【作者】 乔红波；

【导师】 程登发；

【作者基本信息】 中国农业科学院 ， 病虫测报学， 2007， 博士

【摘要】 农作物病虫害是影响粮食生产安全和产量的重要因素之一，防治病虫害大量使用的杀虫和杀菌剂可导致生态环境恶化，影响食品安全。病虫害频繁发生除与作物品种，气候变暖等因素有关外，病虫害监测及预警技术手段滞后仍是病虫害大面积发生的主要原因。遥感实时、动态、无损的技术特点在病虫害监测预警中具有重要作用。本文在实地调查和田间高光谱测量、对麦蚜和白粉病发生特点及不同源遥感机理深入理解的基础上，研究遥感反演病虫害发生环境要素可行性，探讨遥感影像光谱与实测病虫害光谱的关系，利用麦蚜和白粉病地物光谱特性在卫星遥感图像上的反映进行危害信息提取。主要内容如下：利用ASD手持式高光谱仪测定不同病虫害的光谱反射率，对测得的反射率数据进行一阶微分变换、对数变换、归一化变换等变换，利用逐步判别分析、线性判别分析和分层聚类法对不同病虫害进行识别。结果表明：逐步判别分析法选择的波段主要位于红、绿、蓝、和近红外区；分层聚类法选择的波段除了红、绿、蓝、和近红外波段外，还增加了蓝-绿边缘、绿-红边缘和红边区的波段。所选择的波段比原始波段在病虫害识别时具有更高的精度，最高识别精度达90.6％；边缘区波段对病虫害的识别有重要作用；用对数微分变换处理较其他方法处理对病虫害识别有更好的效果。利用手持式高光谱仪和基于数字技术的低空遥感系统，对不同严重度小麦白粉病冠层光谱反射率进行了测定，同时调查病情指数，分析不同时期地面平台光谱反射率与病情指数及低空遥感平台反射率与病情指数、归一化植被指数相关性。结果表明，地面光谱测量冠层光谱反射率和低空遥感数字图像反射率与小麦白粉病病情指数在灌浆期有显著的相关关系，就地面测量结果而言，近红外波段的相关性高于绿光波段，相关系数分别为-0.79和-0.54；低空遥感数字图像红、绿、蓝三波段中，相关性依次降低，相关系数分别为-0.79、-0.75和-0.62；而且低空遥感图像与归一化植被指数也存在较好的相关关系，蓝、红、绿波段，相关系数依次为0.70、0.68和0.54。使用分裂窗算法对NOAA图像进行地表温度反演，将地面实测获得的日最高温度与反演地面温度一一对应进行比较，结果表明遥感反演地温与地面实测的线性关系相关系数最高为0.89，最低为0.51，说明遥感反演地温数据与地面实测地温有显著的相关关系，但反演的值普遍高。对时间序列NDVI分析表明，冬小麦生育期序列曲线具有明显的规律。根据调查地面蚜虫数量和实地光谱测量，建立蚜虫危害与NDVI相关方程：NDVI=-3×1.0^-3X+0.623，R=0.918（P<0.01），达到极显著水平，X为百株蚜量。根据此方程对NOAA-NDVI图像进行分析，对蚜虫危害进行监测。并利用NOAA图像对其它植被要素和冬小麦分布进行研究。根据分裂窗算法对MODIS遥感图像进行地表温度和各种植被要素反演，并与NOAA反演结果对比分析，同时对植被要素和冬小麦面积分布进行监测。从反演的系列变化可以看出，河南省地表温度变化在260-320K之间。MODIS反演地表温度与实测温度相关性分析结果表明，MODIS反演结果更接近于实测值。并对反演结果进行精度分析，结果NOAA和MODIS卫星反演温度的平均误差分别为0.90℃和0.41℃。MODIS-NDVI和NOAA-NDVI直方图分析结果表明，MODIS-NDVI的最大值和平均值均比NOAA-NDVI的大，而且，NOAA-NDVI的动态范围小，因此MODIS-NDVI对植被的响应比NOAA-NDVI更敏感。根据麦蚜和白粉病的发生机理与光谱特征，分别采用TM-NDVI和MPH技术提取危害信息。根据调查点GPS信息在TM-NDVI图像上分别定位健康和受害区两个调查点，信息提取结果表明受害前小麦田块的NDVI相近，而受害后小麦田NDVI值明显降低。对TM图像的DN值进行统计分析，发现健康小麦在第4波段近红外谱段会出现一个反射峰，到第5波段下降，而受病虫害为害的小麦在近红外谱段的光谱值降低，反射峰值出现在第5波段。根据主成分变换的特征向量矩阵，利用第3主分量来提取病虫害危害的特征信息。掩膜图像经主成分变换后的第3主分量在4月30日灰度值接近，而到5月16日健康小麦田的灰度值为比受害小麦田灰度值明显降低。主成分变换的第3分量比较集中地表现病虫为害的特征信息。本论文在对麦蚜和白粉病发生机理和为害特征深入研究基础上，通过实测地面高光谱数据和病虫害发生实地调查，利用不同源的遥感数据对麦蚜和白粉病的危害特征和遥感信息提取技术进行研究。该研究对于运用遥感等高新技术快速、及时获取大面积病虫害发生动态信息，建立病虫害的预测及监测体系，提高病虫灾害防治能力具有重要意义。更多还原

【Abstract】 Plant diseases and insect pests is one of the important factors influencing the agricultural product safety and yield. Excessive application of pesticide and fungicide to control the plant diseases and insect pests led to the deterioration of the ecological and food problems. In addition to the crop varieties and the global warming, one of the main factors for the recent increase in frequency and severity of plant diseases and insect pests is still lacking of adequate surveys to detect incipient damage and plan appropriate control operations. The remote sensing technology has great advantage of plant diseases and insect pests forecasting due to the characteristic of real time, dynamic and non-touch. On the basis of field survey and spectrum measurement, understanding to remote sensing mechanism and plant diseases and insect pests feature spectrum theory thoroughly, this paper focused on the feasibility of retrieval of environment elements, the relationship between the feature spectrum in the field and the remote sensing digital image, and using the plant diseases and insect pests feature spectrum to extract the damage information from the satellite remote sensing digital image. The main contents are as follows:The canopy reflectance of different plant diseases and insect pests was measured by using ASD hand-held spectroradiometers. The reflectance data was transformed by the method of first differential coefficient, logarithm and normalization, then stepwise discriminate analysis and hierarchical clustering were used to identify the different plant diseases and insect pests. The results suggested that bands selected from stepwise discrimination analysis mainly lied along the blue, green, red and near-infrared bands. In addition to the blue, green, red and near-infrared bands, the spectral bands along the blue-green edge, green-red edge and red curves were selected by the hierarchical clustering. Plant diseases and insect pests could be identified more accurate by using selected bands than the original data, the highest recognition accuracy of up to 90.6%, and the bands lying along the edges had important information for discrimination of plant diseases and pests. The spectral data, dealt with the transformation of logarithm and differential coefficient, could achieve better accuracy than others.Using ASD Hand-held Spectroradiometers and low altitude remote sensing system measured the canopy reflectance of wheat damaged by wheat powdery mildew and, at the same time, scored the disease index （DI）. The correlation between reflectance and DI and normalized difference vegetable index （NDVI） and DI was calculated. The results showed that the correlation between reflectance and DI was significant during the stage of milk-filling. The results of in-field measurement showed that the correlation between near infrared reflectance and DI was higher than that of green band with the correlation coefficient of -0.79 and -0.54, respectively. The reflectance correlation between red, green and blue bands decreased in the order with the correlation coefficient of -0.79, -0.75 and -0.62, respectively. The reflectance of low altitude remote sensing was well correlated with NDVI moreover. The correlation coefficients were 0.70, 0.68 and 0.54 for blue, red and green band, respectively. The land surface temperature （LST） was retrieved from NOAA image by the method of split window algorithm, and it was found that there were correlation coefficients maximal to 0.93 （P<0.01） between LSTs from remote sensing and those from ground observation, but the LST generally higher than observations. Analysis of the time sequence NOAA-NDVI image suggested that winter wheat NDVI had the obvious regulation. On the basis of in field investigation and hyperspectral reflectance data established the correlation equation: NDVI = -3×10^-3x + 0.623, R =0.918 （P<0.01）. The aphid damage level from NOAA-NDVI image was retrieved based on above equation, at the same time; the other vegetation elements and winter wheat distribution were retrieved by the NOAA-NDVI image.The LST was retrieved from MODIS 1 B image by the method of split window algorithm and other vegetation elements also retrieved, moreover, the retrieval results were compared to NOAA. The retrieval LSTs sequence results showed the LST ranged from 260k to 320k. The correlation analysis suggested its retrieval result from MODIS was more accurate than that from NOAA, and more close to the ground observation. The error was 0.41°C and 0.90°C, respectively. Analysis of the NOAA-NDVI and MODIS-NDVI histogram showed that their shapes were the same, and the range of their values was wider than that of NOAA, so MODIS-NDVI had more sensitivity to vegetation.Based on the occurrence mechanism and feature spectrum of wheat aphid and wheat powdery mildew, the methods of NDVI and Masking, Principal component transformation and Hue adjust （MPH） technique were applied in extraction of the damage information from TM multi-spectral data. The healthy and damaged field samples were located in terms of the GPS information, the extraction information results showed that the NDVI value almost the same before injury, but the damaged sample NDVI value obviously decrease after injury. The statistics of DN value from TM image suggested that the healthy sample winter wheat had a peak at TM4 and decrease at TM5, but the damaged sample was contrary to healthy. From the PCA eigenvector matrix, the healthy and damaged samples in PC 3 of masked imageries before injury were almost the same, but after injury obviously increased. The results indicated that the PC 3 of masked imagery could reveal wheat aphid and wheat powdery mildew feature information apparently.This paper on the basis of understanding to wheat aphid and powdery mildew occurrence mechanism and their feature spectrum thoroughly and in field investigation, using of multi-source remote sensing image retrieved the LST, vegetation and other elements and extraction the damage information. It is prospective for using remote sensing technology archive large scale and dynamic plant diseases and insect pests occurrence information in real time and quickly, at the same time, it is also important to establish the forecasting system and increase the ability of plant diseases and insect pests control.更多还原