Advancing hyperspectral imaging techniques for root systems: a new pipeline for macro- and microscale image...

Area of Science:

• Plant science
• Agricultural science
• Environmental science

Background:

• Understanding environmental impacts on root growth and health is crucial for agriculture and environmental management.
• Hyperspectral imaging (HSI) offers non-destructive analysis of plant tissues but is underutilized for root systems and the root-soil interface.
• Standardized guidelines for HSI acquisition and data analysis for root studies are lacking.

Purpose of the Study:

• Investigate HSI techniques for analyzing rhizobox-grown root systems at macro- to micro-scales.
• Evaluate the influence of image acquisition parameters and data processing on spectral signatures of root, soil, and root-soil interface.
• Compare classification methods (SAM, K-Means) and machine learning approaches (RF, SVM) for automated root system image classification.

Main Methods:

• Utilized imec VNIR SNAPSCAN camera for HSI across various configurations.
• Focused on three graminoid species with distinct root architectures.
• Compared Spectral Angle Mapper (SAM), K-Means clustering, Random Forest (RF), and Support Vector Machine (SVM) for classification.

Main Results:

• A Random Forest (RF) model trained with SAM classifications and reduced wavelengths achieved 88-91% accuracy in differentiating root and soil.
• The root-soil interface was not clearly resolved but improved root-soil distinction.
• The approach highlighted spectral differences due to configurations, acquisition settings, and species, facilitating root biomass monitoring.

Conclusions:

• Addressed key challenges in HSI acquisition and data processing for root system analysis.
• Laid groundwork for broader VNIR HSI application in root system studies.
• Provided a data analysis pipeline as a Python-based tool for semi-automated analysis of root-soil HSI data.