Infrastructure is best surveyed from the air – and in infrared!

by Ryan Wicks

At the UMass Amherst campus we regularly use UAS to conduct surveys of key infrastructure; whether it be to monitor and document stages of new construction on campus or to survey and inspect existing infrastructure. One of our more recent additions to our array of capabilities is the capacity to develop thermal orthomosaics from long-wave infrared (LWIR) imagery. This can help us map heat sources and thermodynamic processes of buried infrastructure, or look at heat loss in structures.

Fig. 1 – Example LWIR Thermal Image: In this LWIR thermal image temperature is represented in a linear white-hot grayscale; that is to say that black in the image represents the lowest apparent temperature (-12.5 degrees Celsius as indicated in the scale on the right of the image) and white represents the highest apparent temperature (5.5 degrees Celsius as indicated in the scale on the right of the image), and temperatures inside this range are represented with varying shades of gray that are assigned in a linear fashion. The temperatures are only “apparent” because other factors besides temperature can effect the emitted radiation that the camera detects, such as the varying emissivities of materials in the image field of view. This image is tuned to an emissivity of 0.98. The point “Sp1” in this image is shown to have an apparent temperature of -0.8 degrees Celsius. The mostly vertical white streak in this image is actually sewage line buried under the ground, but the heat from it reach the surface and the emitted thermal radiation is visible by a LWIR camera.

In a recent survey project we surveyed roofing of some campus buildings to inspect for suspected leaks. In general, if water leaks into a roof if can compromise the insulation of the building and conduct heat more readily. In the winter time when the outside temperature is significantly cooler than the inside temperature areas of compromised insulation will show up as relatively hotter than the surrounding area on a surface because it is conducting the heat from the inside to the outside more readily.

Fig 2 – RGB Orthomosaic of Survey Area: This image shows a red-green-blue three-color orthomosaic of a roof that was being inspected for water leaks. Gray patches on the roof are visible, but it is unclear what they are. Note that the marked GCPs in the area that are used to constrain the orthomosaic reconstruction to increase accuracy.
Fig. 3 – LWIR Orthomosaic: This image is a grayscale, white-hot (i.e. areas of greater temperature are whiter while areas of lower temperature are darker) orthomosaic of the same region as shown in Fig. 1. The bright regions on the roof are hotter because of ventilation ports exuding warm air from inside the building; dark regions on the roof either indicate either significantly lower temperatures or a material with an emissivity that is much less than the material of the roofing (rubber, e ~ 0.91). Close – up images show that the dark regions are covered in pools of water; these regions are cooler areas due to convection. Warmer regions under these areas of pooled water would not be visible because water attenuates any signal from behind it very well. Note the GCPs used to increase accuracy of the thermal orthomosaic.
Fig. 4 – RGB Close-Up Image: In this close-up image of the surveyed roof area vertical structures on the roof can be seen in reflections on the roof, indicating that there is pooled water on the surface, and those regions correspond to the darker regions in the LWIR.

Working with thermal imagery can be especially tricky for several reason. Unlike many other types of aerial imagery which captures images of reflected radiation, thermal images are images of emitted radiation, that is to say radiation emitted by warm objects. This radiation depends on the emissivity of a material, which is the proportion of how much energy it emits at a given temperature compared to an ideal “black-body” radiator. Consequently all images have to be adjusted and tuned to look at specific materials, and be tuned to compensate for reflected radiation from the background (which is the sky in this case), as well radiation signal loss through the atmosphere due to attenuation by both water and carbon dioxide.

Additionally, just as with orthomosaics constructed from red-green-blue images, we use ground control points (GCPs) to help constrain the reconstruction of LWIR thermal orthomosaics. GCPs for thermal surveys need to be designed slightly differently; while GCPs used for surveys in the visible bands need to have “color” contrast with respect to the ground on which they are placed, GCPs placed for thermal surveys need to have either temperature or emissivity contrast with the ground on which they are placed.

The use of an RTK GNSS receiver or other accurate survey tool allows us to measure this GCPs to within a few centimeters, and consequently a similar level of accuracy can carry over to final data products. This is important so that our thermal orthomosaics can align accurately with RGB orthomosaics, as well as other data layers such and surveys that might show areas of infrastructure; this enables us to relate the thermal signals we see to specific structures or features measured by other means.

Trimble Tech Used:

  • Trimble R10

People Involved:

  • Ryan Wicks, UMass Air
  • Hunter Apteker, BCT undergraduate student
  • Tom Fydenkevez, UMass Amherst Utility Electrical and Mechanical Maintenance Manager
  • Jenna Rostek, Asst. Director Capital Operations

Leave a Reply