基于GF-1/WFV数据的冬小麦条锈病遥感监测

条锈病是冬小麦常见病害,利用遥感影像对条锈病病害区域进行准确监测具有重要意义。该文利用GF-1/WFV影像,结合条锈病地面光谱数据分析,采用冬小麦条锈病遥感监测指数(wheat stripe rust index,WSRI)对河南西华县冬小麦条锈病发病范围进行了估测。首先,利用冬小麦NDVI加权指数(weighted NDVI index,WNDVI)获取冬小麦种植区域。其次,利用影像4个波段反射率之和提取不同冬小麦品种的分布范围,值较高的为条锈病高抗品种(郑麦系列),较低的则是条锈病易感品种(矮壮系列)。再次,构建冬小麦条锈病指数(wheat stripe rust index,WSRI),结合地面实地调查的条锈病分布数据,通过设定合理的WSRI指数划分阈值,提取条锈病染病区域并进行精度验证。结果表明,研究区内小麦条锈病空间分布识别的总体精度在84.0%以上,具有区域监测应用的潜力。该方法简单,可操作性强,表明宽波段GF-1影像结合WSRI指数的技术,是一种比较可行的小麦条锈病遥感监测方案。

Winter wheat stripe rust remote sensing monitoring based on GF-1/WFV data

Abstract: Stripe rust is a common disease of winter wheat, and accurate monitoring of stripe rust disease has great significance. By using the GF-1/WFV images on April 18, 2017 and combined with the analysis on stripe rust ground spectral data, this paper conducted estimation on the scope of winter wheat stripe rust in Xihua County of Henan Province with wheat stripe rust index (WSRI). The main contents of this study included identifying winter wheat area, identifying distribution of winter wheat varieties, calculating winter wheat stripe rust monitoring index, identifying distribution of disease, and verifying accuracy. Identification of winter wheat area was achieved by using weighted normalized differential vegetation index (WNDVI), and computation of WNDVI used images of 7 time phases, with the time scope from October, 2016 to April, 2017, one image each month. The distribution of winter wheat varieties was identified by dividing the thresholds of spectral brightness index (SBI). SBI is the sum of reflectances of 4 wave bands of WFV images. The areas with high thresholds were taken as the distribution areas of high stripe rust resistant varieties (Zhengmai series) and the areas with low thresholds were taken as the susceptible varieties of stripe rust (Aizhuang series). The acquisition of the threshold took the sample points of the ground observation as its basis. The identification accuracies of the variety distribution of different SBI points were tested respectively, and the node with the highest accuracy was taken as the threshold. By using observed spectrum of the ground observation, WSRI of the infected areas was calculated based on the average value of the reflectance of winter wheat observed with the same wave band as GF-1/WFV. The WSRI value of the winter wheat of the normal sample points was 0, and all the values of the infected sample points were larger than 0. The WSRI value was increasing with the increase of the infection degree of the disease, which was consistent with the actual observation results. It indicates that WSRI index has indicative function on winter wheat stripe rust, and it can be used in the remote sensing monitoring for the disease. WSRI index of WFV was calculated by using the methods and parameters specified in the National Industrial Standard of the People''s Republic of China, Technical specification on remote sensing monitoring for crop diseases. And the scope of the WSRI index was between 0.15 and 20.73. The WSRI indices of the images were divided into 100 values with equal intervals, and then 101 node values were obtained. The images were divided 2 parts by using node value, and the accuracy was verified by using ground observation results. The node value with the highest accuracy was taken as the critical threshold between disease and non-disease, which was identified as 4.2 in this study. The pixels with the value higher than the threshold were the disease infected pixels. By doing so, the spatial distribution of the winter wheat infected with stripe rust could be obtained. The study results showed that, the method could objectively reflect the scope of occurrence of winter wheat stripe rust, and the extraction accuracy on infected area was higher than 84.0%. The user accuracy and mapping accuracy of extracting disease point of stripe rust were 86.4% and 82.1% respectively, and the user accuracy and mapping accuracy of extracting healthy point were 79.2% and 88.5% respectively. This method can meet the requirement of disease monitoring. This method is simple and easy to operate, and it shows the application potential of GF-1 images and WSRI indices in winter wheat stripe rust remote sensing monitoring.