智能手机原位牧草生物量估算

doi:10.3864/j.issn.0578-1752.2021.05.006

摘要/Abstract

摘要：

【目的】生物量是草地生态系统物质和能量基础,是最基本的生态参量。以往基于卫星和航空遥感定量反演草地生物量过于专业化,难以在牧民间推广。因此,本文提出一种用手机近距离拍摄的真彩色图像估算牧草生物量方法,构建牧草生物量估算模型,为牧民方便、快捷、无损地掌握牧场牧草长势提供理论依据和技术支撑。【方法】首先,利用手机超高分辨率真彩色图像,分别基于植被指数、纹理特征以及联合植被指数和纹理特征构建牧草生物量估算特征集合。其次,为防止过多的特征提取带来维度灾难,提出一种XGBoost与序列前向选择相结合的特征选择算法（XGB-SFS）,进行特征筛选及最优子集构建。最后,使用随机森林回归和留一法交叉验证对比不同特征集合构建模型的生物量估算效果,分析不同类型特征及XGB-SFS算法在牧草地上生物量（Above Ground Biomass,AGB）估算中的作用。【结果】（1）对比单类型特征构建的模型,基于空间纹理特征的估算模型（R²=0.76）要优于基于光谱植被指数估算模型（R²=0.73）,表明纹理特征在超高分辨率牧草AGB估算中具有一定作用;（2）对比特征选择后的模型,联合空谱多类型特征构建模型优于任何一种单类型特征模型（R²=0.83,RMSE=127.57 g·m^-2,MAE=81.25 g·m^-2）,表明使用多类型特征构建模型,可一定程度上提高牧草AGB估算精度。（3）对比特征选择前后构建的模型,特征选择后的模型估算AGB效果要明显好于未进行特征选择的模型,且筛选出的特征与牧草生物量之间都存在较高的相关性,表明XGB-SFS能够很好降低数据维度的同时提高牧草AGB估算精度。【结论】手机超高分辨率真彩色图像可以对牧草生物量进行准确估算,本文提出的XGB-SFS算法也能从众多特征中筛选出与牧草生物量相关性较高的特征并提高模型估算精度。与以往专业遥感定量反演草地生物量相比,本文方法具有面向大众、成本低廉、使用方便等优势,研究将手机现场采集的数据与遥感和机器学习方法相结合,可开辟新的视角,支持农业信息化发展。

关键词: 生物量, 智能手机, 纹理特征, XGBoost, 牧草

Abstract:

【Objective】Biomass is the material and energy basis of grassland ecosystem and the most basic ecological parameter. In the past, the quantitative grassland biomass retrieval based on aerospace and aerial remote sensing was too specialized to be popularized among herders. Therefore, this paper proposed a method for estimating grassland biomass by using true color images taken on the phone near the ground, and constructing a grassland biomass estimation model, which provided a theoretical basis and technical support for herders to easily, quickly and non-destructively grasp the growth of grassland in their own pasture. 【Method】 Firstly, the feature sets of grassland biomass estimation were constructed based on vegetation index, texture features and combined vegetation index and texture features by using the ultra-high resolution true color images of mobile phones. Secondly, in order to prevent dimensional disaster caused by excessive feature extraction, this paper proposed a feature selection algorithm (XGB-SFS) that combined XGBoost and sequence forward selection to perform feature selection and optimal subset construction. Finally, random forest regression and leave-one-out cross-validation were used to compare the biomass estimation effects of different feature sets to build models, and analyze the role of different types of features and XGB-SFS algorithm in grassland AGB estimation.【Result】 (1) Compared with the model constructed by single-type features, the estimation model based on spatial texture features (R² = 0.76) was better than the estimation model based on spectral vegetation index (R² = 0.73), indicating that texture features had a certain role in the ultra-high-resolution grassland AGB estimation; (2) Compared with the model after feature selection, the combined spatial spectrum multi-type feature construction model was superior to any single-type feature model (R² = 0.83, RMSE = 127.57 g·m ^-2, MAE= 81.25 g·m ^-2), indicating that multi-type feature construction model could improve the accuracy of grassland AGB estimation to a certain extent. (3) Comparing the models building before and after feature selection, the model after feature selection by estimating the AGB effect was significantly better than the model without feature selection, and there was a high correlation between the selected features and grassland biomass, indicating that XGB-SFS could reduce the data dimension and improve the accuracy of grassland AGB estimation.【Conclusion】The ultra-high-resolution true color images of mobile phones could accurately estimate the grassland biomass. The XGB-SFS algorithm proposed in this paper could also select the features with high correlation with the grassland biomass from many features and improve the model estimation accuracy. Compared with the previous professional remote sensing quantitative inversion of grassland biomass, this method had the advantages of facing the public, low cost, and easy to use. The study combined the data collected on the phone with remote sensing and machine learning methods, which could open up new perspectives and support the development of agricultural informatization.

Key words: biomass, smart-phone, texture features, XGBoost, grassland

陶海玉,张爱武,庞海洋,康孝岩. 智能手机原位牧草生物量估算[J]. 中国农业科学, 2021, 54(5): 933-944.

HaiYu TAO,AiWu ZHANG,HaiYang PANG,XiaoYan KANG. Smart-Phone Application in Situ Grassland Biomass Estimation[J]. Scientia Agricultura Sinica, 2021, 54(5): 933-944.

0
/ / 推荐

导出引用管理器 EndNote|Reference Manager|ProCite|BibTeX|RefWorks

链接本文: https://www.chinaagrisci.com/CN/10.3864/j.issn.0578-1752.2021.05.006

https://www.chinaagrisci.com/CN/Y2021/V54/I5/933

图/表 10

图1

表1

表2

图2

图3

表3

图4

图5

图6

图7

参考文献 43

[1]	O'MARA F P. The role of grasslands in food security and climate change. Annals of Botany, 2012,110(6):1263-1270. pmid: 23002270
[2]	安海波, 李斐, 赵萌莉, 刘亚俊. 基于优化光谱指数的牧草生物量估算. 光谱学与光谱分析, 2015,35(11):3155-3160.
	AN H B, LI F, ZHAO M L, LIU Y J. Optimized spectral indices based estimation of forage grass biomass. Spectroscopy and Spectral Analysis, 2015,35(11):3155-3160. (in Chinese)
[3]	姚兴成, 曲恬甜, 常文静, 尹俊, 李永进, 孙振中, 曾辉. 基于MODIS数据和植被特征估算草地生物量. 中国生态农业学报, 2017,25(4):530-541.
	YAO X C, QU T T, CHANG W J, YIN J, LI Y J, SUN Z Z, ZENG H. Estimation of grassland biomass using MODIS data and plant community characteristics. Chinese Journal of Eco-Agriculture, 2017,25(4):530-541. (in Chinese)
[4]	LI H, LI A N, YIN G F, NAN X, BIAN J H. Retrieval of grassland aboveground biomass through inversion of the prosail model with modis imagery. Remote Sensing, 2019,11(13):1597.
[5]	孙世泽, 汪传建, 尹小君, 王伟强, 刘伟, 张雅, 赵庆展. 无人机多光谱影像的天然草地生物量估算. 遥感学报, 2017,22(5):848-856.
	SUN S Z, WANG C J, YIN X J, WANG W Q, LIU W, ZHANG Y, ZHAO Q Z. Estimating aboveground biomass of natural grassland based on multispectral images of unmanned aerial vehicles. Journal of Remote Sensing, 2017,22(5):848-856. (in Chinese)
[6]	LUSSEM U, BOLTEN A, GNYP M L, JASPER J, BARETH G. Evaluation of RGB-based vegetation indices from UAV imagery to estimate forage yield in grassland//The ISPRS Technical Commission III Midterm Symposium on “Developments, Technologies and Applications in Remote Sensing”. Beijing, China, 2018: 5.
[7]	YUE J, YANG G, TIAN Q J, FENG H K, XU K J, ZHOU C Q. Estimate of winter-wheat above-ground biomass based on UAV ultrahigh-ground-resolution image textures and vegetation indices. ISPRS Journal of Photogrammetry and Remote Sensing, 2019,150:226-244.
[8]	吴炳方, 张鑫, 曾红伟, 张淼, 田富有. 资源环境数据生成的大数据方法. 中国科学院院刊, 2018,33(8):804-811.
	WU B F, ZHANG X, ZENG H W, ZHANG M, TIAN F Y. Big data methods for environmental data. Bulletin of the Chinese Academy of Sciences, 2018,33(8):804-811. (in Chinese)
[9]	单杰. 从专业遥感到大众遥感. 测绘学报, 2017,46(10):1434-1446.
	SHAN J. Remote sensing: From trained professionals to general public. Acta Geodaetica et Cartographica Sinica, 2017,46(10):1434-1446. (in Chinese)
[10]	宋德娟, 张承明, 杨晓霞, 李峰, 韩颖娟, 高帅, 董海燕. 高分二号遥感影像提取冬小麦空间分布. 遥感学报, 2018,24(5):596-608.
	SONG D J, ZHANG C M, YANG X X, LI F, HAN Y J, GAO S, DONG H Y. Extracting winter wheat spatial distribution information from gf-2 image. Journal of Remote Sensing, 2018,24(5):596-608. (in Chinese)
[11]	张正健, 李爱农, 边金虎, 赵伟, 南希, 靳华安, 谭剑波, 雷光斌, 夏浩铭, 杨勇帅, 孙明江. 基于无人机影像可见光植被指数的若尔盖草地地上生物量估算研究. 遥感技术与应用, 2016,31(1):51-62.
	ZHANG Z J, LI A N, BIAN J H, ZHAO W, NAN X, JIN H A, TAN J B, LEI G B, XIA H M, YANG Y S, SUN M J. Estimating aboveground biomass of grassland in zoige by visible vegetation index derived from unmanned aerial vehicle image. Remote Sensing Technology and Application, 2016,31(1):51-62. (in Chinese)
[12]	张领先, 陈运强, 李云霞, 马浚诚, 杜克明, 郑飞翔, 孙忠富. 可见光光谱的冬小麦苗期地上生物量估算. 光谱学与光谱分析, 2019,39(8):2501-2506.
	ZHANG L X, CHEN Y Q, LI Y X, MA J C, DU K M, ZHENG F X, SUN Z F. Estimating above ground biomass of winter wheat at early growth stages based on visual spectral. Spectroscopy and Spectral Analysis, 2019,39(8):2501-2506. (in Chinese)
[13]	张爱武, 张帅, 郭超凡, 刘路路, 胡少兴, 柴沙驼. Landsat8光谱衍生数据分类体系下的牧草生物量反演. 光谱学与光谱分析, 2020,40(1):239-246.
	ZHANG A W, ZHANG S, GUO C F, LIU L L, HU S X, CHAI S T. Grass biomass inversion based on Landsat 8 spectral derived data classification system. Spectroscopy and Spectral Analysis, 2020,40(1):239-246. (in Chinese)
[14]	BASTIN J, POLOSAN M, PIALLAT B, KRACK P, BOUGEROL T, CHABARDES S, DAVID O. Changes of oscillatory activity in the subthalamic nucleus during obsessive-compulsive disorder symptoms: Two case reports. Cortex, 2014,60:145-150.
[15]	KELSEY K, NEFF J. Estimates of aboveground biomass from texture analysis of Landsat imagery. Remote Sensing, 2014,6(7):6407-6422.
[16]	LI J, VEERANAMPALAYAMSIVAKUMAR A, BHATTA M, GARST N D, STOLL H, BAENZIGER P S, BELAMKAR V, HOWARD R, GE Y F, SHI Y Y. Principal variable selection to explain grain yield variation in winter wheat from features extracted from UAV imagery. Plant Methods, 2019,15(1):1-13.
[17]	刘畅, 杨贵军, 李振海, 汤伏全, 王建雯, 张春兰, 张丽妍. 融合无人机光谱信息与纹理信息的冬小麦生物量估测. 中国农业科学, 2018,51(16):3060-3073.
	LIU C, YANG G J, LI Z H, TANG F Q, WANG J W, ZHANG C L, ZHANG L Y. Biomass estimation in winter wheat by UAV spectral information and texture information fusion. Scientia Agricultura Sinica, 2018,51(16):3060-3073. (in Chinese)
[18]	王娜, 李强子, 杜鑫, 张源, 赵龙才, 王红岩. 单变量特征选择的苏北地区主要农作物遥感识别. 遥感学报, 2017,21(4):519-530.
	WANG N, LI Q Z, DU X, ZHANG Y, ZHAO L C, WANG H Y. Identification of main crops based on the univariate feature selection in Subei. Journal of Remote Sensing, 2017,21(4):519-530. (in Chinese)
[19]	中华人民共和国农业部畜牧兽医司, 全国畜牧兽医总站. 中国草地资源. 北京: 中国科学技术出版社, 1996.
	Department of Animal Husbandry and Veterinary Medicine, Ministry of Agriculture of the People's Republic of China and National Animal Husbandry and Veterinary Station. China Grassland Resources. Beijing: China Science and Technology Press, 1996. (in Chinese)
[20]	WOEBBECKE D M, MEYER G E, VON BARGEN K, MORTENSEN D A. Color indices for weed identification under various soil, residue, and lighting conditions. Transactions of the ASABE, 1995,38(1):259-269.
[21]	SELLARO R, CREPY M, TRUPKIN S A, KARAYEKOV E, BUCHOVSKY A S, ROSSI C, CASAL J J. Cryptochrome as a sensor of the Blue/Green ratio of natural radiation in Arabidopsis. Plant Physiology, 2010,154(1):401-409. pmid: 20668058
[22]	PENUELAS J, GAMON J A, GRIFFIN K L, FIELD C B. Assessing community type, plant biomass, pigment composition, and photosynthetic efficiency of aquatic vegetation from spectral reflectance. Remote Sensing of Environment, 1993,46(2):110-118.
[23]	HUNT E R, CAVIGELLI M A, DAUGHTRY C S T, MCMURTREY J E, WALTHALL C L. Evaluation of digital photography from model aircraft for remote sensing of crop biomass and nitrogen status. Precision Agriculture, 2005,6(4):359-378.
[24]	HAGUE T, TILLETT N D, WHEELER H. Automated crop and weed monitoring in widely spaced cereals. Precision Agriculture, 2006,7(1):21-32.
[25]	BENDIG J, YU K, AASEN H, BOLTEN A, BENNERTZ S, BROSCHEIT J, GNYP M L, BARETH G. Combining UAV-based plant height from crop surface models, visible, and near infrared vegetation indices for biomass monitoring in barley. International Journal of Applied Earth Observation and Geoinformation, 2015,39:79-87.
[26]	VERRELST J, SCHAEPMAN M E, KOETZ B, KNEUBUHLER M. Angular sensitivity analysis of vegetation indices derived from CHRIS/PROBA data. Remote Sensing of Environment, 2008,112(5):2341-2353.
[27]	MEYER G E, NETO J C. Verification of color vegetation indices for automated crop imaging applications. Computers and Electronics in Agriculture, 2008,63(2):282-293.
[28]	汪小钦, 王苗苗, 王绍强, 吴云东. 基于可见光波段无人机遥感的植被信息提取. 农业工程学报, 2015,31(5):152-159.
	WANG X Q, WANG M M, WANG S Q, WU Y D. Extraction of vegetation information from visible unmanned aerial vehicle images. Transactions of the Chinese Society of Agricultural Engineering, 2015,31(5):152-159. (in Chinese)
[29]	GITELSON A A, KAUFMAN Y J, STARK R, RUNDQUIST D. Novel algorithms for remote estimation of vegetation fraction. Remote Sensing of Environment, 2002,80(1):76-87.
[30]	黄昕. 高分辨率遥感影像多尺度纹理、形状特征提取与面向对象分类研究[D]. 武汉: 武汉大学, 2009.
	HUANG X. Multiscale texture and shape feature extraction and object-oriented classification for very high resolution remotely sensed imagery[D]. Wuhan: Wuhan University, 2009. (in Chinese)
[31]	HARALICK R M, SHANMUGAM K, DINSTEIN I H. Textural features for image classification. IEEE Transactions on Systems, Man, and Cybernetics, 1973 (6):610-621.
[32]	SARKER L R, NICHOL J E. Improved forest biomass estimates using ALOS AVNIR-2 texture indices. Remote Sensing of Environment, 2011,115(4):968-977.
[33]	张爱武, 董喆, 康孝岩. 基于XGBoost的机载激光雷达与高光谱影像结合的特征选择算法. 中国激光, 2019,46(4):0404003.
	ZHANG A W, DONG Z, KANG X Y. Feature selection algorithms of airborne LiDAR combined with hyperspectral images based on XGBoost. Chinese Journal of Lasers, 2019,46(4):0404003. (in Chinese)
[34]	王丽爱, 马昌, 周旭东, 訾妍, 朱新开, 郭文善. 基于随机森林回归算法的小麦叶片SPAD值遥感估算. 农业机械学报, 2015,46(1):259-265.
	WANG L A, MA C, ZHOU X D, ZI Y, ZHU X K, GUO W S. Estimation of wheat leaf SPAD value using RF algorithmic model and remote sensing data. Transactions of the Chinese Society for Agricultural Machinery, 2015,46(1):259-265. (in Chinese)
[35]	赵静, 李志铭, 鲁力群, 贾鹏, 杨焕波, 兰玉彬. 基于无人机多光谱遥感图像的玉米田间杂草识别. 中国农业科学, 2020,53(8):1545-1555.
	ZHAO J, LI Z M, LU L Q, JIA P, YANG H B, LAN Y B. Weed identification in maize field based on multi-spectral remote sensing of unmanned aerial vehicle. Scientia Agricultura Sinica, 2020,53(8):1545-1555. (in Chinese)
[36]	ZHENG H, CHENG T, ZHOU M, LI D, YAO X, TIAN Y C, ZHU Y. Improved estimation of rice aboveground biomass combining textural and spectral analysis of UAV imagery. Precision Agriculture, 2019,20(3):611-629.
[37]	LIU Y, LIU S, LI J, GUO X Y, WANG S Q, LU J W. Estimating biomass of winter oilseed rape using vegetation indices and texture metrics derived from UAV multispectral images. Computers and Electronics in Agriculture, 2019,166:105026.
[38]	SIBANDA M, MUTANGA O, ROUGET M, KUMAR L. Estimating biomass of native grass grown under complex management treatments using worldview-3 spectral derivatives. Remote Sensing, 2017,9(1):55.
[39]	American Society for Photogrammetry and Remote Sensing. Manual of Remote Sensing. New Jersey: John Wiley & Sons, Inc., 1999.
[40]	CHO M A, SKIDMORE A, CORSI F, WIEREN S E V, SOBHAN I. Estimation of green grass/herb biomass from airborne hyperspectral imagery using spectral indices and partial least squares regression. International Journal of Applied Earth Observation and Geoinformation, 2007,9(4):414-424.
[41]	李明诗, 谭莹, 潘洁, 彭世揆. 结合光谱、纹理及地形特征的森林生物量建模研究. 遥感信息, 2006(6):6-9.
	LI M S, TAN Y, PAN J, PENG S K. Modeling forest aboveground biomass by combining the spectrum, textures with topographic features. Remote Sensing Information, 2006,2006(6):6-9. (in Chinese)
[42]	ECKERT S. Improved forest biomass and carbon estimations using texture measures from WorldView-2 satellite data. Remote Sensing, 2012,4(4):810-829.
[43]	陈鹏, 冯海宽, 李长春, 杨贵军, 杨钧森, 杨文攀, 刘帅兵. 无人机影像光谱和纹理融合信息估算马铃薯叶片叶绿素含量. 农业工程学报, 2019,35(11):63-74.
	CHEN P, FENG H K, LI C C, YANG G J, YANG J S, YANG W P, LIU S B. Estimation of chlorophyll content in potato using fusion of texture and spectral features derived from UAV multispectral image. Transactions of the Chinese Society of Agricultural Engineering, 2019,35(11):63-74. (in Chinese)

植被指数 Vegetation index	计算公式 Formula	参考文献 Reference
R	r波段的DN值	—
G	g波段的DN值	—
B	b波段的DN值	—
r	R/(R+G+B)	[20]
g	G/(R+G+B)	[20]
b	B/(R+G+B)	[20]
ExG	2×g - b - r	[20]
GBRI	g/b	[21]
NPCI	(r - b) / (r + b)	[22]
NGBVI	(g - b) / (g +b)	[23]
VEGI	g/ ((r^0.67) ×b^0.33)	[24]
RGBVI	(g² - (b× r)) / (g² + (b ×r))	[25]
RGRI	r / g	[26]
RGMPI	r × g	[26]
RBRI	r / b	—
RBMPI	r × b	—
RBMI	r - b	—
RGMI	r - g	[20]
RGPI	r + g	[27]
RBPI	r + b	—
GBPI	g + b	[27]
GBMI	g - b	[20]
VDVI	(2×g - r - b) / (2×g+r+b)	[28]
VARI	(g - r / (g+r - b)	[29]

波段 Band	纹理特征 Texture features
R	MEA_R,VAR_R,HOM_R,CON_R,DIS_ R,ENT_R,SEC_R,COR_R
G	MEA_G,VAR_G,HOM_G,CON_G,DIS_ G,ENT_G,SEC_G,COR_G
B	MEA_B,VAR_B,HOM_B,CON_B,DIS_ B,ENT_B,SEC_B,COR_B

特征类型 Feature type	数量 Number	入选特征 Selected features
RGBVIs	14	RBMI, R, B, RGMI, NGBVI, RGPI, VDVI, RBPI, G, NPCI, b, RGMPI, GBRI, RBRI
Textures	12	VAR_B, VAR_R, CON_B, MEA_B, HOM_B, VAR_G, MEA_R, DIS_B, SEC_R, MEA_G, SEC_B, ENT_B
VI-Textures	11	R, VAR_G, VAR_B, G, VAR_R, RBRI, RBMI, RBMPI, MEA_G, SEC_B, MEA_B