Poster COAST Bordeaux SPAGNOLI MINGHELLI 2 sc(1)(1) .pdf

Automatic method to extract the Shoreline
using WorldView-2 images, containing foam
J. Spagnoli1, A. Minghelli1, S. Charmasson2
1LSIS,

University of Toulon, La Garde, France
2IRSN, PSE-SRTE/LRTA, La Seyne sur Mer, France

Objective

Results of classifications

• Study the impact of the tsunami of Fukushima (Japan, 2011) on the shoreline
• Find an automatic method to extract the shoreline from satellite images in
presence of foam
• Estimate the evolution of both erosion and accretion areas between 2 dates
: before the tsunami (8 November 2010) and after the tsunami (18 March
2011)

2010
Water
Land
Foam

Multispectral data
• WorldView-2
images

2011

• 2m resolution
2010

2011

Results of classifications and the corresponding land/water maps

Comparisons (false-positive, false-negative)
2011

2010

ED

ED

MLE

SAM

closing

opening

No operations

closing

No operations

closing

opening

No operations

closing

closing

opening
SAM

MLE

MLE gives the best results concerning EP and EN for each date

PIR

Erosion and accretion

Methodology
Multispectral image

Erosion

Classification

Fusion water/foam classes

Morphologic
operation
Segmentation

No operations

EN

opening

EP

No operations

Foam leads to
errors after
thresholding

opening

160 000
140 000
120 000
100 000
80 000
60 000
40 000
20 000
0

160 000
140 000
120 000
100 000
80 000
60 000
40 000
20 000
0

closing

Problem statement

opening

• Color composite :
B : band 1 (425 nm)
G : band 3 (545 nm)
R : band 6 (725 nm)

No operations

• 8 spectral bands
(400 – 1040 nm )

(fermeture ou ouverture)

100%

90%

90%

80%

80%

70%

70%

60%

60%

50%

50%

40%

40%

30%

30%

20%

20%

10%

10%

0%

0%

Supervised-classifications distances

Spectral Angle Mapper
(SAM)

Maximum Likelihood
Estimation (MLE)

SAM

MLE

closing

ED

Accretion

SAM

Concerning erosion area, MLE gives the best results
Concerning accretion area, all methods give equivalent results

Land/water map

i : the number of the spectral band
k : the number of the class
Rk : the covariance matrix of class k
x : the spectral profile of the pixel
𝑟𝑘ҧ : the mean spectral profile of class k

Opening

100%

ED

Euclidean Distance (ED)

No operations

𝑁

𝑑 𝑥, 𝑟𝑘 =

෍ 𝑥(i) − 𝑟ഥ𝑘 (i)

2

i=1

𝛼(𝑥, 𝑟𝑘 ) = cos −1

Conclusion
•
•
•
•

Morphologic operations do not enhance the results
Maximum likelihood gives the best results to extract the shoreline
Erosion area : 410 604 m²
Accretion area : 174 804 m²

σ𝑁
𝑖=1 𝑥(i) ∗ 𝑟𝑘ҧ (i)
σ𝑁
𝑖=1 𝑥(𝑖)² ∗

𝑔 𝑥, 𝑟𝑘 = − ln 𝑅𝑘

MLE

− 𝑥 − 𝑟𝑘ҧ

𝑇

σ𝑁
𝑖=1 𝑟𝑘ҧ (i)²

∗ 𝑅𝑘−1 ∗ 𝑥 − 𝑟𝑘ҧ

Reference
• Richards, John A, and Xiuping Jia. 2006. Remote Sensing Digital Image Analysis-Hardback. Springer, Berlin/Heidelberg.
• Kruse, FA, AB Lefkoff, JW Boardman, KB Heidebrecht, AT Shapiro, PJ Barloon, and AFH Goetz. 1993. “The Spectral
Image Processing System (SIPS)—interactive Visualization and Analysis of Imaging Spectrometer Data.” Remote
Sensing of Environment 44 (2–3): 145–163.