Visual Localization Using Mobile Phone
Nousiainen, Joona (2022)
2022
Tieto- ja sähkötekniikan kandidaattiohjelma - Bachelor's Programme in Computing and Electrical Engineering
Informaatioteknologian ja viestinnän tiedekunta - Faculty of Information Technology and Communication Sciences
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2022-02-28
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202202011782
https://urn.fi/URN:NBN:fi:tuni-202202011782
Tiivistelmä
Visual localization has been proposed as a solution for the problems that GPS has, due to GPS signals being blocked or reflected. Visual localization would allow the user to estimate position and orientation in places where GPS does not reliably work, such as indoors, underground or areas with high structures or terrain. The aim of this study was to implement an Android application, that could be used to estimate position and orientation by capturing an image and sending it to a visual localization pipeline. The study also evaluated the accuracy and speed of modern visual localization methods in different localization tasks.
The visual localization pipeline used in this study was offered on behalf of Tampere University's Signal Processing Research Center, and its operation is described briefly in the theory chapter. In the implementation chapter, we go over the development of the Android software and introduce the final product. The implementation chapter also describes the development and operations of the server that allows communication between the Android application and the visual localization pipeline.
The application was tested on Tampere University's Tietotalo building premises in different scenarios. The aim of the experiments was to evaluate which kind of tasks would be easier and which would provide more challenge, while also assessing the accuracy and speed of modern visual localization methods.
The results showed that modern visual localization systems can be used for indoors position and orientation estimation. The delay was short enough that it could be used even for real-time localization tasks. Accuracy was good, especially for scenes that contained unique features for the test area. The problem was two similar looking areas getting mixed up with each other, when the scene did not contain any unique features that could be used to determine the exact place. The performance of the pipeline was also tested in dark, which still gave results, but the accuracy of the position got worse.
The visual localization pipeline used in this study was offered on behalf of Tampere University's Signal Processing Research Center, and its operation is described briefly in the theory chapter. In the implementation chapter, we go over the development of the Android software and introduce the final product. The implementation chapter also describes the development and operations of the server that allows communication between the Android application and the visual localization pipeline.
The application was tested on Tampere University's Tietotalo building premises in different scenarios. The aim of the experiments was to evaluate which kind of tasks would be easier and which would provide more challenge, while also assessing the accuracy and speed of modern visual localization methods.
The results showed that modern visual localization systems can be used for indoors position and orientation estimation. The delay was short enough that it could be used even for real-time localization tasks. Accuracy was good, especially for scenes that contained unique features for the test area. The problem was two similar looking areas getting mixed up with each other, when the scene did not contain any unique features that could be used to determine the exact place. The performance of the pipeline was also tested in dark, which still gave results, but the accuracy of the position got worse.
Kokoelmat
- Kandidaatintutkielmat [8918]