MANTA: Physics-Informed Generalized Underwater Object Tracking

2025-11-28

AI Summary

Underwater object tracking suffers from severe appearance distortion due to wavelength-dependent attenuation and scattering of light, leading to poor generalization of terrestrial-trained models. To address this, we propose a physics-informed universal tracking framework that, for the first time, integrates the Beer–Lambert law into end-to-end learning. Our method jointly optimizes appearance similarity and geometric consistency in feature space via dual-positive contrastive learning and a physics-driven secondary association algorithm. We introduce two novel metrics—Curve Shape Fidelity (CSC) and Geometric Alignment Score (GAS)—to quantitatively evaluate geometric fidelity, and design motion-guided matching and multi-stage temporal modeling mechanisms. Evaluated on four underwater benchmarks, our approach achieves up to a 6% improvement in Success AUC, significantly enhancing long-term tracking robustness and cross-domain generalization across varying depths and water qualities, while maintaining real-time performance.

Technology Category

• Computer Vision: Motion & Tracking
• Search and Optimization: Learning to Search
• Intelligent Robots: State Estimation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Security and Privacy: Tracking, profiling, and countermeasures against them
• User Modeling, Personalization and Recommendation: Metrics for user behavior and evaluating success

Abstract

Underwater object tracking is challenging due to wavelength dependent attenuation and scattering, which severely distort appearance across depths and water conditions. Existing trackers trained on terrestrial data fail to generalize to these physics-driven degradations. We present MANTA, a physics-informed framework integrating representation learning with tracking design for underwater scenarios. We propose a dual-positive contrastive learning strategy coupling temporal consistency with Beer-Lambert augmentations to yield features robust to both temporal and underwater distortions. We further introduce a multi-stage pipeline augmenting motion-based tracking with a physics-informed secondary association algorithm that integrates geometric consistency and appearance similarity for re-identification under occlusion and drift. To complement standard IoU metrics, we propose Center-Scale Consistency (CSC) and Geometric Alignment Score (GAS) to assess geometric fidelity. Experiments on four underwater benchmarks (WebUOT-1M, UOT32, UTB180, UWCOT220) show that MANTA achieves state-of-the-art performance, improving Success AUC by up to 6 percent, while ensuring stable long-term generalized underwater tracking and efficient runtime.

Problem

• Addresses underwater object tracking under wavelength-dependent distortions
• Overcomes physics-driven appearance degradation across depths and conditions
• Solves occlusion and drift challenges through physics-informed re-identification

Innovation

• Physics-informed framework integrating representation learning with tracking
• Dual-positive contrastive learning with temporal and underwater augmentations
• Multi-stage pipeline combining motion tracking with physics-informed re-identification