Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss

Instance segmentation methods for synthetic aperture radar (SAR) ship imaging have certain unsolved problems. 1) Most of the anchor-based detection algorithms encounter difficulties in tuning the anchor-related parameters and high computational costs. 2) Different tasks share the same features witho...

Descripción completa

Guardado en:

Detalles Bibliográficos
Autores principales:	Fei Gao, Yiyang Huo, Jun Wang, Amir Hussain, Huiyu Zhou
Formato:	article
Lenguaje:	EN
Publicado:	IEEE 2021
Materias:	Anchor-free convolutional neural network (CNN) instance segmentation synthetic aperture radar (SAR) Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809
Acceso en línea:	https://doaj.org/article/bb6adce2c0df4aadb275cd581e0f8423
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

id	oai:doaj.org-article:bb6adce2c0df4aadb275cd581e0f8423
record_format	dspace
spelling	oai:doaj.org-article:bb6adce2c0df4aadb275cd581e0f84232021-11-19T00:00:12ZAnchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss2151-153510.1109/JSTARS.2021.3123784https://doaj.org/article/bb6adce2c0df4aadb275cd581e0f84232021-01-01T00:00:00Zhttps://ieeexplore.ieee.org/document/9594445/https://doaj.org/toc/2151-1535Instance segmentation methods for synthetic aperture radar (SAR) ship imaging have certain unsolved problems. 1) Most of the anchor-based detection algorithms encounter difficulties in tuning the anchor-related parameters and high computational costs. 2) Different tasks share the same features without considering the differences between tasks, leading to mismatching of the shared features and inconsistent training targets. 3) Common loss functions for instance segmentation cannot effectively distinguish the positional relationships between ships with the same degree of overlap. In order to alleviate these problems, we first adopt a lightweight feature extractor and an anchor-free convolutional network, which effectively help to reduce computational consumption and model complexity. Second, to fully disseminate feature information, a dynamic encoder–decoder is proposed to dynamically transform the shared features to task-specific features in channel and spatial dimensions. Third, a novel loss function based on centroid distance is designed to make full use of the geometrical shape and positional relationship between SAR ship targets. In order to better extract features from SAR images in complex scenes, we further propose the dilated convolution enhancement module, which utilizes multiple receptive fields to take full advantage of the shallow feature information. Experiments conducted on the SAR ship detection dataset prove that the method proposed in this article is superior to the other state-of-the-art algorithms in terms of instance segmentation accuracy and model complexity.Fei GaoYiyang HuoJun WangAmir HussainHuiyu ZhouIEEEarticleAnchor-freeconvolutional neural network (CNN)instance segmentationsynthetic aperture radar (SAR)Ocean engineeringTC1501-1800Geophysics. Cosmic physicsQC801-809ENIEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Vol 14, Pp 11352-11371 (2021)
institution	DOAJ
collection	DOAJ
language	EN
topic	Anchor-free convolutional neural network (CNN) instance segmentation synthetic aperture radar (SAR) Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809
spellingShingle	Anchor-free convolutional neural network (CNN) instance segmentation synthetic aperture radar (SAR) Ocean engineering TC1501-1800 Geophysics. Cosmic physics QC801-809 Fei Gao Yiyang Huo Jun Wang Amir Hussain Huiyu Zhou Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
description	Instance segmentation methods for synthetic aperture radar (SAR) ship imaging have certain unsolved problems. 1) Most of the anchor-based detection algorithms encounter difficulties in tuning the anchor-related parameters and high computational costs. 2) Different tasks share the same features without considering the differences between tasks, leading to mismatching of the shared features and inconsistent training targets. 3) Common loss functions for instance segmentation cannot effectively distinguish the positional relationships between ships with the same degree of overlap. In order to alleviate these problems, we first adopt a lightweight feature extractor and an anchor-free convolutional network, which effectively help to reduce computational consumption and model complexity. Second, to fully disseminate feature information, a dynamic encoder–decoder is proposed to dynamically transform the shared features to task-specific features in channel and spatial dimensions. Third, a novel loss function based on centroid distance is designed to make full use of the geometrical shape and positional relationship between SAR ship targets. In order to better extract features from SAR images in complex scenes, we further propose the dilated convolution enhancement module, which utilizes multiple receptive fields to take full advantage of the shallow feature information. Experiments conducted on the SAR ship detection dataset prove that the method proposed in this article is superior to the other state-of-the-art algorithms in terms of instance segmentation accuracy and model complexity.
format	article
author	Fei Gao Yiyang Huo Jun Wang Amir Hussain Huiyu Zhou
author_facet	Fei Gao Yiyang Huo Jun Wang Amir Hussain Huiyu Zhou
author_sort	Fei Gao
title	Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
title_short	Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
title_full	Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
title_fullStr	Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
title_full_unstemmed	Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss
title_sort	anchor-free sar ship instance segmentation with centroid-distance based loss
publisher	IEEE
publishDate	2021
url	https://doaj.org/article/bb6adce2c0df4aadb275cd581e0f8423
work_keys_str_mv	AT feigao anchorfreesarshipinstancesegmentationwithcentroiddistancebasedloss AT yiyanghuo anchorfreesarshipinstancesegmentationwithcentroiddistancebasedloss AT junwang anchorfreesarshipinstancesegmentationwithcentroiddistancebasedloss AT amirhussain anchorfreesarshipinstancesegmentationwithcentroiddistancebasedloss AT huiyuzhou anchorfreesarshipinstancesegmentationwithcentroiddistancebasedloss
_version_	1718420689769201664

Anchor-Free SAR Ship Instance Segmentation With Centroid-Distance Based Loss

Ejemplares similares