Building A Thermal Imaging Sensor From Scratch


[Rob] lives in a 100-year-old house, and with these antique lath and plaster walls and old window frames comes a terrible amount of drafts. The usual way to combat this energy inefficiency is with a thermal imaging camera, a device that overlays the temperature of an object with a video image. These cameras are hideously expensive so [Rob] did what any of us would do and built his own.

The build centers around a Melexis MLX90620 far infrared thermopile that can be had for about $80. Basically, this sensor is a very, very low resolution camera (16×4 pixels) that senses heat instead of light. By sticking this sensor on a breadboard with an Arduino Mini and WiFly network adapter, [Rob] is able to pull the data down from the IR sensor to his iPhone and overlay it on the feed from the camera.

The result, as seen in the video above, is a low-resolution but still very useful thermal imaging camera, perfect for looking for cold drafts in an old house or tracking down [Arnie] just like a Predator.

Tip ‘o the hat to [Ronald] for sending this one in.

45 thoughts on “Building A Thermal Imaging Sensor From Scratch

  1. Nice! I didn’t know they had an array like that. I’ve been looking at 8×8 and 32×32 arrays, but they’re pricey (qty 100, approx $100 and $400 respectively).

    For the curious, that was a year and a half ago:

    This one, though, is all digital, not a clocked analog signal like those Heimann sensors I had found. the achieveable framerate is very impressive, too (at a loss of accuracy of course).

    Very nice, though. thermal cameras can be a pain to use because you can lose your frame of reference to find a thermal point. Now, you may not get all the thermal detail of a standard bolometer, but I think this is great for houses. It’s not like you’re looking for the hot wire in a wire bundle.

      1. In all fairness, he traded money for hours. It would be interesting to know how his work equates in $/hr if you compare it to a commercial option.

        But, this is one awesome project and as it’s open source my comment about $/hr is nothing more than an exercise on the back of a napkin.

      1. The FLIR i range are a definite step in the right direction, price wise, but was a bit disapppointed to see they don’t have any sort of video-out connections. A basic composite video connection would be most welcome for connecting to a small DVR/camcorder.

  2. I have always wanted one of thees. This is cool – or hot – depends :)Perhaps with some lenses (ZnSe only ones that would work) and Niplow scanning disk arraignment, the resolution could be improved.

      1. Germanium is very expensive. And so is sapphire. If you want to go “all out” you can use a diamond lens (normally used only in extremely high power CO2 lasers). Using a custom diamond lens with a $65 thermal sensor would be a cool hack…

  3. It’s probably worth pointing out that you can rent a nice $2000 professional version with SD card capability from Home Depot for about $50 for a 24-hour period…

    For smaller spot work I have successfully used non-contact thermometers to locate cold spots on walls near windows/doors, as well as floor/ceiling areas, to pinpoint where wind/air must be sneaking in.

  4. I had to hunt around for it on the pages linked, but apparently this is the hardware setup in question: The sensor is an MLX90620, connected to an Arduino, which is in turn connected to a wifi breakout board. The iPhone gets the sensor data via wifi and overlays it with the live video feed.

    1. No, you will be able to see near infrared (like the one used by TV remotes), but not thermal infrared.

      Near infrared photo can be done by a modified camera, or by a camera with nightshot features. You can also use a filter to block visible light, and get only near infrared: a good filter can be done with a piece of processed photo film. Near infrared photos are very beautiful, but they don’t provide any thermal info.

      Read this for more info:

      1. Late reply, but anyone who happens to be going through comments: this sensor array is a TRUE FIR sensor, and can in fact measure um+ wavelengths. It is not a standard CCD/CMOS array, and they are not using the iPhone camera to measure FIR (which is, as you note, stupid). Spend a bit more time reading before deciding to bash a hack?

  5. During one of my searches for DIY thermal imaging stuff I found a project someone did, it was a ‘thermal torch’ with an IR temperature sensor and a bunch of bright red,green & blue LEDs.
    The idea was whatever you pointed the torch at the LEDs would shine coloured light onto it showing wether it was hot, warm or cold.

    If I can find the page again I’ll link it here.

  6. Since this data is being fed into software on the smartphone, there are loads of applications for this. One that springs to mind would be pointing it at your breadboard, where it could signal an alert if a component became hot enough to burn skin.

  7. That’s very neat. I like that.

    I suggest upping the frame rate a bit to say 30fps to be closer to the iphone camera and then try implementing motion tracking in software. That way he could possibly build up a more detailed heat map by sweeping the device around a little instead of holding it steady.

  8. First off, kudos to a really cool idea well implemented. That is really clean.

    I’ve looked into that sensor and my main gripe with it is not the resolution, it’s the refresh speed. If it could update around 20-30 times a second, I would be using it for all kinds of things. As it stands it’s possible to get stand alone units with 100×100 pix 5-10 hz on ebay for around 500$. Real time gray scale units for around 900$. (broken high end ones for even cheaper)

    For those of you who have never used a highly sensitive thermal imager before, I found an unforeseen useful side effect: the worlds most expensive stud finder. That’s right, the wood support beams can be seen pretty clearly through the wall board. It’s pretty awesome when determining where to install shelf brackets, etc.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.