Motion reconstruction of somersaults using three dimensional model matching method from its silhouette image

Abstract

This study, using computer graphics, explores a movement reconstruction method of human somersaults. While human subjects perform on a trampoline, the researcher converts the action into information with a human model that has sixteen links and twelve joints under thirty-one-degrees-of-freedom. Performance is photographed with an infrared camera, and the captured images are turned into silhouette pictures through binary image processing. The silhouettes determine each joint angle, while modifying a model generated earlier. The continuous changes of joint angles during a flip are also calculated. All the data are finalized with animation.