A Proposed Framework for Small Group Detection in Medium Density Crowd

Abstract

In this paper a framework for relatively small group detection from dense crowd is presented
which can work in low to medium density of crowd. Small groups in a crowded feed are the
basic unit to be considered. To find out groups in a crowded source feed is a tedious task. For
this purpose, threshold distance is used as a parameter to find out the distance between
contours to estimate the density of the crowd. The source feed is converted into frames and
threshold distance is defined on the basis of these frames. The experimental results show the
high accuracy of our framework.