Evaluating expanded-food sensory properties by image analysis

The objective of this study was to ascertain whether sensory properties of extruded corn puff texture can be predicted by digital image analysis. Gradient image statistics, such as third moment and standard deviation, run-length statistics, and features computed from fuzzy edge-detected image were used. Visually-determined sensory properties included cell density, cell wall thickness, cell size uniformity. Crushing force between molars and biting force between incisors were among the hardness-related sensory properties. All the sensory properties could be predicted with a coefficient of determination (r(2)) of 0.89 or higher. In most cases, a polynomial function of an image feature was the best model for predicting sensory scores. Image features derived from the intensity band of a hue, saturation, and intensity color model gave the highest r(2) values for predicting visually-determined sensory properties. For hardness-related sensory properties, the features derived from saturation band resulted in the best r(2) values.