| Home | Teaching | Research | Publications |
|
|
Weikersdorfer, David
Researcher
|
Table of Contents |
After receiving my Diploma in Mathematics from the Ludwig-Maximilians-Universität in München, I worked 1.5 years as freelancer and as consultant for a major IT consulting company. Since 2009 I am a PhD student at the IAS group under the supervision of Prof. Michael Beetz. My research interests are human motion tracking and analysis, with an emphasis on markerless visual tracking of hand motions in manipulation acitivities. In my sparse free time I do ballroom dance sport.
My office is located at the 3rd floor in the CoTeSys Central Robotics Lab at Karlstr. 45, 80333 München.
Research Topics
- Markerless human motion tracking: full body and hand
- Joint tracking of the human hand and manipulated objects
- Graphic card accelerated scientific computation using CUDA and OpenCL
- Character physics and cloth and soft bodies simulations in human/object tracking
- Applications of human motion tracking in sports
- Motion segementation and motion primitives
Current research projects
Human Activity Observer: Visually Tracking Hand and Objects in Manipulation Activities
HumActO stands for "Human Activity Observer" and is an end-to-end framework which visually tracks the human hand and several manipulated rigid objects. We are using image streams from three high definition cameras which are segmented using color histograms. To describe the complex articulations of the humand hand one needs at least 26 degrees of freedom. Additionally the state space has to provide six degrees of freedom for every rigid object. In our approach this high dimensional state space is solved using model based particle filtering. We are using pre-built CAD models of hand hand and objects. The hand model can be adapted to a specific antropometry by scaling bones individually.
HumActO was developed in the IAS Group at the Technische Universitaet Muenchen.
Markerless tracking of the full human body in athletic sports
- Floor Exercises (Man) (AVI 4,6 MB)
- Floor Exercises (Woman) (AVI 4,0 MB)
- Shotput (Man) (AVI 3,4 MB)
Tracking was done with the MeMoMan tracker developed at the IAS Group at the Technische Universitaet Muenchen in a collaboration with the institute of sport sciene at the Universität Augsburg.
Realtime tracking of voxel models of humans and objects
Under construction
Tracking of elderly peoply to analyse movement restrictions
Under construction
Student Projects
I can offer interesting student projects and Bachelor/Master/Diploma thesis topics for motivated students in the area of human motion and object tracking.
Open student projects:- Topic: Biomechanics in human motion tracking (with a possible application in sports)
- Topic: Tracking of soft bodies and liquids (possible using graphic card acceleration)
- Topic: Modeling physical interactions between hand and objects (grasp points and collision detection)
- Topic: GPU accelerated image segmentation and voxelization
- Topic: GPU accelerated background subtraction
- Realtime 3D pose estimation of the human hand to control a media player
- Using graphic card acceleration (CUDA) to improve a human motion tracker
- Accurate tracking of the human hand using a particle based tracker
- Automatic initialization of body dimensions for a human model used in particle filter based tracking
- Enhancing a particle filter based tracker by accurate head tracking
- Visualizing character and object animations (SEP project)
Former Research Projects
Automatische Bildanalyse in der zerstörungsfreien Werkstoffprüfung
Werkstudent 2005-2006
Corporate Technology: Power & Sensor Systems, Siemens AG
www.siemens.com
Projekt "ViGWaM" (Vision Guided Virtual Walking Machine)
Werkstudent 2002-2004
Lehrstuhl für Steuerungs und Regelungstechnik, Technische Universität München
www.lsr.ei.tum.de/team/vigwam
Towards practical Quantum Cryptography
Hilfswissenschaftler 2000/2001
Experimental Quantum Physics, Ludwig-Maximilians-Universität München
www.dpg-tagungen.de, xqp.physik.uni-muenchen.de
