Figure animation through human motion tracking

Abstract

The human body is a unique combination of elegance and power which lends itself to performing a wide variety of tasks. The uniqueness of movement of humans has always been hard to reproduce. Attempts at reproducing the human body in animations have always been limited to inaccurate descriptions due the complexities this task represents. To represent the exact motions of the human body animators started using techniques such as rotoscoping which enabled them to design their animations using underlying images of humans in motion. The great advances made in technology have opened the way for investigation of techniques which can automatically reproduce the motion of a human actor in a scene. This creates a new approach for animators who are now able to produce animations which are more realistic, with applications varying from virtual realities to movies. This dissertation aims to develop a system which tackles this automatic reproduction of human motion. The motion of an actor in a confined space is analysed and reproduced through a number of techniques, such as feature detection and tracking. These techniques are used to identify and track both the human as a whole and certain body parts which enable the building of a skeleton model describing the position of the human actor's body.