一种面向SAR图像多尺度舰船目标的检测算法

BAPT-YOLOv8n: Detection Algorithm for Multi-scale Ship Targets in SAR Images

摘要:

在合成孔径雷达图像的舰船目标检测任务中，不同目标的尺度多样性给检测算法带来了巨大挑战；为了解决多尺度舰船目标检测难题，提出了一种BAPT-YOLOv8n算法，该算法以YOLOv8n为基础框架，通过引入卷积块注意力模块重构颈部金字塔网络，提升了对多层次特征的融合能力与对多尺度目标的特征提取能力；此外，采用基于Transformer的检测头结构，进一步提高特征表示能力和上下文信息利用能力，从而改善了小目标的检测效果；在HRSID数据集和SSDD数据集上的对比实验表明，所提算法在检测精度上分别达到93.6%与98.9%，优于其他对比算法；消融实验进一步验证了算法中各改进部分的有效性；表明该算法能够更好适应多尺度舰船目标检测问题。

Abstract:

In the task of ship target detection in synthetic aperture radar (SAR) images, the diverse scales of different targets pose significant challenges to detection algorithms. To address these challenges, a BAPT-YOLOv8n algorithm is proposed. Built upon the YOLOv8n framework, this algorithm enhances the fusion capability of multi-level features and the feature extraction capability for multi-scale targets by introducing convolutional block attention modules to reconstruct the neck pyramid network. Additionally, employing a Transformer-based detection head structure further improves feature representation and context utilization, thereby enhancing the detection performance of small targets. Comparative experiments on the HRSID and SSDD datasets show that the proposed algorithm achieves detection accuracies of 93.6% and 98.9%, respectively, surpassing other benchmark algorithms. Ablation experiments further validate the effectiveness of each improvement in the algorithm, demonstrating its capability to better adapt to multi-scale ship target detection tasks.