6949*5174 CASI image of Heron Island,
UTM zone 56, 36 millions pixels, ground resolution 1 m, July 1, 2 and 3rd 2002
Heron Reef data provided courtesy of the Center for Spatial Environmental Research at the University of Queensland.
Datasetprovided by HyPhoon.com.

work done end march 2014
Introducing the use of a Panchromatic waveband
with hyperspectral imageries


Producing a CASI panchromatic band
Optical calibration of the Panchromatic band
Profiles Red, Yellow and Bue
Bottom type training sites
Bottom typing process in 4SM

Bottom typing result

The following does not use the "PAN solution"
as I still have some problems in the 0-2 m depth range

(March 30th 2014)

4SM runtime
82 minutes on my Toshiba Intel Core I3 X86 laptop 4GB RAM

Seatruth NOT using the "PAN solution"
tide height has been set to 0.20 m, and added to "seatruth" DTM depths which are tide-corrected (Hedley et al)

Z4SM - ZDTM+20 cm
in centimeters
  • Red tones where  over-estimated  depths
  • Blue tones where under-estimated depths
  • severe and orderly discrepancies are observed: this is not caused by noise.

  • Discrepancies:
    • refer to previous work
    • they seem to be less proeminent here, following a distinct improvement of the calibration as regards the operational wavelengths


ZComputed is Z4SM
ZRecorded  is ZDTM+20 cm

Tide height was ultimately tuned to 0.20 m
in order to achieve best fit of this plot
in the 0-1m depth range

Bottom type training sites
Training site may be
  • a polygon: all pixels inside the polygon
  • a polyline: all pixels along the polyline
Blue/Green spectral range
In 4SM, bottom typing uses water column corrected spectral signatures:
  • It only uses the BlueGreen range, in order to ensure a "fair" palyground over the whole depth range: this includes bands 1, 2, ..., 7
  • Therefore, spectral characteristics in the GreenRed range are disregarded: that's bands 8, 9, ..., 17, which are possibly extinct because of diffuse attenuation of sunlight
Training sites ROIs
for types 10,  11 and 12
background is a modal-fitered version
of bottom typing map

The above colors are not intended
to mean/suggest any particular benthic habitat
Training sites ROIs
for types 13, 14, 15 and 16
background is a histeq-enhanced version
of normalized watercolumn corrected TCC
Training sites ROIs
for types 17, 18, ..., 22, 23
background is a plain
BOA watercolumn corrected TCC

Below: pixel signatures accepted/used for deriving an average signature are colored yellow.
It is disappointing that these signatures exhibit so few contrast in the Blue/Green range.
Refer to what I obtained using CASI at Gezirat Siyul Reef.

type 10

type 11

type 12

type 13

type 14
type 15    type 16

type 17

type 18

type 19

type 20

type 21

type 22   type 23

Bottom typing process in 4SM
as of march 2014

Water column correction  in 4SM yields a signature for each shallow pixel.
This signature is used to open a spectral acceptance window : plus or minus exp(2K).
For a pixel signature to be classified as type_N,   the average type_N signature has to
  • fit inside such window over the Blue/Green range of bands 1 to 7
  • exhibit the closest match over the Blue/Green range of bands 1 to 7
Location of typing

Inverse simplified radiative transfer model at the base of atmosphere (BOA)
  • The cyan.................profile....shows the BOA water volume reflectance                                                 Lw
  • The dark blue.........profile....shows the BOA raw reflectance : image data                                    L
  • The grey..................profiles..show several iterations while increasing depthg in the above model        Z
  • The red...................profile....shows the final water volume corrected BOA signature                           LB
  • The upper..yellow...profile....shows BOA     Lw+exp(log(LB-Lw)+2K)
  • The middle.yellow..profile....shows BOA signature for type_21
  • The lower..yellow...profile....shows BOA    Lw+exp(log(LB-Lw)-2K)
==> This pixel is classified as bottom type_21


Créer un site
Créer un site