条带卷积与方向感知排斥损失用于有向船舶检测
Strip Convolution and Direction-Aware Exclusion Loss for Oriented Ship Detection
- East China Jiaotong University（华东交通大学）
- Jiangxi Vocational University of Foreign Studies（江西外语外贸职业学院）
机构由 AI 辅助整理，请以论文原文为准。
AI总结：
针对超高分辨率遥感图像中有向船舶检测的细长船体和密集目标问题，提出结合条带卷积与类别-方向感知排斥损失的检测器，在HRSC2016和DIOR-R上取得显著性能提升。
AI中文摘要：
在超高分辨率（VHR）遥感图像中，由于船体细长几何形状和复杂港口场景中密集的目标分布，有向船舶检测仍然具有挑战性。现有方法通常分别处理几何表示和重复抑制问题。为联合解决这些问题，我们提出了一种包含两个互补组件的定向船舶检测器。C3k2_Strip模块采用正交条带卷积以更好地捕获细长船体结构，而类别感知方向感知排斥损失（CA-DAEL）利用类别、方向和置信度线索抑制冗余预测。在HRSC2016和DIOR-R上的实验分别达到78.45%和53.71%的mAP50:95，仅需2.91M参数。在HRSC2016上，所提方法相比YOLOv11-OBB基线将mAP50:95提升了6.32个百分点，证明了其对于准确有向船舶检测的有效性。
英文摘要：
Oriented ship detection in very high resolution (VHR) remote sensing imagery remains challenging due to elongated hull geometry and dense target distributions in complex port scenes. Existing methods typically address geometric representation and duplicate suppression separately. To jointly tackle these issues, we propose an oriented ship detector with two complementary components. The C3k2_Strip module employs orthogonal strip convolutions to better capture elongated hull structures, while the Class-Aware Direction-Aware Exclusion Loss (CA-DAEL) suppresses redundant predictions using class, direction, and confidence cues. Experiments on HRSC2016 and DIOR-R achieve 78.45% and 53.71% mAP50:95, respectively, with only 2.91M parameters. On HRSC2016, the proposed method improves mAP50:95 by 6.32 percentage points over the YOLOv11-OBB baseline, demonstrating its effectiveness for accurate oriented ship detection.
