PREDICTING FOOD TEXTURE WITH MULTIMODAL TEMPLATE MATCHING

Abstract

Automated cooking systems, customer satisfaction surveys, and food quality certification are just a few of the areas where accurate estimates of texture are needed. This study shows a novel, all-around approach to food texture prediction by combining complex template matching algorithms with information gathered through sight and touch. With the use of force feedback signals, RGB pictures, and depth maps, the system accurately recognizes and classifies textures including crispy, chewy, squishy, and crunchy. By combining real-time sensory input with a carefully chosen library of known texture patterns, template matching is a technique that enables incredibly accurate texture estimations. Several studies have shown that the suggested approach performs better than traditional one-dimensional texture analysis techniques under various illumination, occlusion, and surface contamination conditions. This approach supports human-computer interaction in culinary robotics and smart food processing systems by enabling food to be graded similarly to humans.