3D Soccer Ball Tracking with Fixed and PTZ Cameras

Please enable javascript in your browser.

Fahmi, Reza | 2015

511 Viewed

Type of Document: M.Sc. Thesis
Language: Farsi
Document No: 48857 (19)
University: Sharif University of Technology
Department: Computer Engineering
Advisor(s): Kasaei, Shohreh
Abstract:
In computer vision literature, tracking problems are considered to be among the open and sophisticated issues. 3D object tracking is also an interesting problem which not only needs knowledge in computer vision, but it also requires knowledge of 3D geometry. Tracking soccer ball as a small and speedy obejct with PTZ cameras has several challenges such as blurring, blocking and getting lost. In this thesis, we aim to solve these issues with detection and learning frameworks. Current methods in tracking soccer balls usually exploit heuristic features, which as a result, creates dependency on the problem and its data type. Also, since ball tracking problem is directly related to the sports industry, data used by researchers mostly have higher quality and can be used in industry. This study attempts to provide a detection framework wherein the process of tracking independent of the heuristic characteristics on a normal dataset
Keywords:
Learning ; Three Dimensional Tracking ; Soccer Match Videos ; Soccer Math ; PTZ Cameras ; Detection Framework

No TOC