Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/149010
Title: ArQVIA: Prototipo WebApp de VSLAM para reconstrucciones virtuales 3D en espacios arqueológicos. Explorando la arqueología con técnicas avanzadas multimedia, CBIR, deconstrucción fotogramétrica e inteligencia artificial
Author: Colmenero-Fernández, Alicia  
Tutor: Julià Lundgren, Daniel
Others: Casado-Martínez, Carlos  
Abstract: The achievement of this work is the development of a prototype mobile web app that implements a novel methodology for the interactive visualization of 3D virtual reconstructions located in archaeological spaces using artificial intelligence. Virtual archaeology enables the visualization of content that helps visitors understand the archaeological site through the three-dimensional reconstruction of the original location. One of the main challenges for interaction in visualization is the so-called VSLAM (visual simultaneous localization and mapping). The fundamental problem lies in achieving an accurate matching between our real position and the virtual replica to scale. The different existing techniques, such as marker recognition, analysis, extraction, and comparison of features from the input image against keyframes or point clouds, present issues of portability, high computational costs (embedded systems, GPUs), or library compilation for calculations. The proposed method allows recognition from a free position by querying images using examples (CBIR) executed on a server's VRAM, so that the device only sends an image and manages the web response using the gyroscope. Through feature extraction, an on-site RGB input is compared with the learned characteristics of renders created from a precise synthetic model of the location. Once a match is found, the server returns a 360x360 panoramic view with the reconstruction of the closest stored camera in a repository, adapting the view to the degree of rotation. ArQVIA performs the query in a request-response mode against the backend (Python, JavaScript), providing the visitor with a three-dimensional visualization that can be controlled using the gyroscope as a "magic window." It is a mobile app accessible from any point of the virtually documented archaeological site, executable on a monocular mobile device without special features.
Keywords: machine learning
VSlam
archaeology app
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: 19-Jun-2023
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
acolmeneroTFG0623memoria.pdfMemoria del TFG3,62 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

This item is licensed under aCreative Commons License Creative Commons