Developing an optical motion capture system for ethological researchment

OData support
Supervisor:
Dr. Umenhoffer Tamás
Department of Control Engineering and Information Technology

The motion of humans, or other living beings can be recorded with motion capture systems. The recorded data can be used to animate a virtual character. Motion capture systems are mainly used in the movie, and video game industry. However motion capture systems are also used in other areas, for example for gait analysis.

Furthermore in some ethological observation the participants' precise position and orientation are required to be known. In this scenario a motion capture system can also be used. Further on the participants' pose can be automatically recognized.

There are many methods used by motion capture systems. In this paper I present the development of an active optical motion capture system. In the paper the following topics are covered in detail.

The performer wears homogenous, illuminating markers attached to each relevant joint. The performer movement is captured by six high speed cameras in order to determine the markers' position in the cameras' image. The cameras operate at 60 frame per second, which generates a lot of data. The image processing is distributed among multiple computers. Luckily most of the image processing task is highly parallel, so the high parallel computational capacity of modern programmable GPUs can be utilized.

The main computer calculates the markers' 3D positions after it receives the 2D data from image processing. A hierarchy can be defined between the points in the point cloud, which can be used to determine the bones hierarchical transformation chain in the character's skeleton.

To achieve satisfying results precise calibration is required. During calibration the cameras' position and orientation are determined. Good calibration is also crucial for accurate tracked point identification and labeling.

I also developed a graphical user interface program, which can be used to access the underlying functions.

Downloads

Please sign in to download the files of this thesis.