基于长短路融合及数据平衡的SAR船舶检测算法

SAR Ship Detection Algorithm Based on Long-Short Path Fusion and Data Balance

针对SAR图像检测船舶任务中的目标小、近岸样本目标检测困难等问题,文章提出一种名为长短路特征融合网络(Long and Short path Feature Fusion Network,LSFF-Net)的船舶检测网络。该网络通过长短路特征融合模块有效协调了大目标与小目标检测,避免小目标特征信息的丢失。网络中应用结构重参数化结构提高了模块学习能力。为了满足多尺度目标检测,加入特征金字塔网络,融合多尺度特征。为了应对近岸样本目标检测,设计数据重分配算法,提高了对近岸样本目标的检测精度。实验结果表明:在公开数据集检测时,算法的平均精度(Average Precision,AP)达到97.50%,优于主流目标检测算法。该方法为提高SAR图像中小目标和近岸样本目标检测精度提供了新的实现方案。

This paper proposes a ship detection network called the Long and Short path Feature Fusion Network(LSFF-Net)to address the challenges of detecting small and inshore samples in SAR image ship detection tasks.In LSFF-Net,the Long and Short path Feature Fusion Block(LSFF-Block)makes the model compatible with different scale target information.The application of structural re-parameterization enriches the module learning ability,and the multi-scale features are fused with the feature pyramid network.To address inshore target detection,a data redistribution algorithm is designed to increase detection accuracy of nearshore targets.The experimental results show that the proposed algorithm fully learns the information of the image and is more in line with the characteristics of SAR images.The average precision(AP)of the algorithm reaches 97.50%in the public data set detection results,which is better than the mainstream target detection algorithm.LSFF-Net provides a new solution for improving the accuracy of small and inshore target detection in SAR images.