Motion Capture System

From Master Projects
Jump to: navigation, search

has title::Motion Capture System
status: finished
Master: project within::Multimedia
Student name: student name::Andrea Patiño Orrego
Start start date:=2012/05/07
End end date:=2012/11/30
Supervisor: Thilo Kielmann
Second reader: has second reader::Ben van Werkhoven
Company: has company::Personal Space Technologies
Thesis: has thesis::Media:Thesis.pdf
Poster: has poster::Media:Posternaam.pdf

Signature supervisor



Design and implement a real time motion capture system, based on the optical characteristics of the current tracker provided by Ps-Tech. It is, find the 3D position and orientation of a dynamic articulated subject as a human body, which moves constantly during a period of time in front of the tracker's cameras, and provide the related information produced at the time the motion is executed.

Human motion complexity is related to the physics of the body. The body carrying the movement is composed of several limbs that are joined together. Motion is produced when a limb translates and rotates in a particular way, constrained by its adjoins in the entire chain. Tracking this movement implies the definition of a subject structure. Defining these: joint positions and degrees of freedom between articulations, are matter of this project as well as the generation of accurate data in terms of angles and translations throughout a period of time.

The implementation requires each body segment to be recognized by the tracker, otherwise the results will be reconstructed incomplete. Additionally, when a dependency chain of segments exists not only the occluded segment will disappear, but also neighboring segments connected to it; the position and the number of the segments lost in the process, are important issues that determine the success of this task. In this case, the info captured by the system and the info of a predefined skeleton model will be exploited to address the problem of missed limbs. Information gabs have to be avoided. And the captured motion has to be consistent, avoiding abrupt motion changes of the subject in data-frame occurrence.

After configuring and defining the body structure, subject tracking has to be executed in real-time, system outputs have to reflect accurately the subject motion and latencies have to be minimal. The system should be able to partly solve the problem of missing information and perform related computations in tracker frame-rate. The system execution time has to be controlled and off-line procedures during the tracking process are not an option.