Development of a vision-based feature extraction for food intake estimation for a robotic assistive eating device

Our research aims at making a mock-up of a multi grip tool for a robotic assistive device and a camera system which enable frail elderly to live more independently and to keep track of their food intake. In this research, the development of a vision-based feature extraction for food intake report for a robotic assistive eating device is introduced. The proposed vision system is composed by a RGB-D camera and the eating aid device Bestic. In this paper, the authors proposed an algorithm for estimating the amount of intake food (e.g. candies) after the end of the meal. The proposed vision-based feature extraction was designed and implemented in order to take pictures of the plate during the meal as well as estimate the amount of intake food. A set experiments were carried out in order to verify the performance of the proposed vision system while estimating the amount of intake food with different amount of multi-coloured candies as well as different light conditions. Based on the experimental results, we could verify the system performance is not affected by the amount of candies as well as the light conditions.