Anomaly detection based weighted combination kernel RX algorithm and its parameter selection

WU Xiangwei; GUO Baofeng; CHEN Chunzhong; SHEN Honghai

doi:10.7510/jgjs.issn.1001-3806.2015.06.003

In order to combine the spectral shape difference information and the polynomial kernel function global information, exploit the object feature fully and improve the accuracy of anomaly detection, anomaly detection method was proposed based on weighted combination kernel RX algorithm. A spectral angle kernel function was added to Gaussian kernel function in the anomaly detection method. Because the kernels' parameter and the weighting parameter will affect the efficiency of the algorithm, the random function selection, the hill climbing method and the particle swarm optimization algorithm were implemented for setting the above parameters. Experiment results show that at a constant false alarm rate, it is the best to set the parameters by means of the particle swarm algorithm. Target detection rate is 83.5% by using the weighted combination kernel RX algorithm, higher than that by means of the traditional kernel RX algorithm.

Anomaly detection based weighted combination kernel RX algorithm and its parameter selection

Corresponding author: GUO Baofeng, gbf@hdu.edu.cn;

1. School of Automation, Hangzhou Dianzi University, Hangzhou 310018, China;

2. Changchun Institute of Optics, Fine Mechanics and Physics, Chinese Academy of Sciences, Changchun 130033, China

Received Date: 2014-10-10

Accepted Date: 2014-12-01

Available Online: 2015-11-25

Abstract: In order to combine the spectral shape difference information and the polynomial kernel function global information, exploit the object feature fully and improve the accuracy of anomaly detection, anomaly detection method was proposed based on weighted combination kernel RX algorithm. A spectral angle kernel function was added to Gaussian kernel function in the anomaly detection method. Because the kernels' parameter and the weighting parameter will affect the efficiency of the algorithm, the random function selection, the hill climbing method and the particle swarm optimization algorithm were implemented for setting the above parameters. Experiment results show that at a constant false alarm rate, it is the best to set the parameters by means of the particle swarm algorithm. Target detection rate is 83.5% by using the weighted combination kernel RX algorithm, higher than that by means of the traditional kernel RX algorithm.

HTML

Reference (17)

[1]	MATTEOLI S, DIANI M, CORSINI G. A tutorial overview of anomaly detection in hyperspectral images [J]. IEEE Transactions on Aerospace and Electronic Systems Magazine, 2010, 25(7): 5-28.
[2]	ECHES O, DOBIGEON N, TOURNERET J Y. Enhancing hyperspectral image unmixing with spatial correlations [J]. IEEE Transactions on Geosciences and Remote Sensing, 2011, 49(11): 4239-4247.
[3]	KWON H, DER S Z, NASRABADI N M. Adaptive anomaly detection using subspace separation for hyperspectral imagery [J]. Optical Engineering, 2003, 42(11): 3342-3351.
[4]	REED I S, YU X L. Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution [J]. IEEE Transactions on Acoustics, Speech and Signal Processing, 1990, 38(10): 1760-1770.
[5]	KWON H, NASRABADI N M. Kernel RX algorithm a nonlinear anomaly detector for hyperspectral imagery[J]. IEEE Transactions on Geosciences and Remote Sensing, 2005, 43(2): 388-397.
[6]	WANG L P, ZHANG L P, HAN J T. Detecting algorithm of moving target in dynamic background based on gray-weighted kernel function [J]. Infrared and Laser Engineering, 2013, 42(12): 3453-3457 (in Chinese).
[7]	CHEN D R, SUN B, TAO P, et al. Spatial neighboring clustering method for hyperspectral imagery based on kernel spectral angel cosine [J]. Acta Electronica Sinica, 2008, 36(10): 1992-1995(in Chinese).
[8]	HAN J, YUE J, ZHANG Y, et al. SAM weighted KEST algorithm for anomaly detection in hyperspectral imagery [J]. Journal of Infrared and Millimeter Waves, 2013, 32(4): 359-365(in Chinese).
[9]	WANG S B. Study on anomaly target detection technology in hyperspectral images [D]. Haerbin: Harbin Institute of Technology, 2010: 22-35(in Chinese).
[10]	WU X M. The research on hyperspectral imagery unmixing technology based on kernel methods [D]. Hangzhou:Zhejiang University, 2011: 15-33(in Chinese).
[11]	WANG Ch. Research on multiple kernel learning based target interpretation technologies in hyperspectral imagery [D].Haerbin: Harbin Institute of Technology, 2011: 13-23(in Chinese).
[12]	GAN P P, WANG R Sh. Spectral remote sensing recognition basis and technology research [J]. Remote Sensing Technology and Application, 2002, 17(3): 53-60(in Chinese).
[13]	SHAWE-TAYLOR J, CRISTIANINI N. Pattern recognition of kernel method [M]. 2nd ed. Beijing: China Machine Press, 2006: 29-50(in Chinese).
[14]	DANG J W, WANG Y P, DI F W, et al. Artificial intelligence [M]. Beijing: Electronic Industry Press, 2012: 178-180(in Chinese).
[15]	EBERHART R, KENNEDY J. A new optimizer using particle swarm theory[C]//Proceedings of the 6th International Symposium on Micro Machine and Human Science. New York, USA : IEEE , 1995: 39-43.
[16]	GAO X J. On classification of hyperspectral remotely sensed imagery based on support vector machines[D]. Hangzhou: Hangzhou Dianzi University, 2012: 42-44 (in Chinese) .
[17]	LIANG Zh J, WANG K F, GU G Q, et al . Digital speckle image correlation method base on particle swarm optimization algorithm[J]. Laser Technology, 2014, 38(5): 603-607 (in Chinese).

Anomaly detection based weighted combination kernel RX algorithm and its parameter selection

Abstract

References

Proportional views

通讯作者: 陈斌, bchen63@163.com

Proportional views