Journal of Geo-information Science

MSDFENet： Change Detection in High-Resolution Optical Remote Sensing Images Based on Multi-Scale Differential Feature Enhancement

PENG Daifeng, LI Yaning, ZHOU Dingwei, GUAN Haiyan

Journal of Geo-information Science, 2025, 27(9): 2250-2267. DOI: 10.12082/dqxxkx.2025.250160

实验模型	F1/%	参数量/M	测试时间/s
FC-EF	84.94	1.35	34
FC-Siam-Conc	87.08	1.55	38
FC-Siam-Diff	86.85	1.35	38
UNet++_MSOF	88.33	9.05	65
DTCDSCN	88.41	41.07	70
DSIFN	89.17	50.44	117
STANet	88.75	16.93	64
BIT	89.43	3.49	63
ICIFNet	89.96	23.84	88
VcT	90.11	3.57	100
PRNet	89.48	3.41	56
本文方法	90.68	5.64	45

Tab. 6 Comparison of performance and complexity metrics for different models

Other figure/table from this article

Fig. 1 Architecture of MSDFENet network
Fig. 2 Structure of APDC module
Fig. 3 Structure of MSFA module
Tab. 1 Comparison of different change detection methods
Tab. 2 Comparison of quantitative metrics for different methods on the LEVIR-CD dataset (%)
Fig. 4 Experimental results of the LEVIR-CD dataset
Tab. 3 Comparison of quantitative metrics for different methods on the CDD dataset (%)
Fig. 5 Experimental results of the CDD dataset
Tab. 4 Comparison of quantitative metrics for different methods on the WHU-CD Dataset (%)
Fig. 6 Experimental results of the WHU-CD dataset
Tab. 5 Quantitative results of ablation study (%)
Fig. 7 Visual results of ablation experiment
Fig. 8 Visualization of four-stage feature heatmap in MSDFENet