Inspiration can strike from the strangest places. Unearthing a forgotten Melexis MLX90614 thermopile from his ‘inbox,’ [Saulius Lukse] used it to build a panoramic thermal camera.
[Lukse] made use of an ATmega328 to control the thermal sensor, and used the project to test a pair of two rotary stage motors he designed for tilt and pan, with some slip rings to keep it in motion as it captures a scene. That said, taking a 720 x 360 panoramic image one pixel at a time takes over an hour, and compiling all that information into an intelligible picture is no small feat either. An occasional hiccup are dead pixels in the image, but those are quickly filled in by averaging the temperature of adjoining pixels.
The camera rig works — and it does turn out a nice picture — but [Lukse] says an upgraded infrared camera to captured larger images at a time and higher resolution would not be unwelcome.
Another clever use of a thermopile might take you the route of this thermal flashlight. if you don’t build your own thermal camera outright.
[Thanks for the tip, Imn!]
10 thoughts on “Thermal Panorama One Pixel At A Time”
I do have one of these sensors. Ijust would actually have an use for a thermal camera, and so it seems just renting one would be much much cheaper than bulding this one. That said, it’s a hack!
The MLX90621 only costs $20 more but you get a 16X4 array that is easy to use.
They have MLX90640 32×24 pixel available from september 2017 on their website. Now, if that will be cheap enough, it will be the start of thermal cameras.
Price? Who knows. https://www.youtube.com/watch?v=WSZ3GGDusTk
*the start of thermal cameras everywhere.
This is fascinating, not the least because it shows the angular resolution of these cameras. Great article.
Earlier thermal cameras also used mechanical scan. Normally with a fast rotating prism mirror (like in a laser printer) for x-deflection and a servo or prism mirror for y-deflection. So you do not need any sliprings and don’t need to move the whole sensor assembly, only mirrors.
Can be done, just don’t have mirrors that work in far IR range by hand. Also scanning 360×180 with mirrors is kind of complicated.
2 mirrors should be all that’s needed for 360×180 range (in theory it could do even more, but the base gets in the way), you just have to use a different geometry of the “arm”…
I am referring 360×180 as full visible sphere (rotation 360*deg, tilt 180*deg). How much more do you speak of? :)
More then 180° tilt :P
The 2nd mirror can theoretically turn 360° without interrupting the optical path as well, but obviously at some point, the base of whatever it’s bolted to will block the view.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)