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.

Remote Sensing Image Change Detection of Buildings Based on Improved Swin-Transformer

Swin Resnetswin Transformers for Change Detection in Remote Sensing Images

Building Instance Change Detection from High Spatial Resolution Remote Sensing Images Using Improved Instance Segmentation Architecture

Siamese Transformer-Based Building Change Detection in Remote Sensing Images

M-Swin: Transformer-Based Multiscale Feature Fusion Change Detection Network Within Cropland for Remote Sensing Images

BUILDINGS CHANGE DETECTION USING HIGH-RESOLUTION REMOTE SENSING IMAGES WITH SELF-ATTENTION KNOWLEDGE DISTILLATION AND MULTISCALE CHANGE-AWARE MODULE

SwinSUNet: Pure Transformer Network for Remote Sensing Image Change Detection

Building Change Detection Based on an Edge-Guided Convolutional Neural Network Combined with a Transformer

SwinV2DNet: Pyramid and Self-Supervision Compounded Feature Learning for Remote Sensing Images Change Detection

A Siamese Swin-Unet for image change detection

A VHR Bi-Temporal Remote-Sensing Image Change Detection Network Based on Swin Transformer

DEST: Difference enhanced-Swin Transformer for remote sensing change detection

Change Detection for High-resolution Remote Sensing Images Based on a UNet-like Siamese-structured Transformer Network

CD-TransUNet: A Hybrid Transformer Network for the Change Detection of Urban Buildings Using L-Band SAR Images

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

Improved Swin Transformer-Based Semantic Segmentation of Postearthquake Dense Buildings in Urban Areas Using Remote Sensing Images

Detail Enhanced Change Detection in VHR Images Using a Self-Supervised Multiscale Hybrid Network

A dual-difference change detection network for detecting building changes on high-resolution remote sensing images

Hybrid transformer-CNN networks using superpixel segmentation for remote sensing building change detection

SFGT-CD: Semantic Feature-Guided Building Change Detection From Bitemporal Remote-Sensing Images With Transformers