基于NIRS的马尾松苗木根部含水量预测模型

doi:10.3969/j.issn.1000-2006.201902028

国家林草科技领军期刊
中国精品科技期刊
中国高校百佳科技期刊
江苏省新闻出版政府奖期刊奖
RCCSE林学权威期刊（A+）
CSCD核心期刊
Scopus数据库收录期刊
中文核心期刊
SCD核心期刊

作者加群：102861116

微信公众号：南京林业大学学报

高级检索

南京林业大学学报（自然科学版） ›› 2019, Vol. 43 ›› Issue (6): 91-96.doi: 10.3969/j.issn.1000-2006.201902028

基于NIRS的马尾松苗木根部含水量预测模型

倪超¹(), 张云¹, 高捍东²

1.南京林业大学机械电子工程学院,江苏南京 210037
2.南京林业大学林学院,国家林业和草原局南京林木种子检验中心,江苏南京 210037

收稿日期:2019-02-26 修回日期:2019-09-08 出版日期:2019-11-30 发布日期:2019-11-30
基金资助:
国家自然科学基金面上项目(31570714);江苏省高校优秀中青年教师和校长境外研修计划

Prediction model of moisture content of masson pine seedling roots based on near infrared spectroscopy

NI Chao¹(), ZHANG Yun¹, GAO Handong²

1. College of Mechanical and Electronic Engineering,Nanjing Forestry University, Nanjing 210037, China
2. Southern Tree Seed Inspection Center, State Forestry and Grassland Administration,College of Forestry, Nanjing Forestry University, Nanjing 210037, China

Received:2019-02-26 Revised:2019-09-08 Online:2019-11-30 Published:2019-11-30

摘要/Abstract

摘要：

【目的】马尾松是我国南方主要造林树种,其根部水分含量是评价树木活力的重要指标。本研究构建了一种基于近红外光谱(near infrared spectroscopy,NIRS)的马尾松苗木根部含水量预测模型。【方法】首先采集根部近红外光谱数据,然后利用可变加权堆叠自动编码器结合支持向量回归构建预测模型。可变加权堆叠自编码器用来逐层提取与输出相关的特征,支持向量回归根据自编码器生成的特征实现了含水量更精确预测。【结果】与其他常用模型的结果相比,提出的模型在马尾松苗木根部水分预测中可以达到最佳性能,校正集中决定系数达到0.970 8,均方根误差为0.635 8;预测集中决定系数达到0.941 3,均方根误差为1.027 0。【结论】基于近红外光谱技术, 可变加权堆叠自动编码器与支持向量回归相结合可实现马尾松苗木根部含水量准确预测。

关键词: 马尾松苗木, 水分含量, 近红外光谱, 自动编码器, 可变加权, 支持向量回归

Abstract:

【Objective】 Masson pine is the main afforestation species in southern China. The moisture content of roots is an important indicator for evaluating the vigour of seedlings. In this study, the method for estimating moisture content of masson pine seedling roots based on near infrared spectroscopy is advanced. 【Method】 The spectrum of masson pine seedling root is obtained. The prediction model which combines variable-wise weighted stacked autoencoder with support vector regression, was proposed to estimate the moisture content in the root of masson pine seedlings. The variable-wise weighted stacked autoencoder is used to extract the features which relating to the output and the support vector regression further accurately predict the moisture contents by using the identified features. 【Result】 Compared with the other commonly used models, our model had the best performance in the root moisture prediction, withR² value of 0.970 8 and RMSE of 0.635 8 in the training dataset and R² value of 0.941 3 and RMSE of 1.027 0 in the validation dataset. 【Conclusion】 The combination of the variable-wise weighted stacked autoencoder and support vector regression is feasible to realize the accurate prediction of moisture content in masson pine seedlings based on near infrared spectroscopy.

Key words: masson pine seedling, moisture content, near infrared spectroscopy, autoencoder, variable-wise weighted, support vector regression

中图分类号:

S725

倪超,张云,高捍东. 基于NIRS的马尾松苗木根部含水量预测模型[J]. 南京林业大学学报（自然科学版）, 2019, 43(6): 91-96.

NI Chao, ZHANG Yun, GAO Handong. Prediction model of moisture content of masson pine seedling roots based on near infrared spectroscopy[J].Journal of Nanjing Forestry University (Natural Science Edition), 2019, 43(6): 91-96.DOI: 10.3969/j.issn.1000-2006.201902028.

图/表 5

图1

图2

图3

图4

表1

参考文献 29

[1]	沈海龙, 丁贵杰, 高捍东, 等. 苗木培育学[M]. 北京: 中国林业出版社, 2009.
	SHEN H L, DING G J, GAO H D, et al. Seedling cultivation [M]. Beijing: China Forestry Publishing House, 2009.
[2]	喻方圆, 徐锡增. 苗木生理与质量研究进展[J]. 世界林业研究, 2000(4):17-24. DOI: 10.13348/j.cnki.sjlyyj.2000.04.006.
	YU F Y, XU X Z. Advances in research on seedling physiology and quality[J]. World Forestry Research, 2000(4):17-24.
[3]	张亚伟, 王书茂, 陈度, 等. 基于近红外的小麦植株含水率检测方法[J]. 农业机械学报, 2017, 48(S1):118-122,261.
	ZHANG Y W, WANG S M, CHEN D, et al. Method for detecting moisture content of wheat plants based on near infrared[J]. Transactions of the Chinese Society for Agricultural Machinery, 2017, 48(S1):118-122,261.
[4]	吴伟斌, 刘文超, 李泽艺, 等. 基于高光谱的茶叶含水量检测模型建立与试验研究[J]. 河南农业大学学报, 2018, 52(5):818-824. DOI: 10.16445/j.cnki.1000-2340.2018.05.024.
	WU W B, LIU W C, LI Z Y, et al. Establishment and experimental study of tea water content detection model based on hyperspectral[J]. Journal of Henan Agricultural University, 2018, 52(5):818-824.
[5]	刘燕德, 邓清, 张光伟. 高光谱技术的赣南脐橙叶片含氮量分析[J]. 中国农机化学报, 2016, 37(9):99-103. DOI: 10.13733/j.jcam.issn.2095-5553.2016.09.022.
	LIU Y D, DENG Q, ZHANG G W. Analysis of nitrogen content in leaves of minnan navel orange by hyperspectral technique[J]. Journal of Chinese Agricultural Mechanization, 2016, 37(9):99-103.
[6]	孔军龙, 赵京音, 杨娟. 基于可见/近红外光谱技术定量分析青菜叶片含氮量[J]. 上海农业学报, 2013, 29(3):36-39. DOI: 10.3969/j.issn.1000-3924.2013.03.009.
	KONG J L, ZHAO J Y, YANG J. Quantitative analysis of nitrogen content in leaves of green cabbage based on visible/near infrared spectroscopy[J]. Acta Agriculturae Shanghai, 2013, 29(3):36-39.
[7]	王志超, 王建军, 戴晓宇, 等. 基于低成本RGB相机和近红外相机的作物叶片叶绿素含量估测方法比较研究[J]. 吉林农业, 2018(24):53. DOI: 10.14025/j.cnki.jlny.2018.24.031.
	WANG Z C, WANG J J, DAI X Y, et al. Comparative study on estimation methods of chlorophyll content in crop leaves based on low cost RGB camera and near infrared camera[J]. Agriculture of Jilin, 2018(24):53.
[8]	裴浩杰, 冯海宽, 李长春, 等. 基于多元线性回归和随机森林的苹果叶绿素含量高光谱估测方法比较[J]. 江苏农业科学, 2018, 46(17):224-230. DOI: 10.15889/j.issn.1002-1302.2018.17.060.
	PEI H J, FENG H K, LI C C, et al. Comparison of hyperspectral estimation methods of apple chlorophyll content based on multiple linear regression and random forest[J]. Jiangsu Agricultural Sciences, 2018, 46(17):224-230.
[9]	AIKEN L S, WEST S G, PITTS S C. Multiple linear regression[J]. Handbook of Psychology, 2003, 4(19):481-507.
[10]	李颖, 李耀翔, 徐浩凯, 等. 基于降噪处理的蒙古栎木材气干密度NIRS定标模型[J]. 南京林业大学学报(自然科学版), 2016, 40(6):148-156. DOI: 10.3969/j.issn.1000-2006.2016.06.023.
	LI Y, LI Y X, XU H K, et al. NIRS calibration model for air-dry density of Mongolian oak wood based on noise reduction [J]. Journal of Nanjing Forestry University(Natural Sciences Edition), 2016, 40(6):148-156.
[11]	刘镇波, 孙凤亮, WANG X M, 等. 基于近红外光谱法的人工林杨木木质素含量预测[J]. 南京林业大学学报(自然科学版), 2013, 37(6):121-126. DOI: 10.3969/j.issn.1000-2006.2013.06.024
	LIU Z B, SUN F L, WANG X M, et al. Prediction of lignin content in poplar plantation based on near infrared spectroscopy[J]. Journal of Nanjing Forestry University (Natural Sciences Edition), 2013, 37(6):121-126.
[12]	DEVOS O, RUCKEBUSCH C, DURAND A, et al. Support vector machines (SVM) in near infrared (NIR) spectroscopy: Focus on parameters optimization and model interpretation[J]. Chemometrics and Intelligent Laboratory Systems 2009, 96:27-33. doi: 10.1016/j.chemolab.2008.11.005
[13]	Üstün B, MELSSEN W, BUYDENS L. Visualisation and interpretation of support vector regression models[J]. Analytica Chimica Acta, 2007, 595:299-309. doi: 10.1016/j.aca.2007.03.023
[14]	BALABIN R M, LOMAKINA E I, SAFIEVA R Z. Neural network (ANN) approach to biodiesel analysis: analysis of biodiesel density, kinematic viscosity, methanol and water contents using near infrared (NIR) spectroscopy[J]. Fuel, 2011, 90:2007-2015. doi: 10.1016/j.fuel.2010.11.038
[15]	沈剑波, 雷相东, 李玉堂, 等. 基于BP神经网络的长白落叶松人工林林分平均高预测[J]. 南京林业大学学报(自然科学版), 2018, 42(2):147-154.
	SHEN J B, LEI X D, LI Y T, et al. Prediction of average height of stands ofLarix olgensis plantation based on BP neural network [J]. Journal of Nanjing Forestry University(Natural Sciences Edition), 2018, 42(2):147-154.
[16]	HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18:1527-1554. doi: 10.1162/neco.2006.18.7.1527
[17]	LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86:2278-2324. doi: 10.1109/5.726791
[18]	VINCENT P, LAROCHELLE H, LAJOIE I, et al. Stacked denoising autoencoders: learning useful representations in a deep network with a local denoising criterion[J]. Journal of Machine Learning Research, 2010, 11:3371-3408.
[19]	GEHRING J, MIAO Y, METZE F, et al. Extracting deep bottleneck features using stacked auto-encoders[C]// Acoustics, Speech and Signal Processing (ICASSP), IEEE International Conference: 2013: 3377-3381.
[20]	TAO C, PAN H, LI Y, et al. Unsupervised spectral-spatial feature learning with stacked sparse autoencoder for hyperspectral imagery classification[J]. IEEE Geoscience and Remote Sensing Letters, 2015, 12:2438-2442. doi: 10.1109/LGRS.2015.2482520
[21]	WANG X, LIU H. Soft sensor based on stacked auto-encoder deep neural network for air preheater rotor deformation prediction[J]. Advanced Engineering Informatics, 2018, 36:112-119. doi: 10.1016/j.aei.2018.03.003
[22]	YU X, TANG L, WU X, et al. Nondestructive freshness discriminating of shrimp using visible/near-infrared hyperspectral imaging technique and deep learning algorithm[J]. Food Analytical Methods, 2018, 11:768-780. doi: 10.1007/s12161-017-1050-8
[23]	ZHANG W, LIU Z, HU J, et al. Near infrared spectroscopy drug discrimination method based on stacked sparse auto-encoders extreme learning machine[J]. Artificial Intelligence and Robotics, 2018, 752:203-211.
[24]	LIU T, LI Z, YU C, et al. NIRS feature extraction based on deep auto-encoder neural network[J]. Infrared Physics & Technology, 2017, 87:124-128.
[25]	XU L, JIANG J H, WU H L, et al. Variable-weighted PLS[J]. Chemometrics and Intelligent Laboratory Systems, 2007, 85:140-143. doi: 10.1016/j.chemolab.2006.06.001
[26]	YUAN X, HUANG B, WANG Y, et al. Deep learning based feature representation and its application for soft sensor modeling with variable-wise weighted SAE[J]. IEEE Transactions on Industrial Informatics, 2018, 14(7):3235-3243. doi: 10.1109/TII.9424
[27]	BALABIN R M, LOMAKINA E I. Support vector machine regression (SVR/LS-SVM)—an alternative to neural networks (ANN) for analytical chemistry? Comparison of nonlinear methods on near infrared (NIR) spectroscopy data[J]. Analyst, 2011, 136:1703-1712. doi: 10.1039/c0an00387e
[28]	AVCI E. Selecting of the optimal feature subset and kernel parameters in digital modulation classification by using hybrid genetic algorithm-support vector machines: HGASVM[J]. Expert Systems with Applications, 2009, 36:1391-1402. doi: 10.1016/j.eswa.2007.11.014
[29]	PAL M, MATHER P. Support vector machines for classification in remote sensing[J]. International Journal of Remote Sensing, 2005, 26:1007-1011. doi: 10.1080/01431160512331314083

模型 model	校正集 calibration set		预测集 prediction set
模型 model	R²	均方根误差 RMSE	R²	均方根误差 RMSE
PLSR	0.838 5	1.436 0	0.839 9	1.461 0
SVR	0.926 7	0.962 9	0.867 2	1.399 0
SAE-ANN	0.950 8	0.844 4	0.883 8	1.398 0
SAE-SVR	0.971 5	0.656 4	0.894 3	1.376 0
VW-SAE-ANN	0.960 5	0.760 5	0.913 1	1.197 0
VW-SAE-SVR	0.970 8	0.635 8	0.941 3	1.027 0

基于NIRS的马尾松苗木根部含水量预测模型

Prediction model of moisture content of masson pine seedling roots based on near infrared spectroscopy

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 5

参考文献 29

相关文章 6

编辑推荐

Metrics

本文评价

作者

读者

审稿

[1]	王珏, 李彦杰, 陈益存, 高暝, 赵耘霄, 吴立文, 黄世清, 张永志, 朱康烁, 汪阳东. 近红外光谱技术在林业领域的应用[J]. 南京林业大学学报（自然科学版）, 2023, 47(3): 237-246.
[2]	孙铭辰, 姜立春. 基于机器学习算法的樟子松立木材积预测[J]. 南京林业大学学报（自然科学版）, 2023, 47(1): 31-37.
[3]	李颖,李耀翔,徐浩凯,姜立春. 基于降噪处理的蒙古栎木材气干密度NIRS定标模型[J]. 南京林业大学学报（自然科学版）, 2016, 40(06): 148-156.
[4]	刘镇波,孙凤亮,WANGXiangming,沈晓燕,王鹏,薛占川,刘一星. 基于近红外光谱法的人工林杨木木质素含量预测[J]. 南京林业大学学报（自然科学版）, 2013, 37(06): 121-126.
[5]	胡晓健1，喻方圆2*，刘建兵3，万劲4. 干旱胁迫对不同种源马尾松苗木针叶内可溶性糖含量的影响[J]. 南京林业大学学报（自然科学版）, 2009, 33(05): 55-.
[6]	喻方圆1，刘建兵1，胡晓健2. 马尾松苗木中松醇的分布及其对干旱胁迫的响应[J]. 南京林业大学学报（自然科学版）, 2009, 33(02): 13-.