Logo site CYBER-II
Home pageResultsWP1. Acquisition and modelling
News items
Results
Wednesday 26 July
We are updating the results
Geometric Modelling of the Virtual Human
Thursday 27 July 2006
Thursday 27 July 2006

Geometric Modelling of the Virtual Human

Technical Report written by Soha Maad

Objective

This report elaborates on the various aspects of the geometric modelling of the virtual human to be incrusted in the CYBER-II augmented scene. The report describes the research and practical work undertaken to fulfill the requirements of the study of the movements of a real human to be incrusted in a virtual world. These requirements were identified in the document accessible here. They involve:

  • The adoption of an articulated skeleton-based model to represent the human in the virtual scene. The model should be compliant with the h-anim standard.
  • The study of the dependencies between the prameters of the model.
  • The identification of a hirerarchy of the movements of the different members of the human skeleton.
  • The estimation of the movements of the human while respecting the dependencies between the parameters of the model and the hierarchical structure of the skeleton representation of the human
  • The exploration of the use of cluster of PCs to respect the real time constraint of the rendering.

This report is divided into five parts. My research work on the geometric modelling of the virtual human initiated with a survey of various practices for modelling a virtual human. This survey is briefly overviewed and summarised in the first part of this report. Following the survey, a particular skeleton based model and a synthetic VRML-based h-anim compliant human models were considered. These models helped in identifying important parameters and dependencies governing virtual human models. The technical work related to these models and the identification of the corresponding parameters and dependecies are described in the second part of this report. The third part of this report describes a technique to estimate human motion based on matching the images of an animated human skeleton (a synthetic human model) with the captured images of a real moving human. Part 4, foresees some applications of the CYBER-II project in particular domains. Part 5 concludes with recommendations and future works related to the geometric modelling of virtual humans to meet the various constraints imposed by the CYBER-II project.

1. Survey of Articulated human models and standards

You can read the results of the survey in the following publication. The following links allow you to read the paper.

“Geometric Modelling of Moving Virtual Humans: A Survey of Various Practices”
Soha Maad, Saida Bouakaz
Journées GTMG - Groupe de Travail en Modélisation Géométrique - 2004

2. Geometric Modelling of a Virtual Human: Some Practical Work

This part allows you to view and download two programs:

  • The first program is a JAVA applet that allows you to manipulate a 3D model of a virtual human created using geometric primitives: spheres and cylinders. The model is compliant with the h-anim standard. The applet gives you insight on the manual manipulation of various parameters of the geometric model of the virtual human. To view the applet click on:
    java applet [*] (Note: your brower must have java enabled, and java3D (available here) must be installed.)
  • The second programs allows you to manipulate a 3D model of a virtual human created using VRML 97. To interact with the VRML model, the latter was imported in a JAVA program using the JAVA3D VRML loader. To download the program click on:
    java applet [*]

3. Estimation of Human Motion Based on Image Matching Techniques

In the context of my work, the estimation of human motion involves the estimation of the parameters for animating a synthetic human skeleton. This estimation is based on the analysis of captured multi-views videos of a real moving human. This analysis relies on an image matching technique and it involves the following steps:

  • Silhouette extraction [1] for the images of the the real moving human
  • Silhouette extraction for the images of the animated synthetic human skeleton
  • Matching [2] the extracted silhouette of the moving human with that of the human skeleton
  • Finding the best parameters that animate the human skeleton according to the human motion

Have a look at the JAVA class documentation for the human motion estimation program by clicking at:
java applet [*]

4. CYBERII Applications

You may be curious to see how far CYBERII applications can go. Well ... have a look at this paper. The following links allow you to read the paper.

“From Virtual to Augmented Reality in Financial Trading: A CYBER-II Application”
Soha Maad, Saida Bouakaz
European Conference on Combinatorial Optimization ECCO XVII- 2004

5. Future Research Work

  • Future research work involves the development of an XML based 3D Moving Human (X3DMH) description. The X3DMH script is not confined solely to the description of the skeleton and geometry of the moving human but includes as well the description of the characteristics of the surrounding scene (light, texture, and material). As such the script would consist of 3 nodes: the Human node conform to the h-anim specification for a standard humanoid
  • The texture node describing the texture applied to the humanoid
  • The environment node describing the lighting of the surrounding scene
  • The X3DMH script would be rendered through a JAVA and JAVA3D based viewer.

[*] Sorry but the Webmaster has lost this applet!

[1] Silhouette extraction involves the substraction of the foreground image of the moving human from the background image which is constant.

[2] The image matching technique uses an optimisation algorithm to estimate the best fit parameters that match the skeleton position to the moving human position.