Sentinel-2与Landsat 8数据组合下的多特征冬小麦面积提取

Extraction of multi-feature winter wheat area based on Sentinel-2 and Landsat 8 data

遥感卫星的波段设置、信噪比及传感器观测角度等因素都会影响作物提取精度。为充分挖掘与发挥Sentinel-2卫星多光谱成像仪(MSI)与Landsat 8陆地成像仪(OLI)在冬小麦信息提取方面的优势,本文以商河县为研究区,基于两数据源的光谱特征、纹理特征、植被指数特征组合数据,利用随机森林(RF)与支持向量机(SVM)对冬小麦进行提取。结果表明:基于单一影像的最优Kappa系数与最优OA分别为0.89和95.13%,基于组合数据源的最优Kappa系数为0.92,最优OA为95.28%,两数据源组合的精度优于单一数据源提取精度;数据组合效果与分类器的性能有关,RF的Kappa系数相对于SVM分别提升0.04、0.20和0.11,OA分别提升2.41%、11.31%和6%,RF对冬小麦提取精度优于SVM。本文研究结果对于构建中高分辨率影像组合的典型农作物分类提取体系具有重要意义。

Remote sensing satellite band setting,signal to noise ratio and sensor observation angle will affect the accuracy of crop extraction.In order to fully tap the advantages of Sentinel-2 satellite multispectral instrument and Landsat8 land imager in winter wheat information extraction.this study takes Shanghe County as the research area.Based on the combination data of spectral characteristics,texture characteristics and vegetation index characteristics of the two data sources,random forest classification and support vector machine are used to extract winter wheat.Experiments show that the optimal Kappa coefficient and optimal OA based on a single image are 0.89 and 95.13%,respectively.The optimal Kappa coefficient based on the combined data source is 0.92 and the optimal OA is 95.28%.The accuracy of the combination of two data sources is better than that of the single data source.The data combination effect is related to the performance of the classifier.The kappa coefficient of RFC is increased by 0.04,0.20 and 0.11 compared with SVM,and OA is increased by 2.41%,11.31%and 6%,respectively.The extraction accuracy of RF for winter wheat is better than that of SVM.This study is of great significance for constructing a typical crop classification and extraction system based on medium-high resolution image combination.