Scholarlink

<p>Three-dimensional (3D) head scans are crucial for head modelling and fit evaluation, yet obtaining accurate scalp shape is challenging due to hair interfering with optical scanners. This paper presents a novel method for creating digital head models with truthful scalp shapes. We utilized a custom-designed scalp rig to digitize the scalp surface beneath the hair. By integrating these digitized scalp contour points with the original head scans, we compiled a ‘digital bald’ female head dataset (<i>N</i> = 180). This new dataset, in conjunction with data from a U.S. Air Force female scalp surface shape collection initiative, was used to train and assess four statistical shape modelling methods for scalp surface prediction: Multivariant Linear Regression, Principal Components Regression, Partial Least Squares Regression, and Posterior Shape Modelling. The results indicate that all four methods yielded comparable error margins, with a surface distance RMSE difference of approximately 1&#xa0;mm. However, this seemingly minor 1&#xa0;mm RMSE variation had a discernible impact on the fidelity of the predicted scalp shapes.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Female scalp shape acquisition and prediction

  • 期刊论文

  • 发表日期：

    2026年7月10日

查看全文
  • Peng Li,
  • Hyegjoo E. Choi-Rokas,
  • Robin L. Carufel,
  • Todd N. Garlie,
  • K. Blake Mitchell

摘要

Three-dimensional (3D) head scans are crucial for head modelling and fit evaluation, yet obtaining accurate scalp shape is challenging due to hair interfering with optical scanners. This paper presents a novel method for creating digital head models with truthful scalp shapes. We utilized a custom-designed scalp rig to digitize the scalp surface beneath the hair. By integrating these digitized scalp contour points with the original head scans, we compiled a ‘digital bald’ female head dataset (N = 180). This new dataset, in conjunction with data from a U.S. Air Force female scalp surface shape collection initiative, was used to train and assess four statistical shape modelling methods for scalp surface prediction: Multivariant Linear Regression, Principal Components Regression, Partial Least Squares Regression, and Posterior Shape Modelling. The results indicate that all four methods yielded comparable error margins, with a surface distance RMSE difference of approximately 1 mm. However, this seemingly minor 1 mm RMSE variation had a discernible impact on the fidelity of the predicted scalp shapes.