End-to-End Method With Transformer for 3-D Detection of Oil Tank From Single SAR Image

Ma, Chao; Zhang, Yueting; Guo, Jiayi; Hu, Yuxin; Geng, Xiurui; Li, Fangfang; Lei, Bin; Ding, Chibiao

doi:10.1109/tgrs.2021.3127986

Cited by 8 publications

(6 citation statements)

References 49 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This marks a paradigm shift in object detection from a multi-step process to an end-to-end solution. DETR simplifies the process of object detection and has achieved significant results in optical object detection [ 33 , 34 , 35 ]. Additionally, researchers have also explored the application of DETR in SAR object detection.…”

Section: Related Studiesmentioning

confidence: 99%

“…Additionally, researchers have also explored the application of DETR in SAR object detection. Notable works include OEGR-DETR [ 13 ], which proposes an OEM module and GRC loss for enhancing the localization of rotated objects; Chao Ma [ 34 ] et al propose cylinder IOU and incident angle priors for end-to-end 3D SAR object detection; TSDet [ 35 ] uses an enhanced attention module for precise identification of SAR ship targets. These works validate the effectiveness of DETR for SAR object detection, and our method will further explore the performance of DETR in multi-class SAR object detection.…”

Section: Related Studiesmentioning

confidence: 99%

See 1 more Smart Citation

CCDN-DETR: A Detection Transformer Based on Constrained Contrast Denoising for Multi-Class Synthetic Aperture Radar Object Detection

Zhang,

Zheng,

et al. 2024

Sensors

View full text Add to dashboard Cite

The effectiveness of the SAR object detection technique based on Convolutional Neural Networks (CNNs) has been widely proven, and it is increasingly used in the recognition of ship targets. Recently, efforts have been made to integrate transformer structures into SAR detectors to achieve improved target localization. However, existing methods rarely design the transformer itself as a detector, failing to fully leverage the long-range modeling advantages of self-attention. Furthermore, there has been limited research into multi-class SAR target detection. To address these limitations, this study proposes a SAR detector named CCDN-DETR, which builds upon the framework of the detection transformer (DETR). To adapt to the multiscale characteristics of SAR data, cross-scale encoders were introduced to facilitate comprehensive information modeling and fusion across different scales. Simultaneously, we optimized the query selection scheme for the input decoder layers, employing IOU loss to assist in initializing object queries more effectively. Additionally, we introduced constrained contrastive denoising training at the decoder layers to enhance the model’s convergence speed and improve the detection of different categories of SAR targets. In the benchmark evaluation on a joint dataset composed of SSDD, HRSID, and SAR-AIRcraft datasets, CCDN-DETR achieves a mean Average Precision (mAP) of 91.9%. Furthermore, it demonstrates significant competitiveness with 83.7% mAP on the multi-class MSAR dataset compared to CNN-based models.

show abstract

Section: Related Studiesmentioning

confidence: 99%

Section: Related Studiesmentioning

confidence: 99%

CCDN-DETR: A Detection Transformer Based on Constrained Contrast Denoising for Multi-Class Synthetic Aperture Radar Object Detection

Zhang,

Zheng,

et al. 2024

Sensors

View full text Add to dashboard Cite

show abstract

“…Cui et al [34] introduced a lightweight framework based on threshold neural networks, adaptively determining optimal detection thresholds within sliding windows. Recent research has introduced Transformer into SAR detection tasks, Ma et al [35] improved upon end-to-end Transformers by incorporating incident angles as prior labels and introducing feature descriptor operators based on scattering centers. Zhou et al [36] proposed a lightweight meta-learning approach for small sample SAR object detection.…”

Section: Related Work 21 Object Detection In Sar Imagesmentioning

confidence: 99%

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

Luo,

Zhang,

Guo

et al. 2024

Remote Sensing

View full text Add to dashboard Cite

The unique imaging modality of synthetic aperture radar (SAR) has posed significant challenges for object detection, making it more complex to acquire and interpret than optical images. Recently, numerous studies have proposed cross-domain adaptive methods based on convolutional neural networks (CNNs) to promote SAR object detection using optical data. However, existing cross-domain methods focus on image features, lack improvement on input data, and ignore the valuable supervision provided by few labeled SAR images. Therefore, we propose a semi-supervised cross-domain object detection framework that uses optical data and few SAR data to achieve knowledge transfer for SAR object detection. Our method focuses on the data processing aspects to gradually reduce the domain shift at the image, instance, and feature levels. First, we propose a data augmentation method of image mixing and instance swapping to generate a mixed domain that is more similar to the SAR domain. This method fully utilizes few SAR annotation information to reduce domain shift at image and instance levels. Second, at the feature level, we propose an adaptive optimization strategy to filter out mixed domain samples that significantly deviate from the SAR feature distribution to train feature extractor. In addition, we employ Vision Transformer (ViT) as feature extractor to handle the global feature extraction of mixed images. We propose a detection head based on normalized Wasserstein distance (NWD) to enhance objects with smaller effective regions in SAR images. The effectiveness of our proposed method is evaluated on public SAR ship and oil tank datasets.

show abstract

“…(1) due to SAR sensors' side-looking echolocation manner, oil tanks would not present regular circular shapes but discrete points with layover effects on images, and (2) as SAR sensors coherently record backscattered echoes, oil tanks are difficult to be distinguished from background clutter and speckle. Thus, conventional optical image-targeted approaches are inappropriate for SAR data [2,8,9].…”

Section: Introductionmentioning

confidence: 99%

Automatic Monitoring of Oil Tank 3D Geometry and Storage Changes With Interferometric Coherence and SAR Intensity Information

Tsai,

Huang,

Chen

et al. 2024

IEEE J. Sel. Top. Appl. Earth Observations Remote Sensing

View full text Add to dashboard Cite

Continuous monitoring of oil tanks is vital for analyzing local fuel consumption. Synthetic Aperture Radar (SAR) has been a popular data source as it guarantees day-and-night and all-weather sensing capacity. However, most earlier studies adopt a scene-wise and oil tank-wise scheme, which is inefficient as there can be hundreds of oil tanks on an oil depot, while only a few are dynamic. Also, no study explores both intensity coherence and interferometric coherence for oil tank dynamics mapping. This study proposes a novel three-stage strategy to detect all oil tanks, identify dynamic oil tanks, and estimate their fuel volume changes based on both the intensity and phase information of SAR in both slant-range and geocoded projections. Results indicate that the intensity coherence can perfectly differentiate dynamic and stable oil tanks (a Jeffries-Matusita distance of 1.997) and is less vulnerable to repeat-pass SAR factors, such as baselines and atmospheric conditions. Via evaluating estimations' consistency, our scattering keypoint detection exhibits 0.23 and 0.87 m precision of tank heights and diameters, respectively. By validation with ground truth data, oil tanks exhibiting floatingroof changes larger than 0.23 m are correctly identified. Also, the estimated storage changes agree well with actual changes with an R-squared value of 0.98 and a root-mean-square error (RMSE) corresponding to 1.05 m biases in floating-roof heights. These quantitative assessments confirm the robustness and broad applicability of our non-in situ data-needed approach, highlighting the opportunity to utilize spotlight SAR data to automatically and comprehensively monitor oil tank dynamics in remote sites.

show abstract

End-to-End Method With Transformer for 3-D Detection of Oil Tank From Single SAR Image

Cited by 8 publications

References 49 publications

CCDN-DETR: A Detection Transformer Based on Constrained Contrast Denoising for Multi-Class Synthetic Aperture Radar Object Detection

CCDN-DETR: A Detection Transformer Based on Constrained Contrast Denoising for Multi-Class Synthetic Aperture Radar Object Detection

SAR-CDSS: A Semi-Supervised Cross-Domain Object Detection from Optical to SAR Domain

Automatic Monitoring of Oil Tank 3D Geometry and Storage Changes With Interferometric Coherence and SAR Intensity Information

Contact Info

Product

Resources

About