基于GF-1时间序列影像的引黄灌区农作物种植结构监测

Crop Planting Structure Monitoring with Multi-Temporal GF-1 Wide Field View Data in Yellow River Irrigation Area of Ningxia

利用2018年农作物生长期的GF-1/WFV高空间分辨率卫星影像数据,经时间序列谐波分析法(HANTS)去除云和水汽等因素引起的噪声后,重新构建高质量的NDVI时间序列数据,结合采集的农作物地面样本数据,进行主要农作物关键生育期的特征分析,构建农作物决策分类模型,开展宁夏引黄灌区主要农作物种植结构的监测分析。研究结果显示:研究区主要农作物的分类总体精度为86.5%,Kappa系数为0.77,春小麦、玉米和水稻的分类精度分别为76.9%、88.8%和85.5%,种植面积分别为45451.0 hm2、214703.1 hm2和81472.6 hm2,分别占引黄灌区农作物面积的9.0%、42.3%和16.1%。基于GF-1/WFV多时相影像数据的决策树分类方法可以为研究区农作物信息提取提供参考,主要农作物提取能够获得较高的精度,具有一定的业务监测应用价值。

High quality NDVI time series data are reconstructed after removing the noise caused by cloud and water vapor using Harmonic Analysis of Time Series(HANTS)method based on the GF-1 wide field view(WFV)data of crop growth period in 2018.The characteristics of main crops during key growth periods are analyzed with the ground sampling data,and the crop classification model is constructed to monitor the crop planting structure in Yellow River irrigation area of Ningxia.The results show that the overall classification accuracy of the main crops is 86.5%,and the Kappa coefficient is 0.77.The classification accuracy of spring wheat,maize and paddy rice are 76.9%,88.8%and 85.5%,respectively.The planting area is 45451.0 hm2,214703.1 hm2 and 81472.6 hm2,respectively,accounting for 9.0%,42.3%and 16.1%of the crop area in the Yellow River irrigation area.Based on GF-1/WFV multi-temporal data,the decision tree classification method can be used to obtain highaccuracy main crop information so as to provide reference for the study of main crop information extraction.Therefore,the method is of application value for operational monitoring.