基于稀疏表示的核素能谱特征提取及核素识别

张江梅; 季海波; 冯兴华; 王坤朋

doi:10.11884/HPLPB201830.170435

基于稀疏表示的核素能谱特征提取及核素识别

doi: 10.11884/HPLPB201830.170435

1.
西南科技大学信息工程学院, 四川绵阳 621900
2.
中国科学技术大学信息科学技术学院自动化系，合肥 230026

基金项目:

国家自然科学基金项目 61501385

四川省科技支撑计划项目 2016GZ0210

四川省科技厅应用基础项目 2016JY0242

详细信息

作者简介:
张江梅(1975—), 女，副教授，博士，从事核电子学与探测技术研究；zjm@swust.edu.cn

中图分类号: TL817
计量
- 文章访问数: 1292
- HTML全文浏览量: 338
- PDF下载量: 224
- 被引次数: 0
出版历程
- 收稿日期: 2017-11-06
- 修回日期: 2017-11-30
- 刊出日期: 2018-04-15

Nuclide spectrum feature extraction and nuclide identification based on sparse representation

1.
School of Information Engineering, Southwest University of Science and Technology, Mianyang 621900, China
2.
Department of Automation, School of Information Science and Technology, University of Science and Technology of China, Hefei 230026, China

摘要

摘要: 提出了一种基于稀疏表示的核素能谱特征提取方法，其实质是将核素能谱在区分性最好的稀疏原子上进行投影。利用稀疏分解方法对核素能谱进行稀疏分解，提取分解系数向量作为表征核素的特征向量，通过模式识别分类方法建立分类模型实现核素识别。与传统稀疏分解方法的区别在于：在能谱稀疏分解过程中按照稀疏字典中的原子排列顺序顺次进行分解；其次，分解目的在于特征提取，即最终提取到的特征对不同核素具有可区分性，并不要求核素能谱的重构精度。在²⁴¹Am, ¹³³Ba, ⁶⁰Co, ¹³⁷Cs, ¹³¹I和¹⁵²Eu共6种核素1200个能谱数据上进行了核素识别实验，7种不同分类算法的平均识别率达到91.71%，实验结果的统计分析表明，本文提出的特征提取方法识别准确率显著地高于两种传统核素能谱特征提取方法准确率。
- 伽马能谱 /
- 核素识别 /
- 稀疏表示 /
- 特征提取 /
- 模式识别
Abstract: A sparse representation based method for nuclide spectrum feature extraction is proposed. The essence of this method is to decompose the energy spectrum on the best distinguishable sparse atom. The sparse decomposition method is used to decompose the nuclide energy spectrum, and the decomposition coefficient vector is taken as the feature to represent the energy spectrum. The classification model is established by the pattern recognition algorithm to realize the nuclide identification. The main difference from the traditional sparse decomposition method is that we decompose the energy spectrum in accordance with the sparse atoms in the sequential order in sparse dictionary. In the experiments, 6 kinds of radionuclide including ²⁴¹Am, ¹³³Ba, ⁶⁰Co, ¹³⁷Cs, ¹³¹I and ¹⁵²Eu, 1200 energy spectra are used and the average nuclide identification accuracy on 7 different pattern recognition algorithms is 91.71%. The results of statistical tests show that the proposed algorithm performs significantly better than two traditional nuclide spectrum feature extraction methods.
- gamma spectrum /
- nuclide identification /
- sparse representation /
- feature extraction /
- pattern recognition

HTML全文

图 1 稀疏分解系数

Figure 1. Coefficients of sparse decomposition

下载: 全尺寸图片幻灯片

表 1 三种特征提取方法在模拟核素上的识别结果

Table 1. Classification results of the three feature extraction methods

methods	sparse representation(rank)	SG +derivative(rank)	TS+derivative(rank)
KNN	97.11%(1)	88.75% (2)	72.67% (3)
NavieBayes	88.57%(1)	44.58% (2)	38.92% (3)
SMO	72.86%(1)	19.50% (3)	21.83% (2)
PART	96.00%(1)	91.25% (2)	78.00% (3)
J48	96.86%(1)	90.83% (2)	74.42% (3)
CART	96.29%(1)	89.50% (2)	76.00% (3)
RBFNetwork	94.29%(1)	73.17% (2)	58.00% (3)
mean	91.71%(1)	71.08%(2.14)	59.98%(2.86)

下载: 导出CSV

表 2 Holm检验

Table 2. Holm test

i	methods	z=(R_i-R₁)/SE	p	α/(k-i)
1	TS+derivative peak seeking	(2.86-1)/0.535 4=3.479 7	0.000 5	0.025 0
2	SG+derivative peak seeking	(2.14-1)/0.535 4=2.132 7	0.032 9	0.050 0

下载: 导出CSV

参考文献(15)

[1]	Portnoy D, Fisher B, Phifer D. Data and software tools for gamma radiation spectral threat detection and nuclide identification algorithm development and evaluation[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2015, 784: 274-280. doi: 10.1016/j.nima.2014.11.010
[2]	Burr T, Hammada M. Radio-isotope identification algorithms for NaI γ spectra[J]. Algorithms, 2009, 2(1): 339-360. doi: 10.3390/a2010339
[3]	Uher J, Roach G, Tickner J. Peak fitting and identification software library for high resolution gamma-ray spectra[J]. Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment, 2010, 619(1): 457-459.
[4]	易义成, 宋朝晖, 管兴胤, 等. 溴化镧高剂量率线性响应范围的测定[J]. 强激光与粒子束, 2016, 29: 096002. doi: 10.11884/HPLPB201628.151241 Yi Yicheng, Song Chaohui, Guan Xingying, et al. Measurement of linear response upper limit for LaBr₃ to pulsed gamma radiation. High Power Laser and Particle Beams, 2016, 29: 096002 doi: 10.11884/HPLPB201628.151241
[5]	安力, 何铁, 郑普, 等. 伴随粒子法γ能谱本底测量技术[J]. 强激光与粒子束, 2013, 25(11): 3045-3049. doi: 10.3788/HPLPB20132511.3045 An Li, He Tie, Zheng Pu, et al. Research on γ background spectra in associated particle technique. High Power Laser and Particle Beams, 2013, 25(11): 3045-3049 doi: 10.3788/HPLPB20132511.3045
[6]	何剑锋. 低能量分辨率γ能谱数据解析方法研究[D]. 成都: 成都理工大学, 2013. He Jianfeng. A study for decomposition method of lower-energy resolution gamma-energy resolution gamma-ray spectra data. Chengdu: Chengdu University of Technology, 2013
[7]	Mallat S, Zhang Z. Matching pursuits with time-frequency dictionaries[J]. IEEE Trans on Signal Processing, 1993, 41(12): 3397-3415. doi: 10.1109/78.258082
[8]	Aha D, Kibler D, Albert M. Instance-based learning algorithms[J]. Machine Learning, 1991, 6(1): 37-66.
[9]	George H, Langley P. Estimating continuous distributions in Bayesian classifiers[C]//Proceedings of the Eleventh Conference on Uncertainty in Artificial Intelligence, 1998: 338-345.
[10]	Platt J. Fast training of support vector machines using sequential minimal optimization[M]. Massachusetts: MIT Press, 1999: 185-208.
[11]	Frank E, Witten I. Generating accurate rule sets without global optimization[C]//The Fifteenth International Conference on Machine Learning. 1998: 144-151.
[12]	Quinlan J. C4.5: programs for machine learning[M]. San Francisco: Morgan Kaufmann Publishers, 1993.
[13]	Breiman L, Friedman J, Olshen R, et al. Classification and regression trees[M]. Boca Raton: CRC Press, 1984.
[14]	Bouchoux S, Paindavoine M. Implementation of pattern recognition algorithm based on RBF neural network[C]//Proc of SPIE. 2002, 8(18): 4572-4581.
[15]	Demsar J. Statistical comparisons of classifiers over multiple data sets[J]. Journal of Machine Learning Research, 2006, 7: 1-30.