3D Modeling of Soccer Players from an Arbitrary Viewpoint Using Multi-view video images

JebalBarezi Sarbijan, Elham | 2012

381 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 43605 (19)
  4. University: Sharif University of Technology
  5. Department: Computer Engineering
  6. Advisor(s): Kasaei, Shohreh
  7. Abstract:
  8. For penalty detection and interpretation of sport match scenes (such as soccer), it is sometimes necessary to watch the scene from an arbitrary viewpoint captured by a virtual camera. In this research, we focus on 3D reconstruction of soccer players using videos that are captured from relatively far distance to the soccer field. In capturing videos from sport scenes, such as soccer scenes, an important issue is the problem of capturing videos from a wide environment to cover almost all parts of the field. Using many cameras with overlapped views might be an answer to this problem but it is impractical due to its cost for capturing and transmitting the data. Furthrmore, because of capturing videos from a relatively far viewpoint and having an open environment with uncontrolled lighting and environmental conditions, the quality of input images is not high either. In addition, the body of players has a non-rigid and moving shape with small size and low quality. The overlap of payers’ bodies and the shadow problems make the process even more difficult. These serious challenges make most of current algorithms to be inapplicable for 3D reconstruction of sport scenes. Most of existing algorithms apply some constraints such as having many cameras and using videos that are captured by HD quality cameras. In this research, our proposed method is based on using the prior knowledge on the structure and movement of human body across time to overcome the problems of 3D reconstruction while using only a few cameras with small size and low quality videos. Body pose estimation and tracking at consecutive frames is used to adapt a 3D template model of the body with the estimated pose for computing the final model of player’s body. Our 3D template is an articulated model where each part is non-deformable and corresponds to one part of the body. Body model is generated by using overlapping parameters of correlated parts and our goal is to adapt these parts to input frames. Our experimental results show that by extracting information of 3D template model at synthesis environment and adapting this model with real images with using fewest number of cameras (two) and low quality and low frame rate video images, we can reconstruct a model of player’s body that is more similar to real images relative to previous works based on using an articulated template of body
  9. Keywords:
  10. Three Dimensional Modeling ; Three Dimensional Reconstruction ; Moving Objects ; Multiview Videos ; Deformable Moving Objects ; Three Dimentional Articulated Template

 Digital Object List

  • محتواي پايان نامه
  •   view