Akademik Veri Yönetim Sistemi
FOOD CONTROL, cilt.156, 2024 (SCI-Expanded)
Aflatoxin (AF) contamination of food poses a serious threat to humans and animals. Many crops, including figs, are susceptible to aflatoxin contamination. In fig production facilities, workers manually sort aflatoxin-contaminated figs using the Bright Greenish Yellow Fluorescent (BGYF) method under ultraviolet (UV) light. This manual sorting process requires expertise, and long-term exposure to UV radiation may pose health risks. This study aims to improve the safety of fig production facilities by reducing the potential harm to human health by investigating a method that is more effective than the BGYF method for the automated detection of aflatoxin-contaminated figs. The correlation between aflatoxin and BGYF was investigated, and the detection rate of aflatoxin via BGYF was found to be 73%. We utilized an optical system with light sources of various wavelengths (250-1000 nm) to generate datasets from fig images. Feature vectors from these datasets were also extracted by applying transfer learning. Using SVM to classify feature vectors extracted from images taken under a 365 nm light source using the MobileNetV2, ResNet101V2 and InceptionResNetV2 models, we achieved 100% detection accuracy for contaminated figs and 92.3% for uncontaminated figs. Our proposed method yielded an overall classification accuracy of 96%, indicating that deep learning extracted features can be used for rapid, automatic, and effective detection of aflatoxin-contaminated figs.