Optimizing Multi-Class Change Segmentation in High-Resolution Satellite Imagery With a Siamese Network for Low-Resource Environments

This work focuses on multi-class change segmentation to quantify alterations between temporal images. Unlike traditional binary change detection, which requires post-processing to identify specific transformations, we introduce a Siamese architecture called SiamSegCD. This model is designed for dire...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:2024 IEEE India Geoscience and Remote Sensing Symposium (InGARSS) S. 1 - 4
Hauptverfasser: Srivastava, Noopur, Jain, Kamal
Format: Tagungsbericht
Sprache:Englisch
Veröffentlicht: IEEE 02.12.2024
Schlagworte:
Computational modeling
Computer architecture
Deep learning
Encoderdecoder architectures
Feature extraction
Image segmentation
Multi-class change segmentation
Remote sensing
Residual neural networks
Satellite images
Semantics
Siamese network
Stand-alone self-attention
Testing
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:This work focuses on multi-class change segmentation to quantify alterations between temporal images. Unlike traditional binary change detection, which requires post-processing to identify specific transformations, we introduce a Siamese architecture called SiamSegCD. This model is designed for direct segmentation of land-use and land-cover changes, particularly suited for low-resource environments. SiamSegCD utilizes deep learning models like UNet and DeepLabv3+ with a lightweight ResNet-50 backbone and stand-alone self-attention to yield a fully self-attentional model, implemented in a Siamese twin framework. The model's performance is assessed using the HRSCD dataset, revealing better results with the DeepLabv3+ encoderdecoder compared to the UNet-based version, which demonstrated a 5.98% improvement in mean Intersection over Union (mIoU).
DOI:10.1109/InGARSS61818.2024.10984419