boson radiometric

[bakas90]

I'm using a Boson R camera.
I am getting raw data using flirpy.

However, I'm not exactly sure how to get the metadata.

I want to get the radiometric value.
I also want to create a SEQ file.

I would be grateful if you could tell me how.

[jveitchmichaelis]

Hi,

I want to get the radiometric value.

Flirpy was written before the R variant was released and unfortunately I don't have a radiometric camera to test with.

Do you know what format the raw data from the camera is returned as? It should be straightforward to update the library to include radiometric support. Documentation is here

I'd guess we need to check if TLinear mode is enabled at least, and there seem to be options to control what units are output.

I also want to create a SEQ file.

SEQ files are a sequence of FFF files. In theory you could construct these manually, but this isn't something that Flirpy supports - it's mostly used in the opposite direction (getting data out of SEQ files).

Is there a reason you want an SEQ file?

[jveitchmichaelis]

Assuming you've created a camera object, give this a go. This should change the video mode to TLinear.

function_id = 0x000E000E #sysctrlSetUsbVideoIR16Mode
command = struct.pack(">H", 2) # FLR_SYSCTRL_USBIR16_MODE_TLINEAR = 2
res = camera._send_packet(function_id, data=command)
#res = camera._decode_packet(res) # Not needed, but you can see what res is to check if the command went through

# Now try grabbing a frame

...

This should put your camera into TLinear mode and then multiply your raw values by 0.01 to get to Kelvin.

[jveitchmichaelis]

@bakas90 I see the thumbs up, any luck with this? :)

[olfk1234]

@jveitchmichaelis Hello! I tried the code above and I got a warning stating "WARNING:flirpy.camera.boson:Boson response: bad command ID" when running the decode packet function. Is there another workaround to convert the Boson into TLinear mode or to get the temperatures using flirpy?

[jveitchmichaelis]

If you go to the receive function and print/dump the return from the camera, what do you get?

I don't have a radiometric Boson so I'm unable to check this myself.

[olfk1234]

@jveitchmichaelis After doing print(camera.receive()) after the code that you mentioned before, I am getting b' '

[jveitchmichaelis]

Try with 0x000E000D:

Looks like a typo with the previous function code (that was get, not set).

However you should also send the 0x000E000E command after to check that the setting is correct. I suggest having a read through the document in the link above, it has all the information about which commands you can send and what should be returned. It should be straightforward to figure out how to use Flirpy to implement the ones you need.

[millercommamatt]

I'm just diving into flirpy for my radiometric Boson. I requested the USB Vidieo IR Mode as show below. The returned status value was 2 which should be T-Linear mode. Grabbing an image resulting the an unsigned 8-bit numpy array which doesn't seem correct to me. I was expecting a uint16 array.

I'm sure I'm missing something simple, but this is my first time both with this library and more importantly, a Boson camera.

>>> res=camera._send_packet(0x000E000E)
>>> res=camera._decode_packet(res,receive_size=4)
>>> res
b'\x00\x00\x00\x02'
>>> image=camera.grab()
>>> image
array([[ 93, 117,  93, ..., 117,  86, 117]],
      shape=(1, 655360), dtype=uint8)

>>> res=camera._send_packet(0x003E0002)  # TLinearGetControl()
>>> res=camera._decode_packet(res,receive_size=4)
>>> res
b'\x00\x00\x00\x01'  # 1 = FLR_ENABLE

[millercommamatt]

Please disregard my prior message.

I wasn't understanding the output that was being returned.

I needed to reshape the vector returned and combine the big end and little end of the 8-bit values to construct the 16 bit temperatures.

>>> image_reshaped = np.reshape(image,(-1,640,2))
>>> image_uint16 = np.uint16(image_reshape_2[:,:,1])*256 + np.uint16(image_reshape_2[:,:,0])

[jveitchmichaelis]

Yeah though you shouldn't need to use the low level functions to do this. Interesting that grab is not returning a 16 bit array though, it should handle that conversion for you... If you look at the source for the Boson class, grab should just return an image from opencv's capture interface (which is setup to be 16 bit).

Out of interest is this a radiometric camera?

[millercommamatt]

I'll dig into why OpenCV isn't returning the expected data structure. Issue #105 reported the same phenomenon. Let me know if you have troubleshooting suggestions. I've not tangled with OpenCV before.

I'm using a Radiometric Boson 640.

Thanks for this library. It saved me a ton a work. I might send some PRs to expand some of the functionality for the Boson class.

[jveitchmichaelis]

Yeah it looks like a regression in OpenCV so you could try and bisect to figure out where it was introduced and we could raise an issue directly.

Otherwise I would really welcome PRs for radiometric functionality as this is something I don't have the ability to test at the moment. Or whatever you find is awkward - perhaps we need some higher level functions to switch modes.

Thanks!

[jveitchmichaelis]

By the way you can also try casting the array directly using view, prior to reshaping: https://numpy.org/doc/2.2/reference/generated/numpy.ndarray.view.html

[millercommamatt]

Perhaps this is the origin of the apparent change in OpenCV behavior?
https://github.com/opencv/opencv/pull/18498/files

Note the changes in output data types for case V4L2_PIX_FMT_Y16:...