In-water Algorithm (Neural Network) Motoaki Kishino ( RIKEN ) 8 - - PowerPoint PPT Presentation

▶

Sep 21, 2022 163 likes •317 views

In-water Algorithm (Neural Network) Motoaki Kishino ( RIKEN ) 8 November 2000 Kanazawa PI and Co-i Roland Doerffer Motoaki Kishino Helmut Schiller Tomohiko Oishi Heinz van de Piepen Hiroh Satoh Carsten Brockmann

SLIDE 1

In-water Algorithm (Neural Network)

Motoaki Kishino

(RIKEN)

8 November 2000 Kanazawa

SLIDE 2

PI and Co-i

Motoaki Kishino Tomohiko Oishi Hiroh Satoh Kouji Suzuki Tooru Hirawake Roland Doerffer Helmut Schiller Heinz van de Piepen Carsten Brockmann

J. H. M. Hakvoort

Herbert Siegel

SLIDE 3

NN-Training

SLIDE 4

Joseph (1950)

Radiative Transfer Model

rs s w n

R E L × × × × = = = =

) 2 ( 533 .

bt t t t t rs

b a a k a k a k R Q R R + + + + = = = = + + + + − − − − = = = = × × × × = = = =

SLIDE 5

Model of Optical Properties Absorption Coefficient

y c w t

a a a a + + + + + + + + = = = =

(1989) al. et Roesler Kishino) Japan, Sanriku, (off truth sea the

values average : (1997) Fry and Pope :

) 440 ( 014 . * * − − − − λ λ λ λ − − − −

× × × × = = = = × × × × = = = = e CDOM a a C a a a

y c c c w

SLIDE 6

Model of Optical Properties Backscattering Coefficient

32 . 4

500 * 00144 .

− − − −

                        λ λ λ λ = = = =

w b

b

Morel (1974) Gordon and Morel (1983) Morel (1988) Kronfeld (1988)

                       

λ λ λ λ − − − − + + + + = = = = 550 * )) log * 25 . 5 . ( * 02 . 002 . ( * * 30 .

10 62 .

C C bc b

812 .

550 * * 001848 .

− − − −

                        λ λ λ λ = = = = S bbs

s b c b w b t b

b b b b + + + + + + + + = = = =

SLIDE 7

Parameters and their range for the Neural Network training

Parameter Unit Range Chlorophyll a mg m-3 0.1 - 50 Inorganic Suspension g m-3 0.1 - 50 CDOM m-1 0.01 - 10 Data Set for neural network training 100,000 Data set for validation 50,000

SLIDE 8

Neural Network example for OCTS data

nLw(412) nLw(443) nLw(565) nLw(520) nLw(490) chlorophyll a suspended matter yellow substance input layer

utput layer

hidden layer : unit : link

SLIDE 9

SLIDE 10

SLIDE 11

SLIDE 12

SLIDE 13

SLIDE 14

SLIDE 15

In-water Algorithm (Neural Network) Motoaki Kishino ( RIKEN ) 8 - - PowerPoint PPT Presentation

In-water Algorithm (Neural Network)

Motoaki Kishino

(RIKEN)

PI and Co-i

NN-Training

Joseph (1950)

Radiative Transfer Model

rs s w n

R E L × × × × = = = =

) 2 ( 533 .

bt t t t t rs

b a a k a k a k R Q R R + + + + = = = = + + + + − − − − = = = = × × × × = = = =

Model of Optical Properties Absorption Coefficient

y c w t

a a a a + + + + + + + + = = = =

(1989) al. et Roesler Kishino) Japan, Sanriku, (off truth sea the

values average : (1997) Fry and Pope :

× × × × = = = = × × × × = = = = e CDOM a a C a a a

Model of Optical Properties Backscattering Coefficient

500 * 00144 .

                        λ λ λ λ = = = =

b

Morel (1974) Gordon and Morel (1983) Morel (1988) Kronfeld (1988)

                       

λ λ λ λ − − − − + + + + = = = = 550 * )) log * 25 . 5 . ( * 02 . 002 . ( * * 30 .

C C bc b

550 * * 001848 .

                        λ λ λ λ = = = = S bbs

s b c b w b t b

b b b b + + + + + + + + = = = =

Parameters and their range for the Neural Network training

Parameter Unit Range Chlorophyll a mg m-3 0.1 - 50 Inorganic Suspension g m-3 0.1 - 50 CDOM m-1 0.01 - 10 Data Set for neural network training 100,000 Data set for validation 50,000

Neural Network example for OCTS data

R M S

All Bands 11 Bands

Chl a IOS CDOM 0%: 0.00178 0.00189 0.000168 10%: 0.0302 0.0569 0.00312 20%: 0.0520 0.141 0.00601

Except Saturation Bands 9 Bands

Chl a IOS CDOM 0%: 0.00473 0.00221 0.000212 10%: 0.0218 0.135 0.00828 20%: 0.0577 0.201 0.0207