Intelligent Grading of Green Cardamom Using Data Fusion of Electronic Nose and Computer Vision Methods.

Area of Science:

• Agricultural Science
• Sensory Science
• Data Science

Background:

• Green cardamom quality grading traditionally relies on subjective human assessment.
• Objective and non-destructive methods are needed for consistent and efficient quality control.

Purpose of the Study:

• To develop an intelligent system for grading green cardamom quality.
• To evaluate the effectiveness of electronic nose (e-nose) and computer vision (CV) combined with machine learning (ML) for quality assessment.

Main Methods:

• Cardamom capsules and seeds were analyzed using e-nose and CV techniques.
• Three ML algorithms (Decision Tree, Bayesian Network, Support Vector Machine) were employed for classification.
• Correlation-based feature selection (CFS) was used to optimize feature input.

Main Results:

• The fusion of e-nose and CV data with the CFS-Bayesian Network (BN) model achieved 100% accuracy in grading.
• CFS-BN model showed high accuracy (96.67%) for visual grading of capsules and seeds.
• CFS-Decision Tree (DT) model performed well for e-nose data classification (93.33% accuracy).

Conclusions:

• Data fusion of e-nose and CV offers an intelligent, accurate, and non-destructive approach for cardamom quality grading.
• This integrated system provides a reliable and fast method for assessing both cardamom capsules and seeds.
• The developed system has the potential to significantly improve quality control in the cardamom industry.