Structural Approach for Event Resolution in Cricket Videos

Abstract

Classification of multimedia big data today has become a serious issue for organizations. Therefore, new concepts to mine of multimedia big have emerged and people are doing numerous researches on how to effectively handle different types of multimedia data also known as multi model data. In our research, we focus on how to effectively extract and classify data from a multimedia data related to sports videos and draw conclusions considering all media present in the content. We specifically considered the game of cricket in this research to build a multi-model mining approach to identify specific events. We consider low level details such as color variations in videos and pitch in audio to analyze and distinguish different attributes of a given event. The input can be any cricket video and our approach is to identify events such as a four, a six, a dot ball etc. by extracting low level details of its color variations, edges, camera changes and audio frequency variations. This research for video/audio data extraction provides room for addition of further audio data extraction and textual data extraction for classification of a multimodal data set.