Study on disaster-affected patches extraction based on the segment anything model

Chunxiu He; Xianwen Jing

doi:10.1117/12.3021032

19 January 2024 Study on disaster-affected patches extraction based on the segment anything model

Chunxiu He, Xianwen Jing

Proceedings Volume 12980, Fifth International Conference on Geoscience and Remote Sensing Mapping (ICGRSM 2023); 129800K (2024) https://doi.org/10.1117/12.3021032
Event: Fifth International Conference on Geoscience and Remote Sensing Mapping (ICGRSM 2023), 2023, Lianyungang, China

Abstract

Rapid and accurate extraction of disaster-affected patches is of great importance. This study focuses on the extraction of flooded road patches and proposes a method based on the Segment Anything Model, which utilizes the semantic segmentation results obtained from the BiSeNet V2 model as prompt cues for the model. To better integrate the two models, a Prompt Conversion Module is designed to convert the semantic segmentation results into prompt cues, and the SAM model is fine-tuned accordingly. Additionally, a Semantic Fusion Module is introduced to incorporate semantic information from the segmentation results and BiSeNet V2. To validate the effectiveness of the proposed approach, experiments are conducted on the publicly available FloodNet dataset. The results demonstrate that the proposed method not only achieves faster and more accurate extraction of flood-affected patches but also provides corresponding semantic classification information for the extracted patches.

(2024) Published by SPIE. Downloading of the abstract is permitted for personal use only.

Citation Download Citation

Chunxiu He and Xianwen Jing "Study on disaster-affected patches extraction based on the segment anything model", Proc. SPIE 12980, Fifth International Conference on Geoscience and Remote Sensing Mapping (ICGRSM 2023), 129800K (19 January 2024); https://doi.org/10.1117/12.3021032

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