Crowd Density Estimation System

Crowd density estimation Using Fuzzy Inference System in Monitoring CCTV

Soheil Tehranipour and Hamidreza Rashidy Kanan

Electrical Engineering Department, Qazvin Azad University, Qazvin, IRAN

{s.tehranipour, h.rashidykanan}@qiau.ac.ir

Abstract

Crowd Density Estimation is one of main method of automatic attention control to enhance security of people in public. Despite the beneficial previous works in the literature, reliable detection of number of people in a real-time platform is still a challenging issue. In this paper we focus on crowd density and people counting concentrating on automatic background subtraction and ……

In addition, temporal and spatial criteria of each frame of video have been considered to control the operator’s attention

Violence is very often in public places such as universities and schools, shopping malls and public transport. The analysis of crowd and its densities and flows, can show potential risks and suggest appropriate actions. Therefore, there is a great interest in crowd analysis by computer-vision methods.

Automatic crowd analysis was considered by many researchers and numerous studies. A survey by Jacques and colleagues [4] have survived and classified a wide range of studies in the field of crowd analysis using machine vision techniques. To predict behavior of the crowd, many parameters can be considered and studied, for instance speed of the crowd movement, crowd density and movement of persons against the mass movement of the crowd [5]. In some studies, abnormal behavior of the crowd is recognized by presenting a model [6].

One of the most important factors to evaluate the risk in the public is the crowd density. There are many articles on different methods of crowd density estimation [3, 4, 5, 7, 8]. These methods can be divided in three categories: pixel-based, texture-based and object-based methods which will be explained in details in related works

This fuzzy decision making system considers not only crowd density but also critical parameters such as temporal and spatial importance of crowd image as fuzzy membership functions. We consider temporal and spatial criteria due to biology concepts of attention control and so our system acts more similar to human decision making. In addition, in image processing steps of crowd density estimation, modified pixel-based methods have been used. We improved results of crowd density estimation using pixel-based method by calculating compactness of crowd using a fuzzy Inference system. Risk probability is computed in each method and resulted alarm is compared with each other. Also A simple hue, saturation, and value (HSV) histogram-based color model was used to develop our