海杂波背景下弱目标检测
发布时间：2018-08-31 17:23
【摘要】：海面上小目标的雷达检测技术在港口交通、海浪监测、海难和空难搜救、军事侦察等场合中具有广泛应用。对于海面上的小目标,其雷达信号通常较弱,这在目标检测中是个困难的任务。此外,海面雷达信号是典型的非平稳随机信号,对其进行检测需要选择合适的特征和检测器。为了实现海面上弱目标检测,需要在复杂海杂波背景下,迅速准确地提取雷达回波信号中的有用特征信息,并进行识别分类。本文研究了海面弱目标检测技术,主要工作如下：(1)针对海杂波的非平稳特性,应用三参数的分数阶傅里叶变换来处理海杂波信号,使易混淆的主目标与次目标信号特征差别增大,为后续目标识别奠定基础。(2)根据海杂波的非平稳特征,用Hurst指数、Lyapunov指数、分形维数、多重分形谱、近似熵等对其进行表征,并用遗传算法优化选择选取了一种新的联合特征向量,通过特征互补,使特征向量更好表征海杂波的特性。(3)将深度信念网络与隐马尔科夫模型相结合作为目标分类器,应用联合特征向量作为模式分类器的输入进行训练；训练完成后,将其用于海杂波信号的分类,取得了良好的实验结果。测试数据选用加拿大McMaster大学IPIX雷达数据。仿真实验结果表明,本文的海杂波目标识别方法检测准确度较高,在低信噪比情况下也具有良好的检测性能。
[Abstract]:Radar detection technology for small targets on the sea surface has been widely used in port traffic, ocean wave monitoring, maritime and air disaster rescue, military reconnaissance and other occasions. For small targets on the sea surface, the radar signal is usually weak, which is a difficult task in target detection. In addition, the sea surface radar signal is a typical nonstationary random signal, so it is necessary to select suitable features and detectors to detect it. In order to detect weak targets on the sea surface, it is necessary to quickly and accurately extract the useful feature information from radar echo signals and classify them under the background of complex sea clutter. The main work of this paper is as follows: (1) aiming at the non-stationary characteristics of sea clutter, a three-parameter fractional Fourier transform is applied to deal with sea clutter signals. The difference between the signal characteristics of the main target and the secondary target is enlarged, which lays a foundation for the subsequent target recognition. (2) according to the non-stationary feature of sea clutter, it is characterized by Hurst exponent, fractal dimension, multifractal spectrum, approximate entropy, etc. A new joint feature vector is selected by genetic algorithm. By feature complementation, the feature vector can better represent the characteristics of sea clutter. (3) the depth belief network and hidden Markov model are combined as target classifiers. The joint eigenvector is used as the input of the pattern classifier for the training. After the training is completed, the joint eigenvector is applied to the classification of sea clutter signals, and good experimental results are obtained. The test data are selected from IPIX radar data of McMaster University, Canada. The simulation results show that the detection accuracy of this method is high and the detection performance is good in the case of low signal-to-noise ratio (SNR).
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN957.51
本文编号：2215660
[Abstract]:Radar detection technology for small targets on the sea surface has been widely used in port traffic, ocean wave monitoring, maritime and air disaster rescue, military reconnaissance and other occasions. For small targets on the sea surface, the radar signal is usually weak, which is a difficult task in target detection. In addition, the sea surface radar signal is a typical nonstationary random signal, so it is necessary to select suitable features and detectors to detect it. In order to detect weak targets on the sea surface, it is necessary to quickly and accurately extract the useful feature information from radar echo signals and classify them under the background of complex sea clutter. The main work of this paper is as follows: (1) aiming at the non-stationary characteristics of sea clutter, a three-parameter fractional Fourier transform is applied to deal with sea clutter signals. The difference between the signal characteristics of the main target and the secondary target is enlarged, which lays a foundation for the subsequent target recognition. (2) according to the non-stationary feature of sea clutter, it is characterized by Hurst exponent, fractal dimension, multifractal spectrum, approximate entropy, etc. A new joint feature vector is selected by genetic algorithm. By feature complementation, the feature vector can better represent the characteristics of sea clutter. (3) the depth belief network and hidden Markov model are combined as target classifiers. The joint eigenvector is used as the input of the pattern classifier for the training. After the training is completed, the joint eigenvector is applied to the classification of sea clutter signals, and good experimental results are obtained. The test data are selected from IPIX radar data of McMaster University, Canada. The simulation results show that the detection accuracy of this method is high and the detection performance is good in the case of low signal-to-noise ratio (SNR).
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TN957.51
【参考文献】
相关期刊论文 前7条
1 范一飞;罗丰;李明;胡冲;陈帅霖;;海杂波AR谱多重分形特性及微弱目标检测方法[J];电子与信息学报;2016年02期
2 行鸿彦;张强;徐伟;;海杂波FRFT域的分形特征分析及小目标检测方法[J];物理学报;2015年11期
3 田玉芳;尹志盈;姬光荣;郑海永;;基于SVM的海面弱目标检测[J];中国海洋大学学报(自然科学版);2013年07期
4 刘宁波;王国庆;包中华;关键;;海杂波FRFT谱的多重分形特性与目标检测[J];信号处理;2013年01期
5 牛奉高;刘维奇;;分数布朗运动与Hurst指数的关系研究[J];山西大学学报(自然科学版);2010年03期
6 刘宁波;关键;;海杂波的多重分形判定及广义维数谱自动提取[J];海军航空工程学院学报;2008年02期
7 李颖新;李建更;阮晓钢;;肿瘤基因表达谱分类特征基因选取问题及分析方法研究[J];计算机学报;2006年02期
，本文编号：2215660
本文链接：https://www.wllwen.com/kejilunwen/xinxigongchenglunwen/2215660.html
