AI-driven tool for high-dimensional phenotyping of floral health in roses using hyperspectral imaging

Authors: MANJUNATHA, P - The University Of Texas At Dallas; SUNDARAVADIVEL, P - University Of Texas; BORAH, S - University Of Texas; TAMIL, L - University Of Texas; Kumpatla, Siva Prasad; KNIGHT, P - Mississippi State University; Torbert Iii, Henry

Submitted to: Meeting Abstract
Publication Type: Proceedings
Publication Acceptance Date: 12/20/2025
Publication Date: N/A
Citation: N/A

Interpretive Summary: Plant health assessment in high-value ornamentals such as roses (Rosa spp.) has relied on subjective visual inspection, which is slow, labor-intensive, and unable to detect early stress. Hyperspectral imaging (HSI) offers a powerful, non-invasive alternative by capturing fine-grained spectral information, but its large data volume and complexity limit practical adoption. This manuscript discusses an AI-driven tool for automated rose health analysis using HIS. This work demonstrates a scalable, high-throughput phenotyping approach for precision floriculture, advancing early stress detection, and non-invasive plant monitoring.

Technical Abstract: Plant health assessment in high-value ornamentals such as roses (Rosa spp.) has relied on subjective visual inspection, which is slow, labor-intensive, and unable to detect early stress. Hyperspectral imaging (HSI) offers a powerful, non-invasive alternative by capturing fine-grained spectral information, but its large data volume and complexity limit practical adoption. We present an AI-driven tool for automated rose health analysis using HSI. The tool supports multi-format data ingestion, intelligent preprocessing, advanced AI-based analysis, and an interactive Health Dashboard, enabling researchers to translate high-dimensional spectral data into actionable decisions. Objectives include development of: (1) a sophisticated algorithm for computing diverse spectral vegetation indices (SVIs) for floral phenotyping analysis, (2) a spatially informed health scoring model with dynamic thresholding, and (3) an automated multi-panel “Health Dashboard” for translating high-dimensional data into actionable insights. Applied to fresh rose samples captured with a Pika L hyperspectral camera (~281 bands), the AI-powered platform quantitatively evaluated plant health, mapped spatial stress patterns, and extracted unique spectral signatures tied to physiological stressors such as chlorophyll degradation and water loss. By combining detailed spectral depth with RGB-like visual summaries, the system bridges complex spectral data with practical decision-making. This work demonstrates a scalable, high-throughput phenotyping approach for precision floriculture, advancing early stress detection, and non-invasive plant monitoring.