Optical calibration, bathymetry, water column correction and bottom typing of shallow marine areas, using passive remote sensing imageries
water column correction since 1994
4SM : an operational suite of tools 
written in C under Ubuntu Linux (60,000 lines)
  • no need for field data
  • no need for atmospheric correction
  • works with DNs
  • "same day" service
  • from raw spectral data to final deliverable products 
  • through one single commandline
The simplified RTE https://www.watercolumncorrection.com/sld016.php
JERLOV wrote https://www.watercolumncorrection.com/sld036.php
KIRK       wrote https://www.watercolumncorrection.com/sld037.php
A family of curves https://www.watercolumncorrection.com/4sm-illustrations-calibration-process.php#porquerolles
The SL and BPL https://www.watercolumncorrection.com/4sm-illustrations-calibration-process.php
Calibration diagram https://www.watercolumncorrection.com/4sm-illustrations-calibration-process.php#yes
Deglinting https://www.watercolumncorrection.com/4sm-illustrations-the-deglinting-process.php#lsi
Modeling https://www.watercolumncorrection.com/waimanalo-pollution.php#profiles
Seatruthing https://www.watercolumncorrection.com/waimanalo-pollution.php#regression_west
RMSE https://www.watercolumncorrection.com/waimanalo-pollution.php#Lmyellow=4
Bottom typing https://www.watercolumncorrection.com/waimanalo-bottom-typing.php#July_25th
The 4SM command line  https://www.watercolumncorrection.com/waimanalo-pollution.php#commandline


slide 1

  4SM: a ratio method 
for water column correction of shallow spectral images

Following work by Lyzenga, Philpot, Maritorena and Jerlov
 Simplified TOA RTE : Ls – Lsw =          (LsB – Lsw) / exp (2K * Z)
 Inverted    BOA RTE : LB           = Lw + (Ls    – Lw  ) *exp (2K * Z)
where Lsw=La+Lw, LsB=LB+La, Ls=L+La
No need for atmospheric correction: works with Dns
From raw data to final deliverables “on the fly” in one single commandline
Aims at “Same day” service 

slide 2
Deglinting in 4SM
Hedley's method  
  • Get the linear regressions of Lsi vs LsNIR
    • for all pairs of bands i and NIR over an optically deep area
  • The amount of glint at the current pixel is LsNIR-LswNIR
  • Use the regressions to estimate the amount of glint in other bands,
For WV2 images

slide 3

WV2 at Waimanalo Bay, Oahu, Hawaii

LIDAR seatruth
Red tones for over-estimated depths
Blue tones for under-estimated depths
Dark_blue tones delineate artifacts caused by ”pollution”                   
Profile_black                                                                                          Enhanced TCC
of water column corrected image

                     Deep purple tones signal "pollution"

slide 4

WV2: Profile_black at Waimanalo Bay, Oahu, Hawaii

slide 5


slide 6
WV2 at waimanalo Bay 
Reducing the error on estimated depth

LIDAR DTM is in plain meters: I need a DTM in centimeters

Away from suspected pollution
from row_1 to row_1000

from 0 to 10 m along profile_black

96.2% of all estimated depths  
are within +-2 m of DTM

RMSE=0.84-0.50=0.34 m

average ZComputed-ZLidar = 0.61-0.50=0.11 m

This image is a "worst case"
fully glinted image
Sensor is looking 37° sideways
right into the sun

Thanks you, Ron!
 September 1rst 2012 


Créer un site
Créer un site