A Non-Contact Computer Vision Based Analysis of Color in Foods

Abstract

Since commercial colorimeters measure small area with a fixed geometry, the result of color measurement is usually unrepresentative for heterogeneous materials as in many food items. This paper describes a computer vision based approach for the measurement of color in a user defined polygonal area on the digital image of a food product. The algorithm used for color measurement converts the RGB values of the image captured by a digital camera to monitor L*a*b* values using the standard equations. The RGB responses for a captured image vary from one case to another, so, the direct transformation from RGB to L*a*b is not useful to obtain meaningful information about the color. Here, an artificial neural network (ANN) model was used to convert the monitor L*a*b* values into spectrophotometric L*a*b* values. The ANN model was calibrated by using the IT8 color chart consisting of 288 different colored squares which reflect all possible variations in the color space. The ?E values for the estimated values and the real spectrophotometric values were less than 0.45.