Sistema de conteo automático de flujo de personas por medio de visión artificial

Abstract

A typical problem in such populated as it is Bogotá city is the density of people attending sites public, many sites lack control systems capacity of people because the cost of implementation and maintenance, so think of a solution from non-mechanical devices. Since the area of digital signal processing and artificial vision we set to implement a versatile counter, able to work in real time, and any type of video source, based on open source solutions so that there is no restriction to use, modify, or implement in market the solution. Two different algorithms for identifying pedestrians were implemented, the algorithms are programmed in Python and libraries OpenCV where by performance tests the most optimal algorithm identified regarding use of hardware resources and precision for a count of people on a real environment checked. Eventually results in two identification algorithms that can lead to other applications such as security and surveillance on multiple platforms thanks to the programming language that can be easily ported without any restrictions