Visual Localization with Environment Outline Prior

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F68407700%3A21230%2F22%3A00358250" target="_blank" >RIV/68407700:21230/22:00358250 - isvavai.cz</a>

Výsledek na webu

<a href="http://poseidon2.feld.cvut.cz/conf/poster/proceedings/POSTER_2022.pdf" target="_blank" >http://poseidon2.feld.cvut.cz/conf/poster/proceedings/POSTER_2022.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

Visual Localization with Environment Outline Prior

Popis výsledku v původním jazyce

Visual localization is the task of estimating the camera position and orientation relative to a given environment representation, which is used in applications such as mobile robotics, autonomous driving, and augmented reality. Currently, the environment has to be scanned by a human operator or explored by an autonomous system to create a precise and detailed map, often taking a large amount of storage space. Our approach aims to circumvent the scanning phase and decrease the memory footprint by using a publicly available lightweight environment representation in the form of building floor plans or human-readable digital maps. We extract the outline of the environment visible in images and use hierarchical localization on top of the layout representation, using image retrieval for coarse pose estimate and refining the pose by ICP algorithm. We also try to entirely replace the images in the retrieval step with the extracted outline geometry. The initial experiments show that the approach might achieve reasonable performance while keeping the storage needed for environment and image representation at a minimum.

Název v anglickém jazyce

Visual Localization with Environment Outline Prior

Popis výsledku anglicky

Visual localization is the task of estimating the camera position and orientation relative to a given environment representation, which is used in applications such as mobile robotics, autonomous driving, and augmented reality. Currently, the environment has to be scanned by a human operator or explored by an autonomous system to create a precise and detailed map, often taking a large amount of storage space. Our approach aims to circumvent the scanning phase and decrease the memory footprint by using a publicly available lightweight environment representation in the form of building floor plans or human-readable digital maps. We extract the outline of the environment visible in images and use hierarchical localization on top of the layout representation, using image retrieval for coarse pose estimate and refining the pose by ICP algorithm. We also try to entirely replace the images in the retrieval step with the extracted outline geometry. The initial experiments show that the approach might achieve reasonable performance while keeping the storage needed for environment and image representation at a minimum.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2022

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Proceedings of the International Student Scientific Conference Poster – 26/2022

ISBN

978-80-01-06992-9

ISSN

—

e-ISSN

—

Počet stran výsledku

6

Strana od-do

146-151

Název nakladatele

Fakulta elektrotechnická

Místo vydání

Praha

Místo konání akce

Praha

Datum konání akce

12. 5. 2022

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—