Research on non‐contact infrared imaging technique for multilayer storage identification of oil tanks based on an improved edge‐detection algorithm

The Canadian Journal of Chemical Engineering, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 2, 2025

Abstract Detection of internal storage objects in tanks is crucial for production the petrochemical industry and chemical raw material storage. Compared to traditional methods, infrared detection provides benefits like non‐contact operation, safety, efficiency. In image processing, utilizing edge obtain information an advanced approach. By analyzing thermal texture tank images extracting boundaries between different regions, it possible predict volume To address issues noise, lack clarity, discontinuity existing a novel algorithm called wavelet transform mathematical morphological fusion improve (WMF‐IED) proposed. Roberts, Prewitt, Sobel, Laplacian Gaussian (LOG) WMF‐IED offers several advantages. It not only clear continuous edges but also exhibits minimal mean squared error (MSE). Additionally, achieves maximum signal‐to‐noise ratio (SNR) peak (PSNR). These factors show proposed algorithm's superior performance. Moreover, experimental platform was designed constructed analyze contents using algorithm. The results demonstrate that has strong universality can detect various prediction errors are less than 4% 6% liquid level sludge detection, respectively. Based on analysis results, recommended sampling value proposed, which be selected minimum error.

Language: Английский

---

Ship_YOLO: General ship detection based on mixed distillation and dynamic task-aligned detection head

Ocean Engineering, Journal Year: 2025, Volume and Issue: 323, P. 120616 - 120616

Published: Feb. 17, 2025

Language: Английский

---

ACDF-YOLO: Attentive and Cross-Differential Fusion Network for Multimodal Remote Sensing Object Detection

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(18), P. 3532 - 3532

Published: Sept. 23, 2024

Object detection in remote sensing images has received significant attention for a wide range of applications. However, traditional unimodal images, whether based on visible light or infrared, have limitations that cannot be ignored. Visible are susceptible to ambient lighting conditions, and their accuracy can greatly reduced. Infrared often lack rich texture information, resulting high false-detection rate during target identification classification. To address these challenges, we propose novel multimodal fusion network model, named ACDF-YOLO, basedon the lightweight efficient YOLOv5 structure, which aims amalgamate synergistic data from both infrared imagery, thereby enhancing efficiency imagery. Firstly, shuffle module is designed assist extracting features various modalities. Secondly, deeper information achieved by introducing new cross-modal difference fuse been acquired. Finally, combine two modules mentioned above an effective manner achieve ACDF. The ACDF not only enhances characterization ability fused but also further refines capture reinforcement important channel features. Experimental validation was performed using several publicly available real-world datasets. Compared with other advanced methods, ACDF-YOLO separately 95.87% 78.10% mAP0.5 LLVIP VEDAI datasets, demonstrating deep different modal effectively improve object detection.

Language: Английский

---

VIOS-Net: A Multi-Task Fusion System for Maritime Surveillance Through Visible and Infrared Imaging

Journal of Marine Science and Engineering, Journal Year: 2025, Volume and Issue: 13(5), P. 913 - 913

Published: May 6, 2025

Automatic ship monitoring models leveraging image recognition have become integral to regulatory applications within maritime management, with multi-source co-monitoring serving as the primary method for achieving comprehensive, round-the-clock surveillance. Despite their widespread use, existing predominantly train each data source independently or simultaneously multiple sources without fully optimizing integration of similar information. This approach, while capable all-weather detection, results in underutilization features from related and unnecessary repetition model training, leading excessive time consumption. To address these inefficiencies, this paper introduces a novel multi-task learning framework designed enhance utilization diverse information sources, thereby reducing training time, lowering costs, improving accuracy. The proposed model, VIOS-Net, integrates advantages both visible infrared meet challenges all-weather, all-day under complex environmental conditions. VIOS-Net employs Shared Bottom network architecture, utilizing shared specific feature extraction modules at model’s lower upper layers, respectively, optimize system’s capabilities maximize efficiency. experimental demonstrate that achieves an accuracy 96.20% across spectral datasets, significantly outperforming baseline ResNet-34 which attained accuracies only 4.86% 9.04% data, respectively. Moreover, reduces number parameters by 48.82% compared baseline, optimal performance multi-spectral monitoring. Extensive ablation studies further validate effectiveness individual framework.

Language: Английский

---

Improved RT-DETR for Infrared Ship Detection Based on Multi-Attention and Feature Fusion

Journal of Marine Science and Engineering, Journal Year: 2024, Volume and Issue: 12(12), P. 2130 - 2130

Published: Nov. 22, 2024

Infrared cameras form images by capturing the thermal radiation emitted objects in infrared spectrum, making them complex sensors widely used maritime surveillance. However, broad spectral range of band makes it susceptible to environmental interference, which can reduce contrast between target and background. As a result, detecting targets marine environments remains challenging. This paper presents novel enhanced detection model developed from real-time transformer (RT-DETR), is designated as MAFF-DETR. The incorporates backbone integrating CSP parallelized patch-aware attention enhance sensitivity imagery. Additionally, channel module employed during feature selection, leveraging high-level features filter low-level information enabling efficient multi-level fusion. model’s performance on resource-constrained devices further incorporating advanced techniques such group convolution ShuffleNetV2. experimental results show that, although RT-DETR algorithm still experiences missed detections under severe object occlusion, has significantly improved overall performance, including 1.7% increase mAP, reduction 4.3 M parameters, 5.8 GFLOPs decrease computational complexity. It be applied tasks coastline monitoring search rescue.

Language: Английский

---

Real-Time Long-Distance Ship Detection Architecture Based on YOLOv8

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 116086 - 116104

Published: Jan. 1, 2024

Language: Английский

---

Dynamic Tracking Matched Filter with Adaptive Feedback Recurrent Neural Network for Accurate and Stable Ship Extraction in UAV Remote Sensing Images

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(12), P. 2203 - 2203

Published: June 17, 2024

In an increasingly globalized world, the intelligent extraction of maritime targets is crucial for both military defense and traffic monitoring. The flexibility cost-effectiveness unmanned aerial vehicles (UAVs) in remote sensing make them invaluable tools ship extraction. Therefore, this paper introduces a training-free, highly accurate, stable method UAV images. First, we present dynamic tracking matched filter (DTMF), which leverages concept time as tuning factor to enhance traditional (MF). This refinement gives DTMF superior adaptability consistent detection performance across different points. Next, rigorously integrated into recurrent neural network (RNN) framework using mathematical derivation optimization principles. To further improve convergence robust RNN solution, design adaptive feedback (AFRNN), optimally solves problem. Finally, evaluate methods based on accuracy specific evaluation metrics. results show that proposed achieve over 99% overall KAPPA coefficients above 82% various scenarios. approach excels complex scenes with multiple background interference, delivering distinct precise while minimizing errors. efficacy extracting was validated through rigorous testing.

Language: Английский

---

An Algorithm for Ship Detection in Complex Observation Scenarios Based on Mooring Buoys

Journal of Marine Science and Engineering, Journal Year: 2024, Volume and Issue: 12(7), P. 1226 - 1226

Published: July 20, 2024

Marine anchor buoys, as fixed-point profile observation platforms, are highly susceptible to the threat of ship collisions. Installing cameras on buoys can effectively monitor and collect evidence from ships. However, when using a camera capture images, it is often affected by continuous shaking rainy foggy weather, resulting in problems such blurred images rain fog occlusion. To address these problems, this paper proposes an improved YOLOv8 algorithm. Firstly, polarized self-attention (PSA) mechanism introduced preserve high-resolution features original deep convolutional neural network solve problem image spatial resolution degradation caused shaking. Secondly, introducing multi-head (MHSA) neck network, interference background weakened, feature fusion ability improved. Finally, head model combines additional small object detection heads improve accuracy detection. Additionally, enhance algorithm’s adaptability scenarios, simulates including blur, rain, conditions. In end, numerous comparative experiments self-made dataset show that algorithm proposed study achieved 94.2% mAP50 73.2% mAP50:95 various complex environments, which superior other advanced algorithms.

Language: Английский

---

Infrared Bilateral Polarity Ship Detection in Complex Maritime Scenarios

Sensors, Journal Year: 2024, Volume and Issue: 24(15), P. 4906 - 4906

Published: July 29, 2024

In complex maritime scenarios where the grayscale polarity of ships is unknown, existing infrared ship detection methods may struggle to accurately detect among significant interference. To address this issue, paper first proposes an image smoothing method composed Grayscale Morphological Reconstruction (GMR) and a Relative Total Variation (RTV). Additionally, considering uniformity integrating shape spatiotemporal features established for detecting bright dark in scenarios. Initially, input images undergo opening (closing)-based GMR preserve (bright) blobs with opposite suppressed, followed by relative total variation model reduce clutter enhance contrast ship. Subsequently, Maximally Stable Extremal Regions (MSER) are extracted from smoothed as candidate targets, results channels merged. Shape then utilized eliminate interference, yielding single-frame results. Finally, leveraging stability fluctuation clutter, true targets preserved through multi-frame matching strategy. Experimental demonstrate that proposed outperforms ITDBE, MRMF, TFMSER seven sequences, achieving accurate effective both targets.

Language: Английский

---

Improved lightweight infrared road target detection method based on YOLOv8

Infrared Physics & Technology, Journal Year: 2024, Volume and Issue: 141, P. 105497 - 105497

Published: Aug. 15, 2024

Language: Английский

---