Automatic Food Leftover Estimation in Tray Box Using Image Segmentation

Abstract

Nutritional food is the most important aspect for people to fulfil their daily needs, but nowadays, the change of lifestyle can affect the style of food consumption, for instance leaving food when they take a meal. Besides, food leftover causes loss of consumed nutrients. We took an observation in a canteen in Faculty of Agricultural Technology in University of Brawijaya produces lost of food about 1,142.5 gram per day. To prevent the amount of food waste, this paper proposes an early stage framework to create an automatic leftover estimation method using image segmentation based on color channel. So that, we can optimize of the amount of food served and avoid loss food nutritions hereinafter. We experimented in two types of background of tray box: black and gray, then an automatic cropping process based on each sub area of tray box is implemented. Each part of them contains an item of food, so we detect the area using color channel segmentation and implements B component in LAB color space. After getting the main area of food, then two areas between segmented food image are compared before and after being consumed. The calculation of the predicted leftover portion is calculated based on original weight of food item. The result shows that tray box with black background with constant (cd) is the best to project leftover, in which reaches 2.37 of Root Mean Square Error (RMSE). It proves that the propose method is sufficient to handle the leftover prediction of food.