摘要
[目的]针对黄瓜植株极易染病且部分病害症状相似的问题,利用叶绿素荧光成像系统研究黄瓜植株不同病害区分及早期病害监测的可行性。[方法]采用叶绿素荧光成像系统采集全植株冠层图像,以褐斑病和炭疽病胁迫下的黄瓜植株为试验材料,分析植株生理状态,建立基于叶绿素荧光参数的病害分类和病情诊断模型。首先通过图像的分割得到病斑区域;然后采集植株氮含量、叶绿素含量与荧光参数,并分析其变化趋势;最后基于叶绿素荧光强度和动力学参数对黄瓜褐斑病和炭疽病进行分类和早期监测,分别采用支持向量机(SVM)算法和极端梯度提升(XGBoost)算法对不同程度病害植株进行分类。[结果]与对照植株相比,染病植株叶绿素含量及氮含量呈逐渐下降趋势,最大光化学量子产量(Fv/Fm)、实际光化学效率(ΦPSⅡ)降低,非光化学淬灭(NPQ)、非光化学淬灭系数(qN)和光化学淬灭系数(qP)上升。对于植株的病害与病情分类,利用XGBoost算法进行分类的结果整体较好。对2种病害单独分类的准确率达到90%以上,对2种病害同时分类准确率达到85%以上,对病情和病害种类同时监测的准确率接近80%。[结论]基于叶绿素荧光成像系统监测黄瓜病情和区分其病害种类是可行的,具有良好的应用前景。
[Objectives]Cucumber plants are highly susceptible to disease and some of the disease symptoms are similar,the feasibility of using chlorophyll fluorescence imaging system to distinguish and detect different diseases of cucumber plants was studied in this experiment.[Methods]Cucumber plants under the stress of brown patch and anthracnose were used as experimental materials to design a chlorophyll fluorescence imaging system that could collect canopy images of the whole plant,and the physiological status of the plants were analyzed to establish a diagnostic model based on chlorophyll fluorescence parameters.Firstly,the disease spots were obtained by image segmentation.Secondly,nitrogen content,chlorophyll content and fluorescence parameters of plants were collected,and the variation trend was analyzed in this experiment.Finally,based on chlorophyll fluorescence intensity and kinetic parameters,early detection of cucumber brown spot was conducted,and support vector machine(SVM)algorithm and extreme gradient boosting(XGBoost)algorithm were adopted respectively to classify disease plants of different degrees and detect early diseases.[Results]Compared with the plants in control group,the chlorophyll content and nitrogen content of infected plants decreased gradually,maximum photochemical quantum yield(Fv/Fm)and actual photochemical efficiency(ΦPSⅡ)of infected plants reduced,while the non-photochemical quenching(NPQ),non-photochemical quenching coefficient(qN)and photochemical quenching coefficient(qP)increased.And the results showed that the XGBoost algorithm for the classification of plant diseases and the study of disease severity prediction were better overall.The accuracy rate for the classification between control group and a single disease plants reached 90%,and the accuracy rate for the classification among control group and two kinds of disease plant reached 85%.The accuracy rate for the classification among disease severity prediction,two kinds of disease and control group reached nearly 80%.[Conclusions]
作者
王迎旭
孙晔
李玉花
孙国祥
汪小旵
WANG Yingxu;SUN Ye;LI Yuhua;SUN Guoxiang;WANG Xiaochan(College of Engineering/Engineering Laboratory of Modern Facility Agriculture Technology and Equipmentin Jiangsu Province,Nanjing Agricultural University,Nanjing 210031,China)
出处
《南京农业大学学报》
CAS
CSCD
北大核心
2020年第4期770-780,共11页
Journal of Nanjing Agricultural University
基金
中国博士后科学基金项目(2018M642263)
国家自然科学基金青年基金项目(61701242)
江苏省自然科学基金项目(BK20150686,BK20170727)。
关键词
叶绿素荧光
黄瓜
病害分类
病情监测
支持向量机(SVM)
极端梯度提升(XGBoost)
chlorophyll fluorescence
cucumber
classification of diseases
disease severity prediction
support vector machine(SVM)
extreme gradient boosting(XGBoost)
