Analyzing MicaSense data processed with Agisoft Metashape – MicaSense Knowledge Base

After you have processed your data using Agisoft Metashape, you may have a few more questions on how to use the outputs from this process.

Agisoft helpfully exports orthomosaic reflectance maps as multiband TIFF files. This makes the process of creating composites and index maps in GIS easier.

By default, Agisoft exports reflectance maps with 16-bit integer values. That is, the pixel values will range from 0 to 65,535. For 12-bit images, that would be 0-4096.

You may see something like this when importing the orthomosaic into QGIS:

However, it looks better after changing the band order.

NOTE: 3/2/1 is the correct RGB band order for all RedEdge and Altum cameras. However, if you are using the 10-band camera (Dual Camera System), the band order for RGB is 6/4/1.

While these values still represent radiometrically calibrated data, you may be expecting to see a range from -1 to 1, as you would expect from other software. This output makes creating NDVI maps more intuitive.

You may use GDAL (specifically gdal_calc) to convert the 16-bit (UInt16) data to 32-bit floating point numbers). Keep in mind converting between 16 bit and 32 bit will roughly double the file size. However, you can also perform this calculation in QGIS.

gdal_calc.py --type=Float32 -A input.tif --allBands=A --outfile=output.tif --calc="(A*1.0)/32768.0"

Reflectance TIFF file after importing into QGIS:

Notice that the values range from 0 to nearly 2. Values higher than 1 (100% reflectance) are possible so that abnormally reflectant surfaces like white cars or bodies of water don't wash out the rest of the image. After changing band order for RGB and normalizing min/max values:

To create an NDVI map, you can use the raster calculator in QGIS to perform the calculation:

(NIR - Red) / (NIR + Red)

Remember that Agisoft exports the bands in wavelength order, so the band order is:

Blue, Green, Red, Red Edge, NIR

Thus:

After running this command, QGIS shows the following:

This isn't very helpful, so we want to change the layer style to Singleband pseudocolor:

Notice we have changed the min/max values to 0/1. Technically NDVI ranges from -1 to 1, but using this scale would make the map all green, since values under 0 are unusual in most cases.

You may experiment with changing the min/max values to fit your needs, but it is easy to distinguish bare ground and human-made features from vegetation with this map, despite it not being an agricultural area.