In recent years, geolocation technologies have evolved so fast, that today almost every technological mobile device can be traceable through a simple internet connection. A few years ago, geolocation system was almost another term for Global Positioning System (GPS) receivers, because that was almost the only known technology of that kind. With the technological evolution over the years, new geolocation technologies emerged, and the existing ones became more accurate, accessible and portable. Today it is possible to find cheaper GPS Navigation Systems, and much more accurate and faster than a decade ago. Most smartphones already have an incorporated GPS receiver, sometimes combined with network data, and web browsers can already track (even though with low accuracy) the actual location of some IP address.
Despite all this evolution, geolocation by GPS can fail, due to lack of visibility to the satellites: without direct view from the receiver to a set of at least 4 satellites is not possible to correctly execute the necessary triangulation. To resolve this problem, GPS might be combined with network information (A-GPS), which allows a faster triangulation and an accurate response.
Given this context, this dissertation follows an innovative approach in geolocation that combines GPS information with a computer vision component. The main goal is to demonstrate that geolocation can sometimes be more accurate with the help of an image analysis system, which adds value to the coordinates read from the GPS by reducing the error through an identification of georeferencial entities in captured frames.
With this approach, it would be possible for a device to know its relative position, even in locations where the GPS information is not available, fixing some known problems like the positioning inside Urban Canyons, where the GPS information is unavailable for moments leaving geolocation softwares with no clue about their position.
To demonstrate the validity of this concept a prototype was implemented and used for a series of tests. This prototype consists in an application aimed for public transports, developed for Fraunhofer AICOS, institution focused in the improvement and usability of the information and communication technologies (ICT), mainly of the senior population, with application in mobility. The implemented prototype has the goal to improve the visibility of a public transport passenger to the exterior of the vehicle, using a mobile application that reproduces the exterior landscape, signalling a sequence of points of interest (POI) and adding textual information about that POI to the image.
With the help of computer vision, the lack of geolocation may be compensated and, as can be seen in the evaluation chapter, the system, if able to, know its location even when GPS information lacks or is outdated.
For any additional information regarding this project, please contact us using the inquiries form.