Abstract:Although deep learning-based change detection (CD) methods achieve great success in remote sensing images, they still suffer from two main challenges. First, popular convolutional neural networks (CNNs) are weak in extracting discriminated features focusing on changed regions, since most methods ignore the multifrequency components of bitemporal images. Second, although existing CD methods employ the Transformer structure to capture long-range dependency for global feature representation, it is difficult for them to simultaneously take into account the long-range dependency of changed objects at various scales. To address the above issues, we propose a triple CD network (TCD-Net) via joint multifrequency and full-scale swin-transformer (FST). The proposed TCD-Net has two main advantages. First, we propose a multifrequency channel attention (MFCA) module to boost the ability of modeling the channel correlation, which can compensate for the problem of insufficient feature representation caused by only performing global average pooling (GAP). Furthermore, a joint multifrequency difference feature enhancement (JM-DFE) guiding block is proposed to improve the boundary quality and the position awareness of truly changed objects, which can effectively extract channel features of multifrequency information and thus improve the discriminative ability of features. Second, unlike Siamese-based structures, we propose an FST module as the third branch to model and aggregate the long-range dependency of multiscale changed objects, which can alleviate the missed detections of small objects and achieve more compact changed regions effectively. Experiments on three public CD datasets exhibit that the proposed TCD-Net achieves better CD accuracy with smaller model complexity than state-of-the-art (SOTA) methods. The code is publicly available at https://github.com/RSCD-mz/TCD-Net .

Fully Convolutional Networks with Multiscale 3D Filters and Transfer Learning for Change Detection in High Spatial Resolution Satellite Images

A Multiscale Graph Convolutional Network for Change Detection in Homogeneous and Heterogeneous Remote Sensing Images

Transfer Learning-Based Bilinear Convolutional Networks for Unsupervised Change Detection

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

Remote Sensing Image Change Detection Based on Deep Learning: Multi-Level Feature Cross-Fusion with 3D-Convolutional Neural Networks

Global-Local Collaborative Learning Network for Optical Remote Sensing Image Change Detection

Unsupervised Change Detection in Satellite Images Using Convolutional Neural Networks

A Semisupervised Siamese Network for Efficient Change Detection in Heterogeneous Remote Sensing Images

Triple Change Detection Network Via Joint Multifrequency and Full-Scale Swin-Transformer for Remote Sensing Images.

Change Detection for High-resolution Remote Sensing Images Based on a Siamese Structured UNet3+ Network

Dual-Attention-Guided Multiscale Feature Aggregation Network for Remote Sensing Image Change Detection

DASNet: Dual Attentive Fully Convolutional Siamese Networks for Change Detection in High-Resolution Satellite Images

Ultralightweight Spatial–Spectral Feature Cooperation Network for Change Detection in Remote Sensing Images

Multiscale Change Detection Method for Remote Sensing Images Based on Online Learning Framework

A Deep Siamese Network with Hybrid Convolutional Feature Extraction Module for Change Detection Based on Multi-sensor Remote Sensing Images

An Enhanced and Unsupervised Siamese Network With Superpixel-Guided Learning for Change Detection in Heterogeneous Remote Sensing Images

TChange: A Hybrid Transformer-CNN Change Detection Network

Deep Multiscale Siamese Network With Parallel Convolutional Structure and Self-Attention for Change Detection

A Dual Neighborhood Hypergraph Neural Network for Change Detection in VHR Remote Sensing Images

Change Detection Method with Spatial and Spectral Information from Deep Learning