Empirical Analysis of Visual Localization Accuracy

The paper is devoted to empirical accuracy analysis of the visual localization problem. The accuracy is considered with attention to the design of a maintenance system for an industrial manufacturing cell. Three cases of localization are considered that constitute the most relevant acquisition condi...

Full description

Saved in:
Bibliographic Details
Published in2022 International Russian Automation Conference (RusAutoCon) pp. 354 - 360
Main Authors Iusupova, Anastasiia, Titov, Viktor
Format Conference Proceeding
LanguageEnglish
Published IEEE 04.09.2022
Subjects
Cameras
Error analysis
localization error
Location awareness
Maintenance engineering
PnP accuracy
position estimation
Virtual environments
Visualization
Online AccessGet full text

Cover

More Information
Summary:The paper is devoted to empirical accuracy analysis of the visual localization problem. The accuracy is considered with attention to the design of a maintenance system for an industrial manufacturing cell. Three cases of localization are considered that constitute the most relevant acquisition conditions: single view, dual view, and partial occlusions. A virtual environment is developed for simulation in close to realistic conditions that allows one to account for feature point detection algorithms, camera resolution, etc. Important results have been obtained for the maintenance system design. The simulation revealed that the accuracy is strongly dependent on the distance to the target object and the number of visible points. The orientation error shows little dependence on the altitude of the camera relatively to the target object. The position error exhibits a clear minimum at a particular angle between the cameras in dual view observation. The accuracy under occlusions also indicates a number of patterns.
DOI:10.1109/RusAutoCon54946.2022.9896298