.png){kind=icon}
.png){kind=icon}
Return
Dynamic Evolution: Continuous Change Monitoring in Time-Series Remote Sensing Images
.png){kind=icon}
PRE
AI
J
L
B
L
DOI：10.1109/tgrs.2026.3717221.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The Earth’s surface experiences continuous changes across multiple spatial and temporal scales, and time-series remote sensing images (TSIs) provide continuous observation data for monitoring and analyzing these changes. However, most existing time-series change detection (TSCD) methods based on deep learning primarily focus on binary change classification, lacking the ability to capture the temporal evolution of changes. This limitation hinders a comprehensive understanding of complex, time-dependent change processes. To address this issue, we define continuous change monitoring as the dynamic tracking of changes across multiple time points, which enables not only the detection of change occurrence but also the precise determination of when these changes occur in TSIs. To bridge the methodological gap in deep learning-based continuous change monitoring, we propose a novel continuous change monitoring network (CCM-Net) designed to capture temporal change dynamics in TSIs. CCM-Net employs a U-Net-based convolutional neural network to extract spatial features and generate difference features that emphasize temporal variations. Furthermore, a spatial-spatiotemporal LSTM (SSL) module is incorporated to associate similar change patterns across different spatial locations and time steps. A monitoring inference moment (MIM) module is also introduced to infer the moment of changes, integrating continuous change monitoring and change moment identification into a unified framework. The proposed approach is evaluated on two global-scale datasets, DynamicEarthNet and SpaceNet7. Experimental results demonstrate that CCM-Net achieving <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$F1$ </tex-math></inline-formula>-score improvements of 7.31% and 2.84% over the best competing method on DynamicEarthNet and SpaceNet7, respectively, indicating its effectiveness in tracking temporal evolution of changes in TSIs. Our source code will be made publicly available at: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/lijialu144/CCM-Net</uri>
Keywords:
Change moment identification
continuous change monitoring
deep learning
time-series remote sensing images (TSIs)
Journal
IF:
8.6
Papers:
2.1W
Citations:
10.7W
Organization
Cited Papers