Characterization of Hazelnut Trees in Open Field Through High-Resolution UAV-Based Imagery and Vegetation Indices

Area of Science:

• Agricultural Science
• Remote Sensing
• Plant Pathology

Background:

• Global hazelnut cultivation is increasing, driven by demand for kernels.
• Climate change and pests pose significant threats to hazelnut yields.
• Precision agriculture technologies are crucial for managing crop health and optimizing production.

Purpose of the Study:

• To develop a straightforward method for monitoring hazelnut tree physio-pathological status using drone-based multispectral imaging.
• To evaluate the effectiveness of various vegetation indices (VIs) in predicting tree health.
• To support farmers with data-driven decision-making for sustainable hazelnut farming.

Main Methods:

• Acquired multispectral aerial images (RGB, Red Edge, near-infrared) of 185 hazelnut trees across two orchards.
• Divided each tree image into nine quadrants for enhanced accuracy and reduced false negatives.
• Computed nine vegetation indices (VIs) for each quadrant and classified them as healthy/unhealthy using supervised algorithms.

Main Results:

• Five vegetation indices (GNDVI, GCI, NDREI, NRI, GI) proved effective in predicting tree health.
• Achieved approximately 65% model accuracy with 13% false negatives, largely independent of the classification algorithm used.
• Demonstrated that certain VIs can infer the physio-pathological condition of hazelnut trees.

Conclusions:

• Drone-captured multispectral images provide a rapid, non-destructive method for physiological characterization of hazelnut trees.
• This approach supports sustainable hazelnut cultivation by enabling timely interventions.
• The findings contribute to advancing precision agriculture techniques for crop monitoring and management.