Back to search

University of Alabama Birmingham

Increasing Accuracy of Dietary Assessments by Regular-Shape Recognition and Photogrammetry

Abstract

dc:description.abstract

There is interest in determining the amount of food consumed during dietary studies using technology. Interviews of patients and/or participants can result in inaccuracies in reported food intake. Using food journals or diaries can be more accurate, but depends on participants' motivation to record food intake during meals. Technology can make it easier to record food intake. Previous work utilizing technology has included using single images and a virtual model for food estimation, projecting a circular light on the image to use as a reference, and using a device with a laser to determine distance. Another method uses a wearable device that takes continuous images of meal settings. Previous work utilized image processing and photogrammetry to improve the accuracy of portion size estimation. Brasher's method was successful for many food shapes, but does not work well for some food shapes, e.g. box-shaped food items. Brasher's work showed errors less than 20% for several food items including a pile of corn, a pile of green beans, and a pile of mashed potatoes. Box-shaped food items, such as chocolate cake, were troublesome, resulting in errors above 19%. This dissertation investigated methods of improving Brasher's algorithm when estimating volume of box-shaped food items by applying regular-shape recognition. Smoothing edge pixels improved the error in volume estimation of box-shaped food items, resulting in errors below 12%. Other algorithms developed that use the regular-shape algorithm also showed improvement in the volume estimation. Finding the corners of box-shaped food items using the regular-shape recognition algorithm resulted in errors below 15%. Using an image distance to world distance unit conversion to approximate food volume of box-shaped food items resulted in errors less than 7%. Other methods explored to reduce errors included an algorithm to approximate the plate height, which resulted in plate height estimation errors less than 11%. The accuracy of the computed camera position can play a role in the food volume estimation. A method for improving the accuracy of the computed camera position was explored.

Degree

thesis:*
Level thesis:degree_level
Dissertation

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Hakima, Ihsan K
Contributors dc:contributor
  • Gregg L Vaughn
  • Thomas C Jannett
  • Dalton S Nelson
  • James M Shikany
  • B Earl Wells

Subjects

dc:subject × 6

Identifiers

dc:identifier.*

Chain of custody

source
Harvested from
University of Alabama Birmingham
Base URL
digitalcommons.library.uab.edu/do/oai/
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Hakima, Ihsan K. Increasing Accuracy of Dietary Assessments by Regular-Shape Recognition and Photogrammetry. Dissertation thesis, https://digitalcommons.library.uab.edu/etd-collection/1840