Stanford Markerless Motion Capture Project

The present page contains video objects which can be visualized with Quicktime. Download it for free here

Markerless Motion Capture:

The Natural Motion of Human Body is Captured and Analyzed without attaching markers or straps.

Top left: One of the 8 video streams used for the motion capture process. Captured motion courtesy of Kath Boyer, PhD.

Top right: Visual Hull (green) and overlaid matched model (yellow).

Bottom right: The captured motion can be displayed, re-synthesized or re-targeted automatically in 3D without needing further processing.

Introduction

The purpose of human motion capture systems in biomechanics is to measure the motion of bony segments during various activities. Most current motion capture systems rely on cumbersome skin based markers for tracking. The system being developed by the BioMotion Lab uses only high speed video cameras and does not use skin markers, capturing the subject's motion in a natural and easier way.

Algorithm Steps
The following describes the important steps involved in the algorithm:

1. Data Acquisition	2. Reconstruction of 3D Representation

3. Model Matching	4. Kinematic Extraction

In the accurate analysis of human motion, including both kinematics and kinetics variables, the quality of the kinematic models implemented in markerless motion capture plays a major role. A first approach uses machine learning techniques to perform dimensionality reduction of human shape variability and learns the optimal location of the joint centers with respect to the shape mesh. Morphologic and kinematic model are generated.


The SCAPE database of human shapes developed by Anguelov et al. SIGGRAPH 2005 is used to automatically generate the subject specific model used in the tracking process.	The Shape model is integrated with a joint centers model (Corazza et al. ASB 2007) and a subject specific model of the musculoskeletal system is then defined (Opensim1.0) and used for the biomechanical analysis.

The Markerless Motion Capture System developed at the Biomotion Lab provides the full body calculation of joint angles and joint centers. The system is synchronized with force plate input to allow the calculation of forces and moments at the joints. Biomechanical variables are obtained at a frame rate up to 200 Hz.

The markerless motion capture system developed at the Biomotion Lab focuses on biomechanical applications, spanning from clinical gait analysis to sport performances and injury prevention. Below some examples of motion capture of sport activities.

When greater accuracy is needed functional methods are applied for amore accurate estimation of the joint centers location. Below an example for the hip joint is given. The sequence is tracked first and then the joint center location refined up to a sub-voxel accuracy (i.e. error<1cm).

Although the main focus is biomechanical applications, the markerless system developed at the Biomotion Lab can output in standard BVH format to allow the direct animation of virtual characters in 3D rendering engines.

Biomotion Lab, Stanford University.

For Information Contact: stefano.corazza stanford.edu

Publications:

Corazza S, Muendermann L, Chaudhari A, Demattio T, Cobelli C, Andriacchi T: A markerless motion capture system to study musculoskeletal biomechanics: visual hull and simulated annealing approach Annals of Biomedical Engineering, 2006,34(6):1019-29.

Muendermann, L, Corazza, S, Andriacchi, T: The Evolution of methods for the capture of human movement leading to markerless motion capture for biomechanical applications. Journal of NeuroEngineering and Rehabilitation, 3(1), 2006.

Corazza S., Mündermann L., Andriacchi T., A Framework For The Functional Identification Of Joint Centers Using Markerless Motion Capture, Validation For The Hip Joint, Journal of Biomechanics, 2007.

Mündermann L., Corazza S., Andriacchi T., Accurately measuring human movement using articulated ICP with soft-joint constraints and a repository of articulated models, CVPR 2007.

Patent Pending