Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Global-Local Detail Guided Transformer for Sea Ice Recognition in Optical Remote Sensing Images

Zhanchao Huang, Wenjun Hong, Hua Su

TL;DR

The paper addresses sea ice recognition in optical remote sensing imagery, a task complicated by large scale variation and fine edge details. It introduces GDGT, a UNet-like model with a CNN encoder and Transformer decoder, augmented by a Global-Local Feature Fusion (GLFF) module and a Detail-Guided Decoder (DGD). GLFF blends global structural cues with local texture details, while DGD preserves high-resolution edge information through wavelet-guided decoding. On a GF-2 derived sea-ice dataset, GDGT outperforms strong baselines in key metrics and demonstrates improved delineation of thin and edge-rich ice regions, with implications for climate monitoring and safe navigation.

Abstract

The recognition of sea ice is of great significance for reflecting climate change and ensuring the safety of ship navigation. Recently, many deep learning based methods have been proposed and applied to segment and recognize sea ice regions. However, the diverse scales of sea ice areas, the zigzag and fine edge contours, and the difficulty in distinguishing different types of sea ice pose challenges to existing sea ice recognition models. In this paper, a Global-Local Detail Guided Transformer (GDGT) method is proposed for sea ice recognition in optical remote sensing images. In GDGT, a global-local feature fusiont mechanism is designed to fuse global structural correlation features and local spatial detail features. Furthermore, a detail-guided decoder is developed to retain more high-resolution detail information during feature reconstruction for improving the performance of sea ice recognition. Experiments on the produced sea ice dataset demonstrated the effectiveness and advancement of GDGT.

Global-Local Detail Guided Transformer for Sea Ice Recognition in Optical Remote Sensing Images

TL;DR

The paper addresses sea ice recognition in optical remote sensing imagery, a task complicated by large scale variation and fine edge details. It introduces GDGT, a UNet-like model with a CNN encoder and Transformer decoder, augmented by a Global-Local Feature Fusion (GLFF) module and a Detail-Guided Decoder (DGD). GLFF blends global structural cues with local texture details, while DGD preserves high-resolution edge information through wavelet-guided decoding. On a GF-2 derived sea-ice dataset, GDGT outperforms strong baselines in key metrics and demonstrates improved delineation of thin and edge-rich ice regions, with implications for climate monitoring and safe navigation.

Abstract

The recognition of sea ice is of great significance for reflecting climate change and ensuring the safety of ship navigation. Recently, many deep learning based methods have been proposed and applied to segment and recognize sea ice regions. However, the diverse scales of sea ice areas, the zigzag and fine edge contours, and the difficulty in distinguishing different types of sea ice pose challenges to existing sea ice recognition models. In this paper, a Global-Local Detail Guided Transformer (GDGT) method is proposed for sea ice recognition in optical remote sensing images. In GDGT, a global-local feature fusiont mechanism is designed to fuse global structural correlation features and local spatial detail features. Furthermore, a detail-guided decoder is developed to retain more high-resolution detail information during feature reconstruction for improving the performance of sea ice recognition. Experiments on the produced sea ice dataset demonstrated the effectiveness and advancement of GDGT.
Paper Structure (9 sections, 8 equations, 5 figures, 2 tables)

This paper contains 9 sections, 8 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. The Proposed GDGT Model
  3. Global-Local Feature Fusion
  4. Detail-Guided Decoder
  5. Experiments and Discussions
  6. Experimental Conditions
  7. Experiments and Discussions
  8. Conclusions
  9. Reference

Figures (5)

  • Figure 1: The framework of the proposed GDGT for seaice segmentation and recognition.
  • Figure 2: The Global-Local Feature Fusion (GLFF) Module.
  • Figure 3: The Detail-Guided Decoder (DGD) Module.
  • Figure 4: Visualization of comparative experimental results of predicting image patches.
  • Figure 5: Visualization of prediction results for the entire remote sensing image.