Input image preparation
Make input ROI images
CARS supports the ROI products extracts done with the otbcli_ExtractROI OTB application (raster + geometric model).
First, retrieve the coordinates of the desired extract on the first image (lets call it
img1.jp2
), under the formstartx
,starty
,sizex
,sizey
(in pixels).Perform the extraction of the first image with:
$ otbcli_ExtractROI -in img1.jp2 -out img1_xt.tif uint16 -startx startx -starty starty -sizex sizex -sizey sizey
Extract the same zone on the second image (for example
img2.jp2
), the-mode fit
application option has to be used:
$ otbcli_ExtractROI -in img2.jp2 -out img2_xt.tif uint16 -mode fit -mode.fit.im img1_xt.tif
The application will automatically look for the zone corresponding to img1_xt.tif
within img2.jp2
.
It is possible to use the -elev.dem srtm/
option to use the DEM during this search in order to be more accurate.
It is to be noted that the -mode.fit.vec
option also exists. It accepts a vector file (for example a shapefile or a kml) which enables the image extraction from a footprint.
Make a simple pan sharpening
In the case of Pleiades sensors, the XS color isn’t superimposable to the Panchromatic image.
It can be recommended to apply a P+XS pansharpening with OTB.
$ otbcli_BundleToPerfectSensor -inp image.tif -inxs color.tif -out color_pxs.tif
Make a water mask
To produce a water mask from R,G,B,NIR images, it can be recommended to compute a Normalized Difference Water Index (NDWI) and threshold the output to a low value.
The low NDWI values can be considered as water area.
$ gdal_calc.py -G input.tif --G_band=2 -N input.tif --N_band=4 --outfile=mask.tif --calc="(G-N)/(G+N)<0.3" --NoDataValue=0
See next section to apply a gdal translate to convert the mask with 1bit image struture.
Convert image to binary image
To translate single image or multiband image with several nbits per band to 1bit per band, it can be recommended to use gdal_translate as follows:
$ gdal_translate -ot Byte -co NBITS=1 mask.tif mask_1nbit.tif