|Table of Contents|

A Soybean Variety Identification Algorithm Based on Hyperspectral Image and Neighborhood Rough Set Theory and Its Comprehensive Performance Evaluation(PDF)

《大豆科学》[ISSN:1000-9841/CN:23-1227/S]

Issue:
2018年04期
Page:
596-605
Research Field:
Publishing date:

Info

Title:
A Soybean Variety Identification Algorithm Based on Hyperspectral Image and Neighborhood Rough Set Theory and Its Comprehensive Performance Evaluation
Author(s):
LIU Yao1;  LI Zi-nan1;  WU Tao1;  LIU Lian2;  MENG Xiang-li1
(1.School of Information Engineering, Lingnan Normal University, Zhanjiang 524048, China； 2.College of Information and Communication Engineering, Harbin Engineering University, Harbin 150001, China)
Keywords:
Soybean;  Hyperspectral image;  Neighborhood rough set;  Band selection;  Comprehensive performance
PACS:
-
DOI:
10.11861/j.issn.1000-9841.2018.04.0596
Abstract:
The rapid, efficient and nondestructive identification for soybeans varieties can be realized by using hyperspectral image technology. Hyperspectral image data is large and contains hundreds of bands, therefore it is difficult for data transmission, storage and processing. It is necessary to use band selection method for dimension reduction. At present, band selection algorithms of hyperspectral image for variety identification mainly take classification performance as evaluation criteria, and the stability of the algorithms is ignored. To solve the problem of variety identification for soybeans, hyperspectral band selection algorithms based on the dependence, consistency and information entropy criteria in neighborhood rough set theory were studied in this paper. By introducing Jaccard index as a metric of stability, the changes of stability of the algorithm with the dataset perturbation and the subset size were explored. Since the stability measurement and classification model of band selection algorithm are independent of each other, it couldn’t pursue high stability only and ignored the classification effect.To assess the comprehensive performance of algorithms, for the subsets with the same size, Pareto optimal solutions was proposed. For the subsets with different size, a comprehensive evaluation function of classification performance, stability and subset size was proposed. The results have certain theoretical and application value for obtaining the best band subset of comprehensive performance.

References:

［1］魏利峰. 玉米种子高光谱图像品种检测方法研究［D］. 沈阳：沈阳农业大学, 2017. (Wei L F. Research on detection method of maize variety based on hyperspectral image［D］. Shenyang：Shenyang Agricultural University, 2017.)?
［2］朱大洲, 王坤, 周光华, 等. 单粒大豆的近红外光谱特征及品种鉴别研究［J］.光谱学与光谱分析, 2010, 30(12): 3217-3221．(Zhu D Z, Wang K, Zhou G H, et al. The NIR spectra based variety discrimination for single soybean seed［J］. Spectroscopy and Spectral Analysis, 2010, 30(12): 3217-3221.)
［3］杨冬风, 朱洪德. 基于近红外透射光谱分析和BP神经网络的大豆品种识别［J］. 大豆科学, 2013, 32(2): 249-253. (Yang D F，Zhu H D. Recognition of soybean varieties based on near infrared transmittance spectroscopy and BP neural network［J］. Soybean Science, 2013, 32(2): 249-253.)
［4］柴玉华, 侯升飞, 彭长禄. 基于高光谱图像技术的大豆分级识别方法研究［J］.东北农业大学学报, 2014, 45(4): 107-112. (Chai Y H, Hou S F, Peng C L. Identification of different soybean grades based on hyperspectral imagery［J］. Journal of Northeast Agricultural University, 2014, 45(4): 107-112.)
［5］Tan K, Chai Y, Song W, et al. Identification of soybean seed varieties based on hyperspectral image［J］. Transactions of the Chinese Society of Agricultural Engineering, 2014, 30(9): 235-242.
［6］Liu Y, Xie H, Chen Y H, et al. Neighborhood mutual information and its application on hyperspectral band selection for classification［J］. Chemometrics and Intelligent Laboratory Systems, 2016, 157(157): 140-15
［7］Kalousis A, Prados J, Hilario M. Stability of feature selection algorithms: A study on high- dimensional spaces［J］. Knowledge & Information Systems, 2007, 12(1): 95-116.
［8］严衍禄, 陈斌, 朱大洲. 近红外光谱分析的原理、技术与应用［M］.北京: 中国轻工业出版社, 2013. (Yan Y L, Chen B, Zhu D Z. Near infrared spectroscopy-principles, technologies and applications ［M］. Beijing: China Light Industry Press, 2013.)
［9］Dong Y, Xiang B, Geng Y, et al. Rough set based wavelength selection in near-infrared spectral analysis［J］. Chemometrics and Intelligent Laboratory Systems, 2013, 126(126): 21-29.
［10］Chen Y, Xue Y, Ma Y, et al. Measures of uncertainty for neighborhood rough sets［J］. Knowledge-Based Systems, 2017, 120(C): 226-235.
［11］Pan X, Zhang S, Zhang H, et al. A variable precision rough set approach to the remote sensing land use/cover classification［J］. Computers & Geosciences, 2010, 36(12):1466-1473.
［12］Hu Q, Pedrycz W, Yu D, et al. Selecting discrete and continuous features based on neighborhood decision error minimization［J］. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2010, 40(1): 137-150.
［13］Liu Y, Chen Y, Tan K, et al. Maximum relevance, minimum redundancy band selection based on neighborhood rough set for hyperspectral data classification［J］. Measurement Science and Technology, 2016, 27(12): 125501.?
［14］季金胜. 高分辨率遥感影像典型地物目标的特征选择及其稳定性研究［D］. 上海：上海交通大学, 2015.( Ji J S. Feature selection and its stability for typical geoobjects of the high-resolution remote sensing image［D］. Shanghai: Shanghai Jiao Tong University, 2015.)
［15］Kuncheva L I. A stability index for feature selection［C］//Artificial Intelligence and Applications, 2007: 421-427.
［16］俞研, 黄皓. 面向入侵检测的基于多目标遗传算法的特征选择［J］. 计算机科学, 2007, 34(3):197-200. (Yu Y, Huang H. Feature selection using multi-objective genetic algorithms for intrusion detection［J］. Computer Science, 2007, 34(3):197-200.)
［17］Saeys Y, Abeel T, Peer Y V D. Robust feature selection using ensemble feature selection techniques［J］. Lecture Notes in Computer Science, 2008, 5212: 313-325.
［18］刘瑶, 谭克竹, 陈月华, 等. 基于分段主成分分析和高光谱技术的大豆品种识别［J］. 大豆科学, 2016, 35(4): 672-678. (Liu Y, Tan K Z, Chen Y H, et al. Variety recognition of soybeans using segmented principal component analysis and hyperspectral technology［J］. Soybean Science, 2016, 35(4): 672-678.
［19］Liu Y, Xie H, Wang L, et al. Hyperspectral band selection based on a variable precision neighborhood rough set［J］. Applied optics, 2016, 55(3): 462-472.

Memo

Memo:
-
Last Update: 2018-08-01