Video Tracking of Humans in Robotic Environments

Video tracking is the basis for a number of applications in medical procedures, media production, surveillance, gaming and other fields. Profactor is a research company that develops projects in the fields of machine vision and robotics. One of their projects deals with understanding and learning from human behavior for use in robotic environments. This requires recording human actions and splitting them up into individual movements that can be interpreted. Computing the location and state of the object of interest at each point in time, with the use of video input only, is regarded as video tracking. Therefore, video tracking applied to humans is needed. The proposed goal for this project is to provide an assessment of the performance of video tracking programs applied to humans in industrial environments. In this thesis, we have analyzed the tracking algorithms publicly available and have identified their strong and weak points, and which approaches could have potential use in real-life situations. We have searched for video tracking programs that are publicly available. We have analyzed how tracking programs are evaluated, have identified the criteria that can be evaluated and have developed our own set of tests. Using these tests we have designed experiments to evaluate the performance of video tracking software in production environments. We have attempted to perform experiments with the tracking programs available and have identified the restrictions imposed by such software .