Scientific Exploration for Density Estimation and Crowd Counting of Crowded Scene

Abstract

Abstract Crowd density estimation model is a typical concept which compute the counting of the people in the crowded image. There are many of research papers are written in this area to solve the different kind of real world problems. This paper shows the state of art and different types of framework proposed on density estimation and crowd counting techniques. The population of our country increases year by year, so the purpose of this paper is to explore different techniques available for crowd density estimation. There are lot of applications existing based on this technique like retail system, surveillance system etc., where the foot traffic is very vital in retail system to organize merchandise in aisles, optimize store layout, understand peak times and potentially even protect against theft. In surveillance system, this is mainly used for security purpose. In this, the researcher captures the crowd using the camera and the technique of counting the crowd which is divided into two parts unsupervised and supervised learning. These techniques are further divided into parts which is shown in the paper. The survey presents the different types of datasets of the crowded image which is useful for simulating the people counting in the crowded scenes. In detection based approach, this technique counts the people with the help of face detection technique. In this researchers face the problems of count the people in crowded scenes and also face the problem in surveillance applications. The cluster based approach data works well on the scattered crowd. This method needs image frames; it does not work on still images. In regression approach the model trained is subject to the point of view map. In the model were to be used in a different scene of a changed perspective map, it will have many inaccuracies in its result. The current research work intends to provide a general idea on crowd density estimation and counting approaches employed in visual surveillance in the perception of the study of computer vision. The analysis of review categorizes and frameworks quite a lot approximation related to crowd density and calculating approaches which are beneficial in providing the analysis related to the crowd places of interest.