基于小波域NMF特征提取的SAR图像目标识别方法

宦若虹; 杨汝良

doi:10.3724/SP.J.1146.2007.01889

基于小波域NMF特征提取的SAR图像目标识别方法

doi: 10.3724/SP.J.1146.2007.01889

宦若虹^{①②;杨汝良},
杨汝良

计量
- 文章访问数: 3545
- HTML全文浏览量: 85
- PDF下载量: 1321
- 被引次数: 0
出版历程
- 收稿日期: 2007-12-10
- 修回日期: 2008-04-08
- 刊出日期: 2009-03-19

Synthetic Aperture Radar Images Target Recognition Based on Wavelet Domain NMF Feature Extraction

Huan Ruo-hong^{①②;杨汝良},
Yang Ru-liang

摘要

摘要: 该文提出了一种基于小波域非负矩阵分解特征提取的合成孔径雷达图像目标识别方法。该方法对图像二维离散小波分解后提取低频子带图像，用非负矩阵分解对低频子带图像提取特征向量作为目标的特征，利用支持向量机进行分类完成目标识别。将该方法用于对MSTAR数据中三类目标识别，识别率最高可达97.51%，明显提高了目标的正确识别率。实验结果表明，该方法是一种有效的合成孔径雷达图像特征提取与目标识别方法。
- 合成孔径雷达;特征提取;识别;非负矩阵分解;小波
Abstract: This paper presents a method for synthetic aperture radar images target recognition based on wavelet domain non-negative matrix factorization feature extraction. Low-frequency sub-band image is obtained by 2-D discrete wavelet decomposition of a SAR image. Non-negative matrix factorization is used for extracting feature vectors from the low-frequency sub-band image as the feature of the target. Support vector machine is used to perform target recognition. The method is applied for recognizing three-class targets in MSTAR database and the highest correct probability of recognition arrives at 97.51% which is enhanced obviously. It is concluded that the method proposed in this paper is an effective method for SAR images feature extraction and target recognition.

HTML全文

参考文献(1)

Sandirasegaram N and Englisth R. Comparative analysis offeature extraction (2D FFT and wavelet) and classification(Lp metric distances, MLP NN, and HNeT) algorithms forSAR imagery. Proc. SPIE, 2005, 5808: 314-325.[2]Lee D and Seung H. Learning the parts of objects bynon-negative matrix factorization[J].Nature.1999, 401:788-791[3]Lee D and Seung H. Algorithms for non-negative matrixfactorization. Neural Information Processing Systems (NIPS).Denver, CO, USA, 2000, 7.[4]Tsuge S, Shishibori M, and Kuroiwa S, et al.. Dimensionalityreduction using non-negative matrix factorization forinformation retrieval. IEEE Conf. Systems, Man, andCybernetics. Tucson, USA, 2001, Vol.2: 960-965.[5]Monga V and Mihcak M K. Robust and secure image hashingvia non-negative matrix factorizations[J].IEEE Trans. onInformation Forensics and Security.2007, 2(3):376-390[6]Kotsia I, Zafeiriou S, and Pitas I. A novel discriminantnon-negative matrix factorization algorithm withapplications to facial image characterization problems[J].IEEETrans. on Information Forensics and Security.2007, 2(3):588-595[7]Benetos E, Kotti M, and Kotropoulos C. Musical instrumentclassification using non-negative matrix factorizationalgorithms. IEEE Proceedings International Symposium onCircuits and Systems. Island of Kos, Greece, 2006:1844-1847.[8]Kaarna A. Non-negative matrix factorization features fromspectral signatures of AVIRIS images. IEEE Conf. Geoscienceand Remote Sensing Symposium. Denver, Colorado, USA,2006: 549-552.[9]Ross T D, Worrell S W, and Velten V J, et al.. Standard SARATR evaluation experiments using the MSTAR public releasedata set. Proc. SPIE, 1998, 3370: 566-573.[10]Zhao Q and Principe J C. Support vector machines for SARautomatic target recognition[J].IEEE Trans. on Aerospace andElectronic Systems.2001, 37(2):643-654[11]Nilubol C and Pham Q H. Translational and rotationalinvariant hidden Markov model for automatic targetrecognition. Proc. SPIE, 1998, 3374: 179-185.

施引文献

资源附件(0)

访问统计