cs.ROMar 19, 2026

ROFT-VINS: Robust Feature Tracking-based Visual-Inertial State Estimation for Harsh Environment

Authors: Sanghyun Park, Soohee Han

Organizations: Department of Convergence IT Engineering, POSTECH, Pohang, 37673, Korea · Department of Electrical Engineering and of Convergence IT Engineering, POSTECH, Pohang, 37673, Korea

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

SLAM (Simultaneous Localization and Mapping) and Odometry are important systems for estimating the position of mobile devices, such as robots and cars, utilizing one or more sensors. Particularly in camera-based SLAM or Odometry, effectively tracking visual features is important as it significantly impacts system performance. In this paper, we propose a method that leverages deep learning to robustly track visual features in monocular camera images. This method operates reliably even in textureless environments and situations with rapid lighting changes. Additionally, we evaluate the performance of our proposed method by integrating it into VINS-Fusion (Monocular-Inertial), a commonly used Visual-Inertial Odometry (VIO) system.

Figures & tables

Figure 1: The architecture of our odometry system.
Figure 2: The pipeline of our outlier rejection module.
Table 1: Mean of RPE [m] in the EuRoC MAV datasets
Figure 3: Trajectory in MH05. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 4: Trajectory in V103. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 5: Trajectory in V203. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 6: RPE results: (Left) RPE in MH05. (Center) RPE in V103. (Right) RPE in V203. Green box is the result of ours, purple box is the result of VINS-Fusion.
Figure 7: Trajectory in UMA_parking_csc2. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 8: Trajectory in UMA_parking_eng1. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 9: Trajectory in UMA_fantasy_csc1. Green line is result of ours, purple line is result of VINS-Fusion.
Figure 10: Ours feature tracking image in challenging environment. As the feature points change from blue to red, it indicates that they have been tracked in more frames.
Figure 11: VINS-Mono feature tracking image in challenging environment. As the feature points change from blue to red, it indicates that they have been tracked in more frames.
More in this topic

Visual-Inertial Odometry

Visual-inertial odometry estimates camera motion by combining image geometry with inertial measurements for metric navigation.

See all 54 papers

Explore similar work

CardsList

  1. MFVINS: Multiple Fisheye Camera-Based Visual Inertial System

    Sep 8, 2026Eunseong Jang, YuJin Chung, Sang Jun Lee +2Visual-Inertial SLAMVisual-Inertial Odometry

    Division of Electronic Engineering, Jeonbuk National University, Jeonju, South Korea · BSTAR Robotics, Inc., Palo Alto, CA, USA

  2. KLTNet: Learning Sparse Feature Tracking for Robust and Accurate Monocular Visual-Inertial Odometry

    Aug 25, 2026Renbiao Jin, Danping Zou, Wenxian YuPoint TrackingVisual-Inertial Odometry

    Shanghai Jiao Tong University, Shanghai 200240, China

  3. DL-VINS-Factory: A Modular Framework for Learned Visual Front-Ends in Visual-Inertial SLAM

    Jul 2, 2026Shoon Kit Lim, Melissa Jia Ying Chong, Ting Yang LingVisual-Inertial SLAMVisual-Inertial Odometry

    University of Southampton Malaysia, Iskandar Puteri, Johor, Malaysia · University of Southampton, Southampton SO17 1BJ, U.K.