Intelligent Safety Assessment of Rural Roadways Using Automated Image and Video Analysis

Roadside safety is a critical aspect of transportation management, with elements like rigid obstacles, guardrails, clear zones, and side slopes significantly impacting accident outcomes. The Federal Highway Administration (FHWA) provides a valuable rating system for Department of Transportation (DOTs), but the manual rating process is time-consuming and prone to inconsistencies. This project introduces an innovative solution employing computer vision and machine learning algorithms to automate the roadside safety evaluation process. Utilizing pre-trained models such as VGG16 and images captured from Utah roadways, the research team develops a robust algorithm for automated safety evaluation that aligns with the FHWA rating system, providing a comprehensive and efficient method for assessing roadside conditions. Tailored computer vision algorithms detect specific features, enhancing the accuracy of safety evaluations. Pre-trained models for clear zone detection and roadside slope classification further contribute to a nuanced understanding of roadside elements. The project's outcome is a shapefile containing safety rankings for road segments on five state roads. This tool empowers traffic engineers with data-driven insights, enabling informed decision-making for prioritizing improvement projects and enhancing road safety. The automated approach showcased in this research offers a promising avenue for strengthening roadside safety measures and preventing potential accidents. While acknowledging challenges such as periodic retraining and potential false positives, this approach stands as a promising addition to existing methods. The study culminates in a shapefile encompassing safety rankings, roadside features, and illustrative sample images, providing a tangible tool for Utah Department of Transportation (UDOT) in optimizing road safety strategies.


  • English

Media Info

  • Media Type: Digital/other
  • Edition: Final Report
  • Features: Figures; Photos; References; Tables;
  • Pagination: 46p

Subject/Index Terms

Filing Info

  • Accession Number: 01907833
  • Record Type: Publication
  • Report/Paper Numbers: MPC-669, MPC 23-510
  • Created Date: Feb 12 2024 10:31AM