Indian Journal of Science and Technology

The PDF file you selected should load here if your Web browser has a PDF reader plug-in installed (for example, a recent version of Adobe Acrobat Reader).

If you would like more information about how to print, save, and work with PDFs, Highwire Press provides a helpful Frequently Asked Questions about PDFs.

Alternatively, you can download the PDF file directly to your computer, from where it can be opened using a PDF reader. To download the PDF, click the Download link above.

Fullscreen Fullscreen Off


Objectives: In order to achieve convenient sports video accessing without sequential scanning, automated sports video categorization is presented in this study. Methods/Analysis: In order to build efficient sports video categorizing system, edge features obtained from Non Subsampled Shearlet Transform (NSST) are taken into account. Then, sports genre categorization is done by Nearest Neighbor (NN) classifier due to its discriminative learning approach. The five sports category; tennis, cricket, volleyball, basketball and football are considered. Findings: To validate the proposed system based on NSST, experiments are carried using internal database video at frame level. Totally, 500 video clips are collected in which 100 video clips are gathered for each sports genre. The proposed system achieves maximum average classification accuracy of 94.80% at 4 directions of 2-scale NSST features while using city block distance measure in KNN classifier. For the same NSST features, the Euclidean, cosine and correlation distance measures gives an accuracy of 93.20%, 92.80% and 92% respectively. Conclusion/Application: The effectiveness of the system is clearly demonstrated by the experimental assessment. The proposed framework can adequately classify the sports video into one of the five predefined genre.

Keywords

Edge Features, Nearest Neighbor Classifier, NSST, Shearlet Transform, Sports Video Classification.
User