Student Name: Misbah Ur Rehman
Research Topic: Design of a Probabilistic Marker-less Motion Capture System for Bio-centered Applications
Research Interests: Machine Vision
Supervisor: Dr. Mohsin Rizwan
Current Progress: In-process
Description
The vision-based tracking of humans is desirable in many scenarios including character motion capture, pose recognition, sports bio-mechanic applications, clinical gait analysis and human-computer interaction. Currently there are two main techniques that are used to capture human motion data – Marker-based and marker-less.
For the marker-less techniques there is generally a lack of information with-in the images to reconstruct the complete set of body kinematic links, including self-occlusion. The complexity of the human body itself plays a role with its many degrees of freedom, and the non-linear and non-rigid nature of human gait can induce several problems
The proposed research tends to accomplish the task of human motion capture using simple 2D cameras based on some a-priori information about the human body and defining a probabilistic approach for the location of joints and inference of joint angles. The tracking is done using particle filters. The advantages include no specific requirements on the environment or special markers on the target. The results will be compared with a marker-based system from Qualisys