Enhanced Spatiotemporal Remote Sensing Fusion: A Comprehensive Evaluation of DASCNN Against Deep Learning Baselines

DOI

10.2991/978-94-6239-616-6_33How to use a DOI?

Keywords

Deep learning; generative adversarial networks; image fusion; remote sensing; residual networks; spatiotemporal fusion; spectral preservation; transformers; U-Net

Abstract

Timely integration of heterogeneous spatiotemporal and spectral information from various satellite observations is crucial to the remote sensing domain for advanced applications such as land cover classification, change detection, and environmental monitoring. As an extension of our previously proposed Deep Artistic Sequence-based CNN for fusion (DASCNN) with Point-Plane Distance (PPD) fusion, this paper offers a comprehensive comparative study of DASCNN against four state-of-the-art deep learning alternatives: encoder-decoder architectures (U-Net), residual networks, vision transformers, and generative adversarial networks (GANs). A systematic assessment via six complementary quality metrics with various levels of fusion significance is applied: structural similarity (SSIM), pixel-level authenticity (MSE, PSNR), spectral authenticity (SAM), general quality (Q-Metric), and human perceptual authenticity (PIQE). Extensive experiments with MODIS satellite data demonstrate that DASCNN possesses the highest structural authenticity (SSIM: 0.9980) while GANs and DASCNN possess the highest spectral authenticity (SAM: 1.5247). Although DASCNN is beaten pixel-wise by U-Net (MSE: 0.0471 vs. DASCNN’s 0.1073), the compromise established in this study is on structural and spectral authenticity - the needed attributes in remote sensing applications. Practical validation of the land cover classification supports the dominance of DASCNN as it achieves 87.4% accuracy while U-Net and its transformer counterparts reach 84.6% and 85.9%, respectively. The importance of the additions of the Deep Artistic Sequence and PPD fusion is confirmed through ablation studies.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Swathi Nallagachu
AU  - R. Sandanalakshmi
PY  - 2026
DA  - 2026/03/31
TI  - Enhanced Spatiotemporal Remote Sensing Fusion: A Comprehensive Evaluation of DASCNN Against Deep Learning Baselines
BT  - Proceedings of the International Conference on Artificial Intelligence and Secure Data Analytics (ICAISDA 2025)
PB  - Atlantis Press
SP  - 427
EP  - 440
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-616-6_33
DO  - 10.2991/978-94-6239-616-6_33
ID  - Nallagachu2026
ER  -