IoT Capable Mechanism for Crowd Analysis

Abstract

This paper describes a crowd analysis of different activities using surveillance videos is an important topic for communal security. This paper also describes the detection of dangerous crowds if the weapon is present in the crowd. In our study, we are using raspberry pi 3 board for the development of a system that consists of ARMv8 CPU that detects the human heads and provides a count of humans in the region using Open CV-Python. The direction of the movement of the person can be achieved by human tracking. Generally, there are three different stages algorithm for computer-based crowd analysis, (1) people counting, (2) people tracking, and (3) crowd behavior analysis. This project is made for security purposes where there is a possibility of a dangerous crowd, for example, mall, railway station, shopping center. In our method, we are used CNN to trained dangerous weapons and DNN used for human detection. This method not only detects the direction of the crowd but also detects if the crowd is dangerous or not. In this method, also count the total number of human and it also gives confidence score that means, in how many percents it is related to original people. In this way, we could have prevented many deaths and injuries.