Thermal cameras hold an enduring fascination as well as being a useful tool for the engineer. After all, who wouldn’t want to point one at random things around the bench, laughing with glee at finding things warmer or colder than expected? But they’ve always been so expensive, and a lot of the efforts that have sought to provide one for little outlay have been rather disappointing.
This has not deterred [Offer] though, who has made an extremely professional-looking thermal camera using an M5Stack ESP32-based computer module and an AMG8833 thermal sensor array module in a 3D-printed case that copies those you’d find on a commercial unit. The modular approach makes it a simple prospect for the constructor, the software can be found on GitHub, and the case files are hosted on Thingiverse. You’ll be finding warm and cold things on your bench in no time, as the video below shows.
Most of the thermal cameras we’ve seen have centred upon the FLIR Lepton module, but that’s a component that remains expensive. This project shows us that thermal cameras are a technology that is slowly becoming affordable, and that greater things are to come.
Thanks [CYK] for the tip.
The AMG8833 is an 8×8 pixel sensor which is a good start. Upgrading to a MLX90640 32X24 pixel sensor ($45) would make it really nice – https://www.digikey.com/product-detail/en/melexis-technologies-nv/MLX90640ESF-BAB-000-SP/MLX90640ESF-BAB-000-SP-ND/8638466
That would seem to be a nice upgrade, but it looks like it’s not available yet.
Oops. Mouser has them in stock presently, or Sparkfun has breakouts for $60. https://www.mouser.com/Search/Refine.aspx?Keyword=MLX90640
thanks for sharing, the bare part price on digikey is only 1.5 times as expensive but has 12x the pixel count, also refresh rates up to 64Hz? at noisier temperature readings of course.
The article is missing that the sensor is only 8 by 8 pixels (and what’s displayed is only an interpolated picture).
Do we really need an amazing resolution to look at thermal images, though?
I agree that 8×8 is not much, but I’m pretty sure most people would be more than happy with something around 64×64 pixels or even 256×256. There’s a point where adding more pixels won’t really help for something like thermal imaging.
What’s needed is a faster scanning rate, the image is that video seemed to lag a bit behind what was being pointed at – unless it’s caused by the interpolation and the speed of the ESP32.
Your comparison doesnt make sense, 256×256 pixels is already very good and amongst the contemporary thermal cameras <= 1k€ (Thermal Seek, Thermal Expert, Flir One…). Its like saying "I agree, a Hotwheels toycar isnt that nice, but most people would be more than happy with a Ford Mustang."
Bench is one. Merging with a regular image works great around the house.
FLIR holds patent for that, right?
A patent on compositing? I doubt it.
FLIR say patent pending for their MSX tech which overlays visible light data ontop of thermal data.
looks like it has:
https://youtu.be/Ey-MI66a7jE?t=20
I hope thermal sensors won’t become commonplace like say microcontrollers. Had they become ubiquitous like Arduino, terrorists would use them to build SAM missiles and MH17 would happen every month or so.
If they have enough money and knowledge to build a missile, I think the least of their trouble would be sourcing thermal sensors.
And considering IR missle guidance is 1950’s level tech and something like a 4 cell sensor….
That is why it has a 10Hz refresh rate to keep it from being a good guidance system. So what if it is 1950s tech so are ICBMs and Thermonuclear warheads. That does not make it simple to reproduce.
What’s the problem with 10Hz refresh rate? That seems plenty fast enough to me. Jumbo jets aren’t known for thrust-vectoring around like mosquitoes.
First sidewinder had 1 sensor and something like 6 tubes. It was a masterpiece of mechanical and electronic integration.
Much easier to ask spy agencies for a few crates. Between the US, Russia, China, & the Saudis someone will be happy to oblige.
*sigh*
‘THE TERRORISTS!!!!!one’ have been able to get thermal sensors ever since they were first sold commercially. The reason they’re not useful is that all non-military sensors are hardware limited to 9Hz refresh rate making them worthless for missile tracking. Whether a “TERRORIST’ has one or a million of those sensors is completely irrelevant.
Well, if they had an array of 9 they could get 81 frames a second, would that work? ( still cheap) A million might be overkill, processing nine million frames a second would be pretty challenging.
Consider that, just for starters, we live in a society that sells a very powerful explosive at every filling station, and everything needed for simple chemical weapons at every hardware store and pool supply. No questions asked. If we were being told the truth about the origins and motives of terrorism, you’d expect our cities to look like the set of Escape From New York by now. Yet, the knuckle draggers you’re worrying about never manage to do much better than machetes and rental trucks without some government’s intelligence agency leading them by the hand, and can’t seem to figure out that the airport and its RADAR systems are a much higher value target than individual aircraft. Think about what’s wrong with that picture for a minute, and you’ll realize that the real terrorists generally wear Armani, hold public office or sit in board rooms, and can buy their toys right off of the international arms market ready made.
Even with the 10hz refresh rate, that’s decent cheap thermal imaging. I wonder if some active cooling would boost the sensitivty in a useful way?
Nice. But my $200 FLIR ONE is a lot nicer :)
that’s not even close to the point here… it’s a *HOME MADE* thermal camera! who the F!$# cares about your pre-made unit? about all that’s good for here is validation and calibration.
very nice project.
I also found this video to be very informative and inspiring: https://www.youtube.com/watch?v=A9F1ezGgaC4
I just ordered my from MLX90640 w/ breakout board from Sparkfun last week and it was delivered last night. Here’s a thermal selfie: http://tinyimg.io/i/0xlyKx3.png Processing keeps giving me errors about the yellow highlighted portion of my snapshot though. It keeps saying something about it being out of bounds, but if i stop and start it enough the image starts streaming again. It must be an issue with timing, but I didn’t have time to work on it much. The breakout has 4 wires (using i2c) that i have going to a Teensy 3.5 connected via usb. It works really well and was very responsive! I was pointing it at all sort of things like my processor and a/c outlet in my room. Now I’m trying to evolve into Predator and move to the jungle.
Judging from your “selfie”, you are bald, have no beard, and your left arm is way smaller than your right arm.
B^)
Haha funny enough I have a full head of hair and just a little bit of a beard. If you look at my head you can almost make out my glasses the have polycarbonite lenses (with an antireflective coating — not sure if that matters) that are blocking the thermal sig a bit. Also my left arm and body was turned in weird way while I was waving at the camera so I think a lot of my upper left arm was tucked close to my torso and might have been blocked a little by it. That and my evolution into Predator probably was starting. roflcopter :P
Hello, very good project !
I printed the 3D parts and i bought the M5stack and the sensor. But, i am not very familiar with the coding part. I see a cpp file and a ino file. But, there is no makefile or soemthing like this in order to compile it ? Could you explain a little bit more how to compile/push the generated binary to the the m5stack ? probably with a microsd card, no ?
thank you for your help.
Again, they good concept.
Jean
maybe someone can help me with some hints.
I want to use mlx90640 as thermal camera for a smart home project. my ideea is to place it in the corridor and count how many people are in house and in which room. i am willing to use the module with raspberry pi zero w but i can not decide which variant should I buy, 55° or 110°. my corridor is about 2x2m and I want to place it near the bulb light. any advice would be highly appreciated.
Hey radu! You might want to reconsider placing the thermal camera so close to the light because the heat from the light will make a bright spot at one corner of the thermal cameras sensor. Not only that but as the light stays on the air around the camera and the camera itself will get hotter and hotter and I’m not sure you will get a quality picture from the camera. I’m not sure if it can be damaged by the heat from the light so you may want to look at the spec sheet for the thermal camera too. The camera can adjust “exposure” based on the temperature difference between the objects on the camera but as that range goes from one extreme (hot light to much cooler person) to another you will not get as much range in terms of sensitivity to someone’s face or hands in comparison to their clothing or something which could be useful for counting people passing by. Also try to keep the light (or other hot things) out of the view of the camera in general if possible to prevent these extremes. As far as field of view go I’d try to go with the wider fov sensor which I’m assuming is the 110 fov model but I’m not certain. Keep in mind the wider the fov is the less resolution per degree you’ll get which means working out details will be more difficult. I got the 55 degree one myself I believe and I probably needed the wider fov one myself.
Hi Jon, thanks for reply. It is a very good observation about placing the sensor too close to the light bulb, but my wish is to have a view from the top and also connection to mains power.
I will buy the 110 degrees as resolution is not my biggest wish, but only to count persons and track their movement
Is this detect a water leakage in wall and floor.?
It could work. Depends how deep the water leakage is in the wall and how much water ist leaking.