An Empirical Study on Big Video Data Processing: Architectural Styles, Issues, and Challenges

Abstract

Video data contributes to the majority of big data, henceforth, how to efficiently and effectively discovering knowledge from large-scale video data becomes a crucial challenge. In this paper, we propose multiple architectural styles for the domain of large-scale video data analytics services. These styles include online combined with offline processing style, distributed shared repositories, image mining and prediction services with deep learning techniques. These architectural styles are successfully implemented and examined in a number of domains including smart traffic and smart drones, as demonstrated in a middleware developed specifically for large-scale continuous video data processing.