An offside soccer detection system using ontology and deep learning

Abstract

Nowadays, the Soccer events detection domain has become a more critical issue that attracts many researchers due to the enormous volume of available soccer video data worldwide. Consequently, it was a complicated task to recognize events using the video object detection process. This challenge leads us to propose an approach based on deep learning supplied by the ontology paradigm. This article develops a soccer offside detection system divided into two parts: applying deep learning algorithms to extract both visual and audio low-level features like balls, players, referee whistle sound, Etc. The second one considers these results and runs some ontology SWRL rules to identify events like offside or not offside players. Our final experiments demonstrate that the proposed approach reached better results than the other ones in the state-of-the-art.