叶面型水稻病虫害高光谱响应特性及监测机理研究

Rice Bacterial Blight (RBB, Xanthomonas ryzae) and Rice Leaf Folder (RLF, Cnaphalocrocis medinalis (Güenée)) are two typical foliar disease and insect pests in rice crop. The damage symptoms of RBB and RLF are similar. When the rice leaves are infected by the bacteria of Xanthomonas oryzae and fed by the larva of Cnaphalocrocis medinalis (Güenée), the rice leaves exhibit chlorosis and leaf water content decreases. It's difficult to discriminate RBB and RLF using optical remote sensing technique because the same spectral characteristics are caused by different pests in remote sensing images. Generally, the occurrence of disease and insect pests are influenced by many physiological factors in paddy rice fields, which was ignored by most of the current researches. The single and multiple physiological factors (i.e. water, nitrogen and planting density) are considered in the application project.The hyperspectral reflectance of imaging spectroradiometer, physiochemical parameters of single leaf and population parameters of rice crops are measured at the leaf scale under laboratory, the field canopy scale and the regional scale. Firstly, same spectral wavebands and exclusive spectral wavebands from visible to shortwave spectral regions (400-1700nm) are determined among RBB, RLF and water supply, nitrogen condition and planting density by using multiple spectroscopy analysis techniques. The new hyperspectral feature parameters and vegetation indices will be developed according to the spectral profile characteristics and exclusive spectral wavebands. Secondly, the inversion models will be built to estimate physiochemical parameters of single leaf and population parameters of rice crops with statistical models, electromagnetic radiation transfer models and data mining algorithms. Finally, a series of image classification signatures such as endmember spectra, texture features and damage symptoms are extracted from the hyperspectral images, meanwhile the objects are classified. Then the hyperspectral remote sensing models based on phenology of RBB and RLF, habitat and experts knowledge are developed to discriminate the damage symptoms of RBB and RLF, to difference pests between RBB and RLF, and quantitatively inverse the damage severity of RBB and RLF through direct and indirect methods. The former is to input the abovementioned hyperspectral parameters, and the latter is to input the physiochemical parameters of single leaf and population parameters of rice crops which indicate the damage symptoms and severity of RBB and RLF. The overall goal of application project is to provide reference for monitoring the disease and insect pest of other crops with similar damage symptoms using hyperspectral.

稻白叶枯病和稻纵卷叶螟是典型的叶面型水稻病虫害，具有近似的"失绿"又"失水"为害状，呈现 "同谱异害" 的特征更加剧了遥感识别的难度。本项目结合病虫害通常受多因素影响的现状，设置水分、肥力和密度等复合影响实验，在室内叶片、田间冠层和田块三个尺度上，获取相应的成像光谱和个体理化指标、群体参数。先采用多种光谱分析技术，确定前述病虫害和生理性胁迫在400-1700nm相同的和可能的专属性光谱响应段，构建新型高光谱特征参数和植被指数；再运用统计模型、电磁波辐射传输机理模型和数据挖掘算法估算个体理化指标和群体参数；最后在高光谱图像端元光谱、为害状、纹理特征提取和分类的基础上，通过输入高光谱参量或反映病虫为害的理化指标、群体参数两种模式，构建基于病虫害物候信息、生境信息和专家知识的高光谱遥感模型，从而精确地判别和定量地反演稻白叶枯病和稻纵卷叶螟的为害程度，为为害状近似的农作物病虫害遥感监测提供借鉴。

病虫害是农作物产量和品质的重要生物性影响因素，病虫害类型精准识别和危害程度监测是病虫测报和预防的基础，尤其是具有同谱异物特质的近似的病害和虫害，例如稻白叶枯病和稻纵卷叶螟。本研究通过ASD光谱辐射仪和Headwall高光谱成像仪测定了缺水、缺肥影响条件下，叶片和冠层尺度下的反射辐射光谱。研究发现：(1)与健康光谱相比，水稻受到稻纵卷叶螟取食和稻白叶枯病菌侵染后，反射光谱值在橙红光谱段(650-730 nm)增高，在近红外谱段(850-1250 nm)和短波红外波段(1550-1700 nm、2050-2250 nm)降低。(2)病虫害胁迫反射光谱具有明显的“红移”现象。(3)叶绿素、胡萝卜素、叶片含水量降低30-75%。(4)比值光谱指数R1080/R705、R1675/R705能够定量反演色素(叶绿素、胡萝卜素)和叶片含水量，精度达到95%左右。(5)稻白叶枯病、稻纵卷叶螟与健康水稻在叶片尺度的识别精度均为100%，在冠层尺度的识别精度分别为94.29%和91.43%，稻白叶枯病和稻纵卷叶螟在叶片和冠层的区分度分别为91.43%和85.71%。研究表明：高光谱遥感能够有效识别危害症状近似的病害和虫害，为生产者精准施药提供可靠的病虫危害程度和危害类型信息。