A Slice-based Method for Food Volume Estimation

Abstract

Researches on self-dietary assessment have aroused interest in scientists in the field of food and health. User self-reports have long been a traditional way for individual dietary assessment. However, the consumed weight or volume from individuals is highly dependent on their subjective judgment and may lead to biased dietary analysis results. Therefore, direct estimation of food volume tends to be a more appropriate way for self-dietary assessment. 3D measurement is an effective method for object volume estimation. At present, there are some good strategies based on 3D measurement for estimation of the object volume, such as virtual reality based method and convex hull-based method. However, due to complexity and diversity of the surface of food items, traditional methods for object volume estimation often fail to achieve good results in food volume estimation. Therefore, a slice-based method is proposed in this paper, which is a calculus-based strategy that obtains an estimation of food volume via segmentation and integration of point cloud. Slice-based method is aimed to solve the difficulties caused by diverse food shapes and complex food surfaces in volume estimation and can provide a more accurate volume estimation for subsequent nutrient intake analysis.