New Flir products

[frenky]

I did a bit of work in finding the real resolution of new Flir One. I drew a line in the middle of each "bump" of a wire and then aligned grid so that it matched all the lines perfectly.
My conclusion is that Flir One v2 outputs images with approx 94x125 resolution. Perhaps the sensor is really 120x160 but the images are not.

[tomas123]

It would be nice to be able to scrub this data, either in the Flir One app, Flir Tools Mobile, or exiftool.

Phil from exiftool decided that the flir tags are non editable.
But Exiftool can show you (-v5) the hex code and then you can reset the values to zero with a hex editor.
$ exiftool -v5 IMG_1854_untouched.JPG | grep -C10 "52 00 00 00 01"
    06a6: 00 00 00 00 01 00 00 00 01 00 00 00 44 35 41 34 [............D5A4]
    06b6: 43 32 31 43 37 36 44 30 41 38 36 45 41 46 37 35 [C21C76D0A86EAF75]
    06c6: 46 36 31 43 46 46 42 33 34 32 31 45 00 00 07 00 [F61CFFB3421E....]
    06d6: 00 00 01 00 04 00 00 00 02 02 00 00 01 00 02 00 [................]
    06e6: 02 00 00 00 4e 00 00 00 02 00 05 00 03 00 00 00 [....N...........]
    06f6: 22 07 00 00 03 00 02 00 02 00 00 00 57 00 00 00 ["...........W...]
    0706: 04 00 05 00 03 00 00 00 3a 07 00 00 05 00 01 00 [........:.......]
    0716: 01 00 00 00 00 00 00 00 12 00 02 00 06 00 00 00 [................]
    0726: 52 07 00 00 00 00 00 00 1b 00 00 00 01 00 00 00 [R...............]
    0736: 34 00 00 00 01 00 00 00 54 00 00 00 64 00 00 00 [4.......T...d...]
    0746: 52 00 00 00 01 00 00 00 15 00 00 00 01 00 00 00 [R...............]
    0756: 02 07 00 00 64 00 00 00 57 47 53 38 34 00 07 00 [....d...WGS84...]
    0766: 03 01 03 00 01 00 00 00 06 00 00 00 12 01 03 00 [................]
    0776: 01 00 00 00 00 00 00 00 1a 01 05 00 01 00 00 00 [................]
    0786: b2 07 00 00 1b 01 05 00 01 00 00 00 ba 07 00 00 [................]
    0796: 28 01 03 00 01 00 00 00 02 00 00 00 01 02 04 00 [(...............]
    07a6: 01 00 00 00 c2 07 00 00 02 02 04 00 01 00 00 00 [................]
    07b6: 6d 37 00 00 00 00 00 00 48 00 00 00 01 00 00 00 [m7......H.......]
    07c6: 48 00 00 00 01 00 00 00 ff d8 ff c0 00 11 08 00 [H...............]
    07d6: 78 00 a0 03 01 22 00 02 11 01 03 11 01 ff c4 00 [x...."..........]
    07e6: 1f 00 00 01 05 01 01 01 01 01 01 00 00 00 00 00 [................]
--
--
  | |         06ea: 4e 00                                           [N.]
  | | 2)  GPSLatitude = 27 52 0.84 (27/1 52/1 84/100)
  | |     - Tag 0x0002 (24 bytes, rational64u[3]):
  | |         072e: 1b 00 00 00 01 00 00 00 34 00 00 00 01 00 00 00 [........4.......]
  | |         073e: 54 00 00 00 64 00 00 00                         [T...d...]
  | | 3)  GPSLongitudeRef = W
  | |     - Tag 0x0003 (2 bytes, string[2]):
  | |         0702: 57 00                                           [W.]
  | | 4)  GPSLongitude = 82 21 17.94 (82/1 21/1 1794/100)
  | |     - Tag 0x0004 (24 bytes, rational64u[3]):
  | |         0746: 52 00 00 00 01 00 00 00 15 00 00 00 01 00 00 00 [R...............]
  | |         0756: 02 07 00 00 64 00 00 00                         [....d...]
  | | 5)  GPSAltitudeRef = 0
  | |     - Tag 0x0005 (1 bytes, int8u[1]):
  | |         071a: 00                                              [.]
  | | 6)  GPSMapDatum = WGS84
  | |     - Tag 0x0012 (6 bytes, string[6]):
  | |         075e: 57 47 53 38 34 00                               [WGS84.]


I think the simplest way is writing a short regex/awk srcipt to zero around the WGS84 marker.
There are two entries for gps coordinates.
First entry is the regular jpg exif header.
Second entry you found in the proprietary Flir exif header.

Code: [Select]

$ hexdump -C IMG_1854_untouched.JPG | grep -B5 S84
00000710  00 00 05 00 01 00 01 00  00 00 00 00 00 00 12 00  |................|
00000720  02 00 06 00 00 00 52 07  00 00 00 00 00 00 1b 00  |......R.........| 
00000730  00 00 01 00 00 00 34 00  00 00 01 00 00 00 54 00  |......4.......T.| 
00000740  00 00 64 00 00 00 52 00  00 00 01 00 00 00 15 00  |..d...R.........|
00000750  00 00 01 00 00 00 02 07  00 00 64 00 00 00 57 47  |..........d...WG| !!!
00000760  53 38 34 00 07 00 03 01  03 00 01 00 00 00 06 00  |S84.............|
--
00029300  00 01 00 00 00 00 00 00  00 4e 00 57 00 00 00 00  |.........N.W....|
00029310  00 89 1c ac 6c ed dd 3b  40 15 15 38 19 b8 96 54  |....l..;@..8...T|
00029320  40 00 00 80 bf 00 00 00  00 00 00 00 00 00 00 00  |@...............|
00029330  00 00 00 00 00 00 00 00  00 00 00 00 00 00 00 00  |................|
00029340  00 00 00 80 bf 00 00 00  00 00 00 00 00 00 00 80  |................|
00029350  bf 00 00 80 bf 00 00 80  bf 57 47 53 38 34 00 00  |.........WGS84..| !!!

image source https://www.eevblog.com/forum/testgear/new-flir-products/msg712432/#msg712432

[tomas123]

I did a bit of work in finding the real resolution of new Flir One. I drew a line in the middle of each "bump" of a wire and then aligned grid so that it matched all the lines perfectly.
My conclusion is that Flir One v2 outputs images with approx 94x125 resolution. Perhaps the sensor is really 120x160 but the images are not.

You are right, after correction of offset from a half pixel the image looks better.
Don't forget that only the Flir Exx takes sharp images from a wire. The Flir Ex images are also poor.

[cynfab]

thomas123,

Thanks for the heads up on the exiftool -v5 option, most useful!!

Looks like Flir has added some new tags and another GPSInfo SubDirectory:

  | | FLIR Record 0x2b, offset 0x1b730, length 0x00b8
  | | 6)  GPSInfo (SubDirectory) -->
  | |     - Tag 0x002b (184 bytes):


Lat & Lon references as ascii chars N/S E/W at offset 0x09 & 0x0b.
Latitude & Longitude as double precision floats at offset 0x10 & 0x18.
Room for other info as well, total of 184 bytes.
 
using exiftool -v5 <filename> I examined the 2 areas of GPSInfo in the following 3 images:

FLIR0002_sample.jpg  Included with the Flir One App for Android
IMG_1854_untouched.JPG   image source https://www.eevblog.com/forum/testgear/new-flir-products/msg712432/#msg712432
flir_20150908T144845.jpg  I took this on 9/8/2105.

FLIR0002_sample.jpg has a creation date of 5/16/2014 shows a GPSLongitude of E 18 in Flir Tools Mobile.
The normal GPSInfo tags also show E 18 and the sign of the double float is + in the additional Flir gps data, which is correct.
Lat & Lon ref in the additional Flir gps data are correct.
Both GPSLatitude & GPSLongitude Tags are present in the normal exif GPSInfo section.

I don't have any early samples from W Lon or S Lat so it's hard to know what Flir Tools Mobile would show.

IMG_1854_untouched.JPG has a creation date of 7/2/2015 shows GPSLongitude of W 82 in Flir Tools Mobile.
The normal exif GPSInfo tags shows the same, BUT the sign of the double float is + in the additional Flir gps data which should be -.
Lat & Lon ref in the additional Flir gps data are correct.
Both GPSLatitude & GPSLongitude are present in the normal exif GPSInfo section.

flir_20150908T144845.jpg has a creation date of 9/8/2015 shows a GPSLongitude of E 107 in Flir Tools Mobile which should be W 107.
The normal exif GPSLongitudeRef tag correctly shows W. The sign if the double float is - in the additional Flir gps data, which is correct.
Lat & Lon ref in the additional Flir gps data are correct.
The GPSLongitude Tag is missing from the normal exif GPSInfo section.


I conclude that Flir fixed the sign problem between 7/2 and 9/8, BUT Flir tools Mobile hasn't caught up, AND Flir
messed up the normal exif GPSInfo section by leaving out the GPSLongitude Tag.
So there are issues in both the Flir One App, and in Flir Tools Mobie App (for Android).


Please refer to the gps.txt file that is attached that shows the details of the exiftool data dumps.
also are attached screenshots of the Flir Tools Mobile info screen for each of the image files.


   ...ken...

[FrankenPC]

So, do you think we can expect a resolution doubling in another 24 months?  I'm being snarky.  The resolution is ridiculous.  I know Seek has terrible calibration and consistency issues, but someone needs to pick up the ball and make high-rez smart phone modules that are precise.  This is capitalism we're talking about here.  Where are the competitors?

[frenky]

There is therm-app module with high thermal resolution but it's 4 times the price of flir or seek.

Sent from my LG-D855 using Tapatalk

[in.Vitro]

The Downside of the F1 in Europe/Germany is, that is very duifficult to purchase one...

Look when they announced the second gen. Its still not availoble here, which causes that I would take the SeeK.
Has someone an idea why there is such a massive Delay? Thats Downside ridiculous

I dont care if the the Image is a "little" bit more noisy but bla bla bla...you get the Point.

[mct75]

I wish they'd make a ~30 Hz one (for the US).  I don't own a F1, but from all the videos I have seen, the refresh rate makes anything other than static (non moving camera and subject) observation nauseating.   Plus, with a higher frame rate, it might be possible to do some sub-sampling interpolation type tricks to get a higher amount of pixel data.

[kaz911]

Hardware Revisions

I got on F1v2 about a month ago and I do like the size and usability so I was lucky? and picked up another one on eBay UK at a discount (GBP 145'ish / $220'ish) -

First one - reports hardware revision D
2nd one - reports hardware Revision G

So they a quickly stepping through the revisions. I don't know which one is best - but one of them is my dads "pre-Christmas" present

I have a Flir Franken4 for the serious stuff.. - but since I'm in the middle of packing up in the Middle East and moving the the wet cold UK - I have not had time to test so much between them. But they are great for HVAC use and testing the underfloor heating in our new house

One issue - I kept target on a very matte object - but every time it ran the internal calibration - temperature of the target shot up by about 1 - 1 1/2 deg c. Then it would slowly settle down to a stable temperature and stay there - until the thing re calibrated again. (This was all after the ~ sign had gone)

[rjardina]

Firstly, I understand the images below are above the temperature specs of the Flir One, But does anyone know what the black blob is??

Sorry for using fahrenheit, we walk backwards in the US.

[encryptededdy]

Firstly, I understand the images below are above the temperature specs of the Flir One, But does anyone know what the black blob is??

Sorry for using fahrenheit, we walk backwards in the US.

I think the sensor is just too overexposed - those spots read -40C according to FLIR Tools, which I think is as low as it goes.

The same thing happens on my Therm-App - if I point at a source over ~500C, the image/temperature just reads -20C.

I assume this is similar to what happens on some digital cameras (esp. cheaper ones) which display a black spot if you point them at the sun.

[tomas123]

But does anyone know what the black blob is??

Great! 
A sensor overload show us the real pixel size and the resize algorithm to 200% without influence of a unsharpen lense

Code: [Select]

exiftool -b -RawThermalImage flir_20150914T001022.jpg > test.png
convert  -define png:swap-bytes=on test.png -auto-level -scale x1280 -gamma 0.6 -depth 8 swap.png
convert -size 8x8 xc:none -stroke red -strokewidth 0  -draw "line 0,0 0,7"  -draw "line 0,0 7,0"  8x8.png
convert -size 960x1280 tile:8x8.png -transparent white 960x1280.png
convert swap.png 960x1280.png -composite overlay.png

here a resized crop:

one square = one sensor pixel of Flir One G2 (160x120)

[in.Vitro]

ok... and whats the result? 160x120? 



btw. I've called FLIR in Frankfurt and asked them why there is a delay in the shipping --> high request in Europe/Germany

[tomas123]

ok... and whats the result? 160x120? 

btw. I've called FLIR in Frankfurt and asked them why there is a delay in the shipping --> high request in Europe/Germany

see self: one red square = one sensor pixel of Flir One G2 (160x120)

I've called FLIR in Frankfurt and asked them why there is a delay in the shipping --> high request in Europe/Germany

Germany: I called with messgeraete-chemnitz.de and they ordered 100 x Flir One Android and  100x Flir One Iphone and can deliver at week 40.
Half the amount has already been sold...

[frenky]

With this method the 120x160 does look more "right":

[amyk]

So, do you think we can expect a resolution doubling in another 24 months?  I'm being snarky.  The resolution is ridiculous.  I know Seek has terrible calibration and consistency issues, but someone needs to pick up the ball and make high-rez smart phone modules that are precise.  This is capitalism we're talking about here.  Where are the competitors?

ITAR gets in the way. Remember that this is ex-military tech.

[bktemp]

Both resolutions do not seem to fit perfectly, but is hard to tell because they seem to use a sophisticated interpolation algorithm.
Maybe this inverted area is generated in software (integer overlow?) and not in the sensor itself?
On CMOS sensors the bright light effects the readout gates next to each pixel directly. Based on Mike's teardown the thermal sensor is a big array of passive NTCs, so an inversion of the signal should never happen.

[Fraser]

The microbolometer is an array or thermistors and the FLIR units are sun safe. Unlike a CCD that can suffer a charge well overflow, the microbolometer has no such issues.

An overload situation is most likely to occur in either the A to D due to limited dynamic range, or the software that interprets the A to D output receiving an unexpected value.

I have not seen this over range (not an overload) artefact in my industrial cameras. The resolution displayed in the artefact may not be from the microbolometer pixels and may come from a later processing stage.

As to the resolution..... Good work but I am mindful that what is displayed and stored on the phone has gone through a lot of processing to improve its appearance. Such processing can 'muddy the waters' somewhat. I could ask Bill about this as he seemed to welcome questions.

Fraser

[Fraser]

I have asked Bill Terre  (VP & GM FLIR) for confirmation of the active pixels on the LEPTON 3. Hopefully he will provide a definitive comment on this matter.

Fraser

[Fraser]

OK peeps,

The definitive answer from Bill is that the LEPTON 3 microbolometer is in fact an array of 164x128 pixels. Bill answered that question so fast ! Great chap and he is the VP for heavens sake. Talk about approachable !

Fraser

[mct75]

OK peeps,

The definitive answer from Bill is that the LEPTON 3 microbolometer is in fact an array of 164x128 pixels. Bill answered that question so fast ! Great chap and he is the VP for heavens sake. Talk about approachable !

Fraser

But are they all active and/or are they are all exposed to the application?  Has anyone probed one over SPI to see what the actual data is?  That would make things 1000x easier.

[mikeselectricstuff]

OK peeps,

The definitive answer from Bill is that the LEPTON 3 microbolometer is in fact an array of 164x128 pixels. Bill answered that question so fast ! Great chap and he is the VP for heavens sake. Talk about approachable !

Fraser

But are they all active and/or are they are all exposed to the application?  Has anyone probed one over SPI to see what the actual data is?  That would make things 1000x easier.

the datasheets have already been posted here, no need to probe

[Fraser]

When you are fighting  a resolution war against SEEK you tend to roll out your full capability. FLIR have stated the presence of 4X the resolution of the original Lepton and I think I will believe them.

Bill said he could not understand why a test showed a lower resolution. Good enough for me.

Fraser

[mct75]

OK peeps,

The definitive answer from Bill is that the LEPTON 3 microbolometer is in fact an array of 164x128 pixels. Bill answered that question so fast ! Great chap and he is the VP for heavens sake. Talk about approachable !

Fraser

But are they all active and/or are they are all exposed to the application?  Has anyone probed one over SPI to see what the actual data is?  That would make things 1000x easier.

the datasheets have already been posted here, no need to probe

Surely the datasheet gives the resolution then.  If not, counting the pixel data as it leaves the Lepton would be the best way to determine resolution, certainly better than overlaying grids on a hyper-processed image.   From your Seek teardown we saw that there were a lot of elements that didn't make up the actual matrix of microbolometers used in imaging.   Maybe the Lepton uses some of its pixels to gather uniformity data or something, such that the array is a well and true 164x128 but the data that is exposed over SPI is a matrix of smaller size.  This would explain the image inconsistencies we see on pictures from it, while FLIR is technically correct about the number of elements.   That is all I meant by it.

[tomas123]

Lepton 3:
The weakest link in the chain is the lens quality and not the resolution of the sensor.
(see my images from Flir Exx with a great manual focus lens)

by the way, independent of quadrupled pixel count we have the same artefacts around the heat wire and the magic patterns (on a RAW image!!) 
80x60  https://www.eevblog.com/forum/testgear/flir-one-thermal-imaging-camera-teardown-and-hacks/msg551882/#msg551882
160x120 https://www.eevblog.com/forum/testgear/new-flir-products/msg752281/#msg752281 (see right side)