A Better, Cheaper Smartphone Thermal Imager

thermal

For the last few years, the prices of infrared thermal imaging devices have fallen through the floor, down from tens of thousands of dollars a decade ago, to just about a grand for a very high-resolution device. This dramatic drop in price was brought about by new sensors, and at the very low-end, there are quite a few very inexpensive low resolution thermal imaging devices.

The goal now, it seems, is to figure out some way to add these infrared devices to a smartphone or tablet. There have been similar projects and Kickstarters before, but [Marius]’s entry for The Hackaday Prize is undercutting all of them, and doing it in a way that’s far, far too clever.

Previous ‘thermal imagers on a smartphone’ projects include the Mu Thermal Camera, a $300 Kickstarter reward that turned out to be vaporware. The IR-Blue is yet another Kickstarter we’ve seen, and something that’s actually shipping for about $200. [Marius] expects his thermal imager to cost just $99. He’s getting away with this pricing with a little bit of crazy electronics, and actually designing a minimum viable product.

Both the Mu Thermal Camera and the IR-Blue communicate with their smartphone host via Bluetooth. [Marius] felt radio modules were unnecessary and inspired by the HiJack system where low-power sensors are powered and read through a headphone jack, realized he could do better.

Always the innovator, [Marius] realized he could improve upon the HiJack power harvesting solution, and got everything working with a prototype. The actual hardware in the sensor is based on an engineering sample of the Omron D6T-1616L IR array module, a 16×16 array of IR pixels displaying thermal data on a portable device at 4 FPS.

It’s interesting, for sure, and half the price and quadruple the resolution of the IR-Blue. Even if [Marius] doesn’t win The Hackaday Prize, he’s at least got a winning Kickstarter on his hands. Video of the 8×8 pixel prototype below.


SpaceWrencherThe project featured in this post is an entry in The Hackaday Prize. Build something awesome and win a trip to space or hundreds of other prizes.


35 thoughts on “A Better, Cheaper Smartphone Thermal Imager

  1. It’s cool to see more thermopile array sensors on the market (FLIR has had a patent monopoly for long enough, thank you) but this sensor only has a range of 5-50C, unlike the Melexis sensor in the IR-Blue that’s -50 to 300C. I’ll stick with my IR-Blue for the moment :)

    Also, the IR-Blue can have a second sensor added to increase the res to 16×8, I should see about getting one…

    (auuuugh! AUD$100 in single quantities! Glad I got in on the kickstarter when they were half the price…)

    1. The OMRON sensor (at least the D6T-44L-06 4 x 4 pixels, that I’ve tested so far) has a better measurement range in practice than ‘specified’ (note they say ‘detection range’ in the datasheet). I’ve successfully tested it from (-5 to 100 C). May even go lower than -5 C, but certainly not higher than 100 C.

      – will post a video comparing D6T-44L-06, AMG8832 and MLX90620 performance later this week on the project page

      In the end, it all depends on what you’re actually trying to measure :)

      1. That’s usually way specs in datasheet are written, so essentially all bets are off as far as the manufacturer is concerned. It is fine for a one off or if you screen the parts individually or large enough statistical sample from a production lot.

        Don’t rely on the performance (as in linearity, accuracy, noise level etc) of a part operating outside of its tested/designed range.

        1. No, in this case I don’t think it’s a matter of individual characteristics of the part as much as it’s a matter of the purpose intended by the manufacturer (people detection) and the range that it is calibrated for. For that range the performance is as specified by the datasheet.

          As the most usual applications for a thermal imager roughly fall within this specified range (5-50*C), I find it suitable.

          On top of that, an extended range in which temperatures are not measured with the same accuracy but still measured within what’s considered an acceptable error, I can live with that and it’s still useful.

          For any other application that needs temperature measurement over 100*C, well you would need a different tool :)

    2. Yeah, this sensor is meant for people tracking/detection, as the datasheet says: “High Sensitivity Enables Detection of Stationary Human Presence” but i still think it is a cool project.

      I am actually waiting to get hyper spectral imaging in smartphones. Can you imagine the possibilities for instagram filters ?

    3. > IR-Blue can have a second sensor added to increase the res to 16×8

      this sounds pretty TERRIBLE
      two 4×4 sensors should give you instant 9×9.
      magic allowing this is called interpolation, you wont be able to align two 4×4 sensors to only cover separate parts of the measured space, instead you make them cover same area and exploit misalignment to extract additional information

  2. Very interested in getting a fully made 16×16 setup for my tablet.

    It does beg the question though, how long until phones/tablets have a 16×16 thermal sensor built into them?

  3. Hijack is damn nifty, but don’t you get USB OTG in pretty much every tablet/smartphone? I don’t understand pinching pennies that hard on the device connection for a ~100 dollar gadget, unless I’m wrong about the USB prevalence and bluetooth really is the only other choice.

    1. Personally I wouldn’t want a USB-OTG connection because I might want to have the device charging or communicating with other peripherals while using it. Odd that bluetooth represents such a large cost increase, though.

        1. Implementing the same audio-solution for Android phones would be a major hurdle, mostly because of the variance in audio circuits used for headphones driving. Some phones will simply deliver insufficient driving voltage or current for my hardware to work.

          USB for Android is a safer bet for device compatibility.

        2. Take a look at NXP’S AN11552 “Smartphone Quick-Jack Solution” datasheet , chapter 8.1 and it explains that the vast diversity of hardware that runs android is so different , drivers and all that it is difficult to ensure that apps will work on all devices, especially using low level I/O and re-purposing the phone’s hardware.

          I would of pasted the whole explanation why their product can’t be guaranteed to work on androids, but their datasheet does not want to paste properly and I’m not too lazy to reformat it.

          1. Thanks, I just had a look.
            I guess it makes more sense to use OTG, I think it is one of the great things about mobile phones.

            It would be interesting if some sort of standard of communication over the headphone jack would develop, but I suspect it is not that much interest in it due to the low cost of the USB.

    2. As far as iOS devices are concerned, it has to do with the licensing costs and royalties (Apple MFi program) of getting an accessory officially supported (app. in App Store).

      The only ways to dodge that (at this point) is either audio comm. or BLE.

      For Android, PC there will be an USB connection.

  4. hmm, a wii remote has a HD video camera with an IR filter on it, and bluetooth for <$40

    you would have to defeat it's internal software that just reports the location of the brightest IR spots, but I wouldn't be surprised to find that the wiimote hacking community has done this already

    now, the sensor this article is about may be a lot more sensitive, but it would be good to get a comparison.

        1. “Infrared” is actually an awful big swath of the spectrum.
          http://en.wikipedia.org/wiki/Infrared says 700nm to 1mm. Compare that to the range we can see, 300nm to 700nm.
          The IR you get from the LEDs in remote controls and the Wii’s “sensor” bar is 900-1,000nm, something like that. The IR you’re interested in for thermal imaging is more like 8,000-15,000nm, not even in the ballpark. I think you actually need exotic lens materials to handle that, not just an exotic sensor.

  5. The problem with this project and the similar ones before it is that the resolution is so limited, making the device pretty much useless.

    Personally I’ll wait for the Flir One. High resolution and fairly reasonable price.

          1. Thanks for the link. That answers one big question I had, how do they handle ITAR, and it is the low frame rate.I have several other cameras, DRS and FLIR that are up to 640×480 and definitely fall under ITAR.

  6. Very cool. Nice to see more projects in the thermal imaging space. Problem is – as always – sensor resolution/price and integrating the components into a usable device.

    I have done a much simpler project with mechanical scanning (very slow, but reasonable resolution!) and a Raspberry Pi (Works as WiFi Hotspot/Server). The thing with a Raspberry Pi is that you can easily generate nice looking images with python libraries and do the complete UI/image generation via an embedded Webserver over WiFi.

    See

    http://www.µc.net/2013/12/warmebildkamera-mit-dem-raspberry-pi-4.html
    (unfortunately it’s in german though :-( )

    Sourcecode is available on github
    https://github.com/hermann-kurz/thermography-raspberrypi

    1. Nice resolution. The mechanical scanning reminds me of laser projectors. Aluminum is reflective to thermal IR. You could buy or create some aluminum mirrors. They’ll be lighter than the sensor and you don’t have to stack the servos. That should make it faster without adding much cost.

  7. You do need to combine a regular cam image with the IR image for good results.
    But I get it’s early days.

    Also we need a PC version of such a thing since android and apple are just too damn nosy to use their products. And a PC is still easier to control.

Leave a Reply to polCancel 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.