基于比值导数法的棉花蚜害无人机成像光谱监测模型研究

蚜虫是棉花的主要害虫之一，我国棉花产量每年因蚜虫危害造成的损失高达5%～10%。田块尺度的棉花蚜害空间分布监测可以辅助精准定量施药，减少环境污染。利用无人机搭载成像光谱仪获取的“图谱合一”的遥感数据因其具有分辨率高、时效性高、成本低等优势，可为作物病虫害监测提供了重要数据源。比值导数法模型简洁，运行效率高，结果精确，可以有效的应用于遥感反射率光谱解混处理，提取对目标信息较为敏感的波段，为构建虫害监测模型提供了有效的手段。因此本研究选择棉花典型生产区新疆库尔勒地区为实验区，开展以下工作: （1）以低空无人机搭载成像光谱仪获取棉花蕾期冠层成像光谱影像，结合地面调查数据，获取76个样点光谱数据及蚜害严重度(包含健康植株16个，蚜害严重度1～4级每级选取15个)；(2)分析不同蚜害严重度棉花冠层光谱的特征，并利用比值导数法筛选出对蚜害胁迫敏感的光谱波段，分别为514，566和698 nm波段; (3)构建基于三个敏感波段的光谱反射率、比值导数光谱值的一元线性回归和偏最小二乘法的蚜害严重度估测模型。结果表明：(1)蚜害对棉花冠层的光谱反射率有显著影响。棉株受蚜害胁迫越严重，其在可见光区域的反射率越高，近红外波段反射率越低，发生红边区域“蓝移”；（2）比值导数法可有效提取蚜害棉花冠层光谱敏感波段，所筛选的514，566和698 nm三个波段与相关系数法所筛选的敏感波段一致；(3)利用敏感波段比值导数光谱值所构建的蚜害严重度估测模型精度优于敏感波段光谱反射率所构建的模型, 其中698 nm波段构建的模型精度最佳（R2=0.597, RMSE=0.91）; (4)三个敏感波段的比值导数光谱值所构建的偏最小二乘多元回归模型精度优于单个波段比值导数光谱值所构建的模型（R2=0.612, RMSE=0.89）；(5)基于比值导数法的棉花蚜害无人机成像光谱监测模型可以获取田块尺度的不同严重度蚜害空间分布图，对于精准定量施药有重要的指示意义。

Cotton Aphid Damage Monitoring Using UAV Hyperspectral Data Based on Derivative of Ratio Spectroscopy

Aphids(Aphis gossypii)are one of the main pests of cotton.The annual loss of China’s cotton yield due to aphids is as high as 5%~10%.Monitoring the grading profile of field-scale aphid damage can assist the precise application of quantified pesticide and reduce environmental pollution.The hyperspectral imaging data obtained by unmanned aerial vehicle(UAV)-mounted imaging spectrometer has the advantages of high resolution,high timeliness and low cost.The method based on derivative of ratio spectroscopy has the advantages of simple,efficient and highly accurate,which can be effectively applied to the remote sensing spectrum unmixing process,and extract more sensitive bands to the target information,providing an effective means for the establishment of pest monitoring model.Therefore,in this study,the Korla region of Xinjiang,a typical cotton production area,was selected as the experimental area to carry out the following work.(1)to use a low-altitude-unmannedaerial-vehicle-based hyperspectral imaging instrument to acquire hyperspectral images of cotton(Gossypium)canopies.Spectral data of 76 sample points and severity of aphid damage were obtained(including 16 healthy plants,and 15 were selected from each grade of 1~4 severity of aphid damage.(2)to use the derivative of ratio spectroscopy(DRS)to select sensitive spectral bands from cotton canopy spectra to detect aphid damage at the bud stage,band 514 nm,band 566 nm band 698 nm;(3)to construct unary-linear-regression and partial-least-squares models based on the sensitive bands of reflectance spectra and derivative of ratio(DR)spectra for rating aphid damage.The rexperiments’results revealed the following:(1)aphid damage had a significant effect on the spectral reflectance of the cotton canopy.The more seriously cotton plants were affected by the aphid,the higher the reflectivity in the visible region and the lower the reflectivity in the near-infrared band,and the"blue shift"occurred in the red envelope region.(2)sensitive bands for detecting aphid damage were effectively extracted from the DR spectra of the cotton canopy,and the three selected bands(with wavelengths of 514,566,and 698 nm)were consistent with the sensitive bands extracted by using the correlation coefficient method;(3))the precisions of the aphid-damage-rating models constructed using DR spectra of the sensitive bands were better than those of the models constructed using of the sensitive bands from the reflectance spectra,among which the model constructed with 698 nm band had the best accuracy(R2=0.612,RMSE=0.89);(4)based on derivative of ratio spectroscopy method,the UAV imaging spectral monitoring model of cotton aphid infestation can obtain the spatial distribution map of different severity aphid infestation on the field scale,which is of great significance for precisely quantified pesticide.