplishments, the team designed and implemented a drone-based incident detection framework. This framework first extracts vehicle trajectories at fixed intervals from thermal video data and generates corresponding trajectory images. A customized convolutional neural network (CNN) based deep learning model is then developed and trained to extract traffic features from these images and classify them into three categories: incident, recurrent congestion, and normal traffic. Real-time detection was achieved through continuous processing of incoming thermal video segments. Finally, the research team developed a drone-based, AI-powered, real-time freeway incident detection system, featuring a user-friendly web-based graphical user interface (GUI) for initiating and terminating detection process, visualizing results, and reviewing historical records. The system was tested in six detection flights across three different test sites in Florida. During an incident scenario, test results demonstrated that the system was able to accurately and promptly detect the incident approximately 12 minutes earlier than the local Transportation Management Center (TMC). The thermal video containing the incident scene was displayed on the GUI to support immediate verification and severity assessment by the TMC, thereby facilitating rapid emergency response and potentially saving lives. Additionally, the system extracted and displayed the length of the incident-induced non-recurrent congestion during each flight and its propagation speed across multiple flights, providing valuable information for the TMC to implement effective incident management strategies and mitigate the overall impact.
• Format:
  PDF
• Funding:
  69A3551947136, 79075-32, 79075-33
• Collection(s):
  University Transportation Centers
• Main Document Checksum:
  urn:sha-512:32a88fdefe8c5d94951f59c5863a6a826dfd88f2858e927845200525927ec37950f84be970b4c0766fa3950663163410487c580c5970cabc05c655ab22a83614
• Download URL:
  https://rosap.ntl.bts.gov/view/dot/86215/dot_86215_DS1.pdf
• File Type:
  [PDF - 7.13 MB ]