Underwater image restoration via attenuated incident optical model and background segmentation DOI Creative Commons
Sen Lin,

Yuanjie Sun,

Ning Ye

et al.

Frontiers in Marine Science, Journal Year: 2024, Volume and Issue: 11

Published: Oct. 10, 2024

Underwater images typically exhibit low quality due to complex imaging environments, which impede the development of Space-Air-Ground-Sea Integrated Network (SAGSIN). Existing physical models often ignore light absorption and attenuation properties water, making them incapable resolving details resulting in contrast. To address this issue, we propose attenuated incident optical model combine it with a background segmentation technique for underwater image restoration. Specifically, first utilize features distinguish foreground region from region. Subsequently, introduce layer improve account effects non-uniform light. Afterward, employ new maximum reflection prior estimation achieve restoration Meanwhile, contrast is enhanced by stretching saturation brightness components. Extensive experiments conducted on four datasets, using both classical state-of-the-art (SOTA) algorithms, demonstrate that our method not only successfully restores textures but also beneficial processing under lighting conditions.

Language: Английский

Underwater image captioning: Challenges, models, and datasets DOI
Huanyu Li, Hao Wang, Ying Zhang

et al.

ISPRS Journal of Photogrammetry and Remote Sensing, Journal Year: 2025, Volume and Issue: 220, P. 440 - 453

Published: Jan. 5, 2025

Language: Английский

Citations

4
Oil Spill Drift Prediction Enhanced by Correcting Numerically Forecasted Sea Surface Dynamic Fields with Adversarial Temporal Convolutional Networks DOI
Peng Ren, Qing Jia, Qing Xu

et al.

IEEE Transactions on Geoscience and Remote Sensing, Journal Year: 2025, Volume and Issue: 63, P. 1 - 18

Published: Jan. 1, 2025

Timely and accurate representation of sea surface dynamic fields is crucial for oil spill drift prediction. Numerically forecasted are available in a timely manner, but their accuracy limited. Conversely, reanalysis offer superior suffer from time delays. To enhance the performance prediction, we propose deep learning-based approach to correcting numerically fields, aligning them more closely with fields. Our introduces an adversarial temporal convolutional network (ATCN) framework, consisting (TCN)-based corrector discriminator. The TCN can characterize field sequences both spatially temporally. In this scenario, processes outputs corrected that approximate Adversarial training discriminator further refines corrector. This enhances prediction using We also provide dataset drifts Symphony Sanchi accidents, including related data remote sensing data, establishing baseline evaluating Experiments on validate ATCN framework's effectiveness enhancing

Language: Английский

Citations

2
Cascaded frameworks in underwater optical image restoration DOI
Bincheng Li, Ziqian Chen, Lin Lu

et al.

Information Fusion, Journal Year: 2024, Volume and Issue: 117, P. 102809 - 102809

Published: Nov. 30, 2024

Language: Английский

Citations

8
Marine Equipment Siting Using Machine-Learning-Based Ocean Remote Sensing Data: Current Status and Future Prospects DOI Open Access
Dapeng Zhang, Yunsheng Ma, Huiling Zhang

et al.

Sustainability, Journal Year: 2024, Volume and Issue: 16(20), P. 8889 - 8889

Published: Oct. 14, 2024

As the global climate changes, there is an increasing focus on oceans and their protection exploitation. However, exploration of necessitates construction marine equipment, siting such equipment has become a significant challenge. With ongoing development computers, machine learning using remote sensing data proven to be effective solution this problem. This paper reviews history technology, introduces conditions required for site selection through measurement analysis, uses cluster analysis methods identify areas as research hotspot ocean sensing. The aims integrate into Through review discussion article, limitations shortcomings current stage are identified, relevant proposals put forward.

Language: Английский

Citations

4
IMC-YOLO: a detection model for assisted razor clam fishing in the mudflat environment DOI Creative Commons

Jianhao Xu,

Lijie Cao,

Lanlan Pan

et al.

PeerJ Computer Science, Journal Year: 2025, Volume and Issue: 11, P. e2614 - e2614

Published: Jan. 10, 2025

In intertidal mudflat culture (IMC), the fishing efficiency and degree of damage to nature have always been a pair irreconcilable contradictions. To improve razor clam at same time reduce natural environment, in this study, burrows dataset is established, an intelligent method proposed, which realizes accurate identification counting by introducing object detection technology into activity. A model called culture-You Only Look Once (IMC-YOLO) proposed study making improvements upon You version 8 (YOLOv8). firstly, end backbone network, Iterative Attention-based Intrascale Feature Interaction (IAIFI) module was designed adopted model's focus on advanced features. Subsequently, effectiveness detecting difficult targets such as with small sizes, head network refactored. Then, FasterNet Block used replace Bottleneck, achieves more effective feature extraction while balancing accuracy size. Finally, Three Branch Convolution Attention Mechanism (TBCAM) enables specific region interest accurately. After testing, IMC-YOLO achieved mAP50, mAP50:95, F1best 0.963, 0.636, 0.918, respectively, representing 2.2%, 3.5%, 2.4% over baseline model. Comparison other mainstream models confirmed that strikes good balance between numbers parameters.

Language: Английский

Citations

0
Region gradient-guided diffusion model for underwater image enhancement DOI
Jinxin Shao, Haosu Zhang, Jianming Miao

et al.

Machine Vision and Applications, Journal Year: 2025, Volume and Issue: 36(2)

Published: Jan. 17, 2025

Language: Английский

Citations

0
HUTDNet: A joint unmixing and target detection network for underwater hyperspectral imagery DOI Creative Commons
Qi Li,

Xingyuan Zu,

Ming Zhang

et al.

International Journal of Applied Earth Observation and Geoinformation, Journal Year: 2025, Volume and Issue: 136, P. 104374 - 104374

Published: Jan. 18, 2025

Language: Английский

Citations

0
MUFFNet: lightweight dynamic underwater image enhancement network based on multi-scale frequency DOI Creative Commons
D. J. Kong, Yandi Zhang, Xiaohu Zhao

et al.

Frontiers in Marine Science, Journal Year: 2025, Volume and Issue: 12

Published: Feb. 11, 2025

Introduction The advancement of Underwater Human-Robot Interaction technology has significantly driven marine exploration, conservation, and resource utilization. However, challenges persist due to the limitations underwater robots equipped with basic cameras, which struggle handle complex environments. This leads blurry images, severely hindering performance automated systems. Methods We propose MUFFNet, an image enhancement network leveraging multi-scale frequency analysis address challenge. introduces a frequency-domain-based convolutional attention mechanism extract spatial information effectively. A Multi-Scale Enhancement Prior algorithm enhances high-frequency low-frequency features while Information Flow module mitigates stratification blockage. Joint Loss framework facilitates dynamic optimization. Results Experimental results demonstrate that MUFFNet outperforms existing state-of-the-art models consuming fewer computational resources aligning enhanced images more closely human visual perception. Discussion generated by exhibit better alignment perception, making it promising solution for improving robotic vision

Language: Английский

Citations

0
State-of-the-art techniques for optical underwater image enhancement DOI
Navya Joseph,

S. N. Kumar,

S. Kannadhasan

et al.

International Journal of Image and Data Fusion, Journal Year: 2025, Volume and Issue: 16(1)

Published: March 10, 2025

Language: Английский

Citations

0
Joint Luminance-Saliency Prior and Attention for Underwater Image Quality Assessment DOI Creative Commons
Zhiqiang Lin, Zhouyan He, Chongchong Jin

et al.

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(16), P. 3021 - 3021

Published: Aug. 17, 2024

Underwater images, as a crucial medium for storing ocean information in underwater sensors, play vital role various tasks. However, they are prone to distortion due the imaging environment, which leads decline visual quality, is an urgent issue marine vision systems address. Therefore, it necessary develop image enhancement (UIE) and corresponding quality assessment methods. At present, most (UIQA) methods primarily rely on extracting handcrafted features that characterize degradation attributes, struggle measure complex mixed distortions often exhibit discrepancies with human perception practical applications. Furthermore, current UIQA lack consideration of perspective enhanced effects. To this end, paper employs luminance saliency priors critical first time effect global local achieved by UIE algorithms, named JLSAU. The proposed JLSAU built upon overall pyramid-structured backbone, supplemented Luminance Feature Extraction Module (LFEM) Saliency Weight Learning (SWLM), aim at obtaining multiple scales. supplement aims perceive visually sensitive luminance, including histogram statistical grayscale positional information. reflects variation both spatial channel domains. Finally, effectively model relationship among different levels contained multi-scale features, Attention Fusion (AFFM) proposed. Experimental results public UIQE UWIQA datasets demonstrate outperforms existing state-of-the-art

Language: Английский

Citations

3