Performance Metrics on Hyperspectral Images in Fuzzy Contextual Convolutional Neural Network for Food Quality Analyzer

Abstract

The quality of food and the safety of consumer is one of the major essential things in our day-to-day life. To ensure the quality of foods through their various attributes, different types of methods have been introduced. In this proposed method, three underlying blocks namely Hyperspectral Food Image Context Extractor (HFICE), Hyperspectral Context Fuzzy Classifier (HCFC) and CNN for Food Quality Analyzer (CFQA). Hyperspectral Food Image Context Extractor module is used as the preprocess to get food attributes such as texture, color, size, shape and molecular particulars. Hyperspectral Context Fuzzy Classifier module identifies a particular part of the food (zone entity) is whether carbohydrate, fat, protein, water or unusable core. CNN for Food Quality Analyzer module uses a Tuned Convolutional layer, Heuristic Activation Operation, Parallel Element Merge Layer and a regular Fully Connected Layer. Indian Pines, Salinas and Pavia are the benchmark dataset to evaluate hyperspectral image-based machine learning procedures. These datasets are used along with a dedicated Chicken meat HSI dataset is used in the training and testing process. Results are obtained that about 7.86% of average values in various essential evaluation metrics such as performance metrics such as accuracy, precision, sensitivity and specificity have improved when compared to existing state of the art results.