Recognition of semantic basketball events based on optical flow patterns
This paper presents a set of novel features for classifying basketball video clips into semantic events and a simple way to use prior temporal context information to improve the accuracy of classification. Specifically, the feature set consists of a motion descriptor, motion histogram, entropy of the histogram and texture. The motion descriptor is defined based on a set of primitive motion patterns which are derived form optical flow field. The event recognition is achieved by using kernel SVMs and a temporal contextual model. Experimental results have verified the effectiveness of the proposed method.