ISOPoT: Imaging Sonar Odometry by Point Tracking
2026-06-22 • Robotics
Robotics
AI summaryⓘ
The authors address the problem of reliable navigation for underwater robots using forward-looking sonar, which works well in murky waters but produces noisy and unclear images. They created a new method called ISOPoT that tracks points over multiple frames to better match sonar data and improve navigation accuracy. Their approach includes simple optimization steps to make the system more reliable. Tests on different datasets showed that ISOPoT works better than existing methods in both sonar-only and combined sensor settings.
underwater navigationforward-looking sonarsonar odometrypoint trackingimage noisekeypoint detectionmarine roboticssensor fusion
Authors
Jaša Samec, Vid Rijavec, Marko Peljhan, Aleksander Grm, Andrej Androjna, Danijel Skočaj, Matej Dobrevski
Abstract
Reliable navigation in underwater environments remains a key challenge in marine robotics. In such scenarios, forward-looking sonars are a natural choice for long-range perception, offering wide coverage even in turbid, low-visibility conditions. However, sonar images are inherently noisy, contain artifacts, and lack rich semantic structure, causing standard computer vision methods for keypoint detection and matching to perform poorly. In this paper, we introduce ISOPoT, an imaging sonar odometry method based on modern point tracking techniques. We propose a sonar odometry pipeline that uses multi-frame point tracks as its primary correspondence representation, augmented with lightweight optimizations to improve robustness. We evaluated the proposed method on the Aracati 2017 dataset, as well as on an internal sonar dataset collected in real-world underwater environments. Our results show that ISOPoT outperforms previous state-of-the-art methods consistently in both sonar-only scenarios and in multi-sensor settings.