Automatic portion estimation and visual refinement in mobile dietary assessment
Abstract
A challenging problem for dietary assessment and computer engineering is determining food portion size and nutrient information from a photograph taken with a mobile digital camera (e.g., mobile phone camera). This thesis describes an end-to-end software system for automatically calculating portion volumes and nutrients for specified foods from single images. Details of image acquisition, camera parameter estimation, model reconstruction, and interaction techniques are described. Specific contributions of this thesis include theory and evaluation of spherical and prismatic food models, employing error visualization techniques to provide feedback in model adjustment. Performance constraints inherent to mobile hardware are accounted for. The system is thoroughly evaluated for accuracy using real and simulated images. Drawbacks of said models and techniques are also discussed, along with proposed extensions.
Degree
M.S.E.C.E.
Advisors
Ebert, Purdue University.
Subject Area
Electrical engineering|Computer science
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.