4SM shallow water column correction: 
demonstration using a CASI image

from import of raw data all the way to seatruth

only 54 minutes on my new machine

Demonstration is now final as of May 2nd 2014

 
 
4SM demonstrator step 0
Archiver
  1. Remove/erase all image_specific files and directories
    1. apart from a few shapefiles
    2. apart from the 4SM script
    3. apart from the following directories
      1. /4sm.calls/...
      2. /data/...
      3. /seatruth/... 
      4. /OPF/...



4SM demonstrator step 1
Create database structure
  1. write PCIDSK header file in FILE format
  2. write void U8, U16, S16 and R32 channels




4SM demonstrator step 2
Importer
  1. Import the raw data from /data/... for all bands into raw data channels
 





4SM demonstrator step 3
Importer
  1. Ensure that raw radiance is >=1 for all data pixels
  2. Create a Panchromatic band (if appropriate),
    1. and write it in place of designated band





4SM demonstrator step 4
Scaler
  1. Scale the raw spectral data to the 8 bits range 0-255
    1. for image display purpose only
    2. and write into designated U8 channels





4SM demonstrator step 5
AutoCalibrator
  1. create a mask in designated channel
    1. map all NoData pixels to 239 in the special effects mask
    2. use Lsw.shp shapefile to extract spectral deep water radiance Lsw
    3. use LsM.shp shapefileb to extract spectral radiance LsM of bright bare soil like bright beach sand
    4. use glint.sh to extract the sea-surface glint regressions for all pairs of bands
    5. use various other shapefiles to populate the special effects mask as required





4SM demonstrator step 6
AutoCalibrator
  1. Now that the special effect mask is available
    1. extract calibration data from the image
    2. then produce and display a provisory commandline in textfile AutoCAL.txt
    3. then produce and display a provisory AutoCalibration plot





4SM demonstrator step 7
Importer
  1. Import seatruth DTM, if available





4SM demonstrator step 8
Extractor
  1. Perform a proper extraction of calibration data from the image





4SM demonstrator step 9
Calibrator
  1. Produce and display the image_specific optical calibration plot 
    1. for bands NIR, RED, GREEN and BLUE 
    2. for specified value of the ratio Ki/Kj
    3. at specified wavelengths WLi and WLj  





4SM demonstrator step 10
Typer
not using the panchromatic band
  1. Use training site shapefiles to extract and display bottom type signatures
  2. Write bottom type signatures text file
 





4SM demonstrator step 11
Modeling
Bottom typing
Seatruth evidence
not using the panchromatic band yet
 
Now that all parameters and bottom type signatures are specified
  1. Modeler: process the image
    1. Classify: during this run, also arrange to classify bottom types "on the fly"
    2. CalibratePan: during this run, also arrange to extract/optimize the optical calibration of the panchromatic band "on the fly"
  2.  Modal_filter: then run a Modal_filter on the bottom types map and write it to specified channel
  3. Seatruth: then compute the seatruth difference image ZRecorded+Htide-ZComputed
    1. and display a seatruth evidence plot
 





4SM demonstrator step 12
as of step 12, 4SM now uses the panchromatic band

At this stage, the processing is finished
  1. retrieved depth image is available for viewing
  2. bottom typing   image is available for viewing
  3. the optical calibration of the panchromatic band has been estimated

 

Now that we have secured
proper results and seatruth evidence, 
we take a chance that results
might be improved thanks to the panchromatic band

 
Calibrator
  1. Compute and display a calibration plot involving the panchromatic band
    1. that's for bands NIR, PAN, GREEN and BLUE
    2. for specified value of the ratio Ki/Kj
    3. at specified wavelengths WLi and WLj  





4SM demonstrator steps 13 to 16
Modeling
Bottom typing
Seatruth evidence
using the panchromatic band
 
exactly the same as step 11, except for the panchromatic band
Now that all parameters and bottom type signatures are specified
  1. Modeler: process the image
    1. Classify: during this run, also arrange to classify bottom types "on the fly"
  2. Modal_filter: then run a Modal_filter on the bottom types map and write it to specified channel
  3. Seatruth: then compute the seatruth difference image ZRecorded+Htide-ZComputed
    1. and display a seatruth evidence plot
 

e tutto finito May 2nd 2014
 



Créer un site
Créer un site