大斑病是一种对玉米危害严重的病害, 迫切的需要一种可以快速了解玉米大斑病病情的方法。 以无人机遥感作为新的技术平台, 探究玉米冠层受到大斑病胁迫时的光谱响应情况, 并利用无人机高光谱成像技术对大斑病病情进行监测和可视化研究。 采集玉米多生育期(抽雄期、 灌浆期、 完熟期)冠层500～900 nm的高光谱影像, 根据采集影像的原始光谱和一阶微分光谱特征, 提取出12个大斑病敏感波段位置, 12个波段位置分别为: 514, 532, 553, 680, 714, 728, 756和818 nm, 近红外、 红、 绿波段及红边位置。 根据前人提出的植物病害监测参数结合提取的敏感波段位置, 构建13组针对玉米冠层大斑病的监测光谱参数, 研究不同波段对大斑病病情指数(DI)值的敏感性, 并构建玉米冠层大斑病的监测模型, 验证利用无人机遥感监测大斑病DI值的精度及稳定性。 结果表明: 随病情指数增加, 一阶微分光谱图出现典型的“蓝移”现象, 病害冠层DI值与红光(680～714 nm)和近红外(770～818 nm)的反射率及一阶微分光谱图的红边位置(680～756 nm)相关性更显著, 与绿光波段相关性较低。 在13组监测光谱参数中, 8组与建模样点冠层大斑病实测DI值达到极显著相关水平, 决定系数(R2)均达到0.8以上, 选取各生育期R2达到0.8以上的光谱参数用于玉米冠层大斑病监测模型的构建, 将检验样本的实测值与监测模型的预测值进行相关性分析。 检验表明, 在抽雄期, 模型DI-NDVI(SDλi, SDλj)的回归斜率(0.829 3)和决定系数(R2=0.842 7)都最接近1, 均方根误差(RMSE=4.59)和相对误差(RE=12.3)更小, 说明模型DI-NDVI(SDλi, SDλj)的预测能力和精度更高。 各生育期对应模型均取得较好监测效果, 说明本研究利用无人机遥感对植物病害监测具有指导意义, 对精准农业的发展具有一定的借鉴价值。

Blight disease is a serious disease for corn. Therefore, there is an urgent need for a method for quickly understanding the condition of corn blight disease. In this study, UAV remote sensing is used as a new technology platform to explore the spectral response of corn canopy under the stress of blight disease, and UAV hyperspectral imaging technology is used to monitor and visualize the disease of blight disease. Therefore, this study collected data on corn growth stages (the tasseling period, the filling period, the maturity period), and used the UAV hyperspectral instrument to acquire the hyperspectral image of the canopy 500～900 nm. This research based on the original spectra and first-order differential spectral features of the acquired images, the position of the 12 sensitive spots of the blight disease was extracted. The positions of the 12 bands were: 514, 532, 553, 680, 714, 728, 756 and 818 nm, near-infrared, red, green and red edge positions. At the same time, based on the plant disease monitoring parameters proposed by the predecessors combined with the extracted sensitive band positions, 13 sets of monitoring spectral parameters for corn canopy blight disease were constructed. In this way, the sensitivity of different bands to the disease index (DI) value of blight disease was studied, and a monitoring model for monitoring corn canopy blight disease was constructed to verify the accuracy and stability of using the UAV remote sensing technology to monitor the DI value of blight disease. The results show that with the increase of the disease index, the first-order differential spectrum shows a typical “blue shift” phenomenon, and the correlation between the disease canopy DI value and the red (680～714 nm) and near-infrared (770～818 nm) reflectance and the red edge position (680～756 nm) of the first-order differential spectrum is more significantly, the correlation with the green band is low. Among the 13 groups of monitoring spectral parameters, 8 groups and the modeled canopy blight disease measured the DI value reached a very significant correlation level, R2 all reached above 0.8. Therefore, in this study, the spectral parameters of R2 with a growth period of 0.8 or higher were selected for the construction of the corn canopy blight disease monitoring model, and the correlation between the measured values of the test samples and the predicted values of the monitoring models was analyzed. The test shows that in the tasseling period, the regression slope (0.829 3) and the decision coefficient (R2=0.842 7) of the model DI-NDVI(SDλi, SDλj) are closest to 1, and the root mean square error (RMSE=4.59) and relative error (RE=12.3) are smaller, indicating that the prediction ability and accuracy of the model DI-NDVI(SDλi, SDλj) are higher than others. The results show that the corresponding models in each growth period have achieved good monitoring results, indicating that the research using UAV remote sensing has guiding significance for plant disease monitoring, and has certain reference value for the development of precision agriculture.