low marine areas, using passive remote sensing imageries
WorldView 2 image of Waimanalo Beach, Oahu, Hawaii islands
3289x3241, 2 m ground resolution, courtesy of Ron Abileah

return to WV2 at Oahu
next to Using Ron's deglinted image

The webpage below is to let the reader
get a feeling of the difficulties and uncertainties
that I had to go through
because this image is such a unique case!
The final stuff is at  Using Ron's deglinted image

The data
2 m pixel size, courtesy of Ron Abileah
This dataset from Ron Abileah is comprised of
  • WV2 8-bands image, 3289x3241 pixels, rotated by 90 ° (????????)
  • WV2 8-bands images decluttered by Ron
  • SHOALS depth in meters  -  this is the depths from the SHOALS LIDAR re-sampled to the image pixels.
  • WKB depth - this is the depth  solution from Ron's wave kinematics algorithm using 2 images separated 10 seconds

TCC this scene has been rotated by 90°


your previous DTM is in centimeters

USGS's SHOALS Depths in meters:
this DTM is 1, 2, 3... by discrete values of meters

908 nm
White pixels are at 16U=0
  831 nm
White pixels are at 16U=0

  724 nm
White pixels are at 16U=0

  659 nm

April 10th
Some BiDimensional histograms
Some BiDim histogram raw data whole image
2 vs 3, 2 vs 4, 2 vs 5
Some BiDim histogram raw data whole image
5 vs 6, 5 vs 7, 5 vs 8
  • Please note that all histograms start at Zero for all bands

  • This can't be raw data!

  • Why have this data been truncated? It upsets my 4SM routine. 4SM expects raw data from data provider

April 10th
Optical calibration
  • One of the unknowns at this stage is the effective operational wavelength for each WV2 waveband, as regards the use of Jerlov's data for estimating spectral K in m-1 for visible bands

    • by visual inspection of the computed depths profiles, I seem to need WL[yellow]~=603.6 nm, and other visible bands at mid-waveband

  • call this "depth estimates when initialized with SHOALS depths"

  • hopefully, and over time, and after processing some clear-cut WV2 images with fair coverage of bright bottom substrates over the whole depth range, this shall become expert knowledge and field data will not be needed at all for calibration, just as is the case when using narrow bands of CASI and HYPERION hyperspectral images where wavelengths are much easier to specify

  • latest seatruth results seem to consolidate this view: CoefZ~=1.00 fits the situation

  • then see  No need for field data

Calibration for bands 5, 4, 3 and 2

Calibration for bands 5, 4, 3 and 2

Calibration for bands 5, 4, 3 and 1

Calibration for bands 5, 4, 3 and 1

Wavelengths adopted are:

  • Oceanic water is as clear as it can possibly get

  • This calibration seems to confirm that wavelengths must be offset from mid-wavebands towards shorter values

    • this is clearly the case for the yellow bands, as is the case for red and green

Créer un site
Créer un site