Full-surface defect detection of navel orange based on hyperspectral online sorting technology

Area of Science:

• Agricultural Engineering
• Food Science
• Optical Engineering

Background:

• Accurate surface defect detection and quality classification of navel oranges are crucial for the food industry.
• Low light intensity at the edges of navel oranges hinders effective surface defect identification in hyperspectral imaging.

Purpose of the Study:

• To develop a hyperspectral online sorting device for whole-surface defect detection in navel oranges.
• To address challenges posed by uneven light intensity in hyperspectral imaging of navel oranges.

Main Methods:

• Acquired hyperspectral image data of navel oranges using online detection sorting equipment.
• Extracted spectral images at a characteristic wavelength peak of 1655.72 nm.
• Implemented nonuniformity correction based on quadratic curve fitting to enhance edge light intensity and correct illumination variations.
• Segmented corrected images using thresholding to identify surface defects.
• Performed dimensionality reduction on hyperspectral data.

Main Results:

• Successfully corrected nonuniform light intensity across the entire surface of navel oranges.
• Significantly improved the detection of surface defect pixels compared to pre-correction methods.
• Achieved 100% detection accuracy in online sorting tests.
• Demonstrated enhanced sensitivity for defect detection.

Conclusions:

• The proposed hyperspectral online sorting device effectively improves the sensitivity and accuracy of navel orange surface defect detection.
• Nonuniformity correction is vital for accurate defect identification in hyperspectral imaging of spherical produce.
• Dimensionality reduction contributes to the efficiency of online hyperspectral detection systems.