Machine vision study demonstrates high-accuracy detection of water adulteration in pomegranate juice, indicating potential for automated real-time food authentication.
Key Points
To develop a rapid, non-destructive machine vision and deep learning framework for detecting added water adulteration in pomegranate juice using multi-color space image features.
Captured visible images from 625 pomegranate juice samples across five water adulteration levels under controlled conditions, expanding the dataset 16-fold using data augmentation.
Extracted color and textural features across 19 color channels to train an enhanced 1D convolutional neural network comprising Conv1D, MaxPooling, BatchNormalization, and Dense layers alongside baseline machine learning models.
The enhanced convolutional neural network achieved an overall classification accuracy of 99.45%, with precision, recall, and F1-scores approaching 1.00.
The deep learning architecture outperformed standard classification models, including multilayer perceptron (87.75%), random forest (92.45%), and support vector machine (96.95%).