Detection of Apple Proliferation Disease Using Hyperspectral Imaging and Machine Learning Techniques.

Area of Science:

• Plant pathology
• Agricultural engineering
• Remote sensing

Background:

• Apple proliferation is a significant disease impacting European fruit yields.
• Current detection methods rely on human observation, which is labor-intensive and difficult to automate.
• Early and accurate disease detection is vital for effective management and preventing widespread infection.

Purpose of the Study:

• To investigate the potential of hyperspectral imaging combined with machine learning for detecting apple proliferation symptoms.
• To develop an automated method for disease detection based on spectral signatures of leaf samples.
• To assess the efficacy of different spectral ranges and machine learning algorithms for disease identification.

Main Methods:

• Collected 1160 apple leaf samples over two growing seasons (2019-2020).
• Utilized hyperspectral imaging (400-2500 nm) and PCR analysis for reference data.
• Applied machine learning algorithms (e.g., rRBF) for classification and regression after data preprocessing and feature extraction.

Main Results:

• Imaging multiple leaves per tree improved detection accuracy.
• Spectral indices proved robust for identifying diseased trees.
• Machine learning models achieved high accuracy (0.971) in controlled environments and moderate accuracy (0.731-0.751) in field conditions.
• Regression models predicted qPCR results with an RMSE of 14.491 phytoplasma per plant cell.

Conclusions:

• Hyperspectral imaging coupled with machine learning offers a promising automated solution for apple proliferation detection.
• The full spectral range and spatial distribution of data enhance detection capabilities.
• This approach can support farmers in early disease management, reducing economic losses.