| dc.description.abstract | The capturing of human motion plays a very important role in several computer vision applications, such as virtual reality, automatic control and posture recognition. In some game designation sites, designers need to capture human motions to obtain all actions of roles in the game. When directors want to film a 3D-animated movie, he might also use motion capture frequently. However, the equipments of motion capture are very expensive. Moreover, it is not easy to use these equipments. The actors must wear some devices, which results in the inconveniently of their actions. If there are some inexpensive and easy technologies to be employed to capture human motions, they can be soon applied in the real world.
Researches of motion capture had been developed many years. They can be categories into four main groups, which are (1)magnetic motion capture, (2)ultrasonic motion capture, (3)mechanical motion capture, and (4)optical motion capture. Recently, some researchers in investigate image-based motion capture system. However, these methods are usually time-consuming. In this thesis, we develop a system that captures human motions from video sequences. Actors’ head and limbs are tracked and located directly from video sequences. Another characteristic of the proposed system is that it only uses one camera to capture the video. In the way, the cost can be reduced computing with other systems. Experimental results demonstrate the feasibility and validity of the proposed system. | en_US |