Reflectance Estimation

#81

@muzammil360 thanks for the reply nonetheless.

If there is anyone else that has experience of processing Multispec 4C imagery, can they weigh in with their thoughts as it would be much appreciated?

#82

Can someone explain a bit clearer how to get irradiance without the `SensorModel` tag? How do I apply the equation `count = A’ exposure gain irradiance + B’`? I can get irradiance from the sunshine sensor quite easily, but havent figured out how to do it for the Sequioa without the `A`, `B` and `C` coefficients from `SensorModel` tag.

#83

Simply solve for irradiance in your equation. `B` and `B'` are identical when there is no light and the remaining `A'` gets absorbed into `K`.

Chances are you really do not mean to do this, as a two coefficient model will necessarily be a worse fit.

Think of `C irradiance + B` as a more precise `B'` that explicitly says that it depends on `irradiance` .

#84

Hi @jeromeoc
Unfortunately without the SensorModel tag, it is impossible to get irradiance information. However, if you update your Sequoia to the latest firmware, you should get access to this tag.

Best regards

#85

so instead of using `I = f2(p-B/AeY + C)`, where `A,` `B` and `C` are the `SensorModel` tag values, I solve for
`I = f2(p/eY)` where:
`I`: Irradiance at sensor
`f2`: f-number squared (k)
`p`: raw pixel value (value)
`e`: exposure time (t)
`Y`: ISO value (K)

#86

Yes, the camera was updated half way through a survey season so the `A`, `B` and `C` coefficients for each band are the same through the lifetime of the camera? That way I can retrospectively go back and correct the images that were acquired before the firmware update?

#87

You’re right, the A,B and C are calibrated in factory and so remain constant for the lifetime of the camera.

#88

Perfect! Thanks for all your help

#89

To get Sequoia irradiance which is sunshine corrected (i.e. taking into account the values from the sunshine sensor) should I use `I' = Iseq / Isun` to scale it to the values of the sunshine sensor or just use the values that come from `I = f2(p-B/AeY + C)`?

#90

This forum topic has been extremely helpful. I have a sequoia camera that I bought used so I don’t have access to Pix4d or Agisoft that does the radiometric calibration, and I have some experience in coding in open source software so I wanted to do it myself. I’ve got the metadata (including the base64 data) that I need, and have found a K value for my plate calibration image (for both NIR and RED). I’m a bit loss as to what to do next. Do I just apply that K value to all my images and then I’m done or do I have to apply a (K * sequoia/ sunshine) which I can find the sunshine value but how do I find the sequoia image (as its not the average of pixel values in the plate anymore)?

#91

Hi everyone.

I’m trying to calculate reflectance from image taken by Sequoia+.
However, SensorModel of Sequoia+ seems to be incorrect.
Here is example of SensorModel in green band image:
317048289.62674713, 4954.867746006242, 3950.8731948881677
Compared to example in SEQ-AN-01, SensorModel A of Sequoia+ is much larger. (About 1000 times!)
Other value like ISO, f-number, exposure time is almost same as example in SEQ-AN-01.
If I use calculation method written in SEQ-AN-01, irradiance of pixels will be too small.

How can I calculate reflectance from image taken by Sequoia+?
Is there another calculation fomula for Sequoia+?

#92

Hi @yutak
Thank you for pointing out this problem.
After investigation we find that our implementation of the legacy radiometric calibration (in the application notes) from that for Sequoia+ (much more comprehensive, Pix4D only) differs in the time units. As a consequence, the second needs to be divided by a factor of 1000 for correct irradiance estimation.

Best regards

#93

Thank you so much!

1. You mean exposure time `e` in fomula `I = f2(p-B/AeY + C)` needs to be devided by 1000, right?

2. I’m using both Sequoia and Sequoia+.
So, before calculating reflectance from images,
I need to identify which camera (Sequoia or Sequoia+) takes the images,
as I need to change calculation formula (devide `e` by 1000, or not) in response to camera.
How can I identify camera (Sequoia or Sequoia+) from EXIF tag?
Isn’t there naming rule of serial number?

3. Does Pix4D deal with the difference of calculation formula between Sequoia and Sequoia+?

#94

Exactly

You can use the tag Maker Note to know whether you have a SQ or an SQ+. In a standard SQ, this tag should be 0.0 and for a SQ+ with the appropriate sunshine sensor it should be 1.3.
See files attached. SQ tags.zip (4.7 KB)

Yes, Pix4D knows how to deal with both models.

#95

I’ve understood the calculation formula.

I found EXIF tag `Maker Note Unknown Text` in image I took by Sequoia or Sequoia+.
Some of this tag were 1.3, and others were 0.2, not 0.0.
So, you mean that `Maker Note Unknown Text` should be 0.x in Sequoia, right?
And, is this EXIF tag 1.x in Sequoia+?

Also, Some of images don’t have `Maker Note Unknown Text`.
These images are taken by Sequoia with old firmware (before v1.4.1)
or Sequoia+ with old firmware (v1.5.0, which was originally installed when I bought)
Is there another way to identify camera from EXIF tag?

Thanks.

#96

Is it still under investigation?

If my explanation is not enough, please tell me.

Best regards

#97

Hi @yutak

The first digit of the EXIF tag `Maker Note Unknown Text’ corresponds to the version of the camera (+ or not) and the second one to the version of the sunshine sensor (+ or not). So basically if your tag is anything BUT 1.3, then this is not a SQ+ complete system.

If you used an old firmware without this tag, you can also look for the file sequoia_therm.dat in your acquisition folder. If present, then you have a SQ+

Cheers

#98