Forest-CD: Forest Change Detection Network Based on VHR Images

Detection of forest changes is an important part of the restoration of forest land resources. Timely detection of forest changes is helpful for standardized management of government departments. However, due to the interference of seasonal factors, there are a large number of pseudo changes in the f...

Full description

Saved in:
Bibliographic Details
Published in:IEEE geoscience and remote sensing letters 2022, Vol.19, p.1-5
Main Authors: Jiang, Jiawei, Xiang, Jun, Yan, Enping, Song, Yabin, Mo, Dengkui
Format: Article
Language:English
Subjects:
Boundaries
Change detection
Coders
Decoding
Deep learning
Detection
Feature extraction
forest change detection
Forest resources
Forestry
Forests
Land resources
Landsat satellites
Plant growth
Remote sensing
remote sensing (RS) images
Restoration
Semantics
Task analysis
Transformers
vision transformer
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Detection of forest changes is an important part of the restoration of forest land resources. Timely detection of forest changes is helpful for standardized management of government departments. However, due to the interference of seasonal factors, there are a large number of pseudo changes in the forest change detection results. Current methods of change detection focus on extracting more accurate boundaries of change. In order to suppress pseudo changes in the results of forest change detection, we constructed a model Forest-CD based on the encoder-decoder structure with the help of background information. The Forest-CD encoder uses the Swin transformer as a backbone for extracting the change features and effectively simulating global information. The decoder uses the feature pyramid network to recover feature scale and fuse information at multiple scales. Experiments on a large forest change dataset show that Forest-CD has a higher F1 -Score (0.614) compared to other change detection models. The visualization results show that Forest-CD can well suppress pseudo changes caused by seasonal factors in forestry scenes and can more accurately capture the boundaries of change.
ISSN:1545-598X
1558-0571
DOI:10.1109/LGRS.2022.3188636