Sports Video Annotation and Multi-Target Tracking using Extended Gaussian Mixture model
Daneshwari Mulimani1, Aziz Makandar2

1Daneshwari Mulimani*, Research Scholar, Department of Computer Science, Karnataka State Ak kamahadevi Women’s University, Bijapur (Karnataka), India.
2Aziz Makandar, Research Scholar, Department of Computer Science, Karnataka State Ak kamahadevi Women’s University, Bijapur (Karnataka), India.

Manuscript received on March 26, 2021. | Revised Manuscript received on April 16, 2021. | Manuscript published on May 30, 2021. | PP: 1-6 | Volume-10 Issue-1, May 2021. | Retrieval Number: 100.1/ijrte.A55890510121 | DOI: 10.35940/ijrte.A5589.0510121
Open Access | Ethics and Policies | Cite | Mendeley
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)

Abstract: Video offers solutions to many of the traditional problems with coach, trainer, commenter, umpires and other security issues of modern team games. This paper presents a novel framework to perform player identification and tracking technique for the sports (Kabaddi) with extending the implementation towards the event handling process which expands the game analysis of the third umpire assessment. In the proposed methodology, video preprocessing has done with Kalman Filtering (KF) technique. Extended Gaussian Mixture Model (EGMM) implemented to detect the object occlusions and player labeling. Morphological operations have given the more genuine results on player detection on the spatial domain by applying the silhouette spot model. Team localization and player tracking has done with Robust Color Table (RCT) model generation to classify each team members. Hough Grid Transformation (HGT) and Region of Interest (RoI) method has applied for background annotation process. Through which each court line tracing and labeling in the half of the court with respect to their state-of-art for foremost event handling process is performed. Extensive experiments have been conducted on real time video samples to meet out the all the challenging aspects. Proposed algorithm tested on both Self Developed Video (SDV) data and Real Time Video (RTV) with dynamic background for the greater tracking accuracy and performance measures in the different state of video samples.  
Keywords: Line Segmentation, Player Localization, HGT, RCT.