[Maxbot] wanted a inferared themo-cam that he could use to scan an area and display the results on his computer, the result is this Cheap-Thermocam. This little device is a inexpensive thermal sensor mounted on a pair of servo’s for XY scanning, custom java software and an Arudino to mate the two.
The sanner does a limited but still useful resolution of 42×32 pixels. To help out with smoothness of the colors, every other line of the finished product is interpolated against its neighbors. It takes the device about two minutes to measure each of the 1344 points, but what it lacks in speed it more than makes up for in cost, boasting around 100$ build cost, and ease of construction.
While exact details of what parts you need and how to hook everything up is lacking at this moment, there is a page in the works that will let you know all the good details “soon”, though if you are inspired to replicate one of these nifty devices it seems the most difficult part would be tapping into the sensor’s electronics and the custom software.
Cheap, hacky, and even comes with its own roll of duct tape. We have a short video for you after the break.
31 thoughts on “Arduino Thermo-cam”
SANNING COMPLETE CAPTAIN!
wow, that is a pretty solid picture!
what comes to my mind is, maybe scan in a spiral from the outer corner to the center. that way you could save the time for returning the head between scanlines and shave off a few seconds.
Great idea, agree with lordicewind about spiral scanning to shave off time but Jeeze, someone tell the camera guy that a video camera have a right way up and a wrong way up. Please. Oh and about focus. Thanks!
Geez… what’s with the camera roll… ergh.
But ignoring that, nicely done!
Cool. No, hot…
A sanner darkly?
Finally! An Arduino project worth being positive about, this is good stuff.
Wasn’t this done last year a couple of times?
Good implementation though.
Really cool. Couldn’t you use a spinning/rotating mirror to read rows/columns, would be faster than moving entire thermo-cam.
Nice job man. This would be great for finding places where my heat is escaping my house. Save me a few bucks in heating costs.
I like it. Thermovision cameras used to analyze thermal losses are unnecessarily fast == expensive. It is not like temperature of a wall surface can dramatically change between two picture frames. This thing nicely fits for most (civilian) purposes which don’t require imaging in real time.
I hope next thing Maxbot adds is freehand panning shot thermal scanning: wave the “camera” while holding it in hand and read its movement with accelerometer ic’s instead of precisely controlling its movement with rotator – that would make it more compact, robust and versatile.
Where did he find a thermal camera for $100?
Nice, just few questions:
1. How do you tap into the IR thermometer to read the temperature with a microcontroller? That part is the most crucial of this project and isn’t yet documented.
2. Is it scanning too quickly or slowly? I have a Casio wristwatch with an IR thermometer built-in and when it’s in constant-scanning mode it takes about 1.5 seconds per temperature read, I guess the handheld ones are quicker but how do you determine the optimal read tiem?
3. What’s the brand of the IR thermometer used and where did you get one with a spot distance of 30:1 for less than $100?
Any IR thermometer with a spot distance of less than 30:1 just isn’t worth it for this kind of project and ones with a 30:1 or higher can get very expensive.
@Ben, he didn’t, that’s an IR thermometer which he’s using as a single pixel thermal camera, the servos make it scan an area to build up a picture.
A MAJORITY of that picture is interpolated data, from a thermal sensor that wasn’t intended for scanning use (delay, polling intervals, misread data from polling too quickly). This means the temperature was likely not entirely correct for many of these points.
As all it returns is larger blobs of color, I doubt very highly if this would be terribly useful, though for $100 I doubt it matters.
“inferared” “sanner”? I hate to say it as it has been said so many times before by others, but it drives me nuts. Please get someone to proofread your posts!
That said, an inferared sanner for $100 is pretty impressive even if it is interpolating a lot of the data.
I’m to dizzy to think
First, This is an example of what Hacking’s truly about. Making stuff do our bidding. Good Work! Maxbot gets one of my “Leading By Example” awards for having shared the project so far.
Often- Knowing it can be done is MORE than anyone ever gets from just buying a factory gadget. That’s directed at making something as opposed to just operating shrinkwrapped tools.
Judge not beyond the “does it work?” might be a good sanity check on posts otherwise lacking in reason to hit send? Slagging things for a typoe or a shakycam is petty childish and *NOT* how Hackers are supposed to act! If you want to be that petty? Please rant about such things to yourself in a mirror ok? For the love of Rust folks- Why do we keep seeing such petty flaming here?
RE: <100 thermals? Harbor Freight and the Flea Market vendors have <$20 IR units all day long. Even if the ASIC 's output is formatted for a sorta custom LCD. Copypasta of links skipped- they're too prone to link rot. Moving on a bit:
There was a kid's toy "Spy Gear" unit that sold for <$20 and down almost a dozen years ago. In several variations. All basically a PIR Alarm Sensor modded to show on a bar graph scale or analog meter- the thermal read of what it was point at. A cheap toy barely good enough to show that the TV in a room had been on and the Radio had not. Or if the Dog had been lying on the couch then got up when it heard your key in the lock. But- It's the ancestor of this project's core element. And- before that – there were "Thermopile" sensors for Burglar Alarms that edged into "Barely Semiconductor" tech. Basically an array of sensitive thermocouples in an optics cavity that tripped a heavily biased relay when a human's heat entered it's view. A HARDCORE hack would be building a Thermocoouple Welder and recreating those sensors.
How can we convince a major console maker that their next augmented reality add-on has to have a thermal imager inside – just imagine, thermal cameras for 1/10 the normal cost ;)
I have to agree with Oren Beck on the slating of finer details regarding spelling and camera work – this site is about hacking not pulitzer prize winning write-ups and award winning camera work, if you want your hacks videoed to look polished and slick and devoid of the real nitty-gritty details that enable you to copy the hack yourself, you’re welcome to head on over to KipKay…
But, a $20 IR thermometer is going to be pretty shit if you intend to use one for a project like this, those are the ones that have an 8:1 or even 2:1 beam divergence (meaning 2 units away it’s reading an area of 1 unit diameter), absolutely useless for any sort of detail at any distance.
This infrared thermometer
Has an easily accessible computer interface inside.
Cheap too – Usually goes on sale for less than $20 on a regular basis. I had the idea to do this same project last year and bought one of these to play with. When I opened it up I found a header that could be brought to a computer to control and read the device. Full specs were available online. His thermometer probably has the same interface.
And despite the crappy beam divergence of the HF unit I was still able to detect my cat lying on the floor about 10 feet away. Even though the resolution was crappy and the scan was slow this would still be useful for lots of stuff.
Instead of scanning the sensor to make a 2d image, use compressive sensing. http://dsp.rice.edu/cscamera
I thought you were serious, until I noticed you spelled typo wrong and were just being satirical.
The scanner itself should use a swizzling pattern to provide an initial image quickly then fill in the details using multiple passes. It will let you concentrate on scanning areas of interest instead of the whole scene, as well.
two mini servos from ps3 laser blocks, mirrors?
obviously you would need to use front surface mirrors for this but it should work.
An alternate idea, get two surplus small bearing races from old brushless fans and carefully Superglue six or so identical CDROM deck magnets to either side with alternating poles, being careful not to jam the mechanism.
A useful trick here is to get a spare drawing pin, grind it flat and then position it centrally.
Drop baking soda into the middle and drip glue onto it, wait for it to set then repeat until secured.
Makes a useable polygonal mirror which doubles as a motor with a salvaged driver coil and Hall sensor.
The idea has something to it, but if the mirrors have some emission coefficient at all, then the values will depend on the mirror temperature to some degree. Otherwise they would need to be extremely high grade, probably gold-coated or something like that.
Also, thermopile sensors are not super-fast as someone mentioned above already, they need to dissipate the heat after the radiation source has been removed and the other way around. A polygonal mirror driven by a stepper might do the job though, as the movements are very precise and fast and the sensor has some time to settle while the mirror stays still before the next step occurs.
@salec: My friend did this with a Parallax Propeller and (I think) a Wii Nunchuk (for the accelerometer):
He was originally thinking of going mechanical (with either this method or a Nipkow disk) but decided operator control would be easier to implement. It’s cool to see a comparison of the two approaches.
This is a very good idea! I’m definitely going to buy a laser thermometer and build a laser scanner. This will be really handy to spot heat leakages inside my house.
As for the time it takes to sample a single image, I just need to find a laser thermometer able gather the data a little faster. Maybe even try to replace the circuit deals with temporal filtering (I guess) and replace it with spatial filtering on software.
Thanks for this good post
About lordicewind’s remark, it does not need to return, it can easily scan up to down then move one step right then down to up and just zig-zag that way.
Not that that’s the issue from the looks of it, on the video the return to position is very rapid.
sorry i haven’t responded to all this feedback but I missed that my project was published on this site :D In the meantime I created a homepage with an instruction on how to build this device:
The mentioned “spiral scanning” is a very good point, i will try to implement that.
@Ben: The sensor has a maximum field of view of 5 degree. Check out this datasheet: http://www.melexis.com/Asset/IR-sensor-thermometer-MLX90614-Datasheet-DownloadLink-5152.aspx on page 36.
Please be kind and respectful to help make the comments section excellent. (Comment Policy)