Direct Ping-Level Landmark Detection for Side-Scan Sonar SLAM in Feature-Poor Seabed Environments

Jinho Im, Seonghun Hong

Keywords: Side-scan sonars, Ping-level landmark detection, Feature-poor environments, Underwater SLAM

TL;DR: We propose a ping-level landmark detection method for SSS-based SLAM that exploits anomalous backscatter deviations with a range-adaptive threshold, enabling reliable landmark extraction in feature-sparse underwater environments.

Abstract: Acoustic perception in underwater environments presents fundamental challenges distinct from those encountered in terrestrial or aerial domains, including GPS-denied navigation, noisy sensor measurements, and sparse visual features in featureless seabeds. Side-scan sonar (SSS) is a particularly attractive sensing modality for such environments, offering wide-area seabed coverage and reliable long-range acoustic measurements. However, most existing SSS-based SLAM approaches operate in the image domain, extracting visual features from stacked acoustic images, and their performance deteriorates in feature-poor or homogeneous seabed environments where distinctive structures are absent. Furthermore, range-dependent intensity variations and speckle noise inherent in SSS measurements further degrade the reliability of feature extraction and data association. This paper proposes a direct ping-level landmark detection method that exploits anomalous deviations in raw backscatter intensity profiles as landmark features, without relying on image formation or appearance-based descriptors. By modeling the slant-range-dependent attenuation of backscatter intensity and applying a spatially adaptive threshold, the proposed method enables reliable landmark extraction directly from individual sonar pings, even in feature-sparse seabed environments where conventional approaches struggle. The proposed detection method is integrated into a landmark-based SSS SLAM framework that directly incorporates raw ping measurements without intermediate image construction. Real-world field experiments demonstrate improved detection precision over image-level and ping-level baselines and enhanced vehicle localization accuracy in feature-poor seabed environments.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Paper Acceptance: Yes

Submission Number: 13