Modeling of Archimedes Optimization Algorithm with Deep Learning Driven Automated Social Distancing Detection and Classification

Abstract

Social distancing (SD) detection refers to the usage of technology, especially image processing and computer vision techniques, to monitor and identify compliance with SD guidelines in public spaces. The objective is to automatically analyze images or video feeds to identify individuals and measure the distances between them. Leveraging convolutional neural networks (CNN) or similar deep learning (DL) methods, the system analyzes images or video feeds to identify the presence of individuals and measure the distances between them. By learning sophisticated spatial patterns and relationships, the DL algorithm identifies instances where safe distance is not maintained, which provides real-time visualizations or alerts to assist in imposing public health measures. This study develops an Archimedes Optimization Algorithm with Deep Learning Driven Automated Social Distancing Detection and Classification (AOADL-SDDC) technique. The purpose of the AOADL-SDDC technique is to exploit hyperparameter tuned DL model for SD recognition. Primarily, the AOADL-SDDC technique makes use of Wiener Filter (WF) based noise removal and Dynamic Histogram Equalization (DHE) based contrast enhancement. For pedestrian detection process, the UNet++ with RMSProp optimizer is used, which detects the pedestrians accurately and the distance among them can be computed by the use of Manhattan distance. To detect SD, Bi-directional convolutional LSTM (BiConvLSTM) model can be employed and its hyperparameters can be adjusted by the use of AOA. A comprehensive set of experiments was carried out for examining the performance of the AOADL-SDDC technique. The stimulation value stated that the AOADL-SDDC technique reaches superior performance over other models.