• AI-driven, Web-deployed, Low-Cost Visual Sensing of Stormwater Outlet Flow

    NCDOT Research Project Number: 2025-03

 Executive Summary

  • ​Robust and reliable information is needed regarding stormwater infrastructure performance to effectively manage excess flow and pollutants from stormwater runoff. In particular, measuring stormwater outlet flow is important as outlets provide the critical function of regulating flow for structural stormwater control measures (SCMs), thus outlet monitoring aids in evaluating performance and identifying possible design problems and causes for failure. In relatively small watersheds, stormwater flow monitoring often involves the installation of weirs at the outlet of stormwater pipes, and use of in situ in water sensors in harsh environments and difficult places, which renders stormwater flow monitoring difficult at best. 


    Image-based sensing, an active area of research, has the potential to alleviate a lot of the difficulties associated with in situ in water sensors. In a NCDOT sponsored project, we initially tried to use traditional machine vision techniques to images and videos of stormwater flow, with little success. This was until we decided to use Artificial Intelligence (AI) based techniques.  These approaches have been transformative and stunning in many respects. They have made finding the water level in a culvert from an initially ‘impossible task’ from low light, low contrast images a stunningly simple task. The ‘AI-driven, Web-deployed, Low-Cost Visual Sensing of Stormwater Outlet Flow’ project proposes to take advantage of the experienced gained and the talented personnel recruited to go to the next step and all the way to creating user-friendly, user-ready App deployed tools.  

    For this, we propose to build on the existing project and use a suite of additional AI-driven technologies to automatically measure velocities from videos, something never done before, although necessary to measure stormwater flow.  With both water stage and velocity measurements, flow can be calculated. We propose to deploy the computer vision tools produced on mobile/web client applications, as well as on smart cameras. This will largely streamline and will change monitoring stormwater flow from a difficult to a user friendly, low-cost, and low-maintenance task. 



  
Francios Birgand
Researchers
  
Francios Birgand
  
Andy McDaniel
  
John W. Kirby
  
NC State University
  

 Related Documents

No content found

 Report Period

  • August 1, 2024 - July 31, 2027

 Status

  • In Progress

 Category

  • Environment and Hydraulics

 Sub Category

  • Water Quality and Pollutant Discharge

 Related Links

Was this page helpful?