基于特征波段选择和机器学习的陆地棉叶片水分估算

doi:10.12118/j.issn.1000-6060.2022.667

Abstract

Abstract:

It is critical to ensure timely and accurate monitoring of leaf water content (LWC) when assessing the growth status of cotton. To accurately estimate cotton LWC, hyperspectral data, and leaf water data from cotton leaves in the oasis of the Ugan River-Kuqa River Delta, Xinjiang, China, were selected and processed using fractional differentiation of raw spectra. The sample were analyzed through correlation coefficient analysis, competitive adaptive reweighted sampling (CARS), successive projections algorithm (SPA), genetic algorithm (GA), Monte Carlo uninformative variables elimination (MC-UVE), and a combination of CARS and SPA to filter the feature bands. The modeling of the LWC inversion was executed through random forest regression (RFR) based on the whale optimization algorithm (WOA), and independent samples were used for validation analysis. The results show that: (1) The disparities in the number and positions of the feature bands obtained using the different feature band screening methods are different, where the number of feature bands obtained through MC-UVE is 8 while CARS produced 38. The positions of the characteristic bands identified through the SPA, GA, and CARS-SPA methods are considerably consistent and fundamentally concentrated in the near-infrared range of 950-1050 nm. (2) The CARS-SPA-WOA-RFR model has the best inversion with an R² of 0.93 and a root mean square error of 0.032. This model can provide a decision basis for accurate and rapid monitoring of cotton drought and precision irrigation.

Key words: spectral, leaf water content, feature band selection, machine learning

CUI Jintao, Mamat SAWUT. Estimation of leaf water content in upland cotton based on feature band selection and machine learning[J].Arid Land Geography, 2023, 46(11): 1836-1847.

Figures/Tables 16

Fig. 1

Tab. 1

Tab. 2

Fig. 2

Fig. 3

Fig. 4

Fig. 5

Fig. 6

Fig. 7

Fig. 8

Fig. 9

Fig. 10

Tab. 3

Tab. 4

Fig. 11

Fig. 12

References 33

[1]	刘文静, 范永胜, 董彦琪, 等. 我国棉花生产现状分析及建议[J]. 中国种业, 2022(1): 21-25.
	[ Liu Wenjing, Fan Yongsheng, Dong Yan Qi, et al. Analysis and suggestions on the current situation of cotton production in China[J]. China Seed Industry, 2022(1): 21-25. ]
[2]	马春玥, 买买提·沙吾提, 依尔夏提·阿不来提, 等. 新疆棉花种植业地理集聚特征及影响因素研究[J]. 作物学报, 2019, 45(12): 1859-1867. doi: 10.3724/SP.J.1006.2019.94041
	[ Ma Chunyue, Sawut Mamat, Ablet Ershat, et al. Characteristics and influencing factors of geographical agglomeration of cotton plantation in Xinjiang[J]. Acta Agronomica Sinica, 2019, 45(12): 1859-1867. ] doi: 10.3724/SP.J.1006.2019.94041
[3]	孙俊, 丛孙丽, 毛罕平, 等. 基于高光谱的油麦菜叶片水分CARS-ABC-SVR预测模型[J]. 农业工程学报, 2017, 33(5): 178-184.
	[ Sun Jun, Cong Sunli, Mao Hanping, et al. CARS-ABC-SVR model for predicting leaf moisture of leaf-used lettuce based on hyperspectral[J]. Transactions of the Chinese Society of Agricultural Engineering, 2017, 33(5): 178-184. ]
[4]	Junttila S, Hölttä T, Saarinen N, et al. Close-range hyperspectral spectroscopy reveals leaf water content dynamics[J]. Remote Sensing of Environment, 2022, 277: 113071, doi: 10.20944/preprints 202108.0497.v1.
[5]	Li L, Ustin S L, Riano D. Retrieval of fresh leaf fuel moisture content using genetic algorithm partial least squares (GA-PLS) modeling[J]. IEEE Geoscience and Remote Sensing Letters, 2007, 4(2): 216-220. doi: 10.1109/LGRS.2006.888847
[6]	Sun J, Zhou X, Hu Y G, et al. Visualizing distribution of moisture content in tea leaves using optimization algorithms and NIR hyperspectral imaging[J]. Computers and Electronics in Agriculture, 2019, 160: 153-159. doi: 10.1016/j.compag.2019.03.004
[7]	Li X L, Wei Z X, Peng F F, et al. Estimating the distribution of chlorophyll content in CYVCV-infected lemon leaf using hyperspectral imaging[J]. Computers and Electronics in Agriculture, 2022, 198: 107036, doi: 10.1016/j.compag.2022.107036.
[8]	杨宝华, 陈建林, 陈林海, 等. 基于敏感波段的小麦冠层氮含量估测模型[J]. 农业工程学报, 2015, 31(22): 176-182.
	[ Yang Baohua, Chen Jianlin, Chen Linhai, et al. Estimation model of wheat canopy nitrogen content based on sensitive bands[J]. Transactions of the Chinese Society of Agricultural Engineering, 2015, 31(22): 176-182. ]
[9]	张文旭, 佟炫梦, 周天航, 等. 基于高光谱成像的棉花叶片氮素含量遥感估测[J]. 沈阳农业大学学报, 2021, 52(5): 586-596.
	[ Zhang Wenxu, Tong Xuanmeng, Zhou Tianhang, et al. Remote sensing estimation of cotton leaf nitrogen content based on hyperspectral imaging[J]. Journal of Shenyang Agricultural University, 2021, 52(5): 586-596. ]
[10]	易翔, 张立福, 吕新, 等. 基于无人机高光谱融合连续投影算法估算棉花地上部生物量[J]. 棉花学报, 2021, 33(3): 224-234.
	[ Yi Xiang, Zhang Lifu, Lü Xin, et al. Estimation of cotton above-ground biomass based on unmanned aerial vehicle hyperspectral and successive projections algorithm[J]. Cotton Science, 2021, 33(3): 224-234. ]
[11]	陈鹏. 基于无人机多源遥感的马铃薯叶绿素含量反演机理及模型构建[D]. 焦作: 河南理工大学, 2019.
	[ Chen Peng. Retrieval mechanism and model construction of chlorophyll content in potato based on multi-source remote sensing of unmanned aerial vehicle[D]. Jiaozuo: Henan Polytechnic University, 2019. ]
[12]	于雷, 洪永胜, 周勇, 等. 高光谱估算土壤有机质含量的波长变量筛选方法[J]. 农业工程学报, 2016, 32(13): 95-102.
	[ Yu Lei, Hong Yongsheng, Zhou Yong, et al. Wavelength variable selection methods for estimation of soil organic matter content using hyperspectral technique[J]. Transactions of the Chinese Society of Agricultural Engineering, 2016, 32(13): 95-102. ]
[13]	王佳文, 彭杰, 纪文君, 等. 基于电磁感应数据的南疆棉田土壤pH反演研究[J]. 干旱区研究, 2022, 39(4): 1293-1302.
	[ Wang Jiawen, Peng Jie, Ji Wenjun, et al. Soil pH inversion based on electromagnetic induction data in cotton field of southern Xinjiang[J]. Arid Zone Research, 2022, 39(4): 1293-1302. ]
[14]	Li L L, Sun J, Tseng M L, et al. Extreme learning machine optimized by whale optimization algorithm using insulated gate bipolar transistor module aging degree evaluation[J]. Expert Systems with Applications, 2019, 127: 58-67. doi: 10.1016/j.eswa.2019.03.002
[15]	Zhou J, Zhu S Li, Qiu Y G, et al. Predicting tunnel squeezing using support vector machine optimized by whale optimization algorithm[J]. Acta Geotechnica, 2022, 17: 1343-1366. doi: 10.1007/s11440-022-01450-7
[16]	Zhao F, Li W D. A combined model based on feature selection and WOA for PM_2.5 concentration forecasting[J]. Atmosphere, 2019, 10(4): 223, doi: 10.3390/atmos10040223.
[17]	苏毅, 王克如, 李少昆, 等. 棉花植株水分含量的高光谱监测模型研究[J]. 棉花学报, 2010, 22(6): 554-560.
	[ Su Yi, Wang Keru, Li Shaokun, et al. Monitoring models of the plant water content based on cotton canopy hyperspectral reflectance[J]. Cotton Science, 2010, 22(6): 554-560. ]
[18]	王强, 易秋香, 包安明, 等. 棉花冠层水分含量估算的高光谱指数研究[J]. 光谱学与光谱分析, 2013, 33(2): 507-512. pmid: 23697143
	[ Wang Qiang, Yi Qiuxiang, Bao Anming, et al. Discussion on hyperspectral index for the estimation of cotton canopy water content[J]. Spectroscopy and Spectral Analysis, 2013, 33(2): 507-512. ] pmid: 23697143
[19]	赵巧珍, 丁建丽, 韩礼敬, 等. MODIS和Landsat时空融合影像在土壤盐渍化监测中的适用性研究——以渭干河—库车河三角洲绿洲为例[J]. 干旱区地理, 2022, 45(4): 1155-1164.
	[ Zhao Qiaozhen, Ding Jianli, Han Lijing, et al. Exploring the application of MODIS and Landsat spatiotemporal fusion images in soil salinization: A case of Ugan River-Kuqa River Delta Oasis[J]. Arid Land Geography, 2022, 45(4): 1155-1164. ]
[20]	玉苏甫·买买提, 吐尔逊·艾山, 买合皮热提·吾拉木. 新疆渭-库绿洲棉花种植面积遥感监测研究[J]. 农业现代化研究, 2014, 35(2): 240-243.
	[ Mamat Yusup, Hasan Tursun, Gulam Magpirat. Remote sensing of cotton plantation areas monitoring in delta oasis of Ugan-Kucha River, Xinjiang[J]. Research of Agricultural Modernization, 2014, 35(2): 240-243. ]
[21]	刘帆. 分数阶微分算法在医学超声弹性图像去噪中的应用研究[D]. 昆明: 昆明理工大学, 2018.
	[ Liu Fan. Application of fractional differential algorithm in medical ultrasonic elastic image denoising[D]. Kunming: Kunming University of Science and Technology, 2018. ]
[22]	李长春, 施锦锦, 马春艳, 等. 基于小波变换和分数阶微分的冬小麦叶绿素含量估算[J]. 农业机械学报, 2021, 52(8): 172-182.
	[ Li Changchun, Shi Jinjin, Ma Chunyan, et al. Estimation of chlorophyll content in winter wheat based on wavelet transform and fractional differential[J]. Transactions of the Chinese Society for Agricultural Machinery, 2021, 52(8): 172-182. ]
[23]	Li H D, Liang Y Z, Xu Q S, et al. Key wavelengths screening using competitive adaptive reweighted sampling method for multivariate calibration[J]. Analytica Chimica Acta, 2009, 648(1): 77-84. doi: 10.1016/j.aca.2009.06.046 pmid: 19616692
[24]	Zhang J K, Rivard B, Rogge D M. The successive projection algorithm (SPA), an algorithm with a spatial constraint for the automatic search of endmembers in hyperspectral data[J]. Sensors, 2008, 8(2): 1321-1342. pmid: 27879768
[25]	程介虹, 陈争光, 衣淑娟. 最小相关系数的多元校正波长选择算法[J]. 光谱学与光谱分析, 2022, 42(3): 719-725.
	[ Cheng Jiehong, Chen Zhengguang, Yi Shujuan. Wavelength selection algorithm based on minimum correlation coefficient for multivariate calibration[J]. Spectroscopy and Spectral Analysis, 2022, 42(3): 719-725. ]
[26]	宋相中. 近红外光谱定量分析中三种新型波长选择方法研究[D]. 北京: 中国农业大学, 2017.
	[ Song Xiangzhong. Research of three new wavelength selection methods in near infrared spectroscopy quantitative analysis area[D]. Beijing: China Agricultural University, 2017. ]
[27]	Mirjalili S, Lewis A. The whale optimization algorithm[J]. Advances in Engineering Software, 2016, 95: 51-67. doi: 10.1016/j.advengsoft.2016.01.008
[28]	李畸勇, 张伟斌, 赵新哲, 等. 改进鲸鱼算法优化支持向量回归的光伏最大功率点跟踪[J]. 电工技术学报, 2021, 36(9): 1771-1781.
	[ Li Qiyong, Zhang Weibin, Zhao Xinzhe, et al. Global maximum power point tracking for PV array based on support vector regression optimized by improved whale algorithm[J]. Transactions of China Electrotechnical Society, 2021, 36(9): 1771-1781. ]
[29]	吾木提·艾山江, 买买提·沙吾提, 马春玥. 基于分数阶微分和连续投影算法-反向传播神经网络的小麦叶片含水量高光谱估算[J]. 激光与光电子学进展, 2019, 56(15): 251-259.
	[ Hasan Umut, Sawut Mamat, Ma Chunyue. Hyperspectral estimation of wheat leaf water content using fractional differentials and successive projection algorithm-back propagation neural network[J]. Laser & Optoelectronics Progress, 2019, 56(15): 251-259. ]
[30]	Zhang M J, Zhang S Z, Iqbal J. Key wavelengths selection from near infrared spectra using Monte Carlo sampling-recursive partial least squares[J]. Chemometrics and Intelligent Laboratory Systems, 2013, 128: 17-24. doi: 10.1016/j.chemolab.2013.07.009
[31]	Han Z Z, Deng L M. Application driven key wavelengths mining method for aflatoxin detection using hyperspectral data[J]. Computers and Electronics in Agriculture, 2018, 153: 248-255. doi: 10.1016/j.compag.2018.08.018
[32]	Jia M, Li W, Wang K K, et al. A newly developed method to extract the optimal hyperspectral feature for monitoring leaf biomass in wheat[J]. Computers and Electronics in Agriculture, 2019, 165: 104942, doi: 10.1016/j.compag.2019.104942.
[33]	Mohammadi B, Mehdizadeh S. Modeling daily reference evapotranspiration via a novel approach based on support vector regression coupled with whale optimization algorithm[J]. Agricultural Water Management, 2020, 237: 106145, doi: 10.1016/j.agwat.2020.106145.

数据类型	样本类型	样本数/个	叶片水分含量/%
数据类型	样本类型	样本数/个	均值	标准偏差	变异系数
数据集Ⅰ	建模集	70	78.34	0.043	5.54
	验证集	30	77.30	0.042	5.40
	样本集	100	78.03	0.043	5.50
数据集Ⅱ	验证集	38	79.00	0.074	9.35
数据集Ⅱ	样本集	125	77.79	0.080	10.26

变量筛选方法	描述
CC	运算效率高，过程简单。变量间存在共线性。
CARS	有效去除自相关性高的波段，适合高维数据的筛选^[23]。变量间存在共线性，选择波段稳定性低。
SPA	变量间冗余少，共线性最小，缩短建模时间^[24]。没有考虑所有特征波长之间的共线性^[25];挑选特征变量过程中倾向于选择共线性较小的变量点而不是有效变量点^[25]。
GA	具有全局优化能力。但需要多次运算以确定最佳变量子集。
MC-UVE	稳定性较高。需要定义阈值，导致变量数目改变。
CARS-SPA	进一步剔除冗余信息，提取出有效波段且多重共线性较低，运算效率较高^[6]。筛选变量较少，容易丢失关键信息。特征波长集建模效果受粗选算法结果的影响较大^[26]。

微分阶数	算法	表达式	R²	RMSE
0.7	WOA-RFR	y=0.6934x+0.2403	0.894	0.017
	SVR	y=0.4627x+0.4201	0.553	0.029
	RFR	y=0.5463x+0.3546	0.885	0.021
0.9	WOA-RFR	y=0.7086x+0.2290	0.846	0.018
	SVR	y=0.5242x+0.3706	0.650	0.026
	RFR	y=0.5545x+0.3491	0.882	0.021
1.1	WOA-RFR	y=0.7086x+0.2254	0.877	0.017
	SVR	y=0.6291x+0.2896	0.767	0.022
	RFR	y=0.6121x+0.3045	0.897	0.019
1.3	WOA-RFR	y=0.6778x+0.2522	0.927	0.016
	SVR	y=0.7505x+0.1953	0.881	0.016
	RFR	y=0.6206x+0.2976	0.898	0.019
1.7	WOA-RFR	y=0.6381x+0.2816	0.844	0.020
	SVR	y=0.7293x+0.2117	0.889	0.016
	RFR	y=0.5831x+0.3271	0.893	0.020
1.9	WOA-RFR	y=0.7061x+0.2293	0.851	0.018
	SVR	y=0.6169x+0.3022	0.732	0.023
	RFR	y=0.5676x+0.3392	0.880	0.021

变量筛选方法	算法	建模集		验证集
变量筛选方法	算法	R²	RMSE	R²	RMSE
全波段	WOA-RFR	0.573	0.029	0.480	0.030
	SVR	0.375	0.032	0.035	0.037
	RFR	0.583	0.030	0.021	0.032
CC	WOA-RFR	0.927	0.016	0.946	0.017
	SVR	0.881	0.016	0.908	0.016
	RFR	0.898	0.019	0.950	0.022
CARS	WOA-RFR	0.912	0.017	0.929	0.015
	SVR	0.688	0.040	0.977	0.033
	RFR	0.889	0.025	0.916	0.023
SPA	WOA-RFR	0.937	0.016	0.941	0.015
	SVR	0.398	0.034	0.757	0.024
	RFR	0.913	0.022	0.964	0.022
GA	WOA-RFR	0.622	0.028	0.647	0.025
	SVR	0.712	0.024	0.723	0.002
	RFR	0.889	0.024	0.858	0.001
MC-UVE	WOA-RFR	0.847	0.020	0.868	0.016
	SVR	0.882	0.015	0.883	0.015
	RFR	0.852	0.025	0.821	0.024
CARS-SPA	WOA-RFR	0.935	0.017	0.942	0.019
	SVR	0.326	0.036	0.568	0.029
	RFR	0.911	0.023	0.878	0.024

Estimation of leaf water content in upland cotton based on feature band selection and machine learning

RichHTML

PDF (PC)

Like

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 16

References 33

Related Articles 15

Metrics

Comments

Recommended 0

[1]	LI Shiyao, CONG Shixiang, WANG Rongrong, YU Hailong, HUANG Juying. Monitoring of maize canopy SPAD value under drought stress based on UAV multi-spectral remote sensing [J]. Arid Land Geography, 2023, 46(7): 1121-1132.
[2]	WEI Huimin, JIA Keli, ZHANG Xu, ZHANG Junhua. Prediction of soil salinity based on machine learning and multispectral remote sensing in Yinchuan Plain [J]. Arid Land Geography, 2023, 46(1): 103-114.
[3]	ZHAO Shuang,DING Jianli,HAN Lijing,HUANG Shuai,GE Xiangyu. Response analysis and modeling of microwave dielectric properties of typical saline soil in Xinjiang [J]. Arid Land Geography, 2022, 45(5): 1534-1546.
[4]	LIU Xuhui,BAI Yungang,CHAI Zhongping,ZHANG Jianghui,DING Bangxin,JIANG Zhu. Inversion and validation of soil salinity based on multispectral remote sensing in typical oasis cotton field in spring [J]. Arid Land Geography, 2022, 45(4): 1165-1175.
[5]	YIN Hanmin,Guli JIAPAER,YU Tao,Jeanine UMUHOZA,LI Xu. Wheat yield estimation with remote sensing in northern Kazakhstan [J]. Arid Land Geography, 2022, 45(2): 488-498.
[6]	HU Guigui,YANG Fenli,YANG Lian’an,ZHENG Yurong,WANG Hui,CHEN Weijun,LI Yali. Spatial prediction modeling of soil organic matter content based on principal components and machine learning [J]. Arid Land Geography, 2021, 44(4): 1114-1124.
[7]	TIAN Yanjun,SHI Ying,SHUAI Yanmin,YANG Jian,SHAO Congying,FAN Lianlian,MA Honghong. Land cover information retrieval from temporal features based remote sensing images [J]. Arid Land Geography, 2021, 44(2): 450-459.
[8]	LI Hui-rong. Estimation of snow depth based on reflectance and bright temperature in Xilin Gol League [J]. Arid Land Geography, 2020, 43(6): 1567-1572.
[9]	CAO Xiao-yi, DING Jian-li, GE Xiang-yu, LIANG Jing, CHEN Wen-qian, CHEN Xiang-yue, TANG Pu-en. Estimation of soil conductivity based on spectral simulation of different satellites [J]. Arid Land Geography, 2020, 43(1): 172-181.
[10]	BAO Qing-ling, DING Jian-li, WANG Jing-zhe, CAI Liang-hong. Hyperspectral detection of soil organic matter content based on random forest algorithm [J]. Arid Land Geography, 2019, 42(6): 1404-1414.
[11]	ZHANG Xi-wei, WANG Lei, WANG Xi-yuan. Sand information extraction method based on CART decision tree [J]. Arid Land Geography, 2019, 42(5): 1133-1140.
[12]	WANG Tao, YU Cai-li, YAO Na, ZHANG Nan-nan, BAI Tie-cheng. A comparison of the salt content in sandy soil between the MLR model and PLSR model [J]. 干旱区地理, 2018, 41(6): 1295-1302.
[13]	CAO Xiao-yang, MU Yue, CAO Xiao-ming, ZHANG Pu, FENG Yi-ming. Grain size retrieving of gobi surface based on hyperspectral data [J]. , 2017, 40(2): 397-404.
[14]	WANG Shuang, DING Jian-li, WANG lu, NIU Zeng-yi. Remote sensing monitoring of soil salinization based on surface spectral modeling [J]. , 2016, 39(1): 190-198.
[15]	XI A-xing, LIU Zhi-hui, XU Qian, ZHANG Bo. Retrieval method for estimating snow porosity in the northern slope of Tianshan Mountains based on hyperspectral data [J]. , 2015, 38(6): 1253-1261.