Spatial Temporal Fusion Based Features for Enhanced Remote Sensing Change Detection

Researchers at Jomo Kenyatta University of Agriculture and Technology have developed a new approach to remote sensing change detection, leveraging spatial-temporal dependencies to extract contextual understanding from RS images. This innovative method uses parallel encoders and Long Short Term Memory (LTSM) layers to optimize information representation and reduce noise interference. The proposed technique has been evaluated on the EGY-BCD dataset, achieving an overall accuracy of 97.4%, an F1 Score of 89%, and an intersection over union (IoU) of 86.7, outperforming conventional CD methods.

Key Takeaways:

• The proposed technique incorporates spatial-temporal dependencies to create contextual understanding in remote sensing change detection, improving the accuracy and robustness of change detection tasks.
• The use of parallel encoders and LTSM layers enables the extraction of highly representative deep features, which are then concatenated and processed through the decoder to optimize information representation.
• The study evaluated the proposed technique on the EGY-BCD dataset, achieving higher overall accuracy (97.4%), F1 Score (89%), and IoU (86.7) compared to conventional CD methods.
• The results demonstrate the potential of incorporating spatial-temporal dependencies in change detection tasks for remote sensing images.
• The proposed approach aims to maximize authentic information while reducing noise interference by introducing the context of change.

Statistics:

• Overall accuracy: 97.4%
• F1 Score: 89%
• Intersection over union (IoU): 86.7
• Number of datasets used for evaluation: 1 (EGY-BCD)
• Number of publications cited: 1 (Scientific Reports)

Sources:

• Spatial temporal fusion based features for enhanced remote sensing change detection. Scientific Reports, 2025,15(1):1-13. (Scientific Reports - http://www.nature.com/srep/index.html)
• Journal of Engineering (2025) October 20, 2025; p 3959.