Hyperspectral, multimodal fusion, and tinyml for rice crop monitoring: A focused survey and synthesis toward the tinyricenet framework
Smart Agricultural Technology · 2026-12 · 15 · 102716
DOI: 10.1016/j.atech.2026.102716 ↗Kurz gefasst
Fachartikel in der Zeitschrift Smart Agricultural Technology, 2026-12. Verfasst von Uchechi Grace Dike und 2 weiteren Autor:innen. Die Arbeit nennt 73 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0).
Abstract (Originalsprache)
Reliable monitoring of rice disease, stress, growth, and yield requires evidence across sensing scales and time. This focused survey synthesizes hyperspectral and multispectral sensing, multimodal fusion, and machine learning for field deployment in rice monitoring. Here, TinyML denotes inference on microcontrollers under strict memory, computational, and energy constraints; it is distinguished from broader edge computing on more capable devices and field gateways. A structured search and screening process yielded 108 sources, which were classified by evidence scope and methodological role. The synthesis shows that public agricultural hyperspectral benchmarks mainly represent one date and support spectral and spatial modelling, whereas operational yield prediction generally requires multitemporal satellite, UAV, meteorological, or field measurements. Rice disease and physiological studies demonstrate strong performance within individual studies, but comparisons remain limited by heterogeneous targets, sample units, and validation designs. Multimodal studies support complementary sensing, yet few systematically test calibration error, missing or asynchronous inputs, seasonal transfer, or distribution shift. Deployment studies seldom report complete device, memory, latency, energy, quantization, and field testing results. TinyRiceNet is presented as a conceptual synthesis framework rather than a newly trained or benchmarked model. It organizes spectral stress assessment, visual symptom detection, seasonal yield estimation, and fusion that accounts for data quality while preserving distinct labels and validation requirements. The review identifies paired multiscale rice datasets, external validation, evaluation with missing modalities, transfer across sensors, and transparent hardware reporting as priorities for deployable rice monitoring research.
Zitieren
APA
Dike, U. G., Dong, M., & Sato, K. (2026). Hyperspectral, multimodal fusion, and tinyml for rice crop monitoring: A focused survey and synthesis toward the tinyricenet framework. Smart Agricultural Technology, 15, 102716. https://doi.org/10.1016/j.atech.2026.102716
Vancouver
Dike UG, Dong M, Sato K. Hyperspectral, multimodal fusion, and tinyml for rice crop monitoring: A focused survey and synthesis toward the tinyricenet framework. Smart Agricultural Technology. 2026;15:102716. doi:10.1016/j.atech.2026.102716
BibTeX
@article{Dike2026,
author = {Dike, Uchechi Grace and Dong, Mianxiong and Sato, Kazuhiko},
title = {Hyperspectral, multimodal fusion, and tinyml for rice crop monitoring: A focused survey and synthesis toward the tinyricenet framework},
journal = {Smart Agricultural Technology},
year = {2026},
volume = {15},
pages = {102716},
doi = {10.1016/j.atech.2026.102716}
}
EUVIMED