MOCAP Motion Capture

Abstract

The Human Motion Capture or MOCAP are terms used to describe the capturing of actual human motion and then translating it on to CG models. In simple words, Human Motion is captured through multiple cameras in a controlled environment and then this very motion is transferred onto 3D models. The world of Human Motion Detection has progressed and done wonders. The systems available in the market are complex and very expensive as they are dependent on hardware requirement. Most of these systems range into the tens of thousands of dollars. Our aim is not to reinvent the wheel, rather to create a cheaper system, which laid emphasis on the systems software to compensate for the hefty hardware requirements. The Project is divided in to two parts, i) Motion Detection ii) 3d Scene Reconstruction. For motion detection, we used two to three cameras, and tracked the motion of an actor, in a controlled environment: The actor is to wear white markers on black apparel in front of a black background. This was achieved using the techniques of neighborhood matching and region growing. The reconstruction process is subdivided into two parts: first, calibration of cameras being used; second, computing the points in space that project to the tracked image points. Calibration is done by using a chessboard pattern, corner identification, square count and then relating the real world units to the computer system units. The tracked sequence of points is used to reconstruct corresponding set of points in 3d space representing the object surfaces on the scene using method of stereoscopic reconstruction and linear triangulation.