Classification of Pepper Seeds by Machine Learning Using Color Filter Array Images

Area of Science:

• Agricultural Science
• Computer Science
• Image Processing
• Machine Learning

Background:

• Penja pepper (Piper nigrum) from Cameroon is a high-value, sought-after spice, difficult to distinguish from other varieties by seed appearance alone.
• Accurate classification of pepper varieties is crucial for market value and quality control, presenting a challenge for traditional identification methods.

Purpose of the Study:

• To develop and evaluate an automated method for classifying Penja pepper seeds using color filter array (CFA) images.
• To differentiate between white and black Penja pepper seeds and other varieties using image processing and supervised machine learning.

Main Methods:

• Collected 5618 pepper seed samples, including Penja white and black varieties.
• Utilized image processing techniques to extract 18 distinct attributes from Color Filter Array (CFA) images.
• Trained and compared four supervised machine learning models, including a Support Vector Machine (SVM).

Main Results:

• The Support Vector Machine (SVM) model demonstrated superior performance in classifying pepper seeds.
• The SVM model achieved a classification accuracy of 0.87, with precision, recall, and F1-score all at approximately 0.874.

Conclusions:

• Image processing combined with machine learning, specifically SVM, provides an effective method for classifying pepper seeds.
• This automated approach can aid in distinguishing valuable pepper varieties like Penja pepper, supporting quality assurance and market integrity.