Abstract:Change detection (CD) in remote sensing (RS) images is a critical task that has achieved significant success by deep learning. Current networks often employ pixel-based differencing, proportion, classification-based, or feature concatenation methods to represent changes of interest. However, these methods fail to effectively detect the desired changes, as they are highly sensitive to factors such as atmospheric conditions, lighting variations, and phenological variations, resulting in detection errors. Inspired by the transformer structure, we adopt a cross-attention mechanism to more robustly extract feature differences between bitemporal images. The motivation of the method is based on the assumption that if there is no change between image pairs, the semantic features from one temporal image can well be represented by the semantic features from another temporal image. Conversely if there is a change, there are significant reconstruction errors. Therefore, a Cross Swin transformer-based Siamese U-shaped network namely CSTSUNet is proposed for RS CD. CSTSUnet consists of encoder, difference feature extraction, and decoder. The encoder is based on a hierarchical residual network (ResNet) with the Siamese U-net structure, allowing parallel processing of bitemporal images and extraction of multiscale features. The difference feature extraction consists of four difference feature extraction modules that compute difference feature at multiple scales. In this module, Cross Swin transformer is employed in each difference feature extraction module to communicate the information of bitemporal images. The decoder takes in the multiscale difference features as input, injects details and boundaries iteratively level by level, and makes the change map more and more accurate. We conduct experiments on three public datasets, and the experimental results demonstrate that the proposed CSTSUNet outperforms other state-of-the-art methods in terms of both qualitative and quantitative analyses. Our code is available at https://github.com/l7170/CSTSUNet.git.

A Transformer-Based Network for Change Detection in Remote Sensing Using Multiscale Difference-Enhancement

TransUNetCD: A Hybrid Transformer Network for Change Detection in Optical Remote-Sensing Images

Multiscale Spatial-Channel Transformer Architecture Search for Remote Sensing Image Change Detection

TChange: A Hybrid Transformer-CNN Change Detection Network

A hybrid U‐shaped and transformer network for change detection in high‐resolution remote sensing images

An attention-based multiscale transformer network for remote sensing image change detection

DiFormer: A Difference Transformer Network for Remote Sensing Change Detection

Multiscale Fusion CNN-Transformer Network for High-Resolution Remote Sensing Image Change Detection

A CBAM Based Multiscale Transformer Fusion Approach for Remote Sensing Image Change Detection

A Network Combining a Transformer and a Convolutional Neural Network for Remote Sensing Image Change Detection

CSTSUNet: A Cross Swin Transformer-Based Siamese U-Shape Network for Change Detection in Remote Sensing Images

MCTNet: A Multi-Scale CNN-Transformer Network for Change Detection in Optical Remote Sensing Images

TransY-Net: Learning Fully Transformer Networks for Change Detection of Remote Sensing Images

TransY-Net:Learning Fully Transformer Networks for Change Detection of Remote Sensing Images

Structure-Aware Multiscale Hybrid Network for Change Detection of Remote Sensing Images

MTN: A Multi-Scale Transformer Network for Different Resolution Remote Sensing Images Change Detection.

Relating CNN-Transformer Fusion Network for Change Detection

Relating CNN-Transformer Fusion Network for Remote Sensing Change Detection

MFATNet: Multi-Scale Feature Aggregation via Transformer for Remote Sensing Image Change Detection

Remote Sensing Change Detection With Bitemporal and Differential Feature Interactive Perception