A Vision-Based Target Localization System for the Meal Assistance Robot

As the most important kind of daily activities, the dietary activity including moving cutlery is still a hard challenge for the elderly and disabled with upper limbs weakness and most of them are suffering from it. Therefore, research on meal assistance robots (MAR) become more imperative. This paper studies a vision-based meal assistance robot system, which solves the problem of the low success rate of traditional MARs. According to the problem of choosing scooping points, we calculate food points with Gaussian distribution density and then estimate the weights of scooping points. Meanwhile, a facial feature point model is referred in our structure and help our algorithm local and orient the position and angles of mouth, which solve the problem of scavenging food. To verify our algorithm, we build a prototype and present its effect in experiment section.