A Cloud-Based Early Detection of Pushing at Crowded Event Entrance utilizing Central-Smoothing Hyper Graph Neural Network

A large-scale venue crowding at the doors can result in dangerous and sometimes fatal circumstances, especially when attendees begin shoving one another to get inside the event sooner. In this manuscript, A Cloud-Based Early Detection of Pushing at Crowded Event Entrance utilizing Central-Smoothing Hyper Graph Neural Network (CLD-EDP-CEE-CSHGNN) is proposed. In this, the input videos are collected from Real Time Dataset. Then, the input videos are pre-processed using Bayesian Boundary Trend Filtering (BBTF).To crop the ROI key frames. Next, pre-processed videos are given into the feature extraction stage using Separable Synchro Extracting Transform (SSET). It extracts the relevant features, like name, size, form and color. Then, the extracted features are fed into the CSHGNN to detect the Crowded Event Entrances which classifies as pushing and non-pushing. The Geese Migration Optimization Algorithm (GMOA) is employed to optimize the weight parameters of CSHGNN. The performance of the CLD-EDP-CEE-CSHGNN displays the results of the detections accurately. The CLD-EDP-CEE-CSHGNN is implemented and its performance is examined under some performance metrics, such as G-mean, Correlation coefficient, Identification rate. The simulation results display that Performance of proposed CLD-EDP-CEE-CSHGNN approach methods attains25.23%, 16.12%, and 21.27% higher G-mean, 11.02%, 22.11% and 34.23% higher Correlation coefficient compared with existing methods.