Study on Qualitative Impact Damage of Loquats Using Hyperspectral Technology Coupled with Texture Features

Area of Science:

• Agricultural Engineering
• Food Science
• Spectroscopy

Background:

• Post-harvest bruising of 'Zaozhong 6' loquats significantly reduces economic value and can compromise food quality and safety.
• Accurate detection of bruising and its severity is crucial for proper post-harvest treatment and loss reduction.
• Dimensionality reduction techniques are needed to improve the speed and efficiency of loquat bruising detection.

Purpose of the Study:

• To propose and evaluate a multispectral analysis method (MAM) for accurate, rapid, and nondestructive detection of loquat bruising.
• To identify effective spectral regions and features for distinguishing bruised from normal loquats.
• To develop a method for classifying the degree of bruising in loquats.

Main Methods:

• Principal Component Analysis (PCA) was applied to visible and near-infrared (Vis-NIR) spectral data (400-1000 nm) to identify key spectral regions and principal components (PCs).
• A morphological segmentation method (MSM) was developed based on PC2 score images for distinguishing bruised from normal loquats.
• Multispectral image processing involved analyzing weight coefficients, selecting characteristic wavelengths in the NIR region, and extracting texture features using a gray level co-occurrence matrix (GLCM) from K_782/999 ratio images.

Main Results:

• PCA effectively identified spectral regions and PCs for distinguishing bruised loquats.
• The proposed MAM, combining NIR spectral data and texture features from ratio images, achieved a classification accuracy of 91.3% for bruising degree.
• The MSM enabled rapid detection of normal versus bruised fruits, while the MAM classified the severity of bruising.

Conclusions:

• The multispectral analysis method (MAM) is an effective dimensionality reduction technique for loquat bruising detection.
• MAM simplifies the prediction process, improves prediction accuracy, and ensures reliable classification of bruising severity.
• Both MSM and MAM are effective for rapid, nondestructive detection and classification of loquat bruising, offering practical applications in post-harvest processing.