Pitch Frames Classification in a Cricket Video Using Bag-of-Visual-Words

Abstract

We propose a method for classification of pitch frames in a cricket video based on bag-of-visual-words technique. Bag of visual words is a popular and successful technique in image classification and object-based classifications. In this paper, we demonstrate three different techniques based on bag-of-words methodology. The three different techniques use three different set of features for the classification of pitch frames in a cricket video. The three different types of features we use are SIFT (Scale-Invariant Feature Transform), LBP (Local Binary Patterns), and CTE (Color+Texture+Edge) features. We evaluate the three techniques on the dataset of cricket (http://cse.iitk.ac.in/~vision/dipen/), made available online by Mr. Dipen Raghuwani. Our experiments by using the above mentioned three types of features have shown significant results.