• 検索結果がありません。

Segmentation of sea ice and open water from ship radar images using U-net Toshiyuki Takagi

N/A
N/A
Info
ダウンロード
Protected

Academic year: 2021

シェア "Segmentation of sea ice and open water from ship radar images using U-net Toshiyuki Takagi"

Copied!
1
0
0

読み込み中.... (全文を見る)

今ダウンロードする ( 1 ページ )

全文

(1)

Segmentation of sea ice and open water from ship radar images using U-net

Toshiyuki Takagi

1

and Kazutaka Tateyama

2

1

National Institute of Technology, Kushiro College

2

Kitami Institute of Technology

With global warning, there has been a growing economic interest in shipping in the Arctic. Espencially the Northern Sea Route (NSR) has opened up as a possible avenue of trade in containerized products between Asia and Europe. The sea ice has caused significant damage to ships. It is important to accurately obtain the position and size of sea ice from ship radar images.

In order to search a route from the ship radar image, it is important to divide the sea ice and the open water. However, it is difficult to obtain the sea route in the radar images of a ship because there are various noise and distortion such as sea clatter and radio wave attenuation with respect to distance.

Figure.1(a) is a ship radar image that provided at lat. 79 degrees 57,236 N and long. 142 degrees 31,540 W by Canadian Coast Guard Ship (CCGS) Louise S. St-Laurent in August, 2012. In the center of the radar, the brightness is higher due to sea clutter. And the brightness decreases as going away from the center of the radar. It is not possible to easily separate sea and sea ice from the ship radar image.

A U-net which is one of deep learnig, is useful technque for complex image segmentaion. The network is based on the fully convolutional network. The radar image in Figure.1(a) is input to the U-net.In our method, in order to create desired data corresponding to input data, multiple temporally continuous images are superimposed to supplement missing sea ice due to noise in Figure.1(b). Figure.1(c) shows the esitmated image by the U-net. As shown in this figure, sea and sea ice are separated from ship radar image.

( a) input image (b) desired image (c) resulting segmentation from the U-net

Figure 1. Simulation results.

ACKNOWLEDGEMENTS

This work was supported by JSPS KAKENHI Grant Number JP 16K01299.

References

T.Takagi, K.Tateyama and T.Ishiyama, Obstacle Avoidance and Path Planning in Ice Sea using Probabilistic Roadmap Method, the 22nd IAHR international Symposium on Ice, Singapore,2014

M.Choi, H.Yamaguchi and K.Nakagawa, Arctic sea route path planning based on an uncertain ice prediction model, Cold Regions Science and Technology 10961-69, 2015

Olaf Ronneberger, Philipp Fischer, and Thomas Brox,U-Net: Convolutional Networks for Biomedical Image Segmentation

Medical Image Computing and Computer-Assisted Intervention (MICCAI), Springer, LNCS, Vol.9351: 234--241, 2015

Lei Wang,K. Andrea Scott,Linlin Xu,David A. Clausi, Sea Ice Concentration Estimation During Melt From Dual-Pol SAR

Scenes Using Deep Convolutional Neural Networks: A Case Study, IEEE Transactions on Geoscience and Remote

Sensing ,Volume: 54, Issue: 8, Aug. 2016

参照

今ダウンロードする ( PDF - 1 ページ - 102.25 KB )

関連したドキュメント

Preface to the TAC Reprint

The notion of free product with amalgamation of groupoids in [16] strongly influenced Ronnie Brown to introduce in [5] the fundamental groupoid on a set of base points, and so to give

It is a variant of the construction used in [8] and [9]

Thus, in Section 5, we show in Theorem 5.1 that, in case of even dimension d > 2 of a quadric the bundle of endomorphisms of each indecomposable component of the Swan bundle

Such distinctions are however important in the produc- tion of limit cycles close to the foci in perturbations of the systems

We describe a little the blow–ups of the phase portrait of the intricate point p given in Figure 5. Its first blow–up is given in Figure 6A. In it we see from the upper part of

On the nonlinear lower order termg(x, u,∇u), we assume that it is a Carathéodory function having natural growth with respect to|∇u|, and satisfies the sign condition

ELMAHI, An existence theorem for a strongly nonlinear elliptic prob- lems in Orlicz spaces, Nonlinear Anal.. ELMAHI, A strongly nonlinear elliptic equation having natural growth

Preface to the 19th Takagi Lectures

The Mathematical Society of Japan (MSJ) inaugurated the Takagi Lectures as prestigious research survey lectures.. The Takagi Lectures are the first se- ries of the MSJ official

Preface to the 24th Takagi Lectures

The Mathematical Society of Japan (MSJ) inaugurated the Takagi Lectures as prestigious research survey lectures.. The Takagi Lectures are the first series of the MSJ official

to the Geometric Satake Equivalence

I give a proof of the theorem over any separably closed field F using ℓ-adic perverse sheaves.. My proof is different from the one of Mirkovi´c

The exact asymptotic of the time to collision

Keywords: continuous time random walk, Brownian motion, collision time, skew Young tableaux, tandem queue.. AMS 2000 Subject Classification: Primary: