OTPL-VIO: Robust Visual-Inertial Odometry with Optimal Transport Line Association and Adaptive Uncertainty

TL;DR

This work presents a stereo point-line VIO system in which line segments are equipped with dedicated deep descriptors and matched using an entropy-regularized optimal transport formulation, enabling globally consistent correspondences under ambiguity, outliers, and partial observations.

Abstract

Robust stereo visual-inertial odometry (VIO) remains challenging in low-texture scenes and under abrupt illumination changes, where point features become sparse and unstable, leading to ambiguous association and under-constrained estimation. Line structures offer complementary geometric cues, yet many efficient point-line systems still rely on point-guided line association, which can break down when point support is weak and may lead to biased constraints. We present a stereo point-line VIO system in which line segments are equipped with dedicated deep descriptors and matched using an entropy-regularized optimal transport formulation, enabling globally consistent correspondences under ambiguity, outliers, and partial observations. The proposed descriptor is training-free and is computed by sampling and pooling network feature maps. To improve estimation stability, we analyze the impact of line measurement noise and introduce reliability-adaptive weighting to regulate the influence of line constraints during optimization. Experiments on EuRoC and UMA-VI, together with real-world deployments in low-texture and illumination-challenging environments, demonstrate improved accuracy and robustness over representative baselines while maintaining real-time performance.

Figures (6)

• Figure 1: System overview of OTPL-VIO.
• Figure 2: Our data-collection platform.
• Figure 3: Qualitative evaluation on EuRoC sequences V103 and V203.The estimated trajectories (right) are colored by ATE, with red boxes indicating the locations of challenging scenes shown on the left. (a) Abrupt illumination changes in V103. (b) Low-texture regions in V203. Here, red crosses and green dots denote failed and successful triangulations, respectively, with their specific counts provided in the bottom-right corner.
• Figure 4: Qualitative results on the UMA-VI dataset: (a) Corridor-eng and (b) Third-csc1. Estimated trajectories are colored by ATE (m) where ground truth is available; dashed lines indicate segments without ground truth. Black inset boxes visualize the accumulated end-point drift between the start and end of the sequences. Red boxes mark the locations of the challenging views shown on the left. (a) Low-texture corridor. (b) Abrupt illumination changes.
• Figure 5: Qualitative results on real-world experiment. The estimated trajectories are overlaid on the floor plan for Seq: Indoor1 (a) and Seq: Indoor2 (b). The environment contains scenes with low texture and illumination changes.