2026-10-11T07:23:38Z https://api.figshare.com/v2/oai

oai:figshare.com:article/33988485 2026-09-24T17:37:54Z category_146 category_14 category_873 category_45 category_734 category_931 category_811 portal_5 item_type_3 month_year_09_2026

<p>Detailed parameters of the training model.</p> Miao Zhou (1440016) Dezhi Han (4391263) Xiang Shen (127119) Yangshuyi Xu (16462504) Chongqing Chen (14185807) Space Science Molecular Biology Chemical Sciences not elsewhere classified Sociology Biological Sciences not elsewhere classified Information Systems not elsewhere classified Mathematical Sciences not elsewhere classified tiny ship targets target detection scenarios synthetic aperture radar speckle noise interference spatial dependency modeling resolution detection head provide technical support maritime traffic monitoring marine environmental monitoring incorporating channel shuffling improved c3k2 module detecting tiny ships complex marine backgrounds 58 percentage points scale feature representation feature aggregation strategies experimental results show small ship targets extremely small targets effectively improve shallow div >< p 53 %, corresponding yolo achieves values feature perception results validate stem stage ssdd datasets source code proposed algorithm parameterized convolutions paper proposes localization accuracy insufficient accuracy frequency perception filtering mechanism enhance multi <div><p>Synthetic Aperture Radar (SAR) ship target detection holds significant application value in maritime traffic monitoring and marine environmental monitoring. However, due to challenges such as small ship targets, complex marine backgrounds, and speckle noise interference, existing methods still suffer from insufficient accuracy in small-target detection scenarios. To address this issue, this paper proposes a high-resolution detection algorithm named LHR-YOLO, based on YOLO11. In the Stem stage, a Gaussian–Laplacian edge-enhancement and Gaussian-filtering mechanism is introduced to effectively improve shallow-feature perception for tiny ship targets. In the Backbone, an improved C3k2 module is designed by incorporating channel shuffling, re-parameterized convolutions, and feature aggregation strategies to enhance multi-scale feature representation. Meanwhile, in the Neck–Detect stage, high-frequency perception, spatial dependency modeling, and a high-resolution detection head are integrated to further improve the localization accuracy of extremely small targets. Experimental results show that LHR-YOLO achieves values of 71.59% and 73.40% on the HRSID and SSDD datasets, respectively. Compared with YOLO11n, LHR-YOLO improves on HRSID from 53.95% to 60.53%, corresponding to an increase of 6.58 percentage points. These results validate the effectiveness of the proposed algorithm for detecting tiny ships in complex SAR maritime scenes and provide technical support for high-precision maritime target monitoring.The source code is available at: <a href="https://github.com/LRYTH/LHR-YOLO" target="_blank">https://github.com/LRYTH/LHR-YOLO</a>.</p></div> 2026-09-24T17:37:40Z Dataset Dataset 10.1371/journal.pone.0359334.t003 https://figshare.com/articles/dataset/_p_Detailed_parameters_of_the_training_model_p_/33988485 CC BY 4.0