Vol. 12, Issue 11 (2023)

Author(s):
Neeraj Tiwari, Deepoo Meena and Ravi Prakash Pandey

Abstract:
Evaluating the quality of food and spices is essential to ensure human nutrition. Non-destructive methods, such as computer vision, have been explored by researchers for assessing the quality of food and spices. This research specifically highlights the importance of quality assessment for turmeric, given its nutritional value and vulnerability to fraudulent activities. The affordability of low-quality rice powder makes it an enticing option for adulterating turmeric powder due to its lower market price. To address this issue, the study utilizes an enhanced convolutional neural network (CNN) for classifying turmeric powder images and detecting fraud. A dataset comprising 3000 image samples is divided into six categories, representing pure turmeric powder and varying levels of adulteration with rice flour (10%, 15%, 20%, 25%, and 34%). The primary objective is to improve fraud detection capabilities, thereby safeguarding the integrity of turmeric in the market. In the initial image processing stage, unwanted components are removed. The incorporation of data augmentation (DA) proves crucial to address overfitting concerns in the CNN. Specifically, the MobileNet-v2 model architecture is employed for the classification task. In implementing this deep learning approach, the image dataset is randomly divided into two main sets: 90% for training-validation in the CNN and 10% designated for a blind test. The refined model demonstrates an impressive 90% accuracy during the validation phase, with a minimal 7% misclassification rate observed in blind testing. This highlights the effectiveness of the method as a quality and safety control measure for the turmeric industry. The study's findings underscore the potential of computer vision, especially in combination with deep learning (DL), as a valuable tool for assessing quality and exposing fraudulent practices in turmeric powder.