SAR图像自动目标识别研究
发布时间：2018-07-16 16:03
【摘要】：合成孔径雷达自动目标识别(Synthetic Aperture Radar Automatic TargetRecognition, SAR ATR)是国防和民用领域中的具有挑战性的一个问题。近几十年来，SAR自动目标识别技术在预处理、特征提取及识别方面的研究已经取得了很大进展。SAR自动目标识别问题中会涉及到机器学习和模式识别领域的许多技术。随着SAR图像数据采集能力的不断增长，对这些图像进行准确、快速的理解和识别越来越引起人们的关注和重视。针对合成孔径雷达的自动目标识别问题，本文从SAR图像分割、特征提取与识别方面展开了深入的研究。本论文的主要工作如下： 1.针对SAR图像的预处理问题，，分析和讨论了几种常用的滤波和分割方法，提出一种阈值化技术与形态学方法相结合的SAR图像分割方法，可以更好地保持目标和阴影的细节部分，为后续的识别打下基础。 2.提出一种基于SAR目标和阴影特征融合的目标识别方法，该方法充分利用了SAR图像目标和阴影的结构信息；首先根据前面提出的SAR图像分割方法，将目标和阴影从杂波背景中分割出来，抑制背景杂波对后续识别的影响，然后分别提取目标和阴影的特征进行串行融合，最后结合最近邻分类器得到识别结果。给出实验结果，可以看出本方法具有良好的识别性能。 3.针对现有SAR图像目标识别存在的问题，提出一种快速的SAR图像目标识别方法，该方法引入压缩感知的思想，采用随机感知矩阵对多尺度图像特征进行降维，然后在降维后的特征集上利用邻分类器进行分类识别。实验结果表明，该方法可以有效地降低特征维数，减少运行时间。
[Abstract]:Synthetic Aperture Radar Automatic Target recognition (SAR ATR) is a challenging problem in the field of national defense and civil affairs. In recent decades, SAR automatic target recognition technology has made great progress in preprocessing, feature extraction and recognition. The SAR automatic target recognition problem involves many techniques in the field of machine learning and pattern recognition. With the increasing of SAR image data acquisition ability, people pay more and more attention to the accurate, fast understanding and recognition of these images. Aiming at the automatic target recognition problem of synthetic Aperture Radar (SAR), this paper focuses on SAR image segmentation, feature extraction and recognition. The main work of this paper is as follows: 1. Aiming at the problem of SAR image preprocessing, several common filtering and segmentation methods are analyzed and discussed. A new SAR image segmentation method combining thresholding technique and morphological method is proposed, which can better preserve the details of target and shadow. Lay the foundation for subsequent recognition. 2. A target recognition method based on SAR target and shadow feature fusion is proposed, which makes full use of the structure information of SAR image. The target and shadow are separated from the clutter background to suppress the influence of background clutter on the subsequent recognition. Then the features of the target and shadow are extracted for serial fusion and the recognition results are obtained by combining the nearest neighbor classifier. The experimental results show that the method has good recognition performance. In view of the existing problems in SAR image target recognition, a fast SAR image target recognition method is proposed. The method introduces the idea of compressed perception, and uses random perception matrix to reduce the dimension of multi-scale image features. Then, the neighbor classifier is used to recognize the feature set after dimensionality reduction. The experimental results show that this method can effectively reduce the characteristic dimension and the running time.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52
本文编号：2126901
[Abstract]:Synthetic Aperture Radar Automatic Target recognition (SAR ATR) is a challenging problem in the field of national defense and civil affairs. In recent decades, SAR automatic target recognition technology has made great progress in preprocessing, feature extraction and recognition. The SAR automatic target recognition problem involves many techniques in the field of machine learning and pattern recognition. With the increasing of SAR image data acquisition ability, people pay more and more attention to the accurate, fast understanding and recognition of these images. Aiming at the automatic target recognition problem of synthetic Aperture Radar (SAR), this paper focuses on SAR image segmentation, feature extraction and recognition. The main work of this paper is as follows: 1. Aiming at the problem of SAR image preprocessing, several common filtering and segmentation methods are analyzed and discussed. A new SAR image segmentation method combining thresholding technique and morphological method is proposed, which can better preserve the details of target and shadow. Lay the foundation for subsequent recognition. 2. A target recognition method based on SAR target and shadow feature fusion is proposed, which makes full use of the structure information of SAR image. The target and shadow are separated from the clutter background to suppress the influence of background clutter on the subsequent recognition. Then the features of the target and shadow are extracted for serial fusion and the recognition results are obtained by combining the nearest neighbor classifier. The experimental results show that the method has good recognition performance. In view of the existing problems in SAR image target recognition, a fast SAR image target recognition method is proposed. The method introduces the idea of compressed perception, and uses random perception matrix to reduce the dimension of multi-scale image features. Then, the neighbor classifier is used to recognize the feature set after dimensionality reduction. The experimental results show that this method can effectively reduce the characteristic dimension and the running time.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN957.52
【参考文献】
相关期刊论文 前10条
1 倪维平;严卫东;边辉;吴俊政;芦颖;王培忠;;基于MRF模型和形态学运算的SAR图像分割[J];电光与控制;2011年01期
2 杜廷娜,曹云露;数字图像线性滤波分析与实现[J];东华大学学报(自然科学版);2005年04期
3 汪雄良;冉承其;王正明;;基于紧致字典的基追踪方法在SAR图像超分辨中的应用[J];电子学报;2006年06期
4 石光明;刘丹华;高大化;刘哲;林杰;王良君;;压缩感知理论及其研究进展[J];电子学报;2009年05期
5 魏弘博,吕振肃,蒋田仔,刘新艳;图像分割技术纵览[J];甘肃科学学报;2004年02期
6 周建民;何秀凤;;星载SAR图像的斑点噪声抑制与滤波研究[J];河海大学学报(自然科学版);2006年02期
7 郑文;张鑫;;基于CFAR的SAR图像分割[J];廊坊师范学院学报(自然科学版);2011年04期
8 谢志鹏;陈松灿;;CSMP:基于约束等距的压缩感知匹配追踪[J];计算机研究与发展;2012年03期
9 尹奎英;刘宏伟;金林;;快速的Otsu双阈值SAR图像分割法[J];吉林大学学报(工学版);2011年06期
10 袁礼海;宋建社;薛文通;赵伟舟;;SAR图像自动目标识别系统研究与设计[J];计算机应用研究;2006年11期
相关博士学位论文 前1条
1 尹奎英;SAR图像处理及地面目标识别技术研究[D];西安电子科技大学;2011年
本文编号：2126901
本文链接：https://www.wllwen.com/kejilunwen/wltx/2126901.html
