首页 >  , Vol. , Issue () : -

摘要

全文摘要次数: 116 全文下载次数: 182
引用本文:

DOI:

10.11834/jrs.20211354

收稿日期:

2021-05-26

修改日期:

2021-08-25

PDF Free   EndNote   BibTeX
生成式知识迁移的SAR舰船检测
娄欣, 王晗, 卢昊, 张文驰
北京林业大学
摘要:

为解决基于深度卷积神经网络进行SAR舰船检测网络训练过程中数据获取、数据标注等问题,本文提出一种生成式知识迁移的SAR舰船检测框架,该框架由生成式知识迁移网络和舰船检测网络两部分组成。通过知识迁移网络生成与有标注的光学遥感图像空间分布一致且包含SAR图像特征的带标注模拟图像;使用所生成的带标注模拟图像,进一步优化舰船检测网络,以提高基于深度卷积神经网络的舰船检测的泛化性能。SAR-Ship-Detection-Datasets (SSDD)和AIR-SARShip-1.0两个公开数据集上的实验结果表明,该框架有效提高了在仅包含少量标注SAR图像样本情况下的舰船目标检测效果,可显著降低舰船在复杂背景图像中漏检和误检的概率。

SAR ship detection via generative knowledge transfer
Abstract:

To address data acquisition and labeling data in the training process of SAR ship detection network based on deep convolutional neural network, we propose a SAR ship detection framework via generative knowledge transfer of a knowledge transfer network for SAR image generation and a SAR ship detection network. The knowledge transfer network generates simulated SAR images that is consistent with the spatial distribution of labeled optical remote sensing images and has similar the feature distribution of SAR images, and these simulated SAR images can optimize the SAR ship detection network to improve the generalization ability. The results on the two public datasets of SAR-Ship-Detection-Datasets (SSDD) and AIR-SARShip-1.0 show that the framework effectively improves ship detection performance in the case of small train set, and can significantly reduce the probability of missing alarm and false alarm in detecting complex background images.

本文暂时没有被引用!

欢迎关注学报微信

遥感学报交流群 分享按钮