Nearshore Underwater Target Detection Meets UAV-borne Hyperspectral Remote Sensing: A Novel Hybrid-level Contrastive Learning Framework and Benchmark Dataset

TL;DR

The paper tackles nearshore underwater target detection with UAV-borne hyperspectral imagery, where spectral distortions hinder traditional HUTD methods. It introduces HUCLNet, a two-module framework combining reliability-guided clustering and hybrid-level contrastive learning under a self-paced learning regime, plus hyperspectral-oriented data augmentation. A large-scale ATR2-HUTD benchmark with Lake, River, and Sea subdatasets demonstrates that HUCLNet outperforms state-of-the-art methods across detection accuracy, target preservation, and background suppression. The work provides a practical, scalable path toward robust nearshore UTD and releases the dataset and code to advance research in hyperspectral underwater sensing.

Abstract

UAV-borne hyperspectral remote sensing has emerged as a promising approach for underwater target detection (UTD). However, its effectiveness is hindered by spectral distortions in nearshore environments, which compromise the accuracy of traditional hyperspectral UTD (HUTD) methods that rely on bathymetric model. These distortions lead to significant uncertainty in target and background spectra, challenging the detection process. To address this, we propose the Hyperspectral Underwater Contrastive Learning Network (HUCLNet), a novel framework that integrates contrastive learning with a self-paced learning paradigm for robust HUTD in nearshore regions. HUCLNet extracts discriminative features from distorted hyperspectral data through contrastive learning, while the self-paced learning strategy selectively prioritizes the most informative samples. Additionally, a reliability-guided clustering strategy enhances the robustness of learned representations.To evaluate the method effectiveness, we conduct a novel nearshore HUTD benchmark dataset, ATR2-HUTD, covering three diverse scenarios with varying water types and turbidity, and target types. Extensive experiments demonstrate that HUCLNet significantly outperforms state-of-the-art methods. The dataset and code will be publicly available at: https://github.com/qjh1996/HUTD

Figures (13)

• Figure 1: Limitations of RGB imagery and advantages of hyperspectral imagery in underwater depth estimation. (a) In RGB images, the target and background have nearly identical spatial appearances; (b) In hyperspectral images, the target and background exhibit distinct spectral signatures.
• Figure 2: The illustration of opportunities and challenges in hyperspectral nearshore underwater target detection. (a) Opportunities; (b) Challenges.
• Figure 3: Illustration of spectral distortions induced by underwater conditions, with depth as an example. The spectral signature of underwater target diverge from their reference spectra, and this deviation varies with depth. The similar situations can be observed for other underwater conditions Gillis2020.
• Figure 4: The ATR2-HUTD-Lake sub-dataset. (a) Underwater Scene1; (b) Underwater Scene2.
• Figure 5: The flowchart of the proposed underwater target detection framework.