Orthomosaic from RGB and NIR images

[ben73]

Hello,
i have a Phantom 4 and a Raspberry Pi with a NOIR Camera. The NOIR Camera has a NIR Filter infront.
The Raspberry has a GPS Modul. So the NIR images are geotaged.
I already created two orthomosaic images. One for RGB and one for NIR.
How can i align them and create NDVI images?

[ben73]

Hello,
i have a Phantom 4 and a Raspberry Pi with a NOIR Camera. The NOIR Camera has a NIR Filter infront.
The Raspberry has a GPS Modul. So the NIR images are geotaged.
I already created two orthomosaic images. One for RGB and one for NIR.
How can i align them and create NDVI images?

[FlyRight-3D]

I suppose that could be possible if you were able to trigger both cameras at the same time and ensure they have the same ISO settings and whatnot.   Basically, you want the red band of your RGB camera and the NIR band to be collecting the "exact" same reflectance because even half of a second of delay between the two could and most likely will give you erroneous information.  Because a ratio of NIR and Red reflectance are used to create an index from -1 to +1, which indicates good (or poor) healthy vegetation, that index can be skewed one way or another if the two sensors are collecting different reflectances. 

I hope this helps.

[Dave Martin]

Hello,
i have a Phantom 4 and a Raspberry Pi with a NOIR Camera. The NOIR Camera has a NIR Filter infront.
The Raspberry has a GPS Modul. So the NIR images are geotaged.
I already created two orthomosaic images. One for RGB and one for NIR.
How can i align them and create NDVI images?

As they are from different cameras with different lenses, you'll need the two orthomosaics you mention, and you'll have to geo-reference them both.  Then, load the two ge-referenced orthomosaics into Surfer or QGIS or similar and calculate your chosen index.

[ben73]

Thanks Dave. Do you know a tutorial how to do this in QGIS?

[FlyRight-3D]

I'm sorry I didn't address the alignment issue.  I was more fixated on the integrity of the data you'll be using for either your own decision making or for a customer and I want you to be aware of the possible discrepancies in your vegetation indexes

[SAV]

Hi ben73,

You'll need to use the RASTER CALCULATION tool in QGIS to calculate NDVI values.
Here is video which shows the workflow (with Landsat data though, but it is also just a multi-channel image):
https://youtu.be/4YocMAE9FA0

However, make sure that both of your orthophotos (RGB and NOIR) are properly georeferenced/aligned because the raster calculator will calculate NDVI values based on a pixel by pixel basis, hence you want to have the each RGB pixel 'stacked' on top of the associated NOIR pixel.
Therefore, you might need to georeference/align both orthophotos in QGIS based on features that are easy to identify in both datasets RGB and NOIR. You might also need to adapt the resolution of the RGB orthophoto in order to match the lower resolution NOIR orthophoto before calculating the NDVI.

All the best with your project/work.

Regards
SAV

[jessica_tay]

Hey all,

I also have a RGB image and 2 NIR image. All three were shot at the same place and have the same geotagged locations.

I aligned all three set of images separately in different chunks and am I am trying to see if it is possible to align the chunks together on top of each other. The first time I tried it, i aligned the chunks with point cloud reference, but it took 27 hours to complete.

However, only the RGB chunk aligned with one of the NIR chunk.

I know that it would be easily to export the orthomosaic of each chunk separately then analysis it further in GIS, but I was wondering if it was already possible to align the layers in PhotoScan itself.

Thanks.

Jessica 

[Alexey Pasumansky]

Hello Jessica,

At the moment it is not possible merge orthomosaics from different chunks as multispectral image layers. You can create multispectral output directly from PhotoScan for your dataset only in case you process the images using Multispectral cameras approach.

[jessica_tay]

Hello Alexey,

Thank you for the reply!

Is there a tutorial for multispectral camera approach that I can look at?

Thanks!
Jessica