5G-based video summarization: An analysis from a methodological point of view

Abstract

Surveillance is one of the fast-growing applications used for monitoring and watching people, objects, or the environment to collect information and provide security. The surveillance data is in video form, and analyzing large video is challenging because it is essential to do efficient video streaming online. Video summarization comprises selecting, extracting, and aggregating keyframes for creating a synopsis, which is challenging. Though several methods have been proposed for video summarization, most are inconsistent, poor in processing and delivering video content, and do not focus on solving the root problems interlinked with efficient streaming. Thus, video streaming applications require an efficient video summarization model that can overcome existing issues and challenges and improve the overall quality of service integrated with the advanced technology of 5G. This paper has aimed to discuss various methods, approaches, and technologies used for video summarization to design a better model. It also presents various learning models and a taxonomy of available methods and provides a detailed review. The summary of the model used evaluates its outcome and the existing methods for potential future research works. The proposed approach is compared with existing ones to prove the model's efficiency. The result shows that the proposed model achieved a 62.3 and 52.3 F1 score summarizing the TVSum and SumMe datasets, respectively.