This research addresses the challenge of achieving precise quadcopter localization in complex environments. It proposes a monocular visual-inertial odometry system that fuses camera data with inertial measurements to mitigate issues from textureless surfaces and lighting changes. By employing advanced deep learning techniques such as SuperPoint for feature detection and SuperGlue for matching, the system robustly integrates point and line features. An optimization framework minimizes reprojection and inertial errors, yielding enhanced accuracy on benchmark datasets.

Маалла Я. (науч. рук. Колюбин С.А.) Research on robust visual-inertial odometry algorithms for quadcopters localization // Сборник тезисов докладов конгресса молодых ученых. Электронное издание. – СПб: Университет ИТМО, [2025]. URL: https://kmu.itmo.ru/digests/article/14825