SAR ship detection based on resnet and transfer learning

Yong Li; Zegang Ding; Chi Zhang; Yan Wang; Jing Chen

doi:10.1109/IGARSS.2019.8900290

SAR ship detection based on resnet and transfer learning

Yong Li, Zegang Ding^*, Chi Zhang, Yan Wang, Jing Chen

^*Corresponding author for this work

School of Information and Electronics

Research output: Contribution to conference › Paper › peer-review

46 Citations (Scopus)

Abstract

Synthetic Aperture Radar (SAR) ship detection has been a research hotspot and is significant for marine surveillance. Traditional constant false alarm rate (CFAR) detector has the disadvantages of high false alarm and poor adaptability. Deep learning provides a unique solution for SAR ship detection. However, the traditional deep network cannot reach very deep thus the accuracy is limited, and the training speed is slow. In this paper, a very deep network ResNet with higher accuracy and faster training speed is applied to train the SAR ship detection model. Moreover, transfer learning is applied to combat the small dataset. The proposed method is tested on a general SAR ship dataset and achieves 94.7% average precision. Comparative experiments show that our method has the best performance and which verifies the effectiveness of our method.

Original language	English
Pages	1188-1191
Number of pages	4
DOIs	https://doi.org/10.1109/IGARSS.2019.8900290
Publication status	Published - 2019
Event	39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019 - Yokohama, Japan Duration: 28 Jul 2019 → 2 Aug 2019

Conference

Conference	39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019
Country/Territory	Japan
City	Yokohama
Period	28/07/19 → 2/08/19

Keywords

Deep learning
ResNet
SAR
Ship detection
Transfer learning

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1109/IGARSS.2019.8900290

Cite this

@conference{e50f3c9e01504c44a65c025835ec7f58,

title = "SAR ship detection based on resnet and transfer learning",

abstract = "Synthetic Aperture Radar (SAR) ship detection has been a research hotspot and is significant for marine surveillance. Traditional constant false alarm rate (CFAR) detector has the disadvantages of high false alarm and poor adaptability. Deep learning provides a unique solution for SAR ship detection. However, the traditional deep network cannot reach very deep thus the accuracy is limited, and the training speed is slow. In this paper, a very deep network ResNet with higher accuracy and faster training speed is applied to train the SAR ship detection model. Moreover, transfer learning is applied to combat the small dataset. The proposed method is tested on a general SAR ship dataset and achieves 94.7% average precision. Comparative experiments show that our method has the best performance and which verifies the effectiveness of our method.",

keywords = "Deep learning, ResNet, SAR, Ship detection, Transfer learning",

author = "Yong Li and Zegang Ding and Chi Zhang and Yan Wang and Jing Chen",

note = "Publisher Copyright: {\textcopyright} 2019 IEEE; 39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019 ; Conference date: 28-07-2019 Through 02-08-2019",

year = "2019",

doi = "10.1109/IGARSS.2019.8900290",

language = "English",

pages = "1188--1191",

}

TY - CONF

T1 - SAR ship detection based on resnet and transfer learning

AU - Li, Yong

AU - Ding, Zegang

AU - Zhang, Chi

AU - Wang, Yan

AU - Chen, Jing

PY - 2019

Y1 - 2019

N2 - Synthetic Aperture Radar (SAR) ship detection has been a research hotspot and is significant for marine surveillance. Traditional constant false alarm rate (CFAR) detector has the disadvantages of high false alarm and poor adaptability. Deep learning provides a unique solution for SAR ship detection. However, the traditional deep network cannot reach very deep thus the accuracy is limited, and the training speed is slow. In this paper, a very deep network ResNet with higher accuracy and faster training speed is applied to train the SAR ship detection model. Moreover, transfer learning is applied to combat the small dataset. The proposed method is tested on a general SAR ship dataset and achieves 94.7% average precision. Comparative experiments show that our method has the best performance and which verifies the effectiveness of our method.

AB - Synthetic Aperture Radar (SAR) ship detection has been a research hotspot and is significant for marine surveillance. Traditional constant false alarm rate (CFAR) detector has the disadvantages of high false alarm and poor adaptability. Deep learning provides a unique solution for SAR ship detection. However, the traditional deep network cannot reach very deep thus the accuracy is limited, and the training speed is slow. In this paper, a very deep network ResNet with higher accuracy and faster training speed is applied to train the SAR ship detection model. Moreover, transfer learning is applied to combat the small dataset. The proposed method is tested on a general SAR ship dataset and achieves 94.7% average precision. Comparative experiments show that our method has the best performance and which verifies the effectiveness of our method.

KW - Deep learning

KW - ResNet

KW - SAR

KW - Ship detection

KW - Transfer learning

UR - http://www.scopus.com/inward/record.url?scp=85084117320&partnerID=8YFLogxK

U2 - 10.1109/IGARSS.2019.8900290

DO - 10.1109/IGARSS.2019.8900290

M3 - Paper

AN - SCOPUS:85084117320

SP - 1188

EP - 1191

T2 - 39th IEEE International Geoscience and Remote Sensing Symposium, IGARSS 2019

Y2 - 28 July 2019 through 2 August 2019

ER -

SAR ship detection based on resnet and transfer learning

Abstract

Conference

Keywords

UN SDGs

Access to Document

Other files and links

Fingerprint

Cite this