基于无人机多光谱遥感图像的玉米田间杂草识别

doi:10.3864/j.issn.0578-1752.2020.08.005

中国农业科学 ›› 2020, Vol. 53 ›› Issue (8): 1545-1555.doi: 10.3864/j.issn.0578-1752.2020.08.005

• 耕作栽培·生理生化·农业信息技术 • 上一篇下一篇

基于无人机多光谱遥感图像的玉米田间杂草识别

赵静^1,²,李志铭^1,²,鲁力群^2,³,贾鹏^1,²,杨焕波^1,²,兰玉彬^1,²()

¹ 山东理工大学农业工程与食品科学学院,山东淄博255000
² 山东理工大学国际精准农业航空应用技术研究中心,山东淄博255000
³ 山东理工大学交通与车辆工程学院,山东淄博255000

收稿日期:2019-08-29 接受日期:2019-11-04 出版日期:2020-04-16 发布日期:2020-04-29
通讯作者: 兰玉彬
作者简介:赵静,E-mail: zbceozj@163.com。
基金资助:
山东省引进顶尖人才“一事一议”专项经费资助项目;中央引导地方科技发展专项资金“精准农业航空技术与装备研发”资助项目

Weed Identification in Maize Field Based on Multi-Spectral Remote Sensing of Unmanned Aerial Vehicle

ZHAO Jing^1,²,LI ZhiMing^1,²,LU LiQun^2,³,JIA Peng^1,²,YANG HuanBo^1,²,LAN YuBin^1,²()

¹ School of Agricultural Engineering and Food Science, Shandong University of Technology, Zibo 255000, Shandong
² International Research Center of Precision Agriculture Aviation Application Technology, Shandong University of Technology, Zibo 255000, Shandong
³ School of Transportation and Vehicle Engineering, Shandong University of Technology, Zibo 255000, Shandong

Received:2019-08-29 Accepted:2019-11-04 Online:2020-04-16 Published:2020-04-29
Contact: YuBin LAN

摘要/Abstract

摘要：

【目的】为了精确高效识别玉米田间杂草,减少除草剂施用,提高玉米种植管理精准性。【方法】通过六旋翼无人机搭载多光谱相机获取玉米田块多光谱图像。为分离图像中植被与非植被像元,计算了7种植被指数,采用最大类间方差法提取植被指数图像中非植被区域,制作掩膜文件并对多光谱图像掩膜。通过主成分分析对多光谱图像进行变换,保留信息量最多的前3个主成分波段。将试验区域分为训练区域和验证区域,在训练区域中分别选取了675处玉米和525处杂草样本对监督分类模型进行训练,在验证区域选取了240处玉米样本及160处杂草样本评价模型分类精度。将7种植被指数、3个主成分波段的24个纹理特征及经过滤波的10个反射率,共计41项特征作为样本特征参数。利用支持向量机-特征递归消除算法（support vector machines-feature recursive elimination,SVM-RFE）和Relief算法从41项特征中各筛选14项特征构成特征子集,采用支持向量机、K-最近邻、Cart决策树、随机森林和人工神经网络对特征子集进行监督分类。【结果】支持向量机与随机森林对全部特征及2个特征子集分类效果较好,支持向量机总体精度为89.13%—91.94%,Kappa>0.79,随机森林总体精度为89.27%—90.95%,Kappa>0.79。【结论】SVM-RFE算法对数据降维效果优于Relief算法,支持向量机（SVM）模型对区域冠层尺度下玉米与杂草的分类效果最好。

关键词: 杂草识别, 无人机遥感, 多光谱图像, 特征选择, 监督分类

Abstract:

【Objective】In order to reduce the application rate of herbicides and to make the maize planting management more effective, the accurate identification of weeds in maize fields was investigated based on multi-spectral remote sensing of unmanned aerial vehicles (UAV). 【Method】In this paper, a Red Edge-M multi-spectral camera was mounted in a six-rotor UAV to acquire five single-band images of blue, green, red, red edge, and near-infrared, and the application was taken in Zibo, Shandong province, China to acquire multi-spectral images of a maize field in July 14, 2018. In order to separate the vegetation and non-vegetation pixels in the image, 7 vegetation indices were calculated, the OTSU method was used to obtain the non-vegetation area, and the multi-spectral image was masked. Then multi-spectral image was transformed by principal component analysis, retaining the first three principal component bands with the most information. The experimental region was divided into 3 training areas and 1 verification area. 675 maize and 525 weed samples were selected in the training areas to train the supervised classification model, and 240 maize and 160 weed samples were selected in the verification area to evaluate model classification accuracy. The 7 vegetation indices, 24 texture features of the 3 principal component bands and 10 reflectivity of multi-spectral image bands which were filtered, and a total of 41 features were taken as features of maize and weed. Support vector machines-feature recursive elimination (SVM-RFE) algorithm and Relief algorithm were applied to selecting 14 features from 41 features to constitutes a feature subset separately, and supervised classification for weed detection was performed using support vector machine (SVM), K-nearest neighbor (KNN), Cart decision tree (Cart), random forest (RF) and artificial neural network (ANN) .【Result】SVM and RF performed a better classification with all features and SVM-RFE & Relief feature subsets. The overall accuracy of SVM was 89.13%-91.94%, Kappa>0.79, and overall accuracy of random forest was 89.27%-90.95%, Kappa>0.79.【Conclusion】 SVM-RFE feature selection algorithm was better than the Relief feature algorithm for reducing the original features. SVM model had the highest classification accuracy for identification of weed and maize at regional canopy scales.

Key words: weed identification, UAV remote sensing, multi-spectral image, feature selection, supervised classification

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[J]. 中国农业科学, 2020, 53(8): 1545-1555.

ZHAO Jing,LI ZhiMing,LU LiQun,JIA Peng,YANG HuanBo,LAN YuBin. Weed Identification in Maize Field Based on Multi-Spectral Remote Sensing of Unmanned Aerial Vehicle[J]. Scientia Agricultura Sinica, 2020, 53(8): 1545-1555.

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.0578-1752.2020.08.005

https://www.chinaagrisci.com/CN/Y2020/V53/I8/1545

图/表 10

图1

图2

图3

表2

表3

表4

图4

表5

表6

表7

参考文献 38

[1]	翁凌云 . 我国玉米生产现状及发展对策分析. 中国食物与营养, 2010(1):22-25.
	WENG L Y . Status of corn production in China and its countermeasures. Food and Nutrition in China, 2010(1):22-25. (in Chinese)
[2]	LOUARGANT M, VILLETTE S, JONES G, VIGNEAU N, PAOLI J, GÉE C . Weed detection by UAV: Simulation of the impact of spectral mixing in multispectral images. Precision Agriculture, 2017,18:932-951.
[3]	山东省农业农村厅. 山东省到2020年农药使用量零增长行动方案. 山东农药信息, 2015(4):13-14.
	Shandong Provincial Department of Agricultural and Rural Affairs. Shandong province issued the action plan for zero increase of pesticide use in Shandong province by 2020. Shandong Pesticide News, 2015(4):13-14. (in Chinese)
[4]	周志艳, 明锐, 臧禹, 何新刚, 罗锡文, 兰玉彬 . 中国农业航空发展现状及对策建议. 农业工程学报, 2017,33(20):1-13.
	ZHOU Z Y, MING R, ZANG Y, HE X G, LUO X W, LAN Y B . Development status and countermeasures of agricultural aviation in China. Transactions of the Chinese Society of Agricultural Engineering, 2017,33(20):1-13. (in Chinese)
[5]	兰玉彬 . 精准农业航空技术现状及未来展望. 农业工程技术, 2017,37(30):27-30.
	LAN Y B . Current status and future prospects of precision agricultural aviation technology. Agricultural Engineering Technology, 2017,37(30):27-30. (in Chinese)
[6]	何东健, 何勇, 李明赞, 洪添胜, 王成红, 宋苏, 刘允刚 . 精准农业中信息相关科学问题研究进展. 中国科学基金, 2011,25(1):10-16.
	HE D J, HE Y, LI M Z, HONG T S, WANG C H, SONG S, LIU Y G . Research progress of information science-related problems in precision agriculture. Bulletin of National Natural Science Foundation of China, 2011,25(1):10-16. (in Chinese)
[7]	兰玉彬, 王国宾 . 中国植保无人机的行业发展概况和发展前景. 农业工程技术, 2018,38(9):17-27.
	LAN Y B, WANG G B . China's plant protection drone industry development overview and development prospects. Agricultural Engineering Technology, 2018,38(9):17-27. (in Chinese)
[8]	史舟, 梁宗正, 杨媛媛, 郭燕 . 农业遥感研究现状与展望. 农业机械学报, 2015,46(2):247-260.
	SHI Z, LIANG Z Z, YANG Y Y, GUO Y . Status and prospect of agricultural remote sensing. Transactions of the Chinese Society for Agricultural Machinery, 2015,46(2):247-260. (in Chinese)
[9]	金小俊, 陈勇, 孙艳霞 . 农田杂草识别方法研究进展. 农机化研究, 2011,33(7):23-27.
	JIN X J, CHEN Y, SUN Y X . Research advances of weed identification in agricultural fields. Journal of Agricultural Mechanization Research, 2011,33(7):23-27. (in Chinese)
[10]	毛文华, 张银桥, 王辉, 赵博, 张小超 . 杂草信息实时获取技术与设备研究进展. 农业机械学报, 2013,44(1):190-195.
	MAO W H, ZHANG Y Q, WANG H, ZHAO B, ZHANG X C . Advance techniques and equipments for real-time weed detection. Transactions of the Chinese Society for Agricultural Machinery, 2013,44(1):190-195. (in Chinese)
[11]	邓向武, 齐龙, 马旭, 蒋郁, 陈学深, 刘云海, 陈伟烽 . 基于多特征融合和深度置信网络的稻田苗期杂草识别. 农业工程学报, 2018,34(14):165-172.
	DENG X W, QI L, MA X, JIANG Y, CHEN X S, LIU Y H, CHEN W F . Recognition of weeds at seedling stage in paddy fields using multi-feature fusion and deep belief networks. Transactions of the Chinese Society of Agricultural Engineering, 2018,34(14):165-172. (in Chinese)
[12]	孙俊, 何小飞, 谭文军, 武小红, 沈继峰, 路虎 . 空洞卷积结合全局池化的卷积神经网络识别作物幼苗与杂草. 农业工程学报, 2018,34(11):159-165.
	SUN J, HE X F, TAN W J, WU X H, SHEN J F, LU H . Recognition of crop seedling and weed recognition based on dilated convolution and global pooling in CNN. Transactions of the Chinese Society of Agricultural Engineering, 2018,34(11):159-165. (in Chinese)
[13]	何东健, 乔永亮, 李攀, 高瞻, 李海洋, 唐晶磊 . 基于SVM-DS多特征融合的杂草识别. 农业机械学报, 2013,44(2):182-187.
	HE D J, QIAO Y L, LI P, GAO Z, LI H Y, TANG J L . Weed recognition based on SVM-DS multi-feature fusion. Transactions of the Chinese Society for Agricultural Machinery, 2013,44(2):182-187. (in Chinese)
[14]	潘冉冉, 骆一凡, 王昌, 张初, 何勇, 冯雷 . 高光谱成像的油菜和杂草分类方法. 光谱学与光谱分析, 2017,37(11):3567-3572.
	PAN R R, LUO Y F, WANG C, ZHANG C, HE Y, FENG L . Classifications of oilseed rape and weeds based on hyperspectral imaging. Spectroscopy and Spectral Analysis, 2017,37(11):3567-3572. (in Chinese)
[15]	王海华, 朱梦婷, 李莉, 王丽燕, 赵海英, 梅树立 . 基于剪切波变换和无人机麦田图像的区域杂草识别方法. 农业工程学报, 2017,33(S1):99-106.
	WANG H H, ZHU M T, LI L, WANG L Y, ZHAO H Y, MEI S L . Regional weed identification method from wheat field based on unmanned aerial vehicle image and shearlets. Transactions of the Chinese Society of Agricultural Engineering, 2017,33(S1):99-106. (in Chinese)
[16]	肖武, 任河, 吕雪娇, 闫皓月, 孙诗睿 . 基于无人机遥感的高潜水位采煤沉陷湿地植被分类. 农业机械学报, 2019,50(2):177-186.
	XIAO W, REN H, LÜ X J, YAN H Y, SUN S R . Vegetation classification by using UAV remote sensing in coal mining subsidence wetland with high ground-water level. Transactions of the Chinese Society for Agricultural Machinery, 2019,50(2):177-186. (in Chinese)
[17]	NOBUYUKI O . A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man and Cybernetics, 1979,1(1):62-66.
[18]	孙滨峰, 赵红, 陈立才, 舒时富, 叶春, 李艳大 . 基于植被指数选择算法和决策树的生态系统识别. 农业机械学报, 2019,50(6):194-200.
	SUN B F, ZHAO H, CHEN L C, SHU S F, YE C, LI Y D . Identification of ecosystems based on vegetation indices selection algorithm and decision tree. Transactions of the Chinese Society for Agricultural Machinery, 2019,50(6):194-200. (in Chinese)
[19]	刘畅, 杨贵军, 李振海, 汤伏全, 王健雯, 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)
[20]	ROUSE J W, HAAS R W, SCHELL J A, DEERING D W, HARLAN J C . Monitoring the vernal advancement and retrogradation (Greenwave effect) of natural vegetation Final Rep. RSC 1978-4. Remote Sensing Center, Texas A&M University, College Station, 1974.
[21]	JORDAN C F . Derivation of leaf-area index from quality of light on the forest floor. Ecology, 1969,50:663-666.
[22]	PEARSON R L, MILLER L D . Remote mapping of standing crop biomass for estimation of the productivity of the short-grass prairie. Proceedings of the 8th International Symposium on Remote Sensing of Environment. Environmental Research Institute of Michigan. Ann Arbor, MI, USA, 1972: 1357-1381.
[23]	HUETE A R . A soil-adjusted vegetation index (SAVI). Remote Sensing of Environment, 1988,25:295-309.
[24]	RONDEAUX G, STEVEN M, BARET F . Optimization of soil- adjusted vegetation indices. Remote Sensing of Environment, 1996,55:95-107.
[25]	GITELSON A A, KAUFMAN Y J, MERZLYAK M N . Use of a green channel in remote sensing of global vegetation from EOS- MODIS. Remote Sensing of Environment, 1996,58:289-298.
[26]	HUETE A, DIDAN K, MIURA T, RODRIGUEZ E P, GAO X, FERREIRA L G . Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sensing of Environment, 2002,83:195-213.
[27]	周志华 . 机器学习. 北京: 清华大学出版社, 2017.
	ZHOU Z H . Machine Learning. Beijing: Tsinghua University Press, 2017. (in Chinese)
[28]	李慧, 祁力钧, 张建华, 冀荣华 . 基于PCA-SVM的棉花出苗期杂草类型识别. 农业机械学报, 2012,43(9):184-189, 196.
	LI H, QI L J, ZHANG J H, JI R H . Recognition of weed during cotton emergence based on principal component analysis and support vector machine. Transactions of the Chinese Society for Agricultural Machinery, 2012,43(9):184-189, 196. (in Chinese)
[29]	李俊秀, 姜三平 . 基于主成分分析的图像自适应阈值去噪算法. 红外技术, 2014,36(4):311-314, 319.
	LI J X, JIANG S P . Adaptive threshold image denoising algorithm based on principal component analysis. Infrared Technology, 2014,36(4):311-314, 319. (in Chinese)
[30]	HARALICK R M, SHANMUGAM K . Textural features for image classification. IEEE Transactions on Systems, Man and Cybernetics, 1973,3(6):610-621.
[31]	吴文涵, 陶华敏, 肖山竹, 汤朋文 . 灰度共生矩阵纹理特征提取算法的优化与实现. 数字技术与应用, 2015(6):124-126.
	WU W H, TAO H M, XIAO S Z, TANG P W . Optimization and implementation of texture feature extraction algorithm for gray level co-occurrence matrix. Digital Technology and Application, 2015(6):124-126. (in Chinese)
[32]	韩文霆, 孙瑜, 徐腾飞, 陈香维, SU K O . 基于RGB图像处理的玉米叶片含水率监测方法研究. 农业工程技术, 2016,36(12):75.
	HAN W T, SUN Y, XU T F, CHEN X W, SU K O . Detecting maize leaf water status by using digital RGB images. Agricultural Engineering Technology, 2016,36(12):75. (in Chinese)
[33]	侯群群, 王飞, 严丽 . 基于灰度共生矩阵的彩色遥感图像纹理特征提取. 国土资源遥感, 2013,25(4):26-32.
	HOU Q Q, WANG F, YAN L . Extraction of color image texture feature based on gray-level co-occurrence matrix. Remote Sensing for Land & Resources, 2013,25(4):26-32. (in Chinese)
[34]	韩文霆, 李广, 苑梦婵, 张立元, 师志强 . 基于无人机遥感技术的玉米种植信息提取方法研究. 农业机械学报, 2017,48(1):139-147.
	HAN W T, LI G, YUAN M C, ZHANG L Y, SHI Z Q . Extraction method of maize planting information based on UAV remote sensing technology. Transactions of the Chinese Society for Agricultural Machinery, 2017,48(1):139-147. (in Chinese)
[35]	黄晓娟, 张莉 . 改进的多类支持向量机递归特征消除在癌症多分类中的应用. 计算机应用, 2015,35(10):2798-2802.
	HUANG X J, ZHANG L . Modified multi-class support vector machine recursive feature elimination for cancer multi-classification. Journal of Computer Applications, 2015,35(10):2798-2802. (in Chinese)
[36]	王君 . 基于SVM-RFE的特征选择方法研究[D]. 大连: 大连理工大学, 2015.
	WANG J . The research of feature selection algorithm based on SVM- RFE[D]. Dalian: Dalian University of Technology, 2015. (in Chinese)
[37]	蒋玉娇, 王晓丹, 王文军, 毕凯 . 一种基于PCA和ReliefF的特征选择方法. 计算机工程与应用, 2010,46(26):170-172.
	JIANG Y J, WANG X D, WANG W J, BI K . New feature selection approach by PCA and ReliefF. Computer Engineering and Applications, 2010,46(26):170-172. (in Chinese)
[38]	戴建国, 张国顺, 郭鹏, 曾窕俊, 崔美娜, 薛金利 . 基于无人机遥感可见光影像的北疆主要农作物分类方法. 农业工程学报, 2018,34(18):122-129.
	DAI J G, ZHANG G S, GUO P, ZENG T J, CUI M N, XUE J L . Classification method of main crops in northern Xinjiang based on UAV visible waveband images. Transactions of the Chinese Society of Agricultural Engineering, 2018,34(18):122-129. (in Chinese)

监督分类模型 Supervised classifiers	训练集识别率 Accuracy of training set (%)	测试集 Test set
监督分类模型 Supervised classifiers	训练集识别率 Accuracy of training set (%)	准确率 Accuracy(%)	精确率 Precision(%)	召回率 Recall (%)	F₁(%)
支持向量机 SVM	98.76	98.48	98.49	98.49	98.48
K-最近邻 KNN	94.78	93.94	94.14	93.94	93.92
Cart决策树 Cart	95.40	91.41	91.59	91.41	91.39
随机森林 RF	99.13	96.46	96.53	96.47	96.46
人工神经网络 ANN	97.67	99.00	99.02	99.00	99.00

Relief特征子集 Relief subset	训练集识别率 Accuracy of training set (%)	测试集 Test set
Relief特征子集 Relief subset	训练集识别率 Accuracy of training set (%)	准确率 Accuracy (%)	精确率 Precision (%)	召回率 Recall (%)	F₁(%)
支持向量机 SVM	96.77	97.22	97.23	97.22	97.22
K-最近邻 KNN	100	93.43	93.63	93.43	93.41
Cart决策树 CART	93.91	90.66	91.25	90.66	90.58
随机森林 RF	97.14	92.93	93.02	92.93	92.92
人工神经网络 ANN	95.44	96.67	96.69	96.67	96.67

SVM-RFE特征子集 SVM-RFE subset	训练集识别率 Accuracy of training set (%)	测试集 Test set
SVM-RFE特征子集 SVM-RFE subset	训练集识别率 Accuracy of training set (%)	准确率 Accuracy (%)	精确率 Precision (%)	召回率 Recall (%)	F₁ (%)
支持向量机 SVM	98.76	98.23	98.23	98.23	98.23
K-最近邻 KNN	100	95.96	96.07	95.96	95.95
Cart决策树 CART	97.64	94.44	94.65	94.44	94.43
随机森林 RF	99.38	97.22	97.25	97.22	97.22
人工神经网络 ANN	97.78	99.00	99.02	99.00	99.00

全部特征 All features	地物 Objects	玉米 Maize (%)	杂草 Weed (%)	Kappa系数 Kappa coefficient	总体精度 Overall accuracy (%)
随机森林 RF	玉米 Maize	88.47	4.92	0.82	90.95
随机森林 RF	杂草 Weed	11.53	95.08	0.82	90.95
支持向量机 SVM	玉米 Maize	89.69	4.29	0.84	91.94
支持向量机 SVM	杂草 Weed	10.31	95.71	0.84	91.94
人工神经网络 ANN	玉米 Maize	78.21	5.08	0.70	84.47
人工神经网络 ANN	杂草 Weed	21.79	94.92	0.70	84.47

Relief特征子集 Relief subset	地物 Objects	玉米 Maize (%)	杂草 Weed (%)	Kappa系数 Kappa coefficient	总体精度 Overall accuracy (%)
随机森林 RF	玉米 Maize	86.03	5.32	0.79	89.27
随机森林 RF	杂草 Weed	13.97	94.68	0.79	89.27
支持向量机 SVM	玉米 Maize	85.55	4.88	0.79	89.13
支持向量机 SVM	杂草 Weed	14.45	95.12	0.79	89.13
人工神经网络 ANN	玉米 Maize	77.06	7.38	0.67	82.89
人工神经网络 ANN	杂草 Weed	22.94	92.62	0.67	82.89

基于无人机多光谱遥感图像的玉米田间杂草识别

Weed Identification in Maize Field Based on Multi-Spectral Remote Sensing of Unmanned Aerial Vehicle

RichHTML

PDF

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 38

相关文章 3

Metrics

本文评价

推荐阅读 0

SVM-RFE特征子集 SVM-RFE subset	地物 Objects	玉米 Maize (%)	杂草 Weed (%)	Kappa系数 Kappa coefficient	总体精度 Overall accuracy (%)
随机森林 RF	玉米 Maize	87.99	5.13	0.81	90.56
随机森林 RF	杂草 Weed	12.01	94.87	0.81	90.56
支持向量机 SVM	玉米 Maize	87.85	4.36	0.82	90.76
支持向量机 SVM	杂草 Weed	12.15	95.64	0.82	90.76
人工神经网络 ANN	玉米 Maize	82.62	6.30	0.74	86.77
人工神经网络 ANN	杂草 Weed	17.38	93.70	0.74	86.77

[1]	刘庆飞，张宏立，王艳玲. 基于深度可分离卷积的实时农业图像逐像素分类研究[J]. 中国农业科学, 2018, 51(19): 3673-3682.
[2]	张晓东, 毛罕平, 左志宇, 孙俊, 张红涛. 油菜氮素的多光谱图像估算模型研究[J]. 中国农业科学, 2011, 44(16): 3323-3332.
[3]	范德耀,姚青,杨保军,周营烽,管泽鑫,唐健 . 田间杂草识别与除草技术智能化研究进展[J]. 中国农业科学, 2010, 43(9): 1823-1833 .