Nondestructive classification of saffron using color and textural analysis.

Area of Science:

• Agricultural Science
• Computer Vision
• Machine Learning

Background:

• Saffron classification is crucial for quality control.
• Traditional methods rely on expert opinion, which can be subjective.
• Objective, nondestructive methods are needed for accurate saffron grading.

Purpose of the Study:

• To develop and evaluate a machine vision system for objective saffron classification.
• To compare the performance of various machine learning classifiers for saffron grading.
• To determine the accuracy of image analysis techniques in distinguishing saffron quality classes.

Main Methods:

• Acquired 440 color images of saffron (Pushal, Negin, Sargol) using a mobile phone camera.
• Extracted 21 color and 99 textural features using image analysis.
• Employed 22 different machine learning classifiers for classification.

Main Results:

• Support Vector Machine (SVM) and Ensemble classifiers showed superior performance.
• Quadratic SVM and Subspace Discriminant achieved the highest mean classification accuracy of 83.9%.
• Image analysis effectively differentiated saffron classes based on extracted features.

Conclusions:

• Machine vision offers an objective and nondestructive approach to saffron classification.
• The developed system demonstrates high accuracy, potentially improving saffron quality assessment.
• This technology can enhance the reliability and efficiency of saffron grading in real-world applications.