Spatial-Frequency Decoupling Alignment Encoding for Remote Sensing Change Detection

Area of Science:

• Remote Sensing
• Computer Vision
• Signal Processing

Background:

• Existing remote sensing change detection methods struggle with complex targets and subtle textures.
• Accurate boundary detection between change targets and background remains a challenge.

Purpose of the Study:

• To propose a novel method, spatial-frequency decoupling alignment encoding (SDA-Encoding), for improved remote sensing change detection.
• To effectively leverage both spatial and frequency domain information for enhanced detection accuracy.

Main Methods:

• Utilized a Transformer encoder for bi-temporal feature extraction.
• Applied wavelet transform to decouple features into low-frequency and high-frequency components.
• Developed multi-scale high-frequency interaction (MHI) and position-aware low-frequency enhancement (PLE) modules, incorporating spatial pyramid pooling, dual-domain alignment fusion (DAF), coordinate attention, and selective fusion attention (SFA).

Main Results:

• The SDA-Encoding method demonstrated significant performance improvements on three benchmark datasets.
• The approach successfully achieved unified detection of both fine-grained and structural changes.
• Enhanced ability to distinguish change target boundaries from the background was observed.

Conclusions:

• SDA-Encoding offers a robust solution for remote sensing change detection by effectively integrating spatial and frequency domain information.
• The proposed method overcomes limitations of existing techniques in handling complex targets and subtle textural differences.
• Future work may explore further optimizations and applications of this dual-domain approach.