The FLIR One, a heat camera for the iPhone, is now available
flir.comJust a note on UX and visualization: Mapping temperature to a color/wavelength is not the perceptually best way to convey data. Humans are not good at estimating the distance between two data points when using a rainbow color scale. Using luminosity would be much better, or even a color scheme that uses saturation. See e.g. http://colorbrewer2.org/. Different schemes are perceptually good to use depending on whether your data is categorical, scalar or(/and) has a fixed center value that has semantic meaning. It doesn't look like the actual camera uses this mapping, but the demo pictures on the front page do. I'm not sure which color scale is used for the actual camera interface.
A side note to this: Sometimes, users expect a particular color mapping and will object to using a coloring scheme that is perceptually better. E.g. doctors often view diffusion tensor images where each point in the image represents a 3-dimensional value, using the RGB colors for each dimension. This is a perceptually horrific choice, since practical demonstrations would reveal that there is significant perceptual ambiguity when viewing data represented like this. But an engineer once made a prototype that used this mapping, and now the operators are unwilling to change their habits.
It probably wouldn't be hard to make an option that switches between different types of viewing styles, it's all on the software side.
Here is one camera that can switch views: http://nightvisionplanet.com/media/catalog/product/cache/1/i...
It doesn't seem too strange to use a scale that corresponds to commonly-encountered heat levels, as perceived - yellow, orange, red, white, blue or something - it's not exactly logical, but people grasp it quickly because it has a familiar analogue. You could use a plain luminosity based presentation, but that's not the way humans have experienced heat for the last 40,000 years. It may not be mathematically perfect, but it's what we've got.
The style of visualisation is indeed configurable within their app, as noted in Wired's review:
http://www.wired.com/2014/08/a-review-of-the-iphone-infrared...
There are, apparently, also third-party apps available.
I had a chance to try out the flir one for a day. Pretty sweet but not perfect. It has a number of color scales prominently featured in the iPhone UI. Grayscale might match lumosity.
I used "contrast" to quickly identify hot / cold parts of my house, then "spot meter" to measure the actual temp. Found out part of my attic is not insulated. That's money going out through the roof.
Interesting. As a FLIR camera used to be expensive, I built my own thermal IR camera for $120 (2011): http://oi60.tinypic.com/2820yg2.jpg
Do you have a tutorial?
Sorry, not at the moment. I am using an Arduino, two servo motors, a medical grade thermal IR sensor with a small point of view and a webcam, all connected to a notebook. Shooting a thermal picture takes up to 60sec, so it is not completely comparable with a IR-CMOS that FLIR cameras use, but a lot cheaper and the measured temperatures are more precise.
Probably a scanning IR thermometer. Not really the same thing...
you are right, well almost. I am using not an IR thermometer, but medical grade IR sensor. But you can reuse an IR thermometer too (with a bit lower grade results).
I got one of these to debug thermal leakage. I very much like it.
Here's a shot of my laptop: http://bochs.info/img/IMG_0676-20140822-211311.png
If anyone's curious about something specific, I'll gladly take requests for photos. It's super interesting to see things from a temperature perspective.
If I had an iPhone 5 or 5s, I'd seriously consider this to check for hot spots in a lamp I converted from halogen to LED. Bad connections can generate heat. I'm a novice when it comes to electrical construction, and so it would be nice to have some confirmation that I didn't botch any connections. I've used a non-contact IR thermometer to check for hot spots and have found none other than were expected (the heat sinks of the LED bulbs), but this method doesn't have very good resolution.
In case anyone has an old halogen torchiere lamp sitting around and would be interested in such a conversion, here are some photos of mine to give some ideas: http://imgur.com/a/XbtNc#0
The first photo shows the mount I constructed to hold the bulbs. The parts are four 2.5" corner braces, four 4" mending plates, four Leviton 3352-F lamp holders, and several machine screws, nuts, and lock washers. (There are no lock washers in that photo, because that was a test to check fit, not the final assembly). All the parts for the mount, except the lamp holders, are at the big stores like Home Depot or Lowes for a total of under $10. The lamp holders were at Do it Best Hardware for around $10 for four of them. [1]
I did not have to drill any holes to make this work. Everything is using the holes that are built into the corner braces and mending plates. In particular, the distance between the diagonal holes on the bottom turns out to differ by only about 1 mm from the distance between the holes on the lamp bowl that were used to attach the original halogen socket. That's enough overlap for a #6 machine screw to easily go through and securely attach the thing.
The second photo shows the 4 switches I added to the lamp bowl to control the lights. There are two "on-off" switches, and two "on-off-on" switches. Each lamp holder is controlled by one of the switches. The two on the "on-off" switches can simply be turned on and off with their switch. The two on the "on-off-on" switches are wired so one way turns them directly on, and the other way connects them to the dimmer that is in the lamp pole. Yes, I am aware that I am morally obligated to paint this to look like the underside of a cow.
My original plan was to have two 1600 lumen 2700k bulbs, and two 1600 lumen 5000k bulbs. This would give me a lot of flexibility. If I wanted 2700k light, I could switch one bulb to the dimmer to get from low to 1600, and I could turn the other bulb on to go from 1600 to 3200. Same if I want 5000k light. And if I want even more light, I could do both of these at the same time. [2]
The third photo shows it with everything wired up. It is not pretty, as I pretty much just went point to point, but did not do a very good job of picking segment lengths. And yes, I realize I should insulate the backs of the switches. I'll be using electrical tape to cover those exposed terminals. They are not covered yet because I'm still in the testing phase. I also might rewire it to make the wiring cleaner. (I'm also considering putting the switches in a plastic box)
The gray and orange 3 way connectors are Wago Lever-Nut connectors. They are awesome. They can take any mix of 12-28 gauge wire, they connect by simply lifting the lever, inserting the wire, and then lowering the lever. They disconnect by reversing that, and the wire comes back out completely undamaged, and the connector is completely reusable. Rated for 600 volts, 20 amps. I used these for the connections that connect the stuff in the bowl to the wires coming out of the pole so that the bowl can fairly easily be removed from the pole. Their only downside is that they are more expensive than other connectors, like wire nuts. A box of 50 from Amazon worked out to $0.40 per connector.
The yellow connectors are In-Sure push in connectors. It is possible to remove them, but they are not re-usable, so are really meant for permanent connections. A jar of 200 from Amazon was $0.13 per connector [3].
Not shown is the shade. This design places the bulbs above the rim of the bowl, and this leads to harsh shadows in the room. Right now, I have a simple shade I made by taking some white construction paper and making a cylinder of just under the diameter of the bowl. This sits on top and diffuses the light, although it also absorbs quit a bit. I think I want something that only blocks the front quarter. Currently trying to decide mechanically how to make that work...looking at making some kind of frame out of wire to hold the paper (or some more translucent material) in place. I'm also trying to figure out if I can make some kind of adjustable shade that would let me vary the amount of light blocked, which could serve as an alternative to electronic dimming.
[1] The most surprising thing I learned during this project was that buying these parts from my local big chain hardware stores in person was about 1/2 the price I could get them from Amazon.
[2] Another approach I've toyed with is three 1600 lumen bulbs and one 800 lumen bulb. For very low to 800 lumen, I'd use the 800 lumen bulb on the dimmer. For 800 to 1600, I'd use one of the 1600 lumen bulbs on the dimmer. Add in the other two 1600s to bring the range up to 5500. I've also considered getting rid of the dimmer, and going with 450 lumen, 800 lumen, 1600 lumen, 1600 lumen.
LED bulbs do not dim all the way to zero. Their range is something like 20-100% (my estimate from eyeballing it). That's why I did not go with the simplest conversion, which would be four 1600 lumen bulbs all on the dimmer and no separate switches. The dimmest non-zero output would then be something like the equivalent of a 75 watt incandescent.
[3] I should have checked Home Depot. They stock them locally, in 100 packs, for $0.07 per connector.
What's the max image resolution of the thermal-only and thermal with photo overlay pic?
The photos I've been taking come out as 535x401. It looks like the visible light camera it uses is just VGA (640x480), but the thermal camera resolution is probably the bottleneck anyway.
Can you take a thermal-only picture too?
The cheapest FLIR camera so far has a (afaik) 240x240 low resolution IR sensor. So they maybe combine an upscaled IR image over an actual VGA camera photo.
It's worse than that, most likely. The cheapest FLIR cameras have 320x240 IR sensors, but they downsample to 80x60 unless you either pony up several thousand dollars for the E8 model or buy a $1000 E4 and hack it. There's no way they're enabling 320x240 IR imaging in a $350 iPhone peripheral. If they were, they'd be trumpeting it all over their marketing literature, rather than neglecting to mention resolution at all as they're doing.
And yes, launching this thing a few days before the iPhone 6 announcement is about the stupidest goddamn marketing move since the Osborne 1. Somebody needs to lose their job over that.
In FLIR's defense, 80x60 is still very useful for a lot of things. The ability to overlay low-resolution IR and moderate-resolution visual images is sort of a cheesy gimmick, but it makes the low-res IR sensors vastly more useful. (My E4 is hacked for 320x240 support, so I usually turn the MSX overlay off.)
It may be useful for a lot of things, but it's also a form of product segmentation that wouldn't be happening if there was actual proper competition in the market. 320x240 sensors appear to be cheaper than lower-res ones these days, which is probably why they're downsampling, and I think most of the intended applications benefit from the extra resolution, just not enough to justify ponying up several thousand dollars more.
Pick one: either it is a cheesy gimmick or it makes the sensor vastly more useful. Those are pretty much opposites, it can't be both.
MSX is a band-aid to cover up low resolution IR. With 80x60 IR, MSX is a useful feature, because without it you often can't tell what you're looking at. But with 320x240 IR it's not needed in most cases. Worse, because the IR and visual lenses aren't coaxial, the resulting parallax error tends to actively deceive you about the exact location of small hot spots.
If FLIR didn't go out of their way to cripple the 320x240 sensors in their low-end models, MSX wouldn't be worth incorporating at all.
The only photo settings in the current app:
- Color (you can pick how the color scale is rendered)
- Sensitivity (matte, semi-matte, semi-glossy, glossy)
They're releasing a few more apps of their own and an SDK, so it might be possible in the future to get the raw IR image.
What's the maximum distance that the camera can get infrared image from? If you are outside can you see people from a deletion enter distance, 20-30 feet away?
I haven't tested that exactly yet, but here's a support response I got from FLIR on temperature accuracy:
> At about 100m, you can see the heat signature of a person. However, atmospheric conditions can influence the visible range. Dry, clear weather will provide a greater range than rainy, foggy or humid conditions as water in the atmosphere can absorb infrared radiation before reaching the sensor.
> The spot temperature is a close representation reading for conditions testing and monitoring. It may vary 2 degrees from actual temperature.
That seems like a lot of variance. I would like my thermometer to be accurate to at least 1°.
IR imaging isn't the best technique when you need extremely high accuracy. The reason is that you usually don't know the emissivity of the surface that precisely, or for that matter its reflectivity.
The sensor is capable of good resolution, accuracy, and repeatability, but it doesn't always tell you what you assume it's telling you. People who use these things professionally usually need to take courses on how to interpret the results.
Can it see the heat from a bird in a tree? As someone who birdwatches for a hobby, I wonder how good it is to detect wildlife.
What am I seeing on the right below the LCD, is that the backlight driver?
I think it's the CPU's heatsink/fan. The warm cable on the left is a thunderbolt ethernet adaptor.
Oh a cable, of course. I thought I was looking at hot air blowing out from a vent on the side.
Macbook Pro?
It's last year's 13" Retina Pro.
Worth mentioning the "Affordable thermal imaging" Kickstarter project[1] which currently has about 14 days to go. The nice thing about this product is that it is not tied to a specific phone or tablet, is less expensive than the FLIR One, and you don't need to frequently close a "shutter" to re-calibrate to boot.
[1] https://www.kickstarter.com/projects/1075169276/hemaimager-a...
I'm not really thrilled about it being for a specific phone that is likely to have a physical form factor not guaranteed to have any particular lifetime.
Even with that... the non-iphone versions they previously sold were thousands of dollars. This is a huge leap in affordability even if you have to buy an old iphone 5 to go with it.
For instance, the E4 they sell is $1000 and up.
Fine. I bought it. It's amazing.
They address that in the FAQ with something like "you can always keep your iPhone 5 around as long as you need to use this if it doesn't fit the iPhone 6"
In my opinion, it wouldn't be that hard to adapt to a wider iPhone. It's two pieces: a case with a hole in the back for mounting, and the camera brick - which only touches the back and bottom of the phone.
I want to get one but they launched their product disastrously close to the iphone6 launch. I'm going to wait for their iPhone 6 version because I don't intend to keep my iphone5 around.
I still have my iPhone 4 but it is gathering dust and the batteries are dead so I don't want to waste $350 on something similar.
How much? The buy now link says "FLIR ONE™ is Not Currently Available Outside the US"
$349
Wanted something like this the other day to see how heat was moving around a server room...
Most people use an infrared thermometer 'gun' for this. You can buy them at Home Depot...
FLIR captures some interesting stuff in F1 Racing https://www.youtube.com/watch?v=jvuBe6b2iVk
Is there something specific about the camera in the Iphone that would preclude doing a similar thing to android phones? Or is it just a matter of uniformity of the housing?
That is probably precisely the reason why it's not available for Android, there are hundreds of different physical form factors and many more configurations.
Yeah, not so much the phone's camera as figuring out how to strap the thermal camera on and plug it in. It has its own internal battery as well, so it's not the smallest load.
It does not use the iPhone's camera, I think due to parallax problems. It has both an IR and visible light camera side by side. This means it can take both a thermal and visible image and overlay them, so you can compare or something.
Like this? http://therm-app.com/
$999 is a bit pricey I think.
You can buy a pro-sumer device for that money:
Well, the Opgal device has a resolution of 384 x 288 while the E4 lists 80x60. It's probably worth the asking price.
That's pretty nifty, and a much more elegant product design than FLIR's.
The E4 can be modified for 320x240 support (at least the older ones could, don't know if they've "fixed" this or not), but it sounds like Opgal's sensor is even better in that respect.
That's a good point. I wonder how much you could coax out of a regular camera by putting an IR pass filter in front of it to filter out all the other wavelengths and then using a long exposure.
That might be a dirt cheap way to get a passable IR cam.
You can see some IR sources like remote-control IR LEDs with a conventional CCD camera, but you'd do it by removing its existing IR filter rather than adding one. The IR cameras being discussed here respond to much longer wavelengths, though, and yield completely different results. The black-body spectrum for thermal IR has far too little energy at the frequencies that traditional CCD cameras are sensitive to.
I believe I read somewhere that they are working on Android versions.
I used one of these at MacWorld, and asked the rep what distance it was effective at, their answer was less than 15 feet.
What do you mean by effective? FLIR's support told me it can pick up the heat signature of a person at 100 meters. I went outside earlier and it seemed alright at a long distance. It's not like it simply doesn't work after 15'.
I think due to the resolution many of the interesting things you can do with it will end up being at short range.
I also noticed it reports the sun's temperature as "> 212º F" so there's that.
I didn't get a quantitative definition of effective, I asked how far it would detect a heat signature, they said 15 feet. I'm sure it "works" at greater distance, the question is how much noise you get and how reliable it is at greater distances.
They had one on site and the reason I asked the range was because I held it up and looked and it seemed very very local. I'd be happy to see this thing produce a heat signature that you can identify as human at 100 yards, that isn't the same as a human's heat signature at 100 yards though.
Within five years, the iPhone 10 or so will probably have heat sensors (and many others) built in!