Determining crowd density for security and surveillance system

Abstract

Security has always been the main agenda in ensuring the safety and welfare for government and agencies especially in public area where possible threat that could cause a massive damage is intolerable. So, putting CCTV cameras for increasing safety in public and high security areas seems to be essential. Monitoring the crowd for controlling its density or observing people activity is one of the interested topics for surveillance area. Crowd monitoring system is widely used in many areas such as airports, stadiums, and subways. Estimating crowd density may be a good solution for management and control, maintaining the crowd safety, or prevention of riot and high risk activities. This project offers a computational fast and simple method for estimating the density of crowd. This method is according to Local Binary Pattern feature extractor which is appropriate for real-time applications. One of the characteristics of this method is that without using background subtraction and according to the histogram model that obtained in training step, can estimate the density of crowd in interested areas. Although, pixel based methods that were not appropriate for crowd with high density, this method is robust in areas with high density of crowds. This system takes an input image and then computes its histogram according to Uniform Local Binary Patterns. After that compare this histogram with the 5 histogram model and determined this image is belong to which of five categories "Very Low, Low, Moderate, High, and Very High".