基于GF-1卫星遥感数据识别京津冀冬小麦面积

省级尺度冬小麦面积的精准获取技术是农作物面积遥感监测研究的主要内容之一。为了获取省级尺度的冬小麦种植面积,该文以北京市(京)、天津市(津)和河北省(冀)3个省域范围为例,以国家标准地形图分幅为分类的图幅单元,利用国产GF-1/WFV数据,构建冬小麦面积指数,实现了省级尺度冬小麦面积的识别。本文以冬小麦全部9个月生育期的984景影像作为数据源,依次经过数据预处理、标准图幅单元的NDVI合成、样本点选择、冬小麦面积指数构建、冬小麦作物类型确认、省域范围制图及精度验证等步骤完成研究区域内冬小麦面积的提取。采用区域网平差和6S大气校正算法对数据源预处理,以中国1︰10万标准地形图分幅为分类图幅单元构建冬小麦面积指数,将冬小麦面积指数按照1%的比例等分,并将面积指数从0到100%分割为101个提取节点,将提取节点的NDVI值依次与类型确认样本比较,精度最高的则确认为冬小麦面积提取阈值,同时将该阈值应用于图幅单元内冬小麦面积指数影像,获取冬小麦种植分布。最后冬小麦面积识别的精度表明,以标准地图分幅作为计算单元,在GF-1影像基础上,利用冬小麦面积指数能够显著提高冬小麦与其他地物类型的波谱差异,且冬小麦的总体识别精度达到89.6%,用户精度达到89.8%,制图精度96.5%,Kappa系数0.72。在典型区域,本文算法与监督分类算法精度结果较为一致,除制图精度相差4.77%外,总体精度与用户精度差都在1.00%以内,说明本文算法具有精度高、运行效率高、分类单元识别结果一致性强的特点,能够满足省级尺度农情遥感业务监测的需要。

Acguisition of Winter Wheat Area in the Beijing-Tianjin-Hebei Region with GF-1 Satellite Data

Winter wheat area accurate acquisition at provincial scale is an important aspect in remote-sensing monitoring of crop area. This study aimed at estimating winter wheat area in Beijing, Tianjin, and Hebei at provincial scale using classification map units from the national standard topographic map and the winter wheat area index (WWAI) created from the wide field view (WFV) data of GF-1 satellite. A total of 984 satellite monitoring images between October 1, 2013 and June 30, 2014 were used as data sources. The major process steps were data preprocessing, NDVI synthesis of standard map units, selection of samples, creation of winter wheat area index, confirmation of winter wheat crop type, provincial scale mapping, and accuracy verification. Multi-temporal GF-1/WFV data were preprocessed and NDVI value of images were calculated by using block adjustment and 6S atmospheric correction algorithm. By means of the 1:100 000 national standard topographic map framings of China, as the classification map units, WWAI was equally divided at a proportion of 1 percent into 101 extraction nodes from 0 to 100%. The NDVI values of extraction nodes were compared with type confirmation samples, and the most accurate NDVI was adopted as the extraction threshold value of winter wheat area. This threshold was then used into WWAI image in the map units to obtain winter wheat planting distribution. The identification result showed that, by taking standard map framing as calculation units and based on the creation of WWAI of GF-1 images, we remarkably improved the wave spectrum difference between winter wheat and other ground objects, with the overall accuracy of 89.6%, user accuracy of 89.8%, mapping accuracy of 96.5%, and Kappa coefficient of 0.72. In a typical region, the algorithm proposed had a similar accuracy with supervised classification algorithm. Except for 4.77% of the difference in mapping accuracy, the differences in overall accuracy and user accuracy were less than 1.00%. These results indicate that the used in this study is of algorithm high accuracy, efficiency and consistency in classification unit identification and applicable in agricultural monitoring at provincial level.