融合多重机制的SAR舰船检测  

SAR ship detection with multi-mechanism fusion

作　　者：肖振久[1] 林渤翰 曲海成[1] Xiao Zhenjiu;Lin Bohan;Qu Haicheng(College of Software,Liaoning Technical University,Huludao 125105,China)

机构地区：[1]辽宁工程技术大学软件学院,葫芦岛125105

出　　处：《中国图象图形学报》2024年第2期545-558,共14页Journal of Image and Graphics

基　　金：辽宁省高等学校基本科研项目(LJKMZ20220699);辽宁工程技术大学学科创新团队资助项目(LNTU20TD-23)。

摘　　要：目的 针对合成孔径雷达(synthetic aperture radar, SAR)图像噪声大、成像特征不明显,尤其在复杂场景更容易出现目标误检和漏检的问题,提出了一种融合多重机制的SAR舰船检测方法,用于提高SAR舰船检测的精度。方法 在预处理部分,设计了U-Net Denoising模块,通过调整噪声方差参数L的范围来抑制相干斑噪声对图像的干扰。在YOLOv7(you only look once v7)主干网络构建MLAN_SC(maxpooling layer aggregation network that incorporate select kernel and contextual Transformer)结构,加入SK(selective kernel)通道注意力机制至下采样阶段,增强关键信息提取能力和特征表达能力。为解决MP(multiple pooling)结构中上下分支特征不平衡的问题,改善误检情况,融入上下文信息提取模块(contextual Transformer block, COT),利用卷积提取上下文信息,将局部信息和全局信息结合起来,使图像特征能够更有效地提取出来。在头部引入SPD卷积(space-to-depth convolution, SPD-Conv),增强小目标的检测能力。用WIoU(wise intersection over union)损失函数替换CIoU(complete intersection over union)损失函数,运用动态聚焦机制,在复杂图像上加强对目标的定位能力。结果 在SSDD(SAR ship detection dataset)数据集和HRSID (high-resolution SAR images dataset)数据集上进行了实验对比,结果表明,改进后的方法相比于YOLOv7,AP(average precision)可达到99.25%和89.73%,分别提升了4.38%和2.57%,准确率和召回率为98.41%,93.24%和94.79%,81.83%,优于对比方法。结论 本文通过融合多重机制改进YOLOv7方法,提升了对目标的定位能力,显著改善了SAR舰船检测中复杂舰船的误检和漏检情况,进一步提高了SAR舰船检测精度。Objective In recent years,the efficacy of synthetic aperture radar(SAR)has been increasingly recognized inthe fields of maritime surveillance and vessel detection due to its remarkable all-weather and day-to-night imaging capa⁃bility.The ability of SAR systems to penetrate through clouds and fog has enabled high-quality imaging of the sea surfaceunder various weather conditions.However,SAR imaging is frequently hindered by excessive noise and unclear imagingfeatures,which can lead to erroneous detection in complex maritime environments.In response to this challenge,thisstudy presents an innovative approach that combines state-of-the-art deep learning and computer vision techniques toimprove the accuracy of SAR ship detection.By incorporating several critical enhancements into the YOLOv7 algorithm,the proposed method aims to enhance the capability of SAR systems to identify and track vessels accurately on the sea sur⁃face.The potential of this method is significant for maritime security and surveillance systems,because the accurate andreliable detection of vessels is paramount to ensuring the safety and security of shipping lanes and ports worldwide.Method The present study proposes a novel method that offers significant improvements to the YOLOv7 algorithm for SARship detection.In particular,a U-Net denoising module is designed in the preprocessing stage to suppress coherent specklenoise interference by leveraging deep learning techniques to model the range of parameter L.Moreover,the MLAN_SCstructure is built in the YOLOv7 backbone network.To enhance key information extraction and deep feature expressionabilities,the proposed method also introduces the selective kernel(SK)attention mechanism to improve the false detectionrate.The contextual Transformer(COT)block is integrated into the backbone network to solve the problem of unbalancedupper and lower branch features in the multi-processings(MP)structure and improve the false detection situation.TheCOT block uses convolutional operations and combines local an

关 键 词：SAR图像 舰船检测 YOLOv7 注意力机制 上下文信息提取 SPD卷积(SPD-Conv) WIoU损失函数 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] TN911.73[自动化与计算机技术—控制科学与工程]