Analyzing Facades at KSU Marietta for Heat Leakages
Primary Investigator (PI) Name
Adeel Khalid
Department
SPCEET – Industrial and Systems Engineering
Abstract
Universities experience energy loss through facade leakage, but drone thermography often generates images that are difficult to compare across flights and buildings. This study develops a radiometric UAS workflow that generates repeatable, component-level ΔT metrics, and trains a small YOLO detector to automate indexing. Flights are scheduled for either full shade, or uniform overcast; information on emissivity is determined by material; and a small coplanar high-emissivity patch is also included in every scene. Capture geometry is held constant near a 45° oblique view, with constant standoff, ~85% overlap, and with paired RGB images for alignment and labeling. For each facade element marked (windows, spandrels, doors, vents), ΔT is defined as the component minus local patch and summarized from component to facade to building; paired flights per facade assess short-term repeatability and examine if drift occurs. The presentation will report evaluation on buildings at Kennesaw State, detector performance (mAP@0.5), agreement between detector-enabled and manual ΔT, and review-time savings from end to end. The intent of the workflow is to reduce flight-to-insight yet maintain the thermal ΔT results are still comparable, across flights and buildings and produce decision-ready leakage metrics for multiple buildings or facilities on campus.
Disciplines
Aeronautical Vehicles | Environmental Monitoring | Mechanics of Materials | Navigation, Guidance, Control and Dynamics | Structural Materials
Analyzing Facades at KSU Marietta for Heat Leakages
Universities experience energy loss through facade leakage, but drone thermography often generates images that are difficult to compare across flights and buildings. This study develops a radiometric UAS workflow that generates repeatable, component-level ΔT metrics, and trains a small YOLO detector to automate indexing. Flights are scheduled for either full shade, or uniform overcast; information on emissivity is determined by material; and a small coplanar high-emissivity patch is also included in every scene. Capture geometry is held constant near a 45° oblique view, with constant standoff, ~85% overlap, and with paired RGB images for alignment and labeling. For each facade element marked (windows, spandrels, doors, vents), ΔT is defined as the component minus local patch and summarized from component to facade to building; paired flights per facade assess short-term repeatability and examine if drift occurs. The presentation will report evaluation on buildings at Kennesaw State, detector performance (mAP@0.5), agreement between detector-enabled and manual ΔT, and review-time savings from end to end. The intent of the workflow is to reduce flight-to-insight yet maintain the thermal ΔT results are still comparable, across flights and buildings and produce decision-ready leakage metrics for multiple buildings or facilities on campus.