引用本文
  •    [点击复制]
  •    [点击复制]
【打印本页】 【在线阅读全文】【下载PDF全文】 查看/发表评论下载PDF阅读器关闭

←前一篇|后一篇→

过刊浏览    高级检索

本文已被:浏览 3629次   下载 705 本文二维码信息
码上扫一扫!
基于深度卷积网络的SAR图像目标检测识别
李君宝,杨文慧,许剑清,彭宇
0
(哈尔滨工业大学自动化测试与控制系)
摘要:
在SAR图像解译应用领域,目标的自动检测与识别一直是该领域的研究重点和热点,也是该领域的研究难点。针对SAR图像的目标检测与识别方法一般由滤波、分割、特征提取和目标识别等多个相互独立的步骤组成。复杂的流程不仅限制了SAR图像目标检测识别的效率,多步骤处理也使模型的整体优化难以进行,进而制约了目标检测识别的精度。采用近几年在计算机视觉领域表现突出的深度学习方法来处理SAR图像的目标检测识别问题,通过使用CNN、Fast RCNN以及Faster RCNN等模型对MSTAR SAR公开数据集进行目标识别及目标检测实验,验证了卷积神经网络在SAR图像目标识别领域的有效性及高效性,为后续该领域的进一步研究应用奠定了基础。
关键词:  SAR  目标检测识别  CNN  Fast RCNN  Faster RCNN
DOI:
基金项目:
Deep Convolutional Network Based SAR Image Object Detection and Recognition
LI Jun-bao,YANG Wen-hui,XU Jian-qing,PENG Yu
(Automatic Test and Control Institute, Harbin Institute of Technology)
Abstract:
Automatic target detection and recognition has been the focus in SAR image interpretation field. Generally, the target detection and recognition method of SAR image is divide into independent 4 steps, filtering, segmentation, feature extraction and target recognition. Complex process limits the efficiency of SAR image target detection and recognition. Too many steps make it difficult to optimize the whole model, so the accuracy of method is restricted. In recent years, deep learning has been the famous method in many important computer vision challenges. Deep learning has led to a revolutionary change in the field of computer vision. In this paper, we apply deep learning to SAR image automatic target detection and recognition task. And we verify the feasibility and efficiency of deep learning method through experiments on MSTAR SAR image sets
Key words:  SAR  Target detection and recognition  CNN  Fast RCNN  Faster RCNN

用微信扫一扫

用微信扫一扫