A comprehensive methodological framework for 3D head anthropometric shape modeling of a small dataset.

Area of Science:

• Anthropometry
• 3D Imaging
• Data Analytics

Background:

• Characterizing occupation-specific 3D head shapes is crucial for designing effective head-mounted devices.
• Limited data and systematic approaches pose challenges in classifying and modeling 3D head shapes for small populations.

Purpose of the Study:

• To propose and evaluate a streamlined six-step framework for analyzing 3D head shapes.
• To compare the performance of different clustering and shape modeling techniques for anthropometric data.

Main Methods:

• Utilized 3D head scans from 36 firefighters.
• Evaluated K-means and K-medoids clustering algorithms.
• Assessed NURBS, NURBS least squares (LS), Cubic Spline, and Cubic Spline LS shape modeling methods.
• Validated predicted head shapes against established anthropometric databases (NIOSH, ANSUR II, CAESAR, US Army).

Main Results:

• K-means clustering demonstrated superior performance over K-medoids, with 28% lower distances.
• Surface mapping-based clustering achieved 35% greater accuracy compared to PCA-based clustering.
• Cubic Spline LS exhibited the lowest mean squared error (0.70 cm²) and fastest computation time (0.14 s), outperforming NURBS LS.

Conclusions:

• The proposed framework, integrating surface mapping, K-means clustering, and Cubic Spline LS, effectively provides accurate 3D head shape models.
• This approach enhances the characterization of occupation-specific head shapes, particularly for small population groups.
• The findings support the development of better-fitted head-mounted devices for diverse occupational groups.