Best thermal/ir camera (dev kit) for embedded linux board?

[want2cthermal]

Hey guys,

I'm looking for a thermal camera to hook up to an embedded linux board (raspberry, beaglebone, etc).

I found the FLiR dev kit which connects over spi to linux boards, but it's resolution is only 80x60. I've been reading as many posts as I can, admittedly not all since many threads are well over 100+ pages, but I see there is some firmware flash to get the flir to a better resolution? Would that work for this? - https://www.sparkfun.com/products/13233

I'm also considering the Seek thermal cameras, which appear to have USB connectors and have SDK's that come with them so I can start coding out of the box. Are there any Seek cameras which have codes in C or C++ & work with ARM embedded boards?

So shortly said, I'm a C coder (but I can live w/C++ too) and I want to connect it to any embedded linux board, but need something w/working sample codes as I won't be able to write a driver from scratch for that.

The application is to monitor wildlife at out animal refuge. I saw youtube videos that the higher resolution flir e4 can easily see 150+ meters with zoom. That's what I'd really like to get!

Do these thermal cameras normally come with a /dev/camera interface on linux like other cameras do? I'd really like to use it with opencv.

What would your commendation be? My post is kind of all over the place, sorry about that, just looking for some guidance really.

Thanks

[tomas123]

there are some raspberry pi projects in this forum

Flir One 160x120

load the site:

Code: [Select]

http://192.168.1.70

Flir Ex 320x240

A few hundred lines of code later (simply record all frames with a max temp of higher than some threshold, re-mapping to 8-bit, then encoding as mp4), I got a good picture of the beast:

[want2cthermal]

Hey Tomas, thanks a lot for your help!

Do you know where I can buy that Flir Ex 320x240 that ships to US? I would also need to buy the breakout board that accepts the lens & allows me to connect the SPI pins directly w/out soldering, right? My normal suppliers, adafruit and sparkfun, don't seem to carry that.

I'm still reading through all your posts, but it looks like you have gotten a good control of the camera, and it looks like you did all that without help from Flir. Nice!

[tomas123]

there is a long thread about buy a cheap E4 camera and hack it to 320x240

FLIR is offering a $100 discount on the E4 for Black Friday: http://flir.com/blackfriday2016

Just an FYI.

[want2cthermal]

Thanks, but I'm not looking to buy the entire hand held device, just the lens sensor (no screen, buttons, etc) and the breakout board. This all looks pretty vaguely documented and kind of "may or may not work", heh. I'm probably just going to get the dev kit from sparkfun and hope it can be hacked to higher resolution.

I saw in one of your posts you had a breakout board for the flir and rpi camera connected to the board. Where did you buy that breakout board from?

[Fraser]

Sparkfun FLIR Dev kit uses FLIR LEPTON 2 80x60 pixels core. That core cannot be hacked to a higher resolution as it is only 80x60 physical pixels. LEPTON cores run at their native physical resolution unless cropped in post processing software external to the core.

The FLIR LEPTON 3 is the higher resolution core at 160x120 physical pixels.

It was once the case that it was easier/cheaper to buy a FLIR ONE Gen 2 camera, that contains a socketed LEPTON 3, than buying just the core from FLIR. The FLIR ONE Gen 2 also has the FFC shutter option fitted to the LEPTON 3. IIRC the LEPTON 2 Dev PCB is compatible with the LEPTON 3 but you would need to check this. The data coming from the LEPTON 3 also differs to cope with the higher resolution.

Look for LEPTON development boards as there is at least one other supplier of such with the option to include a LEPTON core. LEPTON 3 cores have been infamously hard to obtain in one off purchases from FLIR however.

Fraser

[tomas123]

I think, you can't enable the radiometric mode of the Lepton 3 outside the FlirOne camera case ;-)

page 27
https://www.eevblog.com/forum/thermal-imaging/flir-one-thermal-imaging-camera-teardown-and-hacks/?action=dlattach;attach=206832

http://www.flir.com/cvs/cores/knowledgebase/index.cfm?view=67336

Can I make use of Lepton's Radiometry Module to make absolute temperature measurements?

Availability of Radiometry commands is restricted to Original Equipment Manufacturer (OEM) customers for Lepton, and requires that an active non-disclosure agreement (NDA) be in place with FLIR.
 
The OEM quantity threshold is 1,000 units for Lepton.

edit 02.12.2016
there are some new infos about Lepton and radiometric mode

http://wiki.brandoncurtis.com/w/Non-Contact_Temperature_Measurement

There is a document under NDA called "FLIR Lepton Radiometry Application Note"
What I also do is leave B and F at its default values and do a least square fit for R and O. I use 100 calibration samples between several lepton pixels multiplied by a weight matrix to build an average raw value and real temperatures from a MLX90614 ir sensor pointing to that average value as inputs.

https://github.com/maxritter/DIY-Thermocam/blob/master/Hardware/Lepton.h

Code: [Select]

/* Set the radiometry mode (not used) */
void lepton_radiometry(bool enable)
{
//Enable or disable radiometry
Wire.beginTransmission(0x2A);
Wire.write(0x00);
Wire.write(0x08);
Wire.write(0x00);
Wire.write(enable);
Wire.endTransmission();
//Data length
Wire.beginTransmission(0x2A);
Wire.write(0x00);
Wire.write(0x06);
Wire.write(0x00);
Wire.write(0x02);
Wire.endTransmission();
//RAD module with OEM bit and command
Wire.beginTransmission(0x2A);
Wire.write(0x00);
Wire.write(0x04);
Wire.write(0x4E);
Wire.write(0x11);
Wire.endTransmission();
}

[want2cthermal]

Sparkfun FLIR Dev kit uses FLIR LEPTON 2 80x60 pixels core. That core cannot be hacked to a higher resolution as it is only 80x60 physical pixels. LEPTON cores run at their native physical resolution unless cropped in post processing software external to the core.

The FLIR LEPTON 3 is the higher resolution core at 160x120 physical pixels.

It was once the case that it was easier/cheaper to buy a FLIR ONE Gen 2 camera, that contains a socketed LEPTON 3, than buying just the core from FLIR. The FLIR ONE Gen 2 also has the FFC shutter option fitted to the LEPTON 3. IIRC the LEPTON 2 Dev PCB is compatible with the LEPTON 3 but you would need to check this. The data coming from the LEPTON 3 also differs to cope with the higher resolution.

Look for LEPTON development boards as there is at least one other supplier of such with the option to include a LEPTON core. LEPTON 3 cores have been infamously hard to obtain in one off purchases from FLIR however.

Fraser

Thanks Fraser, but writing drivers for cameras from scratch isn't something I know very well nor have time to learn. I'd pay 1000 us for something that had a well defined, stable, API to it and let's programmers start programming out of the box, relatively quickly at a good resolution, but it looks like the closest thing to that is sparkfun's low resolution dev kit w/the code from Flir : (

it looks like the reason good, reliable info is so hard to come by, is that one could easily buy a lens, then assemble their own handheld devices like Flir sells for a whole lot cheaper than they sell them for. Who knows, maybe I'll get lucky and find a crowdsourced large purchase for the lepton 3 and the lepton 3 lens will work with the breakout board sparkfun sells. Last thing I need though is to be fighting programmers tooth and nail to get basic info for my non profit project. In another year or two I'm sure all this will be made much easier anyways, I can continue with IR leds on a night vision camera hooked up to opencv until then, even this works extremely well for me : )

Anybody have experience w/the 80x60 resolution? Would I be able to see 100 meters with it? Thanks for all your helps!

[tomas123]

Anybody have experience w/the 80x60 resolution? Would I be able to see 100 meters with it? Thanks for all your helps!

80x60 is really bad

I shot comparison photos with Flir One G1, Flir One G2 , Flir E4 and Flir E40 from a motherboard.
The temperature scale is auto-level, but Flir tools use a different "gamma correction" for the color scale of E40.

Flir One G1 Lepton 80x60


Flir One G2 Lepton 160x120 Mode: ThermalLinearFlux14BitImage (real Lepton 3 resolution)



Flir One G2 Lepton 160x120 with hacked config file
Flir added a fake noise, but you see it only with minor temperatur spans
a sample: https://www.eevblog.com/forum/thermal-imaging/flir-one-thermal-imaging-camera-teardown-and-hacks/msg801504/#msg801504)


Flir Ex 320x240



Flir Exx 320x240

[JohnG]

Maybe this project might be of use: http://www.diy-thermocam.net/

As I recall, there is an option to buy a completed camera.

John

[Fraser]

Also the Groupgets page contains information on LEPTON cores that may be of interest. There is a maker community finding new applications, and form factors, for LEPTON based thermal cameras.

https://groupgets.com/manufacturers/flir/products/lepton

Fraser

[want2cthermal]

Hey fraser, thanks for that link. They happen to be out of stock, while cheaper I still think that's over priced. I could probably buy all that individually and 3d print my own case for well under half that price.

It's also a good bit cheaper to buy the breakout board and lens from digikey instead of sparkfun

This was a neat video I found -

It looks like 80x60 can kind of, sort of, be useful at around 50m. Good news for me is that opencv will definitely be able to tell the color difference at 50m as there is a lot of contrast. All I need to do is detect general area of body heat over a plain so I think 80x60 + working code in my language, to detect heat then trigger night vision lens + ir led lamp will be acceptable. I'm really trying to get night vision quality photos so thermal resolution, at least right now, isn't a huge deal.