Abstract
<title>Abstract</title> <p>Video summarization plays a crucial role in video analytics especially in surveillance scenarios as it represents the shorter version of large video. This short version contains frames that are unique in nature and the uniqueness is determined on the basis of either the user’s interest or the nature of contents. In case of surveillance system, summarization of abnormal events enables the investigator to analyze the voluminous video rapidly. There are several challenges in summarization which includes processing of complex user queries, preservation of chronological order, and evaluating the quality of summarization. This paper proposes an interactive query-based abnormal events synopsis generation algorithm that applies rule-based classifier to handle complex queries and introduces a novel measure named “improved overlapping ratio for evaluating the quality of synopsis”. The experimental results on PETS09 dataset shows that the proposed algorithm outperforms the existing visualization based video synopsis method in terms of accuracy and quality of summarization.</p>