Comparison and Analysis of Various Approaches of Crowd Density Estimation for Passenger Flow Detection System

Abstract

One of the major issues with a public surveillance system is viewing and evaluating congested settings. Numerous tragic crowd-related incidents have happened at sporting events, concerts, religious ceremonies, and other gatherings throughout the years. More than 2500 individuals perished in fatal crowd mishaps between 2011 and 2020. In addition, 79% of the calamities in India have occurred at religious events or pilgrimages.  Every year, similar incidents result in many fatalities.

The present pandemic scenario inspired us to examine how people behave in social situations. In this dangerous pandemic condition, we use surveillance data to identify social distance violations and the presence of freestanding discussion groups (FCG). We were inspired to combine data from many modalities, including video, audio, accelerometer, and optical flow patterns, by the difficulties of light change and occlusion in video data. This study compares and analyzes diverse crowd behaviours using data from intelligent surveillance systems to forecast behaviour and spot anomalies.