This paper presents several optimization methods for analyzing the video clips of soccer matches in real time. Without allowing the human interactions, for achieving the real-time analysis, the proposed system utilizes simple masking methods to detect...
This paper presents several optimization methods for analyzing the video clips of soccer matches in real time. Without allowing the human interactions, for achieving the real-time analysis, the proposed system utilizes simple masking methods to detect the target objects. Targeting the irregular video clips, in addition, we automatically recognize the playground by applying the novel mapping schemes associated with the on-the-fly field capturing, leading to the user-friendly top-view projections. As a result, the proposed optimization schemes achieve 64% fewer ID changes for tracking players while supporting the processing speed of 27fps.