Abstract
Video classification is one of the rising fields in the video analysis. A large number of videos are accessed by people’s daily from television and internet. It is easy for humans to index the video from the collection of videos which contains news, cartoon, sports, comedy and drama. Among the categories, sports video plays a vital role due to their commercial demand. There is a similarity between the different sports video which makes the classification task difficult. In this study, the sports video categorization for five categories of sports like football, cricket, volleyball, tennis and basketball is presented. The sports video categorization system uses Higher Order Spectra Features (HOSF) for the feature extraction from video frames and multiclass Support Vector Machine (SVM) classifier for the classification of videos. The system gives average classification accuracy of 93.44% using HOSF and multiclass SVM classifier.
Highlights
Sports video categorization by measuring video without thresholding is discussed in [1]
Higher Order Spectra Features (HOSF) is used for feature extraction and multiclass Support Vector Machine (SVM) is used for classification
The performance of the sports video categorization system is evaluated by datasets which are obtained by television broadcast from different sessions
Summary
Sports video categorization by measuring video without thresholding is discussed in [1]. The input video frames are represented with a color histogram and are enhanced to reduce the features. The similarity of the videos is compared by nearest neighbor classifier. Amateur sports video genre categorization is described in [2]. The dense trajectories are calculated by k-means clustering algorithm. The bins and Bag Of Words (BOW) values are calculated by the mean and standard deviation to prevent the trajectories of the players from the audience. Motion boundary histogram and histogram oriented gradients are used as default parameters. The radial basis function based SVM is used for classification
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have