High-performance complex event processing framework to detect event patterns over video streams

Abstract

Complex Event Processing (CEP) is an event processing paradigm capable of detecting patterns over streaming data in real-time. Presently, CEP systems have key challenges to preform matching over video streams due to their unstructured data model and complex video patterns which occurs over time and space. In this paper, I introduce the design, implementation and optimization of the proposed CEP framework, which enables the pattern detection over video streams. The work first proposes a Video Event Query Language (VEQL) motivated from current event query languages to write expressive video queries in CEP scenario. The query discusses how to write event query rules for video patterns and encapsulate them as high-level operators. To perform matching over VEQL queries, Video Event Knowledge Graph (VEKG) is proposed, which is a graph-based structured model of video streams. A complex event matcher is then presented which enable spatiotemporal pattern matching over videos using VEQL and VEKG constructs. Finally, three optimization strategies: state summarization, data-driven windows, and tuning deep model cascades are discussed to improve the CEP system performance which I intend to follow in my ongoing PhD research.