A remote sensing image segmentation model based on multi-scale feature fusion

Ao Liu; Jue Lu; Qiang Ma; Yi Han; Yi Zhong

doi:10.1117/12.3007424

30 August 2023 A remote sensing image segmentation model based on multi-scale feature fusion

Ao Liu, Jue Lu, Qiang Ma, Yi Han, Yi Zhong

Proceedings Volume 12797, Second International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023); 127971F (2023) https://doi.org/10.1117/12.3007424
Event: 2nd International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023), 2023, Qingdao, China

Abstract

Large-scale remote sensing images contain rich terrain information, while small-scale images capture abundant detail information. Effectively extracting and fusing these multi-scale information has grown into a major challenge. To tackle these problems, this paper recommends a novel remote sensing image segmentation model called Multi-scale Vision Transformer (MS-ViT). The MS-ViT model comprises a multi-scale feature extraction module, a feature fusion module based on multi-scale self-attention, and a decoder based on convolutional neural networks. Additionally, the model introduces spatial attention mechanism and a position encoding method considering patch sizes, as well as a feature distribution alignment method and corresponding loss functions to further enhance segmentation performance. Experimental results demonstrate significant improvements in performance of the MS-ViT model compared to several other advanced models.

(2023) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Ao Liu, Jue Lu, Qiang Ma, Yi Han, and Yi Zhong "A remote sensing image segmentation model based on multi-scale feature fusion", Proc. SPIE 12797, Second International Conference on Geographic Information and Remote Sensing Technology (GIRST 2023), 127971F (30 August 2023); https://doi.org/10.1117/12.3007424

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available