Modeling The Level of Aflatoxin Contamination in Cocoa Beans Based on Color and Texture Features

Abstrak

This research aimed to obtain the best model for predicting the aflatoxin contamination level in cocoa beans by analyzing the relationship between color-textural features of contaminated cocoa bean images and the corresponding aflatoxin level. The cocoa beans were inoculated with Aspergillus flavus (6ml/100g) as the source of aflatoxin and incubated for 7 days. Aflatoxin levels were measured and cocoa bean images were taken on 1-7 days. The quantification of the beans was carried out using LCMS, while the images were taken with a Canon camera installed on the image acquisition unit. The images were divided into 3 classes, namely aflatoxin-free cocoa beans, contaminated below the threshold, and contaminated above the threshold based on LCMS results. A Visual Basic-based software was used to extract 289 color-texture features from the images. Each feature was compared to obtain the best model using linear, power, polynomial, and logarithmic models.. Based on the extraction of color-texture features followed by regression modeling, the best model for predicting the relationship between the aflatoxin contamination level and color-texture features is a polynomial model using the CMY_Y mean feature, with 0.93 a coefficient of determination.

Kata kunci

• Aflatoksin
• Biji kakao
• Fitur Warna
• Fitur tekstur
• Pemodelan