Infrared Training Center

Wednesday, January 4, 2012

Analyzing Building Images Acquired at Different Times

A camera user writes:

"I find that images taken of the same structure, not much separated in time, sometimes look very different.  This poses a problem for me as I’m trying to compare the heat images of different houses (to detect homes that need weatherization).  Presently I’m not confident that images of two homes reflect actual differences in the structures, or are caused by minor environmental changes or even by artifacts in the photography.  Here is an example.

Attached are two nighttime images of the front of my house, which faces east.  IR0310 was taken at 10:40 PM, IR0408 at 11:11 PM. (The clock on the camera is two hours fast.)  It was a cold night with little temperature change over the half hour between pictures.  The house thermostat was constant. 

I've set the palette and temperature range to give me good differentiation of houses along the street.  I took the first image of my house as I began imaging houses on my street, and I took the second image when I finished the street scan. 

I am surprised, first, that my house looks so different in the two images. And second, that the outside looks warmer in the later image.  If anything, I'd have expected the outside to have cooled.

Glad for any interpretation of this.  Needless to say, with this kind of variation on a single house, it is hard to get good images for comparing houses. "

Remember that when analyzing buildings with thermography, we are primarily interested in thermal patterns, not individual temperatures, for making diagnoses. When looking at these two images, taken 1/2 hour apart, I would expect the thermal contrast on the latter image to be slightly higher than the contrast on the earlier image. The fact that the temperatures actually appeared to increase leads me to believe that the camera was not fully acclimated and stabilized prior to the survey. I usually suggest that an infrared camera be acclimated to its environment for at least 30 minutes prior to beginning any if serious diagnostic data capture.

To resolve this dilemma, here are the steps I used to analyze these images.

  1. First, I want to make sure that histogram equalization was disabled. I did this by opening these images in the Flir word based reporter program. I right clicked the images, clicked settings, and then clicked advanced. I disabled histogram equalization and confirmed that the linear output was checked.
  2. The next thing I did was choose two items in the thermal image having low thermal mass and temperatures that would represent two extremes that I would not expect to have changed with time. I settled on the walkway lamps to the right of the image and the cold area of vegetation just below them. I adjusted the level and span of each image so these would appear uniform between the images. See figures one and two below.

Now you'll notice that the overall apparent temperature levels have been maintained but figure two indicates slightly higher contrast which is what I would expect as these building components cool. But please note that the general thermal patterns are consistent between the two images and times of data capture.