基于多特征融合的雷达目标识别
|
摘要
雷达目标识别是对雷达探测功能的重要延伸,在现代战争中具有极其重要的意义,也是当前雷达信号处理的一个研究热点。随着城市环境的日益复杂,对低、小、慢目标的检测与识别已经成为一个迫切需要解决的问题。现役雷达大部分是低分辨率雷达,在不影响雷达系统数据处理过程的基础上进行目标识别,尤其是低、小、慢目标的识别具有十分重要的实际意义。
鉴于传统雷达目标检测技术一般是基于雷达接收回波能量的检测方法,该方法设置一个门限,只要过门限的点目标都输出了,这就引起一个问题:输出的点目标中既有感兴趣的目标,也有大量的不感兴趣的目标。为了解决这个问题,本文针对脉冲雷达,提出了一种基于多特征融合的雷达目标检测/识别方法,该方法不影响雷达信号处理过程,只需要存储杂波对消后的数据,将输出的点迹数据进行特征提取,然后进行目标识别。论文内容分为低分辨雷达目标识别分析,分析了不同目标的回波序列,包括时域回波幅度和多普勒谱,并在此基础上进行了特征提取,利用支持矢量机进行分类识别,将单个分类器输出的结果进行融合计算,并对该方法的推广性能进行了评估。
特征提取是目标识别过程中的关键环节,提取有效的特征能大大提高识别率。实测数据试验结果表明,本文提出的基于多特征的雷达目标识别方法能够有效的实现地面目标和空中目标的分类,尤其是对于复杂背景下的地面汽车目标和低、小、慢目标的分类识别,该方法识别性能高,实时性强,易于实现,具有广阔的应用前景。同时,对于其他体制的雷达,也可以用该方法进行目标检测和识别。
Radar target recognition is a hot research topic in radar signal processing field now, which has a very special mean in war at modem times. As the city circumstance becoming more and more complexity, the low altitude, small and slowly flying targets’detection and classification has been a very important problem. The great mass of active service radar is low-resolution radar. To recognize the targets in the ground and in the sky based on the low-resolution radar signal processing has practical signify.
The traditional target detection method often based on the energy of received signal. By setting a proper threshold, those signals which larger than the threshold are regarded as a target. A problem that those signals contain both the interested and the uninterested .To solve this problem, a multi-feature detection/classification method based on the measured data was proposed in this paper. This method requires no extra processing rather than a mere store of the data after MTI processing. The feature extraction has been done based on the output of pro-process and select the effective combinations of those features test them by SVM classifier. This paper is organized as follows: firstly study the basic knowledge of the low-resolution radar target recognition, secondly analyze their properties in time and frequency domain, and then propose several feature-extraction methods and performed some classification experiments based on support vector machine (SVM). The result shows the methods are efficient, finally, the fusion result were implemented using the output of the SVM by several fusion methods. The validity of this method has been proved by measured data.
Feature extraction plays a key role in target recognition. Picking-up effective characteristics can greatly improve the identify performance. The results show that the radar target recognition method based on the multi-feature can recognize the air target from the cars in the complex background, especially. This method can get a high recognition performance, real-time processing and easy to implement. At the same time, this method can also be used in the other radar systems to detect and classify the different targets.
鉴于传统雷达目标检测技术一般是基于雷达接收回波能量的检测方法,该方法设置一个门限,只要过门限的点目标都输出了,这就引起一个问题:输出的点目标中既有感兴趣的目标,也有大量的不感兴趣的目标。为了解决这个问题,本文针对脉冲雷达,提出了一种基于多特征融合的雷达目标检测/识别方法,该方法不影响雷达信号处理过程,只需要存储杂波对消后的数据,将输出的点迹数据进行特征提取,然后进行目标识别。论文内容分为低分辨雷达目标识别分析,分析了不同目标的回波序列,包括时域回波幅度和多普勒谱,并在此基础上进行了特征提取,利用支持矢量机进行分类识别,将单个分类器输出的结果进行融合计算,并对该方法的推广性能进行了评估。
特征提取是目标识别过程中的关键环节,提取有效的特征能大大提高识别率。实测数据试验结果表明,本文提出的基于多特征的雷达目标识别方法能够有效的实现地面目标和空中目标的分类,尤其是对于复杂背景下的地面汽车目标和低、小、慢目标的分类识别,该方法识别性能高,实时性强,易于实现,具有广阔的应用前景。同时,对于其他体制的雷达,也可以用该方法进行目标检测和识别。
Radar target recognition is a hot research topic in radar signal processing field now, which has a very special mean in war at modem times. As the city circumstance becoming more and more complexity, the low altitude, small and slowly flying targets’detection and classification has been a very important problem. The great mass of active service radar is low-resolution radar. To recognize the targets in the ground and in the sky based on the low-resolution radar signal processing has practical signify.
The traditional target detection method often based on the energy of received signal. By setting a proper threshold, those signals which larger than the threshold are regarded as a target. A problem that those signals contain both the interested and the uninterested .To solve this problem, a multi-feature detection/classification method based on the measured data was proposed in this paper. This method requires no extra processing rather than a mere store of the data after MTI processing. The feature extraction has been done based on the output of pro-process and select the effective combinations of those features test them by SVM classifier. This paper is organized as follows: firstly study the basic knowledge of the low-resolution radar target recognition, secondly analyze their properties in time and frequency domain, and then propose several feature-extraction methods and performed some classification experiments based on support vector machine (SVM). The result shows the methods are efficient, finally, the fusion result were implemented using the output of the SVM by several fusion methods. The validity of this method has been proved by measured data.
Feature extraction plays a key role in target recognition. Picking-up effective characteristics can greatly improve the identify performance. The results show that the radar target recognition method based on the multi-feature can recognize the air target from the cars in the complex background, especially. This method can get a high recognition performance, real-time processing and easy to implement. At the same time, this method can also be used in the other radar systems to detect and classify the different targets.
引文
[1]王伟,张汉华等.低分辨雷达的目标特征提取方法.国防科技大学学报2002,24(2):31-35.
[2]吴顺君,梅晓春等.雷达信号处理和数据处理技术.电子工业出版社. 292-304.
[3]宋颍凤,张汉华,姜卫东,等.低分辨雷达的目标回波波形的特征提取和分类方法研究[J].雷达与对抗,2002,(4):1-5.
[4]杜兰,刘宏伟,保铮.一种基于距离-多普勒二维联合的群目标分辨方法.电子学报. 2004,6(32):881-885.
[5] H Leungetal, Intelligent Radar Recognition System for Surveillance, Proceedings of IEEE International Conference on System, Manand Cybernetics,1995:2280-2285.
[6] M Ringgs and A D Robinson. Ship Target Recognition Using Low Resolution Radar and Neural Networks. IEEE Trans on Aerospace and Electronic system, 1999,April,386-393.
[7]袁莉基于高分辨距离像的雷达目标识别方法博士论文西安电子科技大学2007.
[8]马建华,刘宏伟,保铮.固定翼飞机和直升机的分类方法研究.现代雷达, 2004, 26(12):45-48.
[9]郁文贤舰船雷达目标识别的智能化方法研究[D]国防科技大学,1992.
[10]边綮祺,张学工.模式识别.北京:清华大学出版社,2000.
[11]刘宏伟保铮.基于复合特征及分层特征选择的雷达HRRP识别.系统工程与电子技术. 2005, 27(4):596-599.
[12] Duda R O, Hart P E. Pattern classification and scene analysis [M].New York :John Wiley & Sons,1973.
[13]马君国空间雷达目标特征提取与识别方法研究.国防科学技术大学研究生学位论文. 2006.
[14]许小剑,黄培康,利用RCS幅度信息进行雷达目标识别,系统工程与电子技术. 1992,6:1-9.
[15]袁莉刘宏伟保铮.基于中心距特征的雷达HRRP自动目标识别.电子学报. 2004,32 (12) : 2078-2081.
[16]赵谊虹程国华史习智.多分类器融合中的一种新的加权算法.上海交通大学学报. 2002,36(6):765-768.
[17]陈丽陈静.基于支持向量机和k-近邻分类器的多特征融合方法.计算机应用.2009.29 (3):833-835.
[18]刘宏伟杜兰袁莉保铮.雷达高分辨距离像目标识别研究进展.电子与信息学报. 2005, 27 (8):1328-1334.
[19]陈行勇黎湘等.基于旋翼微动雷达特征的空中目标识别.系统工程与电子技术. 2006,28(3):372-375.
[20]李可心徐国鑫肖俊岭.基于主成分分析的空中目标识别.现代防御技术. 2008,36(3):1-5.
[21]温福喜刘宏伟.应用科技. 2007,34(1):1-4.
[2]吴顺君,梅晓春等.雷达信号处理和数据处理技术.电子工业出版社. 292-304.
[3]宋颍凤,张汉华,姜卫东,等.低分辨雷达的目标回波波形的特征提取和分类方法研究[J].雷达与对抗,2002,(4):1-5.
[4]杜兰,刘宏伟,保铮.一种基于距离-多普勒二维联合的群目标分辨方法.电子学报. 2004,6(32):881-885.
[5] H Leungetal, Intelligent Radar Recognition System for Surveillance, Proceedings of IEEE International Conference on System, Manand Cybernetics,1995:2280-2285.
[6] M Ringgs and A D Robinson. Ship Target Recognition Using Low Resolution Radar and Neural Networks. IEEE Trans on Aerospace and Electronic system, 1999,April,386-393.
[7]袁莉基于高分辨距离像的雷达目标识别方法博士论文西安电子科技大学2007.
[8]马建华,刘宏伟,保铮.固定翼飞机和直升机的分类方法研究.现代雷达, 2004, 26(12):45-48.
[9]郁文贤舰船雷达目标识别的智能化方法研究[D]国防科技大学,1992.
[10]边綮祺,张学工.模式识别.北京:清华大学出版社,2000.
[11]刘宏伟保铮.基于复合特征及分层特征选择的雷达HRRP识别.系统工程与电子技术. 2005, 27(4):596-599.
[12] Duda R O, Hart P E. Pattern classification and scene analysis [M].New York :John Wiley & Sons,1973.
[13]马君国空间雷达目标特征提取与识别方法研究.国防科学技术大学研究生学位论文. 2006.
[14]许小剑,黄培康,利用RCS幅度信息进行雷达目标识别,系统工程与电子技术. 1992,6:1-9.
[15]袁莉刘宏伟保铮.基于中心距特征的雷达HRRP自动目标识别.电子学报. 2004,32 (12) : 2078-2081.
[16]赵谊虹程国华史习智.多分类器融合中的一种新的加权算法.上海交通大学学报. 2002,36(6):765-768.
[17]陈丽陈静.基于支持向量机和k-近邻分类器的多特征融合方法.计算机应用.2009.29 (3):833-835.
[18]刘宏伟杜兰袁莉保铮.雷达高分辨距离像目标识别研究进展.电子与信息学报. 2005, 27 (8):1328-1334.
[19]陈行勇黎湘等.基于旋翼微动雷达特征的空中目标识别.系统工程与电子技术. 2006,28(3):372-375.
[20]李可心徐国鑫肖俊岭.基于主成分分析的空中目标识别.现代防御技术. 2008,36(3):1-5.
[21]温福喜刘宏伟.应用科技. 2007,34(1):1-4.