SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images
Boundary pixel blur and category imbalance are common problems that occur during semantic segmentation of urban remote sensing images. Inspired by DenseU-Net, this paper proposes a new end-to-end network—SiameseDenseU-Net. First, the network simultaneously uses both true orthophoto (TOP) images and...
Main Authors: | , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Hindawi Limited
2020-01-01
|
Series: | Mathematical Problems in Engineering |
Online Access: | http://dx.doi.org/10.1155/2020/1515630 |
id |
doaj-9f1a2fb466e041f897d45ef5a32b30ec |
---|---|
record_format |
Article |
spelling |
doaj-9f1a2fb466e041f897d45ef5a32b30ec2020-11-25T02:30:00ZengHindawi LimitedMathematical Problems in Engineering1024-123X1563-51472020-01-01202010.1155/2020/15156301515630SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing ImagesRongsheng Dong0Lulu Bai1Fengying Li2Guangxi Key Laboratory of Trusted Software, School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin 541004, ChinaGuangxi Key Laboratory of Trusted Software, School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin 541004, ChinaGuangxi Key Laboratory of Trusted Software, School of Computer Science and Information Security, Guilin University of Electronic Technology, Guilin 541004, ChinaBoundary pixel blur and category imbalance are common problems that occur during semantic segmentation of urban remote sensing images. Inspired by DenseU-Net, this paper proposes a new end-to-end network—SiameseDenseU-Net. First, the network simultaneously uses both true orthophoto (TOP) images and their corresponding normalized digital surface model (nDSM) as the input of the network structure. The deep image features are extracted in parallel by downsampling blocks. Information such as shallow textures and high-level abstract semantic features are fused throughout the connected channels. The features extracted by the two parallel processing chains are then fused. Finally, a softmax layer is used to perform prediction to generate dense label maps. Experiments on the Vaihingen dataset show that SiameseDenseU-Net improves the F1-score by 8.2% and 7.63% compared with the Hourglass-ShapeNetwork (HSN) model and with the U-Net model. Regarding the boundary pixels, when using the same focus loss function based on median frequency balance weighting, compared with the original DenseU-Net, the small-target “car” category F1-score of SiameseDenseU-Net improved by 0.92%. The overall accuracy and the average F1-score also improved to varying degrees. The proposed SiameseDenseU-Net is better at identifying small-target categories and boundary pixels, and it is numerically and visually superior to the contrast model.http://dx.doi.org/10.1155/2020/1515630 |
collection |
DOAJ |
language |
English |
format |
Article |
sources |
DOAJ |
author |
Rongsheng Dong Lulu Bai Fengying Li |
spellingShingle |
Rongsheng Dong Lulu Bai Fengying Li SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images Mathematical Problems in Engineering |
author_facet |
Rongsheng Dong Lulu Bai Fengying Li |
author_sort |
Rongsheng Dong |
title |
SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images |
title_short |
SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images |
title_full |
SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images |
title_fullStr |
SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images |
title_full_unstemmed |
SiameseDenseU-Net-based Semantic Segmentation of Urban Remote Sensing Images |
title_sort |
siamesedenseu-net-based semantic segmentation of urban remote sensing images |
publisher |
Hindawi Limited |
series |
Mathematical Problems in Engineering |
issn |
1024-123X 1563-5147 |
publishDate |
2020-01-01 |
description |
Boundary pixel blur and category imbalance are common problems that occur during semantic segmentation of urban remote sensing images. Inspired by DenseU-Net, this paper proposes a new end-to-end network—SiameseDenseU-Net. First, the network simultaneously uses both true orthophoto (TOP) images and their corresponding normalized digital surface model (nDSM) as the input of the network structure. The deep image features are extracted in parallel by downsampling blocks. Information such as shallow textures and high-level abstract semantic features are fused throughout the connected channels. The features extracted by the two parallel processing chains are then fused. Finally, a softmax layer is used to perform prediction to generate dense label maps. Experiments on the Vaihingen dataset show that SiameseDenseU-Net improves the F1-score by 8.2% and 7.63% compared with the Hourglass-ShapeNetwork (HSN) model and with the U-Net model. Regarding the boundary pixels, when using the same focus loss function based on median frequency balance weighting, compared with the original DenseU-Net, the small-target “car” category F1-score of SiameseDenseU-Net improved by 0.92%. The overall accuracy and the average F1-score also improved to varying degrees. The proposed SiameseDenseU-Net is better at identifying small-target categories and boundary pixels, and it is numerically and visually superior to the contrast model. |
url |
http://dx.doi.org/10.1155/2020/1515630 |
work_keys_str_mv |
AT rongshengdong siamesedenseunetbasedsemanticsegmentationofurbanremotesensingimages AT lulubai siamesedenseunetbasedsemanticsegmentationofurbanremotesensingimages AT fengyingli siamesedenseunetbasedsemanticsegmentationofurbanremotesensingimages |
_version_ |
1715468864378109952 |