Aniruddha et al. (2026) Evaluating Multimodal Fusion Strategies for Resilient Agricultural Sensing Systems

Identification

• Journal: Lecture notes in networks and systems
• Year: 2026
• Date: 2026-01-01
• Authors: Ponnuri Aniruddha, Abhay Shaji Valiyaparambil, K. Sornalakshmi
• DOI: 10.1007/978-3-032-10783-1_19

Research Groups

Department of Data Science and Business Systems, School of Computing, Faculty of Engineering and Technology, SRM Institute of Science and Technology, Kattankulathur, Chennai, India

Short Summary

This paper evaluates three advanced multimodal data fusion techniques (MDFCL, GSIFN, Perceiver IO) for combining agricultural image and time-series data, assessing their versatility, advantages, and limitations to support resilient precision agriculture.

Objective

• To evaluate the adaptation versatility, advantages, and limitations of Multimodal Data Fusion–based Graph Contrastive Learning (MDFCL), Graph-Structured & Interlaced-Masked Fusion Network (GSIFN), and Perceiver IO for the fusion of agro-sensor and growth-image data, providing actionable recommendations for robust, scalable, and efficient precision agriculture.

Study Configuration

• Spatial Scale: Agricultural fields/plants (implied for agro-sensor and growth-image data).
• Temporal Scale: Time-series data, covering agricultural growth cycles.

Methodology and Data

• Models used: Multimodal Data Fusion–based Graph Contrastive Learning (MDFCL), Graph-Structured & Interlaced-Masked Fusion Network (GSIFN), Perceiver IO.
• Data sources: Agricultural time-series data, agro-sensor data, and growth-image data.

Main Results

• MDFCL achieved 97.66% accuracy with image-only inputs, demonstrating cross-modal robustness.
• GSIFN achieved 100.00% classification accuracy in rigorous 5-fold cross-validation, effectively capturing higher-order interactions and countering redundancy.
• Perceiver IO achieved 97.66% accuracy, similar to MDFCL, but with fewer parameters and reduced computational requirements, offering flexible handling of heterogeneous data streams with near-linear complexity.
• The study provides actionable recommendations for the fusion of agricultural image and time-series data to enhance precision agriculture systems.

Contributions

• This study provides the first comprehensive evaluation of advanced multimodal fusion techniques (MDFCL, GSIFN, Perceiver IO) specifically adapted for agricultural image and time-series data, an area previously less explored.
• It offers practical insights into the adaptation versatility, advantages, and limitations of these models, delivering actionable recommendations for developing robust, scalable, and efficient precision agriculture systems.

Funding

• Not explicitly mentioned in the provided text.

Citation

@article{Aniruddha2026Evaluating,
  author = {Aniruddha, Ponnuri and Valiyaparambil, Abhay Shaji and Sornalakshmi, K.},
  title = {Evaluating Multimodal Fusion Strategies for Resilient Agricultural Sensing Systems},
  journal = {Lecture notes in networks and systems},
  year = {2026},
  doi = {10.1007/978-3-032-10783-1_19},
  url = {https://doi.org/10.1007/978-3-032-10783-1_19}
}