基于叶绿素荧光光谱指数的温室黄瓜病害预测

发布时间：2018-04-26 02:21

本文选题：光谱指数 + 病害预测　；参考：《光谱学与光谱分析》2016年06期

【摘要】：温室蔬菜病害的发生及大面积流行严重影响设施农业的生产管理,大大降低设施农业的经济效益。为了实现温室蔬菜病害的无损准确预测,以黄瓜霜霉病害为例,利用激光诱导叶绿素荧光构建光谱特征指数,建立了温室蔬菜病害的预测模型。在试验中采用对比分析的方法,通过对作物健康叶片接种病菌孢子,分别采集健康、接种2d、接种6d和出现明显病症共4组试验样本的光谱曲线,定性分析了荧光强度随叶片样本感染病菌孢子的变化规律;利用光谱曲线不同波段峰谷值创建了叶绿素荧光光谱指数k_1=F_(685)/F~(512)和k_2=F_(734)/F~(512),根据数值的变化范围,设定k_1和k_2分别为20和10时可以作为判断样本出现明显病症与未出现明显病症的特征值,其判断的准确率分别达到96%和94%;利用构建的光谱指数与样本健康状况的分类结果,选择光谱指数F_(685)/F~(512),F_(685)-F_(734),F_(715)/F_(612)可以定性判断样本健康状况,并选择光谱指数F_(685)/F~(512),F_(734)/F~(512),F_(685)-F_(734),F_(715)/F_(612)作为建立定量分析模型的输入量,以预测集分类准确率作为评价标准,对比判别分析、BP神经网络、支持向量机三种数据建模方法,结果表明支持向量机作为霜霉病害预测的建模方法,其预测能力达到91.38%。利用激光诱导叶绿素荧光构建光谱指数方法,研究植物病害的预测问题,具有很好的分类和鉴别效果。
[Abstract]:The occurrence and prevalence of greenhouse vegetable diseases seriously affect the production and management of facility agriculture and greatly reduce the economic benefits of facility agriculture. In order to predict greenhouse vegetable diseases without damage, a prediction model of greenhouse vegetable diseases was established by using laser induced chlorophyll fluorescence to construct spectral characteristic index of cucumber downy mildew (Cucumber downy mildew). In the experiment, the spectral curves of 4 groups of test samples were collected by inoculating spore on healthy leaves of crops, inoculating for 2 days, inoculating for 6 days and appearing obvious diseases. The variation of fluorescence intensity with spore infection in leaf samples was qualitatively analyzed, and the chlorophyll fluorescence spectral index (k ~ 1 / T / F ~ (512) and K _ 2F _ 4 / F ~ (1 / F) ~ (512) were established by using peak and valley values of different bands of spectral curve, according to the range of values, Setting K _ 1 and k _ 2 as 20 and 10:00 respectively can be used as the characteristic values to judge the obvious and non-obvious symptoms in the sample, and the accuracy of the judgment is 96% and 94% respectively, and the classification results of the constructed spectral index and the health condition of the sample are used to make use of the established spectral index and the classification result of the sample's health. Select the number of numbers that can be judged qualitatively in the health of the sample by selecting the index F685 / FJ / FJ 612) and choose the spectral index F685 / FJ / FCI / 612 as input for the establishment of a quantitative analysis model, using the accuracy rate of the prediction set as the evaluation criterion, comparing and discriminating the BP neural network, comparing the discriminant analysis of the BP neural network with the index of 734 / F512C / FSP / FSP as the input for the establishment of a quantitative analysis model, using the accuracy of the prediction set classification as the evaluation criterion, and comparing the discriminant analysis of the BP neural network with the classification accuracy of the prediction set. The results show that support vector machine (SVM) is the modeling method for forecasting downy mildew and its prediction ability is 91.38. The method of constructing spectral index by laser induced chlorophyll fluorescence was used to study the prediction of plant diseases. It has a good classification and identification effect.
【作者单位】：吉林大学生物与农业工程学院;吉林大学植物科学学院;
【基金】：国家高技术研究发展计划(863)项目(2012AA10A506-4,2013AA103005-04) 中国博士后基金面上项目(2013M541308) 吉林大学基本科研业务费项目(450060491471)资助
【分类号】：S436.421;O657.3

【参考文献】