Fog Assisted Smart Video Surveillance System using Semantic Analysis for Abnormal Motion Detection
R. S. Amshavalli, J. Kalaivani · 2024
Recent Research in creating intelligent smart Closed Circuit Television (CCTV) surveillance systems that can counteract the rising levels of insecurity has increased dramatically as a result of the growing demand for video surveillance systems. However, the storage and processing capacity of analytical applications has been exceeded by the enormous volume of video data produced by these devices. By localizing data to the network's edges, this article suggests a smart surveillance system that exhibits high performance in terms of response time and bandwidth. In modern IoT era, the need for transforming the traditional system into smart surveillance system is high due to its demand for instant detection of anomalous occurrences by processing huge amount of video data on the go. Bringing the smartness into the system, makes it more automated and reliable. The challenging phases in making the system smart is storing and analyzing the video surveillance data since it calls for event recognition, visual interpretation, and the identification of relevant context. We employ sliding technique and adaptive contour based algorithms for detailed preprocessing layer that performs key frame extraction and background elimination respectively. And to address the issue of storage and instant processing, we also propose fog based smart surveillance framework that acts as the intermediator component between cloud and end users. The next major challenge is analyzing the incoming video data on the go. It can be addressed through semantic based algorithm for detecting the abnormal events. We propose an ontology based algorithm for detecting the unusual occurrences. Experiments are conducted on a real-time dataset to evaluate the suggested technique. The computed results demonstrate the superiority of the suggested fog based semantic analytics monitoring system over the traditional cloud-based monitoring solutions by attaining high prediction and accuracy rate coupled with low latency in decision making.