Attractive Events Detection in Soccer Videos Based on Identification of Shots

Abstract

The necessity of efficient identification and classification of shots in a soccer game ascends with the increasing popularity and quantity of soccer videos. This paper proposes an effective and real-time identifying system for attractive events, namely shooting shots in soccer video clips. Instead of directly accessing shooting shots, this system identifies close-up shots and audience shots, which generally accompany with shooting shots. Field shots are concerned in this system as well, since this sort of shots is frequently used to describe formations of a team, indicating its strategy. This system begins with the selection of key frames. The ratio of soccer pitch color pixels in the key frame is then calculated to detect long shots. By counting the amount of edge pixels, the differentiation between close-up shots and audience shots will be completed. In order to increase the accuracy, the approach of Key Region Detection is introduced. This innovative method is proven by experiments to be comparatively precise. Experimental results validate the effectiveness of this system.