Poivre Island, Seychelles
a 1950*12233 CASI image, pixel size 1 m, 17 spectral bands

image courtesy of HDI
the cost of the 4SM licensekey for this image is 2014 Euros
 
return to 4SM Study Cases


 
1 - NO NEED for field data, nor for atmospheric correction
2 - this is demonstrated in this website, using a variety of hyper/multi spectral data
 
Requirements are
1 - homogeneous water body and atmosphere
2 - some coverage of optically deep water
3 - some coverage of dry land
 
Problems are
1 - the precision on estimated depth is found wanting, because the noise-equivalent change in radiance  of accessible data is too high for shallow water column correction work 
2 - radiance data should be preprocessed by the provider at level 1 in order to improve S/N ratio
3 - exponential decay: the deeper/darker the bottom, the poorer the performances
 
So
I keep digging
until suitable data
become available
 
4SM processing january 2006
I want to redo it in 2014: shall be much better: see CASI at Heron Island!!

see Sarah Hamylton's paper on CASI at Alphonse Atoll, Seychelles 2011


4SM processing january 2006

go to bottm typing

SUNGLINT
This image is marred with heavy sunglint,
which should have been avoided by flying into or away from the sun.

Removing most of the sunglint has been a challenge,
as sunglint does not have the same properties as the usual skyglint.
Much has been learned in this regard, and 4SM has been improved accordingly.
DEEP SOUTH
The southern half of this image lies at approx 30 m.
it appears to be scattered with very dark towering mounds at approx 4-6m.
In order to yield some results in spite of adverse sunglint, the model had to be operated beyond reasonable limits. 
AMAZING
Altogether, I think these results are amazing:
  • no fancy gimmicks such as derivative spectroscopy, spectral libraries, neural network classifier, principal components analyzis, you name it
    • just a simple empirical "ratio" method inspired by Lyzenga's and Maritorena's work
    • although sub-pixel analysis is a way forward
  • no atmospheric corrections needed
    • most of heavy sunglint effectively is removed
    • although removal of the adjacency effect would be a way forward
  • no field data needed/used
    • yields both computed depth and water column corrected "low-tide" image which is ready for bottom type classification
    • an exclusivity of 4SM!
    • results down to ~30 m of depth in these clear tropical waters


 
 
from left to right:
  • raw TCC
  • deglinted TCC
  • "low-tide" TCC
  • "low-tide normalized" TCC
  • 10 bottom types
  • computed depth


Below is a zoom window
TCC raw image
TCC deglinted image
TCC bottom reflectance
a "low tide" view
TCC normalized bottom reflectance
enhances the visibility of contrasted bottom signatures
Computed depth Z10
Computed depth Z250
Bottom signatures classified into 10 bottom types
raw image
Bottom signatures classified into 10 bottom types
a modal filter has been applied

Calibration diagram
FAIRLY CLEAR WATERS,
BRIGHT IMAGE, BRIGHT SANDS
excellent calibration
Jerlov optical marine water type ~OII



 

Bottom Typing


STATISTICS OF SHALLOW BOTTOM TYPING in  BOA radiance 0-200
 CoefSTD=3.00 nbMISS=3 -E1/1950/1/12233  on Fri Jan 20 08:57:12 2006

The Classifier tool in 4SM is under development: these results should be taken with precautions
 
Type_0 is Land pixel
Type_1 is Unclassified Shallow pixel
Type_2 is Optically Deep water pixel
Type_3 is No Data pixel
Type_4 is Wave Breaker pixel
Type_5 is Cloud/Shadow pixel
Type_6 and above is for Classified Shallow pixel

  1.40% as type_1   UnClassified shallow pixels

  2.42% as type_6   SIG   4.9   4.7   4.5   3.3   3.7   4.2   4.4   4.5   5.2   5.0   4.7   4.9   4.9   4.0   8.5   0.0   0.0
BGSL   4                  STD   0.4   0.3   0.3   0.3   0.3   0.3   0.4   0.4   0.4   0.4   0.4   0.4   0.5   0.4   0.6   0.0   0.0


  7.45% as type_7   SIG   7.7   7.9   7.7   6.3   6.8   7.1   7.1   7.3   7.6   7.6   7.6   8.7   9.9   9.3  14.2   0.0   0.0
BGSL   7                  STD   0.5   0.5   0.6   0.6   0.6   0.5   0.7   0.6   0.6   0.6   0.6   0.6   0.6   1.3   0.7   0.0   0.0


  6.02% as type_8   SIG  10.4  10.6   9.6   8.8   8.4   8.6   8.7   8.7   9.1   9.0   8.9   9.3   9.6   8.7  15.6  11.4   0.0
BGSL   9                 STD   0.8   0.9   0.9   0.9   1.0   1.0   1.0   1.1   1.1   1.1   1.1   1.0   1.0   1.1   1.4   1.7   0.0


 13.35% as type_9   SIG  15.8  17.9  16.7  15.7  15.7  16.6  15.4  15.5  16.1  15.9  16.1  17.8  21.9  19.3  23.5   0.0   0.0
BGSL  16                  STD   1.2     1.4    1.3    1.6     1.4    1.5     1.3    1.3    1.4     1.3    1.3     1.3   1.5      1.2    0.2   0.0   0.0

 
23.13% as type_10  SIG  23.8  26.2  26.2  25.7  26.0  26.7  26.1  27.4  27.9  27.5  26.3  26.4   0.0   0.0   0.0   0.0   0.0
BGSL  26                   STD    1.9    2.1     2.2    2.3    2.2    2.3     1.9     2.0    2.1    1.9     1.7    0.5   0.0   0.0   0.0   0.0   0.0


 19.25% as type_11  SIG  38.3  41.2  41.9  42.6  42.2  40.5  40.5  40.7  41.8  39.8  31.6   0.0   0.0   0.0   0.0   0.0   0.0
BGSL  41                    STD   2.8   3.1   3.4   3.2   3.0   2.8   2.3   2.2   2.1   1.9   1.3   0.0   0.0   0.0   0.0   0.0   0.0

 
7.02% as type_12  SIG  54.8  57.9  57.0  54.8  57.2  60.1  60.1  62.7  64.3  62.4  65.9  64.1  65.3  59.5  73.0  35.4   0.0
BGSL  58                   STD   2.7   3.0   2.9   2.7   2.8   2.6   2.4   2.4   2.7   2.6   2.8   2.7   3.1   3.3   3.3   6.3   0.0


  9.21% as type_13  SIG  65.9  69.8  74.5  76.9  78.4  76.2  76.7  78.9  81.0  78.5  71.1  70.9  94.0   0.0   0.0   0.0   0.0
BGSL  77                   STD   3.1   3.4   3.8   4.1   3.7   4.9   4.6   3.8   4.4   4.2   3.7   5.0   1.9   0.0   0.0   0.0   0.0


  7.68% as type_14  SIG  88.2 103.0  96.5 107.5 105.2 100.6 104.9 101.9 107.0 103.9  93.3  88.8 108.4   0.0   0.0   0.0   0.0
BGSL 105                  STD   3.7   4.6   4.6   4.9   4.2   4.4   5.5   4.4   5.1   4.8   3.4   2.8   5.3   0.0   0.0   0.0   0.0


  3.07% as type_15  SIG 114.8 124.8 121.1 132.2 131.2 125.7 130.7 137.6 146.5 146.3 112.1   0.0   0.0   0.0   0.0   0.0   0.0
BGSL 130                 STD   6.4   5.9   5.6   5.3   5.7   5.0   5.4   6.0   6.8   6.4   5.0   0.0   0.0   0.0   0.0   0.0   0.0

 
Band setting: EXCELLENT
 
   1 [16U] 434.5nm+/- 10.3nm (rows 278-288) DNSRU:1000.000000 a b        First time I see these near blue bands
   2 [16U] 452.0nm+/-  7.6nm (rows 270-277) DNSRU:1000.000000 a b           having a great radiometric quality
   3 [16U] 467.7nm+/-  6.6nm (rows 262-268) DNSRU:1000.000000 a b
   4 [16U] 481.6nm+/-  7.6nm (rows 254-261) DNSRU:1000.000000 a b
   5 [16U] 497.3nm+/-  8.5nm (rows 245-253) DNSRU:1000.000000 a b
   6 [16U] 513.2nm+/-  7.6nm (rows 237-244) DNSRU:1000.000000 a b
   7 [16U] 528.1nm+/-  7.6nm (rows 229-236) DNSRU:1000.000000 a b
   8 [16U] 543.0nm+/-  7.6nm (rows 221-228) DNSRU:1000.000000 a b
   9 [16U] 558.0nm+/-  7.7nm (rows 213-220) DNSRU:1000.000000 a b
  10 [16U] 573.0nm+/-  7.7nm (rows 205-212) DNSRU:1000.000000 a b         One or two more bands
  11 [16U] 588.0nm+/-  7.7nm (rows 197-204) DNSRU:1000.000000 a b        inside this range 
  12 [16U] 603.0nm+/-  7.7nm (rows 189-196) DNSRU:1000.000000 a b        
should be an adavantage
  13 [16U] 634.1nm+/-  6.8nm (rows 173-179) DNSRU:1000.000000 a b
  14 [16U] 654.9nm+/-  4.9nm (rows 163-167) DNSRU:1000.000000 a b
  15 [16U] 700.4nm+/-  4.9nm (rows 139-143) DNSRU:1000.000000 a b
  16 [16U] 740.3nm+/-  6.8nm (rows 117-123) DNSRU:1000.000000 a b        TWO near-infrared in this range 
  17 [16U] 849.8nm+/-  9.7nm (rows 058-067) DNSRU:1000.000000 a b        are a great help for deglinting