Development of algorithms for detecting citrus canker based on hyperspectral reflectance imaging

Area of Science:

• Agricultural Science
• Remote Sensing
• Computer Vision

Background:

• Automated detection of citrus canker is crucial for the citrus industry's profitability.
• Hyperspectral imaging offers a promising non-destructive technology for agricultural product inspection.

Purpose of the Study:

• To assess the feasibility of hyperspectral imaging for classifying citrus canker.
• To differentiate citrus canker from normal fruit surfaces and nine other peel defects.

Main Methods:

• A combination algorithm using principal component analysis (PCA) and a two-band ratio (Q(687/630)) method was developed.
• The performance of the two-band ratio method alone was also evaluated for rapid multispectral imaging system development.

Main Results:

• The combined PCA and two-band ratio approach achieved high classification accuracy (99.5% for training, 98.2% for testing).
• The two-band ratio method alone showed good performance (84.5% for training, 82.9% for testing).

Conclusions:

• The proposed combination approach is highly efficient for classifying citrus canker using hyperspectral imagery.
• The two-band ratio method is effective for discriminating citrus canker, with exceptions for copper burn and anthracnose.