Link to Report: Coming Soon
Background :
ViewBridgeInEnvironments introduces a pioneering framework that integrates environmental factors and contexts into bridge assessment, leveraging advanced panoptic segmentation technologies, while also incorporating the latest computer vision methods beyond segmentation. Traditional bridge inspections focus primarily on structural integrity, often overlooking the surrounding natural and manmade environments that can influence deterioration, accessibility, and safety. This framework addresses that gap by capturing comprehensive visual data from both bridge structures and their environments, enabling a holistic understanding of bridge health and the interactions between structural elements and surrounding conditions.
By incorporating CV-based models, the framework produces interpretable outputs such as binary masks and quantified features, which support actionable decision-making in bridge monitoring, maintenance, and management. The approach allows for simultaneous assessment of structural and environmental conditions, providing insight into potential vulnerabilities caused by adjacent terrain, vegetation, hydrological factors, and nearby infrastructure. Through these analyses, transportation agencies can identify risks, prioritize interventions, and allocate resources more effectively to enhance bridge safety and functionality.
ViewBridgeInEnvironments is designed to leverage low-cost, widely accessible data collection technologies, including imagery from cell phones, cameras, and affordable drones, making it practical for both state-managed and locally owned bridges. The framework is scalable and adaptable, capable of being applied across diverse geographic regions and bridge types, including those in rural or hard-to-access areas where traditional inspection is challenging. While erosion is one example of a feature that can be monitored, the framework is not limited to this, and the project will identify additional key features for comprehensive bridge assessment.
By integrating structural evaluation with environmental context, ViewBridgeInEnvironments enables bridge owners and agencies to make timely, informed decisions, supporting resilient, safe, and sustainable infrastructure. The project represents a significant advancement in applying computer vision and panoptic segmentation to civil infrastructure, combining precision, environmental awareness, and practical deployment to enhance bridge monitoring and management.
Objectives :
Enhancing bridge assessment methods directly supports the strategic goals of the U.S. Department of Transportation, which seeks to ensure safe, efficient, and sustainable transportation systems. By improving the field evaluation and monitoring of bridge infrastructure, these methods contribute to public safety, promote economic growth through reliable transportation networks, and advance sustainability by extending the service life of critical assets. Furthermore, the integration of advanced technologies into bridge assessment aligns with USDOT’s objectives to foster innovation and modernize infrastructure management practices.
Scope :
Task 1 – Comprehensive Literature Review on Segmentation Techniques
Conduct an in-depth review of state-of-the-art computer vision segmentation methods, including semantic, instance, and panoptic approaches, as well as emerging CV models (e.g., ViT+ LLaVA, MaskDINO, UniSeg, SAM+CLIP hybrids), with an emphasis on applications for infrastructure assessment.
Task 2 – Practical Applications of the ‘ViewBridgeInEnvironments’ Framework
Analyze and select key bridge monitoring features and evaluation tasks where computer vision technologies can be effectively applied within various environmental contexts—such as urban and rural areas, waterways, and other settings—with a focus on bridge inspection, damage detection, and condition monitoring. Initial field data collection and exploratory application studies will be conducted as part of this task.
Task 3 – Development of the ‘ViewBridgeInEnvironments’ Framework
Design and implement a flexible framework that integrates computer vision-based segmentation and recognition models to analyze bridge conditions across diverse environments and viewpoints, providing a consistent approach for processing field data.
Task 4 – Field Validation of Key Features of ‘ViewBridgeInEnvironments’
Integrate the selected key features into field evaluations to validate their performance in data-driven condition analysis using computer vision technologies and automated documentation. Additional field data collection and validation activities will also be conducted as part of this task.
Task 5 – Synthesis and Documentation of the Framework
Compile and document the outcomes from Tasks 1–4, highlighting the capabilities, limitations, and practical guidance for applying the ViewBridgeInEnvironments framework in real-world bridge assessments.
Research Team :
Principal Investigator: Mi G. Chorzepa
Co-Principal Investigator: Bjorn Birgisson, UGA and Atorod Azizinamini, FIU