Multi-level glowworm swarm convolution neural networks for abnormal event detection in online surveillance video

A surveillance camera is one of the most important tools for observing people's movements and stopping unauthorized or unplanned activity. Security management experts now significantly rely on video surveillance to combat crime and avert incidents that have a detrimental influence on human civilization. To monitor public activities, the installation of numerous surveillance cameras has drastically increased in both the public and private sectors. Security may be ensured most effectively through video surveillance. Installing a surveillance camera merely provides security personnel with the recorded video. However, integrating intelligent technology to analyze the videos is the only way to spot irregular actions. As a result, the goal of this study is to construct an Intelligent Video Analytics Model (IVAM), also known as a Human Object Detection (HOD) approach, for analyzing and spotting unusual human activity and abundant objects in videos. The proposed IVAM is designed based on Multi-level glowworm swarm convolution neural networks (ML-GSCNN). The proposed approach consists of two stages namely, frame conversion, and abnormal event detection. The captured video is first divided into segments, and then each segment is changed into a frame. After that, abnormal event detection is performed. For abnormal event detection, a novel ML-GSCNN is designed. Here, the hyper-parameter of CNN and the architecture of CNN both are optimized by the glowworm swarm optimization (GSO) algorithm to improve the detection accuracy. The experimental results show that the proposed approach attained better results compared to existing works.