A type-2 fuzzy logic system for event detection in soccer videos

Sequences classification problems in recorded videos are often very complex and have too much uncertainty. In many application domains, such as video event activity detection, sequences of events occurring over time need to be studied in order to summarize the key events from the video clips. In most existing adaptive sequences classification systems, Dynamic Time Warping (DTW) and Gaussian Mixture Mode (GMM) are used as the core techniques in measuring similarity between two temporal sequences, which may vary in speed. Hence, there is a need to develop video event detection systems capable of classifying important events within long video sequences. This paper presents a novel system based on DTW and Interval Type-2 Fuzzy Logic Systems employing the Big Bang Big Crunch (BB-BC) algorithm for video activity detection and classification of critical events from the large-scale data of soccer videos.