基于机器学习的油菜叶片水分含量高光谱估测

doi:10.13304/j.nykjdb.2022.0977

摘要/Abstract

摘要：

含水率是影响光合作用的重要因素之一，为了构建效果更好、更具普适性的油菜叶片含水量（leaf water content， LWC）定量监测模型，以蕾薹期、初花期油菜叶片为研究对象，采用自然风干法去除叶片水分，同步采集叶片质量和光谱信息。为了降低干扰以及消除噪声，采用标准正态变量变换、Savitzky-Golay卷积平滑算法（SG平滑）、多元散射校正、一阶求导和二阶求导5种方法对光谱数据进行预处理，并结合偏最小二乘法（partial least squares， PLS）分析选取最优预处理方法；采用连续投影算法（successive projections algorithm，SPA）筛选预处理后的光谱特征变量，获得对水分含量变化敏感的特征波长；利用支持向量机（support vector regression， SVR）和BP神经网络（back-propagation neural network， BPNN）方法，以特征波长建立的光谱指数为自变量建立油菜叶片水分含量估算模型。结果表明：采用多元散射校正预处理综合表现最好，2个生育期预测集相关系数均达到0.71以上；通过SPA法选择特征变量，分别筛选出特征波长，其中蕾薹期6个，初花期7个；在蕾薹期和初花期叶片水分含量预测模型中，基于SVR模型和BPNN模型建立的模型预测集决定系数（R² ）均在0.800以上，均能实现油菜叶片水分含量的精准监测，其中SVR模型预测效果优于BPNN模型，R² 分别为0.857和0.827，RMSE分别为1.791和1.521。因此，利用油菜叶片高光谱建模反演油菜叶片含水率能准确监测油菜叶片含水率，可为精准农业水分管理提供理论参考。

关键词: 油菜, 叶片含水量, 高光谱, 机器学习

Abstract:

Leaf water content （LWC） is an important factor affecting the photosynthesis of rapeseed. In order to establish a quantitative monitoring model for LWC with better monitoring effect and universality， rapeseed leaves at the budding and initial flowering stages were selected as the research objects. The leaves were subjected to natural air-drying to remove water， and the mass and spectral information were collected simultaneously. To reduce interference and eliminate noise， 5 methods were used to preprocess the spectral data including standard normal variable transformation， Savitzky-Golay convolution smoothing algorithm （SG smoothing）， multiple scattering correction， first-order derivative， and second-order derivative， and the optimal preprocessing method was selected by combining with partial least squares （PLS） analysis. Successive projections algorithm （SPA） was used to select sensitive feature wavelengths for water content changes from the preprocessed spectra. Support vector regression （SVR） and back-propagation neural network （BPNN） were used to establish LWC estimation models based on spectral indices using the selected feature wavelengths as independent variables. The results showed that the multiple scattering correction method performed the best， and the correlation coefficients of the prediction sets for both growth stages were above 0.71. SPA selected 6 and 7 feature wavelengths for the budding and initial flowering stages， respectively. In the LWC prediction models for the 2 growth stages， the models based on SVR and BPNN had determination coefficients （R² ） above 0.800 for the prediction sets and could achieve accurate monitoring of LWC in rapeseed leaves. The SVR model had better prediction performance than the BPNN model， with R² values of 0.857 and 0.827 and RMSE values of 1.791 and 1.521， respectively. Therefore， using high-spectral modeling to invert LWC in rapeseed leaves can accurately detect LWC and provide theoretical reference for precision agriculture water management monitoring.

Key words: rape, leaf water content, hyperspectral, machine learning

中图分类号:

S126

宋丽芳, 廖桂平, 陈敏, 何罗驭阳. 基于机器学习的油菜叶片水分含量高光谱估测[J]. 中国农业科技导报, 2024, 26(5): 110-119.

Lifang SONG, Guiping LIAO, Min CHEN, Yuyang HE-LUO. Hyperspectral Estimation of Rape Leaf Water Content Based on Machine Learning[J]. Journal of Agricultural Science and Technology, 2024, 26(5): 110-119.

图/表 8

图1 原始光谱预处理结果A：原始光谱；B：SG平滑预处理；C：一阶导数预处理；D：SNV预处理；E：MSC预处理；F：二阶导数预处理

Fig. 1 Pre-treated result of original spectralA：Original spectral； B： SG smoothing pretreated； C：1st-derivative pretreated； D： SNV pretreated； E： MSC pretreated； F： 2nd-derivative pretreated

表1 不同预处理方法的叶片含水量PLS模型评价

Table 1 Evaluation of PLS model of leaf water content with different pretreatment methods

生育时期 Growth stage	方法 Method	训练集 Training set		测试集 Prediction set
生育时期 Growth stage	方法 Method	相关系数r	均方根误差RMSE	相关系数r	均方根误差RMSE
蕾薹期 Budding stage	SG平滑 SG smoothing	0.705	1.293	0.627	1.534
	MSC	0.764	1.172	0.735	1.421
	SNV	0.612	1.266	0.536	1.714
	一阶求导 1st-derivative	0.575	1.663	0.524	1.368
	二阶求导 2nd-derivative	0.323	2.759	0.317	1.975
初花期 Initialflowering stage	SG平滑 SG smoothing	0.625	1.672	0.601	2.124
	MSC	0.721	1.514	0.711	1.421
	SNV	0.597	1.463	0.547	1.842
	一阶求导 1st-derivative	0.624	1.453	0.586	1.265
	二阶求导 2nd-derivative	0.423	2.312	0.328	2.226

图2 SPA的特征变量选择

Fig. 2 Selection of characteristic variables of SPA during the budding stage and initial

表2 SPA的特征选择波长

Table 2 SPA feature selection wavelength

生育时期

Growth stage

变量数量

Variable number

输出波长

Output wavelength/nm

蕾薹期

Budding stage

523，561，1 390，1 481，2 316，2 491

初花期

Initial flowering stage

553，645，709，754，1 000，1 650，2 444

图3 光谱反射率BPNN回归拟合结果A：蕾薹期；B初花期

Fig. 3 BPNN regression fitting result of spectral reflectanceA:Budding stage; B:Initial flowering stage

图4 初花期BP神经网络测试集叶片含水量预测结果对比

Fig. 4 Comparison of leaf water content prediction results of BPNN test set at

表3 油菜叶片水分含量BPNN建模预测效果

Table 3 BPNN modeling and prediction effect of rape leaf water content

生育时期Growth stage

训练集 Training set

（n=315）

预测集Prediction set

（n=105）

R²

RMSE

R²

RMSE

蕾薹期

Budding stage

初花期

Initial flowering stage

表4 油菜叶片水分含量SVR建模预测效果

Table 4 SVR modeling and prediction effect of rape leaf water content

生育时期

Growth stage

训练集 Training set

（n=315）

预测集 Prediction set

（n=105）

R²

RMSE

R²

RMSE

蕾薹期

Budding stage

初花期

Initial flowering stage

参考文献 36

1	白桂萍,谢雄泽,谢捷,等.中国油菜生产布局时空演变及影响因素浅析[J].中国油脂,2022,48(4):1-6.
	BAI G P, XIE X Z, XIE J, et al.. Analysis on the temporal and spatial evolution and influencing factors of
	oilseed rape production layout in China [J]. China Oils Fats,2022,48(4):1-6.
2	黄杰.气候变化对中国油菜生产布局的影响研究[D].武汉:华中农业大学,2019.
	HUANG J. Study on the impact of climate change on the distribution of Chinese rape production [D]. Wuhan: Huazhong Agricultural University, 2019.
3	冷博峰,李先容,陈雪婷,等.2008-2019年中国油菜生产性状变化趋势[J].中国油料作物学报,2021, 43 (2):171-185.
	LENG B F, LI X R, CHEN X T, et al.. Variation trend of rapeseed production in china from 2008 to 2019 [J]. Chin. J. Oil Crop Sci., 2021,43(2):171-185.
4	张玮,王鑫梅,潘庆梅,等.干旱胁迫下雷竹叶片叶绿素的高光谱响应特征及含量估算[J].生态学报, 2018, 38 (18):6677-6684.
	ZHANG W, WANG X M, PAN Q M, et al.. Hyperspectral response characteristics and chlorophyll content estimation of Phyllostachys violascens leaves under drought stress [J]. Acta Ecol. Sin., 2018,38(18):6677-6684.
5	张海威,张飞,张贤龙,等.光谱指数的植被叶片含水量反演[J].光谱学与光谱分析,2018,38(5):1540-1546.
	ZHANG H W, ZHANG F, ZHANG X L, et al.. Inversion of vegetation leaf water content based on spectral index [J]. Spectrosc. Spect. Anal., 2018,38(5):1540-1546.
6	梁亮,张连蓬,林卉,等.基于导数光谱的小麦冠层叶片含水量反演[J].中国农业科学,2013,46(1):18-29.
	LIANG L, ZHANG L P, LIN H, et al.. Estimating canopy leaf water content in wheat based on derivative spectral [J]. Sci. Agric. Sin., 2013,46(1):18-29.
7	ZHANG Q X, LI Q B, ZHANG G J. Rapid determination of leaf water content using VIS/NIR spectroscopy analysis with wavelength selection [J]. J. Spectroscopy.,2012,27(2):93-105.
8	DINESH K M, NITIN B, et al.. Rethinking on the methodology for assessing global water and food challenges [J]. Inter. J. Water Resour. Dev., 2020,36(2-3):325-342.
9	瞿益民,唐合年,葛妹兰,等.油菜需水量试验分析[J].江苏水利,2005,(10):20-21, 23.
	QU Y M, TANG H N, GE M L, et al.. Analysis of rape water requirement test [J]. Jiangsu Water Resour., 2005,(10):20-21, 23.
10	黄纯倩,朱晓义,张亮,等.干旱和高温对油菜叶片光合作用和叶绿素荧光特性的影响[J].中国油料作物学报,2017,39(3):342-350.
	HUANG C Q, ZHU X Y, ZHANG L, et al.. Effects of drought and high temperature on photosynthesis and chlorophyll fluorescence characteristics of rapeseed leaves [J]. Chin. J. Oil Crop Sci., 2017,39(3):342-350.
11	MAMNABI S, NASROLLAHZADEH S, GHASSEMI-GOLEZANI K, et al.. Improving yield-related physiological characteristics of spring rapeseed by integrated fertilizer management under water deficit conditions [J]. J. Saudi Biol. Sci.,2020,27(3)797-804.
12	刘晓静,陈国庆,王良,等.不同生育时期冬小麦叶片相对含水量高光谱监测[J].麦类作物学报, 2018, 38(7):854-862.
	LIU X J, CHEN G Q, WANG L, et al.. Monitoring leaf relative water content of winter wheat based on hyperspectral index at different growth stages [J]. J. Triticeae Crops, 2018,38(7):854-862.
13	徐庆,马驿,蒋琦,等.水稻叶片含水量的高光谱遥感估算[J].遥感信息,2018,33(5):1-8.
	XU Q, MA Y, JIANG Q, et al.. Estimation of rice leaf water content based on hyperspectral remote sensing [J]. Remote Sens. Inform., 2018,33(5):1-8.
14	马岩川.基于高光谱遥感的棉花冠层水氮参数估算[D].北京:中国农业科学院,2020.
	MA Y C. Estimation of water and nitrogen parameters of cotton at canopy scale based on hyperspectral remote sensing [D].Beijing: Chinese Academy of Agricultural Sciences,2020.
15	王浩云,李晓凡,李亦白,等.基于高光谱图像和3D-CNN的苹果多品质参数无损检测[J].南京农业大学学报,2020,43(1):178-185.
	WANG H Y, LI X F, LI Y B, et al.. Non-destructive detection of apple multi-quality parameters based on hyperspectral imaging technology and 3D-CNN [J]. J. Nanjing Agric. Univ., 2020,43(1):178-185.
16	代秋芳,廖臣龙,李震,等.基于CARS-CNN的高光谱柑橘叶片含水率可视化研究[J].光谱学与光谱分析, 2022,42(9):2848-2854.
	DAI Q F, LIAO Q L, LI Z, et al.. Hyperspectral visualization of citrus leaf moisture content based on CARS-CNN [J]. Spectrosc. Spect. Anal., 2022,42(9):2848-2854.
17	潘庆梅,张劲松,张俊佩,等.不同品种核桃叶片含水量与高光谱反射率的相关性差异分析[J].林业科学研究, 2019,32(6):1-6.
	PAN Q M, ZHANG J S, ZHANG J P, et al.. Analysis of correlation and differences between leaf moisture and hyperspectral reflectance among different walnut varieties [J]. Forest Res.,2019,32(6):1-6.
18	张晓东,毛罕平,左志宇,等.干旱胁迫下油菜含水率的高光谱遥感估算研究[J].安徽农业科学, 2011, 39(30):18451-18452, 18487.
	ZHANG X D, MAO H P, ZUO Z Y, et al.. Study on estimation model for rape moisture content under water stress based on hyperspectral remote sensing [J]. J. Anhui Agric. Sci.,2011,39(30):18451-18452, 18487.
19	张晓东,李立,毛罕平,等.基于PCA-BP多特征融合的油菜水分胁迫无损检测[J].江苏大学学报(自然科学版),2016,37(2):174-182.
	ZHANG X D, LIP, MAO H P, et al.. Nondestructive testing method for rape water stress with multiple feature information fusion based on PCA-BP method [J]. J. Jiangsu Univ. (Nat. Sci.), 2016,37(2):174-182.
20	仝春艳,马驿,杨振忠,等.基于角度指数的油菜叶片等效水厚度估算研究[J].核农学报,2019,33(1):187-198.
	TONG C Y, MA Y, YANG Z Z, et al.. Estimation of equivalent water thickness of rapeseed leaves based on angle index [J]. J. Nucle. Agric. Sci., 2019,33(1):187-198.
21	张君,蔡振江,张东方,等.基于机器学习与光谱技术的油菜叶片含水率估测研究[J].河北农业大学学报, 2021,44(6):122-127.
	ZHANG J, CAI Z J, ZHANG D F, et al.. Estimation of water content in rape leaves by spectral reflectance combined with machine learning [J]. J. Hebei Agric. Univ., 2021,44(6):122-127.
22	潘月,曹宏鑫,齐家国,等.基于高光谱和数据挖掘的油菜植株含水率定量监测模型[J].江苏农业学报, 2022,38(6):1550-1558.
	PAN Y, CAO H X, QI J G, et al.. Quantitative monitoring models of plant water content in rapeseed based on hyperspectrum and related data mining [J]. J. Jiangsu Agric. Sci., 2022,38(6):1550-1558.
23	杨冬风,李爱传,刘金明,等.耦合平均影响值-连续投影算法优化种子活力近红外检测模型[J].光谱学与光谱分析,2022,42(10):3135-3142.
	YANG D F, LI A C, LIU J M, et al.. Optimization of seed vigor near-infrared detection by coupling mean impact value with successive projection algorithm [J]. Spectrosc. Spect. Anal., 2022,42(10):3135-3142.
24	张楠楠,张晓,王城坤,等.基于高光谱和连续投影算法的棉花叶面积指数估测[J].农业机械学报, 2022, 53(S1):257-262.
	ZHANG N N, ZHANG X, WANG C Q, et al.. Cotton LAI estimation based on hyperspectral and successive projection algorithm [J]. Trans. Chin. Soc. Agric. Mach., 2022,53(S1):257-262.
25	赵静远,熊智新,宁井铭,等.小波变换结合连续投影算法优化茶叶中咖啡碱的近红外分析模型[J].分析科学学报,2021,37(5):611-617.
	ZHAO J Y, XIONG Z X, NING J M, et al.. Wavelet tranform combined with SPA to optimize the near-infrared analysis model of caffeine in tea [J]. J. Anal. Sci., 2021,37(5):611-617.
26	郭阳,史勇,郭俊先,等.近红外光谱技术结合反向区间偏最小二乘算法-连续投影算法预测哈密瓜可溶性固形物含量[J].食品与发酵工业,2022,48(2):248-253.
	GUO Y, SHI Y, GUO J X, et al.. Prediction of soluble solids content in hami melon by combining near-infrared spectroscopy and BiPLS-SPA technology [J]. Food Ferment. Indust.,2022,48(2):248-253.
27	李锦卫.基于计算机视觉的水稻、油菜叶色—氮营养诊断机理与建模[D].长沙:湖南农业大学,2010.
	LI J W. Studies on the diagnosis mechanism and modeling of leaf color-nitrogen nutrition in rice and rapeseed plant by computer vision [D]. Changsha: Hunan Agricultural University,2010.
28	谢素华,杨明高.人民渠平原灌区油菜需水量及需水规律研究[J].四川水利,2001,(1):33-35.
29	冯鹏.油菜高产灌溉技术[J].农业开发与装备,2011, 6:39.
30	谢素华,杨明高,张清.油菜需水量及需水规律的研究[J].四川水利,1996,2:24-25.
31	王巧娟. 关中地区夏大豆—冬油菜的水分产量效应及灌溉制度模型模拟研究[D].杨凌:西北农林科技大学,2022.
	WANG Q J. Study on water yield effect of summer soybean winter rape and Simulation of irrigation schedule model in Guanzhong Area [D].Yangling: Northwest Agriculture Forestry University, 2022.
32	李岚涛.冬油菜氮素营养高光谱特异性及定量诊断模型构建与推荐追肥研究[D].武汉:华中农业大学,2018.
	LI L T. Research on hyperspectral characteristics, quantitative diagnostic models, and topdressing for winter oilseed rape nitrogen status [D]. Wuhan: Huazhong Agricultural University, 2018.
33	THOMAS J R, NAMKEN L N, OERTHER G F, et al.. Estimating leaf water content by reflectance measurements [J]. J. Agron.,1971,63(6):845-847.
34	SHIBAYAMA M, AKIYAMA T. Seasonal visible, near-infrared and mid-infrared spectra of rice canopiesin relation to LAI and above-ground dryphytomass [J]. Remote Sens. Env.,1989,27(2):119-127.
35	SHIBAYAMA M, TAKAHASHI W, MORINAGA S, et al.. Canopy water deficit detection in paddy rice using a high resolution field spectroradiometer [J].Remote Sens. Env.,1993,45(2):117-126.
36	黄敬峰,王福民,王秀珍.水稻高光谱遥感实验研究[M].杭州:浙江大学出版社,2010:1-315.

[1]	张振飞, 颜安, 郭靖, 赵宇航, 袁以琳, 刘鹏, 曲佐昊, 袁川. 基于无人机遥感的苹果产量估测模型研究[J]. 中国农业科技导报, 2025, 27(9): 110-119.
[2]	杨国涛, 张世杰, 陈超, 刘云, 贺琛, 宁英豪, 张勍. 5种烟草常用农药的高光谱识别技术研究[J]. 中国农业科技导报, 2025, 27(7): 122-132.
[3]	呼斯乐, 包玉龙, 图布新巴雅尔null, 陶际峰, 郭恩亮. 基于无人机高光谱和集成学习的春小麦叶绿素含量反演[J]. 中国农业科技导报, 2025, 27(6): 93-103.
[4]	李广瑶, 杨胜龙, 程田飞, 崔雪森, 周为峰, 张胜茂. 中西太平洋金枪鱼渔场环境特征及预报现状分析[J]. 中国农业科技导报, 2025, 27(5): 203-221.
[5]	陈林涛, 刘兆祥, 蓝莹, 牟向伟, 马旭, 王日俊. 基于高光谱技术与主成分分析的稻种品种识别研究[J]. 中国农业科技导报, 2025, 27(3): 104-111.
[6]	李紫琴, 王家强, 李贞, 邹德秋, 张小功, 罗霄玉, 柳维扬. 基于光谱指数的棉花叶片叶绿素密度估算研究[J]. 中国农业科技导报, 2024, 26(8): 103-111.
[7]	付彦博, 冷冰冰, 扁青永, 董志多, 刘国宏, 李海峰, 温云梦, 郭文博, 张万旭. 生物炭和油菜幼苗对土壤重金属镉污染的钝化效应[J]. 中国农业科技导报, 2024, 26(6): 183-190.
[8]	蒋沛含, 杨晓楠, 杨晨旭, 张爱军. 基于偏最小二乘回归的谷子冠层氮素含量高光谱估测研究[J]. 中国农业科技导报, 2024, 26(6): 91-101.
[9]	陈晨, 程大伟, 李兰, 顾红, 郭西智, 李明, 陈锦永. 油菜素内酯调控植物耐盐机理研究进展[J]. 中国农业科技导报, 2024, 26(2): 1-12.
[10]	刘博, 王旺田, 马骊, 武军艳, 蒲媛媛, 刘丽君, 方彦, 孙万仓, 张岩, 刘睿敏, 曾秀存. 白菜型油菜IPT基因家族鉴定及表达分析[J]. 中国农业科技导报, 2024, 26(2): 56-66.
[11]	杨婷婷, 田效琴, 李卓, 刘永红, 李伟. 不同类型地膜对土壤水热和油菜产量的影响[J]. 中国农业科技导报, 2024, 26(12): 157-163.
[12]	边凤霞, 刘凯歌, 容新民. 基于机器学习构建葡萄霜霉病预测模型及验证[J]. 中国农业科技导报, 2023, 25(8): 126-137.
[13]	陈明迪, 程兴茹, 徐博, 张海文, 王旺田, 王友华. 全球转基因油菜专利信息分析与技术展望[J]. 中国农业科技导报, 2023, 25(11): 8-19.
[14]	栗方亮, 孔庆波, 张青. 基于高光谱的琯溪蜜柚叶片磷素含量估算模型研究[J]. 中国农业科技导报, 2023, 25(1): 100-108.
[15]	燕辉. 基于叶绿素荧光分析的油菜响应低温胁迫机制研究[J]. 中国农业科技导报, 2023, 25(1): 58-64.