Child body shape measurement using depth cameras and a statistical body shape model.

Area of Science:

• Biometrics
• Computer Vision
• Anthropometry

Background:

• Accurate child body shape measurement is crucial for health and apparel.
• Existing methods often require specialized equipment or direct measurements, limiting accessibility.
• Low-cost depth cameras offer a potential solution but generate noisy, incomplete data.

Purpose of the Study:

• To develop a rapid and accurate method for measuring child body shapes from low-cost depth camera data.
• To reconstruct complete 3D body avatars and predict standard anthropometric data.
• To overcome challenges posed by noisy and incomplete depth sensor measurements.

Main Methods:

• Utilized a statistical body shape model (SBSM) developed via principal component (PC) analysis.
• Formulated an optimization problem to fit the SBSM to noisy depth camera data in PC-score space.
• Validated the method by comparing reconstructed shapes and predicted anthropometric data against high-resolution laser scans and manual measurements.

Main Results:

• Achieved a mean unsigned distance of 9.4 mm (SD 5.1 mm) between the depth-camera-fitted model and laser scan data.
• Demonstrated high accuracy for torso measurements, with a mean distance of 2.9 mm (SD 1.4 mm).
• Obtained high correlations (>0.9) between predicted and manually measured standard anthropometric dimensions.

Conclusions:

• The proposed method enables rapid and accurate child body shape measurement using affordable depth cameras.
• The SBSM effectively reconstructs complete body avatars and predicts key anthropometric data from incomplete sensor inputs.
• This technique holds promise for applications in pediatric health monitoring, custom apparel fitting, and ergonomic design.