A Low Cost Indoor Positioning System Using Computer Vision

Abstract

In the era of robotics, positioning is one of the major problems in an indoor environment. A Global Positioning System (GPS), which is quite reliable system when it comes to outdoor environments and its accuracy falls in the range of meters. But for indoor environment, which requires a positioning accuracy in centimeters scale, the GPS cannot achieve this task due to its signal loss and scattering caused by the building walls. Therefore, an Indoor Positioning System (IPS) based on several technologies and techniques has been developed to overcome this issue. Nowadays, IPS becomes an active growing research topic because of its limitless implementations in a variety of applications. This paper represents the development of a low cost optical indoor positioning system solution where a static commercial camera is the only sensor. High accuracy in localization within the range of 1 cm is achieved. Detection, classification, and tracking techniques of an object are tested on mobile robots. The system is ideal for an indoor robotic warehouse application, where minimal infrastructure and cost parameters are required. The resulted positioning data are compared to the real measurement, and sent to the rovers via a lightweight broker-based publish/subscribe messaging protocol called Message Queuing Telemetry Transport (MQTT), where the only requirement between the client publisher and subscriber is the availability of a WLAN connection.