Image-Based Food Quantity Estimation Using Body-Part Calibration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image-based food quantity estimation techniques require specific reference objects or complete visibility of body parts, lack personalization, and are not robust under varying lighting and background conditions, compromising usability and accuracy.
Innovation Solution
An apparatus and method that uses a user's body part, such as a finger, to estimate food quantity by detecting physical attributes like length or width, and associates these attributes with a user profile for personalized estimation, adaptable to different conditions and environments.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a specific reference object (e.g., 3D marker with patterns) is used to estimate food quantity, then measurement precision is improved, but device complexity and ease of operation deteriorate due to requiring specific reference objects that may be difficult for users to use correctly
Solution Approach 1:
The system uses the user's own body parts (fingers, hands) as reference objects instead of requiring external reference objects. The user's body parts are always available and eliminate the need to carry or set up specific reference objects, making the system self-service and highly convenient while maintaining measurement precision through personalized anthropometric calibration
Solution Approach 2:
The system can work with multiple different body parts (fingers, hands, palms) and multiple types of food items without requiring different reference objects for each case. A single calibration process using any body part can be used universally across various food estimation scenarios, eliminating the need for multiple specialized reference objects
2Adaptability or versatility
If anthropometric parameters are used for food quantity estimation, then personalization is improved, but device complexity increases due to requiring post-processing calibration and specific reference points
Solution Approach 1:
The system performs anthropometric calibration in advance by capturing images of the user's body parts and computing personalized dimension parameters before actual food estimation. This preliminary calibration stores the user's specific anthropometric data, eliminating the need for complex post-processing during food estimation and simplifying the operational workflow
Solution Approach 2:
The system extracts and isolates specific anatomical landmarks and body part dimensions from images to create personalized reference data. By separating the calibration process into a distinct preliminary stage that extracts anthropometric parameters, the system avoids complexity in the main food estimation process while maintaining personalization
3Measurement precision
If color features are extracted from images for body part detection, then measurement precision is improved, but reliability deteriorates under varying lighting conditions
Solution Approach 1:
The system transforms the detection approach from relying on color features (which vary with lighting) to using geometric and spatial parameters such as body part shapes, contours, and anatomical landmark positions. These geometric parameters remain stable under varying lighting conditions while still enabling precise identification and measurement of body parts for anthropometric analysis
4Measurement precision
If complete visibility of reference objects is required, then measurement precision is improved, but ease of operation deteriorates due to strict positioning requirements
Solution Approach 1:
The system only requires partial visibility of body parts to perform anthropometric measurements by identifying key anatomical landmarks rather than requiring complete visibility of entire body parts. This partial observation approach maintains measurement precision by focusing on critical reference points while significantly increasing positioning flexibility and ease of operation
Solution Approach 2:
The system divides body parts into segments and identifies specific anatomical landmarks (such as finger joints, knuckles, palm edges) rather than treating body parts as whole objects. This segmentation allows precise measurement using only visible portions of body parts, eliminating the need for complete visibility while maintaining accuracy through landmark-based detection
Data Source
Figure 1
Figure 2
Figure 3
AI summary
There is provided a computer-implemented method for performing image-based food quantity estimation. The method comprises acquiring (202) a first image, wherein the first image depicts a food item and a body part of a first user; detecting (204), based on the acquired first image, a first physical attribute of the body part of the first user; identifying (206), based on the acquired first image, the depicted food item; and estimating (208) a quantity of the food item depicted in the acquired first image based on the identified food item and the detected first physical attribute of the body part of the user.