HBM: Homologous Body Modeling
The Digital Human Research Center developed software for creating a homologous human body model. In previous homologous models of the human body, data points have been defined based on landmarks by defining cross sections and dividing points on each cross section. This method is inconvenient because different software is needed for creating homologous models for different parts of the body. In addition, homologous models created this way are rough, and although sufficient information is present for analyzing shape variations, overall appearance is poor. Examples of such models are shown in Figure 1. If generalized software that can create different homologous models for different parts of the human body were available, wider utilization of 3D human body data for product design would be possible.
Together with Professor Hiromasa Suzuki, Faculty of Engineering, University of Tokyo, we developed software for creating a detailed homologous model using the subdivision surface method.
A broken line can be interpolated into a smooth line by using specific rules (Figure 2). The same applies for a surface consisting of polygons. A polygon surface can be subdivided into a smooth surface using the same rules (Figure 3). However, the smoothed surface will be slightly smaller than the original surface consisting of polygons.
Homologous modeling by HBM
In order to create a detailed homologous model using HBM, a rough generic model and locations of landmarks for the generic model are needed. The easiest method to create a generic model is decimating scanned data. In the following example, we used a rough model created by specialized software.
In the next step, scanned data and landmark locations for the scanned data are required. By using HBM, the generic model is subdivided to fit into the scanned data. Landmark locations of the generic model will be strongly fit into the corresponding landmark locations of the scanned data. Figure 4 shows this process. Initially, the generic model differs from the scanned data (Figure 4, left). After 4 steps of subdivision, the generic model fits well into the scanned data, and the model becomes more detailed, but the fit is poor at the toes (Figure 4, middle). After 16 steps of subdivision, the difference between the subdivided generic model and the scanned data is very small (Figure 4, right).
Use of HBM
When a generic model is created by decimating scanned data, the created generic model usually has sufficient detail because the scanned data is very dense. Therefore, numerous subdivisions of the generic model are not required. When the number of data points is very large, created models may be problematic.
This software does not fit the generic model to scanned data with different joint angles from the generic data. To fit a generic model into whole body data, for example, additional software is necessary to make the joint angles comparable. Such software is now under development.
Statistical analysis of detailed homologous models
We developed additional software for analyzing 3D body shapes, known as HBS. If the purpose of statistical analysis is to examine variations, both MDS and PCA versions of HBS are available. However, if the purpose is to create different body shapes that represent a population, the PCA version is suitable. Virtual shapes may not be calculated by the MDS version when the number of data points is too large, or if the shape is too complex.
Specifications of HBM
- Hardware requirements: Microsoft Windows 2000/XP
- Software library, no GUI available
- Input: scanned point cloud (text format) and landmark coordinates (text format) of a specific person, generic model (geo or obj format)
- Output: homologous model of the subject (geo or obj format)
- Function: Generate a homologous model of a subject automatically by fitting a generic model into the scanned data.
HBM software is available through Digital Human Technology, an AIST venture company.
- Tomohiro Inagaki, Yuki Asada, Joe Kuragano, Hiromasa Suzuki, Masaaki Mochimaru and Makiko Kouchi, 2004: Surface reconstruction of human body from point cloud with anatomical features and extraction of measure of human body. Proceedings of Annual Meeting of The Japan Society for Precise Engineering 2004, 271-272. (In Japanese with English abstract)
- Brett Allen, Brian Curless, and Zoran Popovic, 2003: The space of human body shapes: reconstruction and parameterization from range scans. SIGGRAPH 2003.