Yet another cheap thermal imager incoming.. Seek Thermal

[encryptededdy]

I agree with Frenky on this. The SEEK has potential but I still consider its core sub optimal. FLIR are old masters of thermal camera design and image processing. Love them or hate them, they still make decent thermal cameras, even considering the deliberate 'de-tuning' that they apply to limit performance in order to fit the market.

could you please elaborate on that deliberate 'de-tuning' ?

The FLIR One in the stock app only reads out a center 120x90 portion of the sensor. With a custom app you can read out the full 160x120 sensor (which obviously provides higher resolution, and also a wider field of view).

[Fraser]

I can also add that FLIR uses a software based noise generator to degrade the performance. This is to ensure that budget cameras in their range do not damage sales of their more expensive offerings.

Fraser

[Trax]

I can also add that FLIR uses a software based noise generator to degrade the performance. This is to ensure that budget cameras in their range do not damage sales of their more expensive offerings.

Fraser

is that noise generator in the software or the cammera firmware?
a.k.a. is it gone when usign the sustom softwar eposted on this board?

[Fraser]

On the E4 it was a setting in the configuration file set.

Sorry can't comment on how it is implemented on the F1G2.

Fraser

[encryptededdy]

Also, the FLIR One Gen 1 was limited to 100C, and the Gen 2 is limited to 120C. Beyond this, the FLIR One app will just display ">120"

However, the radiometric JPEGs the FLIR One save actually record the full span temperature data beyond 120C, up to 150C (which is the actual max temp of the Lepton sensor). Initially, the "FLIR One Closeup" app for FLIR One G1 (released by FLIR Themselves!) would also allow you to read up to 150C, however I think they've "fixed" that long ago.

[tomas123]

Sorry can't comment on how it is implemented on the F1G2.

Fraser

a little bit off topic:

the downgrade (cropped image with low sensitivity) is only realized by the Flir One app.
If you use the SKD you get the full lepton resolution and sensitivity with mode "ThermalLinearFlux14BitImage".

I'm not sure, if the downgrade of Flir One App is only a side effect of a radical noise filter and a MSX overlay of a real camera with a undersized FOV. 

compare the images:

here is a real live sample (shot after a small warm up time of about 2 minutes):

I saved with the SDK.app simultaneously a upscaled Flir Radiometric JPG  and a real Lepton ThermalLinearFlux14BitImage.

Afterwards I rebuild with my old panorama script (see my footer) a real size 160x120 Lepton radiometric jpg (a Flir format).
You can load this sample jpg images in Flir Tools and compare the quality.

First a original image shot with the Flir App.
The App crop    the Lepton sensor to about 120x90 Pixel.
Please note the artefacts/patterns!
Flir makes a nice lens distortion correction of the Lepton sensor for best MSX overlaying 




real  Lepton sensor 160x120 (no image postprocessing and with noise/grain because the temperature spread is only 4 Kelvin)

[joe-c]

a really nice comparison.

i hope my flir1G2 will arrive soon (was bought on flir web shop end of year 2015).

[Fraser]

@joe-c,

I really am surprised at FLIR's apparent inability to supply these F1G2 cameras. I do not recall them having such issues with previous camera releases, or at least not to the same extent. I recall that it took ages to get my replacement F1G2 camera from FLIR as they did not hold any spare stock at HQ.....all, were being shipped as soon as received.

Is this due to popularity or limited production capability I wonder ?

Fraser

[miguelvp]

Maybe he ordered it from Seek and that's why it's taking so long, after all we are on the Seek thread and we don't go there and show how good the Seek images can be

Edit: almost the full page is about Flir.

[Fraser]

😈

[hanscomps]

Hi,
Noob thermal camera tinkerer here.
Received my Seek Thermal XR a couple of weeks ago; Android version.  Works fine on my S4.
Now got it hooked up to my PC with Python script grabbing frames and displaying it in QT frame.  All good.
Now starting with some calibration and colour mapping; but my FrameID4 seems to be different as what is mentioned in this thread.  See attached.
Two vertical lines and one horizontal at the bottom?  Any ideas?

Hans

[frenky]

That is odd. On old Seek the ID4 and ID10 are gain compensation. I believe that one is for lower end and other is for higher end of temp ranges.

[ixfd64]

I asked on the official Facebook page and was told that the cameras have a thermal sensitivity of 0.1 °C but only display whole numbers: https://facebook.com/SeekThermal/posts/1023007861105118

On the other hand, I'm sure that's something a little hacking would solve. :-)

[the1snm]

I asked on the official Facebook page and was told that the cameras have a thermal sensitivity of 0.1 °C but only display whole numbers: https://facebook.com/SeekThermal/posts/1023007861105118

On the other hand, I'm sure that's something a little hacking would solve. :-)

I'm certainly not an expert but I don't think the Seek units have shown that level of sensitivity at least not what I've seen and read. I bought a Seek XR and a Flirone V2 and did a number of side by side comparisons and I could not get anywhere near that kind of sensitivity out of it. For in home use it struggled but outside it was better at detecting heat further off as wild be expected from the larger sensor and narrower field of view. Please don't think I'm bashing the Seek I have a lot of respect for that unit as I've owned serveral so far. I really hope they come out with a more refined higher res version for the same price in the near future.



Sent from my iPhone

[hanscomps]

That is odd. On old Seek the ID4 and ID10 are gain compensation. I believe that one is for lower end and other is for higher end of temp ranges.

I see FrameID-10 has the same artifacts (non on the calibration and valid frames ID3?)
ID4 and ID10 frames also seems very close to each other.

[frenky]

This is from my Seek:
ID4:


ID10:


You shouldn't look at this frames as an image but as info about each pixel gain at specific temperature range.
That why there is not much similarities with frame ID3.

[frenky]

BTW if anyone wants to play with seek, this is the basic way to get ok images out:

On ID4 frame (gain):
gainCalArr[n]=averageOfID4/FrameID4[n];

On ID1 frame (calibration):
FrameID1[n]=FrameID1[n]*gainCalArr[n];
offsetCalArr[n] = averageOfID1 - FrameID1[n];

On ID3 frame (real thermal image):
FrameID3[n]=FrameID3[n]*gainCalArr[n];
FrameID3[n]=FrameID3[n]+offsetCalArr[n];

And fix patent pixels and dead pixels...

[hanscomps]

BTW if anyone wants to play with seek, this is the basic way to get ok images out:

On ID4 frame (gain):
gainCalArr[n]=averageOfID4/FrameID4[n];

On ID1 frame (calibration):
FrameID1[n]=Frame1ID[n]*gainCalArr[n];
offsetCalArr[n] = averageOfID1 - FrameID1[n];

On ID3 frame (real thermal image):
FrameID3[n]=Frame3ID[n]*gainCalArr[n];
FrameID3[n]=Frame3ID[n]+offsetCalArr[n];

And fix patent pixels and dead pixels...

Ah, thanks, was looking for a summary through the 90-odd pages of this forum.
I assume the dead and patent pixels must be excluded when generating the averageOfID4 and averageOfID1?

freaky; I will pass my odd ID4 frame through the process and see what happens, thanks.

[frenky]

Yes exclude those pixels.

I'm using this code to get "average" which is really "mode":

Code: [Select]

public static int GetMode(int[] arr)
       {
           int[] arrMode = new int[320];
           int topPos = 0;
           for (int i = 0; i < 32448; i++)
           {
               if ((arr[i]>1000)&&(arr[i] / 100 != 0)) arrMode[(arr[i]) / 100]++;
           }

           topPos = Array.IndexOf(arrMode, arrMode.Max());

           return topPos*100;
       }


[cynfab]

New today from Seek:
http://www.thermal.com/products/revealxr-fastframe?utm_source=marketo&utm_medium=email&utm_campaign=revealxr-available

With a claimed frame rate of > 15fps....

[joe-c]

Hey Guys,
providing last version, still not perfect but it goes to a usable direction. 

[-jeffB]

New today from Seek:
http://www.thermal.com/products/revealxr-fastframe?utm_source=marketo&utm_medium=email&utm_campaign=revealxr-available

With a claimed frame rate of > 15fps....

Yeah, I just saw the email. The lack of an actual framerate anywhere in the specs doesn't exactly inspire confidence, but the video snip does look good. If it's really a 30fps refresh rate for under US$500, that seems like kind of a big deal.

[efahrenholz]

So as we can see from seeks demo of the faster frame rate, they aren't averaging the frames together and thus leads to less blur. You can't really tell from the video how much noise exists but I'm guessing the noise reduction using frame averaging isn't great as it is. If they decide to release the faster frame rate version in the compact editions, might be interested in picking one up...At least now you can fiddle with the image a little more.

[frenky]

Hey Guys,
providing last version, still not perfect but it goes to a usable direction. 

I'm not sure if your using ID4 for gain compensation. If you're not, you should because it really improves images.

Left is without gain and right is with gain compensation:

[frenky]

So as we can see from seeks demo of the faster frame rate, they aren't averaging the frames together and thus leads to less blur. You can't really tell from the video how much noise exists but I'm guessing the noise reduction using frame averaging isn't great as it is. If they decide to release the faster frame rate version in the compact editions, might be interested in picking one up...At least now you can fiddle with the image a little more.

The video looks quite promising. And the contrast also looks better so perhaps they did improve sensor a bit?
But it cold be that they have just "overclocked" sensor a bit for faster readouts.

I must hookup logic analyser to my Seek to see what is real frame rate that comes out of the sensor.