Automated 3D Body Landmark Identification via Population Mesh Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for identifying body landmarks from three-dimensional human body scans are either inaccurate due to reliance on skeletal features not captured by 3D laser/optical scans or require significant human intervention, lacking precision in identifying skeletal features from surface geometry alone.
Innovation Solution
A population-driven method that uses a landmarked population to identify body landmarks by parameterizing individuals through mesh transformation, performing topological body segmentation, and part-wise matching to align and scale three-dimensional part-meshes, employing the iterative-closest-point algorithm to minimize registration error, thus eliminating the need for human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If skeletal features are used to identify body landmarks, then anatomical precision is improved, but 3D laser/optical scans cannot capture these internal features
Solution Approach 1:
The patent uses surface geometry features as an intermediary to represent skeletal landmarks. By establishing statistical relationships between visible surface features and hidden skeletal landmarks through population data, the system can infer skeletal landmark positions from surface scans alone, eliminating the need for direct skeletal feature capture
Solution Approach 2:
The patent creates a virtual model that copies the statistical relationships observed in the landmarked population and applies it to new scans. This virtual population model serves as a template to transfer landmark information from the population to individual subject scans, preserving anatomical precision without requiring actual skeletal feature data
2Measurement precision
If manual marking methods are used to identify body landmarks, then human expertise is utilized, but significant human intervention and time are required
Solution Approach 1:
The patent performs preliminary actions by pre-collecting and processing landmark data from a large population in advance. This landmarked population database is created beforehand, containing statistical relationships between surface geometry and landmark positions. When a new scan is processed, this pre-computed knowledge is rapidly applied through statistical mapping, eliminating the need for time-consuming manual marking while preserving accuracy
Solution Approach 2:
The system enables self-service by allowing the landmark identification process to be performed automatically on new scans using the pre-built population model. The statistical mapping algorithm autonomously identifies landmarks without human intervention, making the system self-sufficient for processing new body scans
3Productivity
If traditional landmarking methods are used, then individual body scans are processed, but accuracy decreases in less salient regions
Solution Approach 1:
The patent creates a universal population model that captures common anatomical patterns across diverse individuals. This single statistical model serves multiple functions: it provides prior knowledge for all landmark locations, constrains the solution space for difficult regions, and ensures consistent accuracy across both prominent and subtle anatomical features. The model transfers this universal knowledge to individual scans through statistical mapping
Solution Approach 2:
The system uses feedback from the landmarked population data to guide the landmarking process. The statistical relationships learned from the population serve as feedback constraints that guide the automatic landmarking algorithm, ensuring that even in less salient regions where local features are ambiguous, the landmark positions remain anatomically plausible based on population-level patterns
Data Source
AI summary
A new method for the identification of body landmarks from three-dimensional (3D) human body scans without human intervention is provided. The method is based on a population in whom landmarks were identified and from whom 3D geometries were obtained. An unmarked body (subject) is landmarked if there is a landmarked body in the population whose geometry is similar to that of the subject. The similarity between the surface geometry of the subject and that of each individual in the population can be determined. A search is performed using the mesh registration technique to find a part-mesh with the least registration error; the landmarks of the best-matched result are then used for the subject.