Appearance and characterization of fruit image textures for quality sorting using wavelet transform and genetic...

Area of Science:

• Agricultural Science
• Computer Vision
• Image Processing

Background:

• Fruit quality assessment is crucial for the food industry.
• Traditional methods are often destructive and subjective.
• Objective, non-destructive methods are needed for efficient quality control.

Purpose of the Study:

• To develop and evaluate an image-based system for non-destructive quality grading of mangoes and guavas.
• To identify optimal color and texture features for fruit classification.
• To compare different feature selection and classification techniques.

Main Methods:

• Extraction and analysis of color and textural features from fruit images.
• Utilized Mahalanobis distance and feature intercorrelation for feature selection.
• Employed wavelet families (e.g., db, bior, rbior, Coif, Sym) for texture analysis.
• Applied genetic algorithms for optimal feature selection.
• Classified fruit quality using Support Vector Machine (SVM) and Artificial Neural Network (ANN) classifiers.

Main Results:

• Identified key color and texture features with high discriminatory power for mangoes and guavas.
• Genetic algorithm-based feature selection proved effective in distinguishing fruit quality classes.
• Achieved high classification accuracies: 97.61% for guava grading (SVM) and 95.65% for mango grading (ANN).

Conclusions:

• Image analysis combining genetic algorithms and wavelet texture features provides an accurate and reliable method for non-destructive fruit quality assessment.
• The proposed approach is suitable for objective grading of mangoes and guavas.
• The method has potential for application in automated, in-line sorting systems.