|M.Sc Student||Pekelny Yuri|
|Subject||Articulated Object Reconstruction and Motion Capture from|
|Department||Department of Computer Science||Supervisor||Professor Chaim Craig Gotsman|
|Full Thesis text|
The aim of this work is to present an algorithm for acquiring the 3D surface geometry and motion of a dynamic piecewise-rigid object using a single depth video camera. The algorithm accumulates a separate point cloud for each rigid part of the scanned object. It begins from the point cloud of the first input frame and for each new frame adds points from the regions that were obscured to the camera in the previous frames. The more different regions are sampled from different viewpoints by the camera the more complete the econstructed model becomes. In the end of the scanning process, each acquired point cloud is converted into a separate surface mesh. All rigid part meshes together construct the final 3D model.
The algorithm also reconstructs the dynamic skeleton of the object. The acquired model geometry and skeleton can be imported into a modeling system, in order to create new poses and animations of the reconstructed object.
During the reconstruction process the algorithm estimates the motion of each rigid part of the scanned object over frames, thus it can also be used for markerless motion capture. After the reconstruction process, the captured motion can be applied for the reconstructed 3D model and therefore can be observed from any viewpoint. The captured motion can also be retargeted to another 3D object skeleton, in order to animate a synthetic model based upon a motion which is captured from a real world object or actor.