Semi-automatic player detection for real-time broadcasting systems
Min-Sung Kang, J. W. Lee, Sangcheol Kang, Y.-C. Lim · 2017
Computer vision technology such as object detection and tracking has been widely applied to sports video analysis. This technology analyzes the physical activity and performance of sports players. With this technology, it is also necessary to perform video analysis immediately after important events. In this paper, we focus on cutting out regions of full-screen images according to the movement of an important player in the image. In our system, an operator only assists by means of a simple input for a key player position in the sports game and our semi-automatic detection scheme accurately estimates the region of interest related to the key player. According to our experimental results, the proposed method can extract regions of interest more accurately and rapidly than conventional pedestrian detectors.