Case Study on Multispectral Land Cover Classification - Open Source Geospatial Tools: Applications in Earth Observation - page 293

-f SQLite
We select the SQLite format for the output vector file.
-o forest_ndvi.sqlite
Name of the output vector file.

We refer to Sect. 12.4 for calculating the mean value and standard deviation of the
extracted NDVI sample. Also the histogram is shown there (Fig. 12.5 ) . We set the
NDVI threshold for the vegetation mask to the mean value minus twice the standard
deviation. For the current Landsat image, this corresponds to a value of 3,000 (which
corresponds to the value 0.3 without the scaling factor of 10,000). We acknowledge
that the NDVI threshold can also be estimated interactively using a desktop GIS (e.g.,
QGIS), but, we prefer to present an approach that can be fully automated.

gdal_calc.py -A LC82070232013_ndvi.tif
ₒ --outfile=LC82070232013_ndvi_masked.tif

ₒ --calc="(A>3000)*A" --type=UInt16

-A LC82070232013_ndvi.tif
Name of single band input raster dataset containing NDVI values.
--outfile=LC82070232013_ndvi_masked.tif
Name of output raster data set.
--calc="(A>3000)*A"
Calculates the mask: keep input values if above 3000. All values less or equal
will be set to 0.
--type=UInt16
Output data type is unsigned (short) integer.

The calculated NDVI image can also be useful as input for the classifier, so we
stack it to the Landsat composite, using the utility gdal_merge.py .

gdal_merge.py -o LC82070232013_features.tif -of GTiff -separate
ₒ

-co INTERLEAVE=BAND -co COMPRESS=LZW

ₒ LC82070232013_ndvi_masked.tif LC82070232013_composite.tif

-o LC82070232013_features.tif
Name of the output raster dataset.
-of GTiff
Output image is a GeoTIFF.
-separate
Place each input file into a separate stacked band.