Yet another cheap thermal imager incoming.. Seek Thermal

[miguelvp]

Nice Frenky, I like your output better, and awesome work marshallh and eneuro!

Here is one of their LUT samples and the raw image with my interpretation of their Hot Iron LUT. Also I took yours and applied the same LUT.

Seems like they are doing a fast equalization, because their image is very close to mine. Other than theirs is a bit streched.


My interpretation of their Hot Iron LUT


From Raw, equalized and applying LUT and magnified x 2 (pixel replication zoom)


From their LUT samples


And Frenky's version that looks more like what I would expect.

[marshallh]

Seems they are doing a lot of filtering which is why all the public released shots are so blurry. The actual sensor data is quite good. This thing has potential

[frenky]

I added some code for iron and rainbow palletes:
(I resized images to 200% with irfanview: resample-Lanczos)

[marshallh]

Iron palette looks great. I was just trying to map this one and you beat me to it.
Seems as soon as we have physical units this thing is pretty much whipped.

[mikeselectricstuff]

Seems they are doing a lot of filtering which is why all the public released shots are so blurry. The actual sensor data is quite good. This thing has potential

How do you know that data is from one of their sensors? Could easily have been from something else, generated for testing before hardware was available.

The Ebay listing is almost certainly a pre-order as there is no pic of the real thing, and shipping is estimated at 17-27 Oct

[marshallh]

Seems they are doing a lot of filtering which is why all the public released shots are so blurry. The actual sensor data is quite good. This thing has potential

How do you know that data is from one of their sensors? Could easily have been from something else, generated for testing before hardware was available.

The Ebay listing is almost certainly a pre-order as there is no pic of the real thing, and shipping is estimated at 17-27 Oct

You may be right, the simu-device reports a much earlier firmware than the one included in the package (0.3 opposed to 0.6) so its possible it may be converted/massaged FLIR data.

[miguelvp]

But the raw data matches their claimed resolution.

[miguelvp]

Here is a paintshop pro LUT of what I think their Hot Iron LUT is.

I renamed it from .pal to .txt

Linear progression of red from 127 to 254 increments by 1.
Linear progression of green in increments by 2 with red set to max.

Maybe it should be from 128 to 255 increments of 1 for the first part, should look pretty much the same.

[Stonent]

I now have Java source of the library. I can see the device startup sequence (series of USB SetFeature requests) and it may be possible to get this running on PC with this information. Contact me if you want it.

I saw some of those references as well, but was unsure if they were device specific or just something normal used in USB.

[mikeselectricstuff]

If it's of any relevance, the Flir palettes are defined as YUV - I have an excel doc somewhere that converts them to RGB

[Stonent]

If it's of any relevance, the Flir palettes are defined as YUV - I have an excel doc somewhere that converts them to RGB

I saw YUV mentioned somewhere in the code as I was digging in it last night but also references to A8R8G8B8, R8G8B8, and R8G8B8A8.

I could have very well been in part of the code for exporting images so it may not really matter.

[miguelvp]

Been playing with the offset and even if my hot metal LUT is not as nice as frenky's (please share that LUT), here is the last frame.



So I'm not sure if the life float ring (I meant life saver just couldnt find the words) would show as hot. But being simulation images we will have to wait and see.

[frenky]

Hi.

The way I did it was to get png image for iron palette from Paletten256.zip on this site: http://u88.n24.queensu.ca/exiftool/forum/index.php?topic=4898.15

But I think you need rgb values so I extracted them into txt file (in the attachement).

[frenky]

There is a man on the boat... 
I really hope this is the actual sensor data, because the details are great.

[miguelvp]

There is a man on the boat... 

That's what she said  (Edit: in the US it has a kinky meaning)

Btw thank you the palette you sent me works a treat! with some image processing I think the boat ID is LA 1981 but the last two digits are blurry, could be 1987, but it's a water taxi

[amyk]

USB device (or at least beta) has VID 0x289D and PID 0x000F. Class/subclass 255 (Vendor specific)

Why didn't they use something like UVC? Do they really want vendor lock-in that much?

[eneuro]

I really hope this is the actual sensor data, because the details are great.

It will be interesting to see if I we make visual light photo in sunny day, convert to gray image, scale to this resolution and add some of those LUTs 
Probably many people will think it is thermal image if you told him and do not show oryginal visual light one 

For the moment I'm more interested if this camera will be supported by any Linux distro like many other crappy visual cams connected to USB, so one can use OpenCV to grab data and make some serious image processing in real time.
Without such simply functionality this thing is useless gadget only and I won't pay hundreds of $'es for it unless I can use it as just another camera but thermal.

[mikeselectricstuff]

USB device (or at least beta) has VID 0x289D and PID 0x000F. Class/subclass 255 (Vendor specific)

Why didn't they use something like UVC? Do they really want vendor lock-in that much?

Because they need to do a lot of processing to produce a visible image, so no particular reason to use a standard USB class.

[Fraser]

I have seen a lot of thermal imagery of vessels on water and sadly the image that you guys are manipulating does not look like a thermal image to my eyes. I am happy to be wrong but that boat is not displaying the thermal signiture that I would expect. The fact that the boats ID is so clear would be very unusual unless it was directly heated ! I have considered whether the boat is exposed to high levels of sunlight but it still looks wrong to me.

Your images do look like those produced by simulated (fake) thermal images. These just translate highlights and low lights to different spectrums to simulate the palettes of a TIC.

Try applying a visible light pallete to the data and I suspect you will find it is a visible light image.

Aurora

[frenky]

In gray it does look quite regular...

[callipso]

Oh yeah. Bummer.

[mikeselectricstuff]

The fact that the boats ID is so clear would be very unusual unless it was directly heated !

..or different emissivity - warm day, dark boat?

[callipso]

The guy does however look visible-spectrum-y...

[David Hess]

I have seen a lot of thermal imagery of vessels on water and sadly the image that you guys are manipulating does not look like a thermal image to my eyes. I am happy to be wrong but that boat is not displaying the thermal signiture that I would expect. The fact that the boats ID is so clear would be very unusual unless it was directly heated ! I have considered whether the boat is exposed to high levels of sunlight but it still looks wrong to me.

I was thinking that the images include a lot of near infrared in them lowering the far infrared contrast and making them look more like a visible light image.

[callipso]

I have seen a lot of thermal imagery of vessels on water and sadly the image that you guys are manipulating does not look like a thermal image to my eyes. I am happy to be wrong but that boat is not displaying the thermal signiture that I would expect. The fact that the boats ID is so clear would be very unusual unless it was directly heated ! I have considered whether the boat is exposed to high levels of sunlight but it still looks wrong to me.

I was thinking that the images include a lot of near infrared in them lowering the far infrared contrast and making them look more like a visible light image.

Aren't chalcogenide lenses nearly opaque to nIR?
This is more likely a dummy imagery converted from visible.