Institutional Repository of Vilnius University
Afiniškai transformuotų ortografinių nuotraukų lyginimas naudojant trejetų neuroninį tinklą /
Abstract
dc:descriptionMost unmanned aerial vehicles use a GPS signal for navigation. There is a risk that GPS signal could be interrupted which can cause navigation failures. An approach based on aerial images and convolutional neural networks can be used to solve this problem. The aircraft computer memory should contain a set of aerial images that are comparable to images taken from the aircraft itself with the camera pointing downwards. Comparing these images would determine the location of the aircraft. In this paper a triplet convolutional neural network is created, which could be adapted for unmanned aerial vehicles navigation and also loss function and network model are selected. In addition, an orthographic dataset is created for this experiment, which is applicable to both network training and testing. Aerial images taken from an unmanned aerial vehicles can often have various distortions - rotations, conversions or scalings, so this scenario is also analyzed in this paper. To verify triplet convolutional neural network accuracy a set of affinity transformed images is created. The results obtained after the experiment show that the network 100% recognizes original data and images with small transformations.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Puodžiūnaitė, Eglė,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD107115763&prefLang=en_US