Player detection using one-class SVM

摘要

In this paper, a novel player detection method via One-Class SVM(OCSVM) is proposed, inspired by both theplayer detection problem and the property of the OCSVM. In this detection method, candidate regions are got by localentropy and local range analysis firstly. Then a set of training samples is obtained by several predefined rules on shapeand area. These samples are used to train two OCSVM models. One model uses color feature, and the other uses gradientfeature. Finally, we locate the regions of player by fusing the detection result of the two models. Extensive experimentsdemonstrate effectiveness and efficiency of the proposed method.© (2012) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.