Prediction of Aflatoxin Contamination in Cocoa Beans Using UV-Fluorescence Imaging and Artificial Neural Networks for Enhanced Detection

This study introduces an innovative approach to predicting aflatoxin contamination levels in cocoa beans by leveraging an optimized Artificial Neural Network (ANN) model coupled with UV-fluorescence imaging. Aspergillus flavus-inoculated cocoa beans underwent a 7-day incubation period, and UV lamp-based image acquisition facilitated data collection. Leveraging 289 color-texture features, the developed ANN model exhibited highly promising predictive capabilities. Validation results indicate a remarkably low Mean Square Error (MSE) of 0.0087 and a high R-value of 0.9910, affirming the efficacy of the proposed model. The seamless integration of UV-fluorescence imaging and ANN presents a viable and accurate alternative for detecting aflatoxin in cocoa beans, thereby enhancing food safety practices within the cocoa industry.