Optimization- and Learning-Based Approaches to Visual SLAM and Relocalization
Übersetzter Titel:
Optimierungs- und lernbasierte Verfahren für visuelles SLAM und Relokalisierung
Autor:
von Stumberg, Lukas Michael
Jahr:
2023
Dokumenttyp:
Dissertation
Fakultät/School:
TUM School of Computation, Information and Technology
Betreuer:
Cremers, Daniel (Prof. Dr.)
Gutachter:
Cremers, Daniel (Prof. Dr.); Scaramuzza, Davide (Prof. Dr.); Fallon, Maurice (Prof., Ph.D.)
Sprache:
en
Fachgebiet:
DAT Datenverarbeitung, Informatik
TU-Systematik:
DAT 760; DAT 770
Kurzfassung:
Visual SLAM and Relocalization are essential for a number of applications like robotics, autonomous driving and augmented reality. The first part of this thesis introduces new techniques for optimization-based visual-inertial odometry, including an optimization technique, IMU initializers and marginalization variants. Afterwards we improve visual odometry and relocalization by employing neural networks which estimate depths, uncertainty, relative poses, and deep features.
Übersetzte Kurzfassung:
Visuelles SLAM und Relokalisierung sind essentiell für Anwendungen wie Robotik, autonomes Fahren und Augmented Reality. Der erste Teil dieser Dissertation stellt neue Techniken für optimierungs-basierte visuell-inertiale Odometrie vor, darunter ein Optimierungsverfahren, IMU-Initialisierer und Marginalierungs-Verfahren. Anschließend verbessern wir visuelle Odometrie und Relokalisierung, indem wir neuronale Netze integrieren, die Tiefenwerte, Unsicherheit, relative Posen und Features berechnen.
In reference to IEEE copyrighted material which is used with permission in this thesis, the IEEE does not endorse any of TUM's products or services. Internal or personal use of this material is permitted. If interested in reprinting/republishing IEEE copyrighted material for advertising or promotional purposes or for creating new collective works for resale or redistribution, please go to http://www.ieee.org/publications_standards/publications/rights/rights_link.html to learn how to obtain a License from RightsLink.
«
In reference to IEEE copyrighted material which is used with permission in this thesis, the IEEE does not endorse any of TUM's products or services. Internal or personal use of this material is permitted. If interested in reprinting/republishing IEEE copyrighted material for advertising or promotional purposes or for creating new collective works for resale or redistribution, please go to http://www.ieee.org/publications_standards/publications/rights/rights_link.html to learn how to obtain a Lic...
»