STeInFormer: Spatial-Temporal Interaction Transformer Architecture for Remote Sensing Change Detection

TL;DR

STeInFormer introduces a novel Transformer-based backbone for remote sensing change detection, effectively capturing spatial-temporal interactions and spectral features, leading to superior performance over existing methods.

Contribution

The paper presents the first general backbone network specifically designed for RSCD, incorporating a spatial-temporal interaction Transformer and a multi-frequency token mixer.

Findings

Outperforms state-of-the-art methods on three datasets

Achieves a favorable efficiency-accuracy trade-off

Validates effectiveness through extensive experiments

Abstract

Convolutional neural networks and attention mechanisms have greatly benefited remote sensing change detection (RSCD) because of their outstanding discriminative ability. Existent RSCD methods often follow a paradigm of using a non-interactive Siamese neural network for multi-temporal feature extraction and change detection heads for feature fusion and change representation. However, this paradigm lacks the contemplation of the characteristics of RSCD in temporal and spatial dimensions, and causes the drawback on spatial-temporal interaction that hinders high-quality feature extraction. To address this problem, we present STeInFormer, a spatial-temporal interaction Transformer architecture for multi-temporal feature extraction, which is the first general backbone network specifically designed for RSCD. In addition, we propose a parameter-free multi-frequency token mixer to integrate frequency-domain features that provide spectral information for RSCD. Experimental results on three datasets validate the effectiveness of the proposed method, which can outperform the state-of-the-art methods and achieve the most satisfactory efficiency-accuracy trade-off. Code is available at https://github.com/xwmaxwma/rschange.