Flir One Android App with full Lepton sensor resolution 160x120

[NathanFowler]

Stumbled across this today:

https://www.reddit.com/r/flirone/comments/4a7465/slow_rendering_ive_tweaked_the_flir_one_example/

http://developer.flir.com/forums/topic/samsung-s7-slow-and-glitchy-stock-app-why-also-a-kind-of-work-around/

Figured I'd share.  All the best and cheers.

Edit: Alternative link, and of course Thomas is there   Old news is exciting I guess... sorry for the thread noise.

[tomas123]

http://developer.flir.com/forums/topic/samsung-s7-slow-and-glitchy-stock-app-why-also-a-kind-of-work-around/
...and of course Thomas is there  

yes, we are a flir one hacker community 
It's the same Sarah, who wrote the review for Georgs App (see attachment)

P.S: I got no feedbacks about SR algorithm

(number of detected edges, max. offset of following images, count of stacked images, exponential moving average etc.).

[ChunkyPastaSauce]

Unfortunately the SR don't improve the thermal sensitivity, but it works great on fine structures.
There are some SR parameters inside the algorithm, which are currently fixed values to simplify the use of the app (number of detected edges, max. offset of following images, count of stacked images, exponential moving average etc.).

I would be glad about comments on the SR ...

Looking at the image output image.... I bet you could pull out some more detail using traditional unsharpen masks, high pass filter or something else maybe in realtime. Deblurring probably would also work well (but I think is usually computationally expensive).

[tomas123]

There is already a gauss sharpen filter implemented.
If you set the gauss filter stronger (like on real images) then you get halos around structures with false thermal values.
(It's by design)

[tomas123]

A sample from a dishwasher.
All images are taken of the same position/distance.

original images from Georgs Thermal Camera+ App without post processing

"native Lepton Sensor Resolution 160x120" compared with "bilineare filtering and multiframe super resolution"  (minor differences from freehand shooting)




original Flir One MSX image shot with Flir One SDK (it's same image distance but Flir cropped the image to 120x90 sensor pixels) 



after removing MSX with Flir Tools for Android (Flir Tools reduces the image size  )


and now after opening the image with Flir Tools for PC (a postage stamp  ):

[tomas123]

next sample:
 
Georgs Thermal Camera+ App  160x120 vs. hacked Flir Ex 320x240

Lepton 160x120

Flir Ex 320x240

- shifted point of views, because the cameras have differents FOV (field of view)
- the Ex image is a Screenshot from Flir Tools because the original radiometric jpg is a stamp 

[Fraser]

@Tomas123,

Thanks for the images.

I must say I am impressed with what can be achieved with the LEPTON 3 based FLIR ONE. The original FLIR app just does not do it justice.

Fraser

[tomas123]

Flir is surely not interested in, to get all image capabilities from a 160x120 thermal camera in a 250USD price segment.

We know, that Flirs hardware costs for a 320x240 E8 are under 1000 USD (that is the price for E4 80x60) and Flir sells the E8 for 4000 USD.

Therefore the Flir One SDK with the 14 bit radiometric sensor values is a nice back door.

[frenky]

You are getting really great images out of Lepton3 core... 

I wonder if Peltier module could improve images even more.
On Seek it really makes a huge signal to noise improvement...

My Seek is dead again with the same blank screen and now I'm angry at myself for not buying Flir One when it was 175€ on amazon.co.uk.

[NathanFowler]

Simply amazing quality!  I wish I were on Android, would either of you two (Georg/Tomas) consider developing a ported version to iOS, I am sure we could start a kickstarter campaign and I would happily donate $100 USD for such amazing work!

[tomas123]

The lepton sensor (lens) has a great sharpness, therefore we can make some great image enhancements.

But the NETD (noise equivalent differential temperature) is not so good as a hacked (!) Ex camera.

The Lepton Data Sheet give us a thermal sensitivity <50 mK, but see the difference:


The Ex image (on the left side ) is the original jpg size 240x320 OOC (out of camera) with the imprinted scale. Georgs temperature scale is outside the thermal image.

And believe me, I tried many software tricks to improve the NETD of the lepton sensor, but without noticeable success.
It seems, that Flir makes a great job with his own image post processing algorithmus (calculate RAW from sensor ADC values).
They are brand leaders with such small (=noisy) uncooled microbolometer without temperature stabilisation.

There is a heating element inside the Ex for temperature stabilisation, but this cost some battery power...

[tomas123]

Georgs app also saved the (cropped) Flir radiometric jpg from the scene above.

Now we can see, how aggressive Flir denoise algorithm works:

Flir Tools for Android


same image saved with Flir Tools for PC (Flit Tools for PC change the palette, therfore we see more contrast)



after enlarging the denoised image, we see the well known denoise patterns

[Fraser]

Tomas123,

A minor correction. The E4 microbolometer is not actively temperature stabilised. It uses generic offset tables and the fact that the microbolometer self heats to a temperature equilibrium of around 30C

This where budget cameras often differ to more expensive models.

Fraser

[tomas123]

sorry, we discussed last year about this topic 

My E40 stabilize the microbolometer temperature also a hour after a simple shutoff (this standby mode for fast restart has also the Ex)
There is a red LED inside (visible trough the SD slot) which indicate this "heating".
After a hour in standby modus (camera off but red LED inside is on) the E40 immediately make correct measurements without a tilde (~24°C).
You say "self generated heat" and I think, there is heat source for faster reach the 30°C for correct measurements.

Certainly I am wrong, but it's so simple to implement  ...
We both now, there is no cooling ;-)

@Tomas123

Not exactly.... The microbolometer is not actively heated. The microbolometer generates its own internal heat as a result of the on die electronics (ROIC etc) this self generated heat normally raises the die temperature to approximately 30 Degrees C. I say approximately as there is no control over the temperature. It just settles at an equilibrium point but this could be 28C or 32C, it cannot be held stable.

The offset table was centred on 30degrees C as that is the natural operating temperature of the die at normal ambient temperatures of around 20 Degrees C.

The reason for the tilde symbol next to the temperature reading after start-up is that the microbolometer die will be at ambient and needs time for its die to rise in temperature to the equilibrium point of around 30C. Until it hits that equilibrium the readings are not considered accurate as the offset table is not able to compensate.

All of the above comes from direct discussions with FLIR 

With regard to the power down sequence on the camera. The camera drops back to a low power standby mode that facilitates a fast start. After a period of time the camera drops back to the full sleep mode that consumes minimal power. It takes longer to start from sleep as it is a cold boot rather than a warm boot from standby. It's very much like modern laptop behaviour.

Aurora

[gansgar]

Hello community,

Flir has updated just recently their SDK with a great new feature: Flir One and Cat S60 support.
I updated today my app to the new SDK, but before I release it to the store I would like to test it with some real Cat S60 devices.

So I invite everybody to try out the beta.
https://play.google.com/apps/testing/georg.com.flironetest_01

Give me in this thread some feedback if it works or not.
Because the Cat S60 only has a Lepton sensor resolution of 80x60 some feedback through images would be nice. (screenshot and test-shot)

- George

[e_sousa]

Thanks for adding s60 support.
The beta version doesn't seem to be working on my device though
It will either get stuck asking for the flir one to be connected or showing a all black image and crashing after a few seconds.
Also i can hear a click sometimes, not sure if it is the thermal camera shutter or the focus on the normal camera.

I hope i managed to install the correct version, in the settings it says "2.0.0 - 13 | release"

[gansgar]

Thanks for adding s60 support.
The beta version doesn't seem to be working on my device though
It will either get stuck asking for the flir one to be connected or showing a all black image and crashing after a few seconds.
Also i can hear a click sometimes, not sure if it is the thermal camera shutter or the focus on the normal camera.

I hope i managed to install the correct version, in the settings it says "2.0.0 - 13 | release"

Could you look into the settings under the app informations. Under the tab permission is the camera one on or off? If off could you activate it and try it again? I'll fix that later but just to try it out.

Could you also tell me if you have a report button after the crash?? If yes please press it that helps me to understand the error in more detail.

[e_sousa]

Camera permission is ok in android settings.
Only sometimes it shows the report button after crash, i tried a few times until it showed the button box and sent a report.

[gansgar]

Only sometimes it shows the report button after crash, i tried a few times until it showed the button box and sent a report.

I saw the error. The problem is that it's at a position I cannot get any useful error informations from, except that the error comes from the Flir SDK.
So I compiled the Flir example application from the SDK. The google drive link contains the apk, you should activate in the settings that you can install apps from unknown sources.
https://drive.google.com/file/d/0B1xrpkRXfbF0VmpCNXktMHBwQ2s/view?usp=sharing

And here is how you can activate unknown sources:
http://www.applivery.com/blog/android-unknown-sources/

If the example app runs I can compare it to my own code and find the error.

[e_sousa]

Only sometimes it shows the report button after crash, i tried a few times until it showed the button box and sent a report.

I saw the error. The problem is that it's at a position I cannot get any useful error informations from, except that the error comes from the Flir SDK.
So I compiled the Flir example application from the SDK. The google drive link contains the apk, you should activate in the settings that you can install apps from unknown sources.
https://drive.google.com/file/d/0B1xrpkRXfbF0VmpCNXktMHBwQ2s/view?usp=sharing

And here is how you can activate unknown sources:
http://www.applivery.com/blog/android-unknown-sources/

If the example app runs I can compare it to my own code and find the error.

I installed and played around a bit with the example app, it seems to work fine.

[gansgar]

Great thanks. That's a big step forward. I'll improve the app and throw out an update for all beta tester if think I fixed the problem.

- George

[e_sousa]

Happy to help, anything else you need just ask.
Keep up the good work.

[gansgar]

Okay. I just upload a new beta version. I essentially just changed one line in hope that the app crashed because I used an deprecated image format. But whatsover, we are currently just trying to figure out what is going wrong.

- George

[e_sousa]

It is working now, good job 
I can post some sample shots later if it helps any.

[gansgar]

The samples would be nice. I still don't know how my app reacts to 80x60 instead of the 160x120 from the Flir One.

And also: No problem. It's fun doing it

- George