Submitted:
20 November 2026
Posted:
20 November 2026
You are already at the latest version
Abstract
Greenhouse climate controllers regulate air temperature, humidity, and irrigation, but these measurements do not always reveal whether a crop is responding as intended. This review examines how physiological responses and image-derived morphological traits may provide additional information for greenhouse management. An integrative narrative synthesis was conducted using literature identified through five scientific databases and search platforms, focusing on publications from 2016 to 2026 while retaining earlier foundational studies. A comparison of 14 machine learning studies revealed promising but uneven evidence. A tomato transpiration model incorporating developmental-stage information achieved an R² of 0.92, compared with 0.54 for a conventional crop-coefficient model, while a separate greenhouse irrigation experiment reported a 20.7% increase in water-use efficiency. Greenhouse stress classifiers achieved accuracies of 83–91%, although their evaluation was limited to defined experimental treatments. Most predictive models were assessed offline rather than as components of operational control systems. Moreover, morphological responses may be delayed, nonspecific to individual stresses, and affected by imaging conditions. Based on these findings, a five-stage framework is proposed to connect environmental sensing, image analysis, plant-state estimation, decision-making, and climate actuation, with provisions for measurement validation, uncertainty, and fallback control. Whether additional plant feedback improves crop performance and resource efficiency remains to be established experimentally.
Keywords:
plant-state-aware greenhouse control
; intelligent greenhouse
; plant phenotyping
; computer vision
; plant-state estimation
; artificial intelligence
; adaptive climate control
; precision agriculture
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.