Hello,
I'm experimenting with the Thermal Expert Q1. By now, I've come to understand what other users mean by microbolometers being unstable.
Although TE Q1 does not have an internal FFC flag, it does allow for manual FFC with a lens cap. However, even after waiting five minutes for stabilization and calibrating with the lens cap, I've noticed that readings still lack accuracy.
To fix this, I was wondering if it would be possible to have a referenced object within each thermal picture. This object would have its temperature continuously known by a thermocouple. For simplicity, the thermocouple would simply be attached to the object with an electric tape of known emissivity (and also somewhat equal to whatever is being measured). What would be left to do is to simply identify the electric tape's temperature on the IR camera and compare it with that of the thermocouple, get the difference and apply that difference to the entirety of the picture.
What are the general thoughts for this approach (or if anybody as ever tried it before?)? As a source of error, I thought of the possibility that the electric tape and the thermocouple might simply be not the same temperature. Should there be any other worthy considerations? Another complication would be if the offset isn't constant throughout the map, so the correction made by the reference couldn't be applied to all the other pixels.
Cheers!
I appreciate the link, i'll take the time to go through the document.
Anything below 0.5 degrees would be desirable, currently with just the FFC and 5 minute stabilization I can get up to 1-2 degrees close to the measured temperature.
What do you mean by wildly exactly?
Thank you!
In my experience most "real" thermal cameras under 5000 USD would be struggling even with the 0.5 degrees.
Some results from nameless "yellow manufacturer" sub 10k USD category :
Model A Room temperature offset variation about +-0.5c varying randomly by every internal FFC
Model B drifting about 10 degrees in first 15 minutes after startup with 0 degree target.
Model C (cheapo, sub 2500usd model) drifts so fast after FFC that within seconds there is 1 degree gradient across the screen.
As an added insult there is large(15x15pixels) round "hot spot" that is about 0.8C hotter than surrounding area. Hot spot shows up only after >1hour use.
Something very funky with the optics too, results vary hugely depending on distance.
Model D shows different offset gradient depending on focus distance. Minimum focus distance gives 2C gradient across view area that FFC is unable to compensate.
Flir cameras could be a bit better but I haven't seen enough many of them to say for sure.
Any of the new low cost thermal cameras haven't really impressed me in terms of accuracy, some 20 years old high-end NEC is still way ahead in terms of accuracy and predictability.
Yes I've noticed the impact of distance as well on some results. 1 degree gradient right after an FFC is astounding! Not only can you not get the proper temperature readings, but you also get unreal temperature patterns, making it hard to know how to know if something is actually cooler or warmer than the surrounding environments.
Are the hot spots usually fixed whenever you do some sort of flat surface calibration?
My tests are also being made outside. Under partially clouded skies or bright sunshine I've noticed that the temperature readings on the camera go even more berserk! I feel like it has something to do with the internal temperature reference of the camera not acclimating itself to the environment fast enough.
Yeah... I'm probably asking a lot for this little camera but hey if at least it can get me somewhere remotely close to the proper data then I want to believe that not all is lost!
My tests are also being made outside. Under partially clouded skies or bright sunshine I've noticed that the temperature readings on the camera go even more berserk! I feel like it has something to do with the internal temperature reference of the camera not acclimating itself to the environment fast enough.
Are you sure its not background radiation temperature related?
Even if your object has high emissivity e=0.95 the apparent temperature can vary a bit depending on outdoor conditions. ASSuming +20Cel object in +20Cel outdoor air:
Open blue sky, background radiation temperature probably below -20Cel ! causing 1.6 C error
Dunno what the effective background radiation temperature would be in sunny day... half cloudy with sun just peeping between clouds would be the worst.
@morningglory747 - The main question that comes to my mind is: what is it that you're trying to image and measure?
Others have made the points about "cheap" cameras not being particularly good for true measurements (though they're fine for indication and relative temperature work). My own experience is that my "cheap" (<$5000) thermal cameras can be way off in terms of absolute reading (the inside of my fridge might read +45°C) but if, as you suggest, I put a thermometer in there and discover it's really 5°C then it's very likely that something in the fridge that reads +46°C is actually (46-(45-5))=6°C.
Mind you, in ten minutes it's likely that the self-same things would read a few degrees different.
It may be instructive to research topics such as
accuracy and precision and/or
observational error to get a feel of the general terms you can and can't compensate for.
@mzzj
Interesting, I must say that I am aware that measuring the sky temperature with an IR device has no point, but I didn't think it could have a considerable effect on an object with a high emissivity (I assumed something like 0.1-0.2C at best). I do not quite understand the reasoning behind the 1.6C error you've obtained though, could you go into more details?
@Ultrapurple
This will probably sound impossible, but I am trying to measure temperature of the reproductive organ on flowers, more precisely the part that is called the stigma. Stigmas are small of about ~1mm diameter. These flowers only bloom once a year for like a week so it's a very rush period. My first attempt was foolish in the sense that I didn't stabilize the camera before using it nor did I frequently calibrated it with a flat surface, making my results unusable. I was more careful in my second year. I took the time to stabilize the camera for at least 2 minutes before using it along with frequent calibrations. I also added the referenced object which I’ve just described for extra precautions, in case the pixels deviated too much from the real data. This was more than needed because without this reference, the results would sadly be all over the place. Thus, I probably should have waited more than 2 minutes before stabilizing it or perhaps this camera just simply isn't suited for these sorts of experiments.
Nonetheless, with this referenced object, my results are the best I have... but for some reason they always appear warmer than what it should actually be. This makes me question where I could have gone wrong in the experiment and if it was in fact related to the referenced object I choose.
Interesting... I have never found any noticeable temperature variations on parts of flower, other than those caused by ambient fluctuations. Any differences in apparent temperature I have always put down to being related to the emissivity.
I suspect you're in for an interesting time arranging the necessary optics to get a good DoF. I did some
thermal imaging of the head of a pin heated by a laser and that was ... interesting. You can see and read what I did by following the link. There's no reason your camera would work any differently from the Therm-App I used.
Although I did not dwell to much on the matter, I've had the chance to communicate with a biologist who worked with IR measurement tools for varying leaves and flowers and he agrees with the assessment that the emissivity with the reproductive organ should be somewhat consistent. For petals though that is probably another matter.
In cloudy days, we can see a somewhat uniform temperature distribution across the flower, but on sunny days we can see a clearer pattern across the reproductive organ in the center. Here's an example that illustrates this well (see IR_Sunny.jpg and IR_Cloudy_1.jpg). I must say that IR_Sunny is one of the few cases where I could get a good quality shot of pattern across the flower. For cloudy days, I would usually see what I’m seeing on IR_Cloudy_1.
I actually remember seeing your experiment a while ago because it concerned viewing small objects with an IR camera. I must say that what you've managed to accomplish is certainly impressive! Especially the setup with the combination of both lens. I can't help but feel like this would be difficult to attempt with the Thermal Expert because of how unstable it is with my Android Phone; it only takes a small perturbation for the camera to slightly detach itself from the phone, shutting it off completely and thus necessitating me to wait another 5 minutes before taking any new pictures!
@mzzj
Interesting, I must say that I am aware that measuring the sky temperature with an IR device has no point, but I didn't think it could have a considerable effect on an object with a high emissivity (I assumed something like 0.1-0.2C at best). I do not quite understand the reasoning behind the 1.6C error you've obtained though, could you go into more details?
1.6 degree error comes from "nasty math".
https://www.elcal.ch/files/emissivity.pdfFirst you calculate spectral radiance based on Planc's law, sum it up with the reflected background radiation and then integrate over the spectral range.
Indoor situations are easier as effective background ratiation temperature is usually closer to ambient air temps.
More difficult to estimate the situation outdoors as blue sky has very low radiation temperature even on cozy summer day.
Indoors the effective background temp might be 20c or 22c depending if you are standing right behind the thermal camera.
Outdoors the effective background temperature could be -20c even on nice sunny day with ambient air temp in +20's
(also wavelength dependent, but most thermal cameras operate close to the atmospheric IR window wavelengths
https://en.wikipedia.org/wiki/Infrared_window )
---
Warming up that you see on flower stigma makes sense in sunshine, stigma is in middle of the petals (crappy reflector) that is warmer than ground behind the petals.
I can't help but feel like this would be difficult to attempt with the Thermal Expert because of how unstable it is with my Android Phone; it only takes a small perturbation for the camera to slightly detach itself from the phone
The secret is to use a short male to female (micro) USB extension lead. It's worth experimenting with a cable up to 1m or so long. Then you can put the thermal camera on a tripod or some other mount and not worry about it wobbling in its socket. The lead plugged into your phone doesn't have much mass so it's far less likely to disconnect; if it's loose because of a worn socket then a bit of Blu-tack or similar will probably make it behave.
The main problem with lead length seems to be the DC resistance in the cable causing a voltage drop at the camera end. If you have a decent cable the long lengths may be possible - I had a Therm-App successfully running on more than 10m of good quality USB cable. Yet my Therm-App Pro won't work with more than a few inches of cable, presumably because it draws (much) more power.
When you're searching for an extension cable, make sure you get one that's rated for power AND data - some only connect the power lines, which is fine for a charger extension but not much good in this application. Generally speaking, the cheaper the cable the less likely it is to contain a decent amount of copper, thus it'll be worse for voltage drop.
Transpiration might also play a role here.
Cloudy might mean moister air, hence less cooling due to evaporation. (Assuming plants are dumb and don't regulate... No biologist here
)
Edit: i don't know for flowers but for humans you can actually see individual pores being cooler. Antiperspirant definitely show they work... You might try on plants
For humans you can actually see individual pores being cooler. Antiperspirant definitely show they work...
Very interesting (and way off the original topic) however that's something I must try (if I can find a willing victim to have antiperspirant sprayed through a stencil onto their flesh. Unfortunately the weather is getting cooler here but we don't have a culture here that revolves around mixed saunas - let alone mixed saunas with some prat waving around a thermal camera...
But it sounds fun, though.
Very interesting project @morningglory747. I had a similar experimental project for measuring leaf transpiration using a Seek thermal camera. Inside the lab, we achieved an accuracy under 0.5 degress, but they weren't tested well out in the open
.I used a flat piece of cardboard with minimal reflectance to grab a frame and correct a Seek thermal camera images. Im guessing the correct term for it is Flat field calibration, as you can see here as well,https://github.com/maartenvds/libseek-thermal. (I think I used an older version of the repo). But I'm quite enthusiastic to see what you find and looking forward to contribute more.