2018
DOI: 10.1109/jstars.2018.2871046
|View full text |Cite
|
Sign up to set email alerts
|

Automatic Extraction of Built-Up Areas From Panchromatic and Multispectral Remote Sensing Images Using Double-Stream Deep Convolutional Neural Networks

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
3
1
1

Citation Types

0
25
0

Year Published

2019
2019
2023
2023

Publication Types

Select...
8
1

Relationship

1
8

Authors

Journals

citations
Cited by 45 publications
(25 citation statements)
references
References 38 publications
0
25
0
Order By: Relevance
“…As the number of pooling layers increases, the spatial resolution of the image decreases. The shape characteristics of buildings also change with a change in the image resolution [54]. To adapt to the appearance of morphological building features at different resolutions, a group of adaptive dilated convolutions is constructed to better map these features.…”
Section: ) Residual Connected Unitmentioning
confidence: 99%
“…As the number of pooling layers increases, the spatial resolution of the image decreases. The shape characteristics of buildings also change with a change in the image resolution [54]. To adapt to the appearance of morphological building features at different resolutions, a group of adaptive dilated convolutions is constructed to better map these features.…”
Section: ) Residual Connected Unitmentioning
confidence: 99%
“…In our previous work [90], a double-stream convolutional neural network (DSCNN) model based on Inception V3 was proposed to extract the built-up area automatically. The network consists of two branches: the upper branches manipulate the panchromatic image providing primary information for object classification, and the lower branch attaches more importance to the multispectral image, providing auxiliary information for improvement in accuracy (with the overall accuracy reaching 98.39%).…”
Section: Taking the Scene As The Primary Unit To Interpret Objects Frmentioning
confidence: 99%
“…Compared with the popular region proposal in computer vision fields, OBIA could handle various shapes and sizes of the real land covers. The multibranch parallel network models based on GoogLeNet [34] and skip-layer architectures based on ResNet [35] are investigated [52], [53].…”
Section: Introductionmentioning
confidence: 99%