Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CD-CTFM: A Lightweight CNN-Transformer Network for Remote Sensing Cloud Detection Fusing Multiscale Features

Wenxuan Ge, Xubing Yang, Li Zhang

TL;DR

Cloud-contaminated optical remote sensing images hinder information extraction, requiring reliable cloud masks. The authors introduce CD-CTFM, a lightweight encoder–decoder that fuses local and global features via a CNN–Transformer backbone, a Lightweight Feature Pyramid Module, and a Lightweight Channel-Spatial Attention module. On 38-Cloud and MODIS datasets, CD-CTFM achieves competitive accuracy with substantially fewer parameters and GFLOPS than state-of-the-art methods. The work demonstrates that careful multiscale feature fusion and efficient attention can deliver accurate cloud detection with lower computational cost, enabling faster preprocessing for large-scale remote sensing pipelines.

Abstract

Clouds in remote sensing images inevitably affect information extraction, which hinder the following analysis of satellite images. Hence, cloud detection is a necessary preprocessing procedure. However, the existing methods have numerous calculations and parameters. In this letter, a lightweight CNN-Transformer network, CD-CTFM, is proposed to solve the problem. CD-CTFM is based on encoder-decoder architecture and incorporates the attention mechanism. In the decoder part, we utilize a lightweight network combing CNN and Transformer as backbone, which is conducive to extract local and global features simultaneously. Moreover, a lightweight feature pyramid module is designed to fuse multiscale features with contextual information. In the decoder part, we integrate a lightweight channel-spatial attention module into each skip connection between encoder and decoder, extracting low-level features while suppressing irrelevant information without introducing many parameters. Finally, the proposed model is evaluated on two cloud datasets, 38-Cloud and MODIS. The results demonstrate that CD-CTFM achieves comparable accuracy as the state-of-art methods. At the same time, CD-CTFM outperforms state-of-art methods in terms of efficiency.

CD-CTFM: A Lightweight CNN-Transformer Network for Remote Sensing Cloud Detection Fusing Multiscale Features

TL;DR

Cloud-contaminated optical remote sensing images hinder information extraction, requiring reliable cloud masks. The authors introduce CD-CTFM, a lightweight encoder–decoder that fuses local and global features via a CNN–Transformer backbone, a Lightweight Feature Pyramid Module, and a Lightweight Channel-Spatial Attention module. On 38-Cloud and MODIS datasets, CD-CTFM achieves competitive accuracy with substantially fewer parameters and GFLOPS than state-of-the-art methods. The work demonstrates that careful multiscale feature fusion and efficient attention can deliver accurate cloud detection with lower computational cost, enabling faster preprocessing for large-scale remote sensing pipelines.

Abstract

Clouds in remote sensing images inevitably affect information extraction, which hinder the following analysis of satellite images. Hence, cloud detection is a necessary preprocessing procedure. However, the existing methods have numerous calculations and parameters. In this letter, a lightweight CNN-Transformer network, CD-CTFM, is proposed to solve the problem. CD-CTFM is based on encoder-decoder architecture and incorporates the attention mechanism. In the decoder part, we utilize a lightweight network combing CNN and Transformer as backbone, which is conducive to extract local and global features simultaneously. Moreover, a lightweight feature pyramid module is designed to fuse multiscale features with contextual information. In the decoder part, we integrate a lightweight channel-spatial attention module into each skip connection between encoder and decoder, extracting low-level features while suppressing irrelevant information without introducing many parameters. Finally, the proposed model is evaluated on two cloud datasets, 38-Cloud and MODIS. The results demonstrate that CD-CTFM achieves comparable accuracy as the state-of-art methods. At the same time, CD-CTFM outperforms state-of-art methods in terms of efficiency.
Paper Structure (13 sections, 1 equation, 6 figures, 2 tables)

This paper contains 13 sections, 1 equation, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Overview of CD-CTFM
  4. Lightweight Feature Pyramid Module (LWFPM)
  5. Lightweight Channel-Spatial Attention Module (LWAM)
  6. Experiments and Results
  7. Experimental Settings
  8. Datasets
  9. Evaluation Metrics
  10. Implementation Details
  11. Ablation Studies
  12. Comparison with Other Methods
  13. Conclusion

Figures (6)

  • Figure 1: Overall framework of CD-CTFM. The encoder contains a lightweight backbone and LWFPM while the decoder is based on depthwise separable convolutions (DSC). The LWAM filters information propagated through the skip connections.
  • Figure 2: The detail of the lightweight backbone.
  • Figure 3: The structure of SD block.
  • Figure 4: The detail of Lightweight Channel-Spatial Attention Module
  • Figure 5: Comparison between the results of different methods in 38-Cloud dataset. White area represents cloud, black area represents non-cloud, red area represents false-positive detection and green area represents false-negative detection.
  • ...and 1 more figures