基于遥感地理特征曲线的主要农作物种植面积大尺度提取

Large scale extraction of main crops growing area based on remote sensing geographic feature curve

【目的】研究利用遥感地理特征曲线数据快速提取内蒙古呼伦贝尔市、兴安盟的主要农作物种植面积的方法。【方法】以2018年46期(MOD09Q1)数据为数据源,计算NDVI,利用NRF滤波方法重建时间序列数据集,并与5月NDVI最低值、DEM数据、坡度数据、5月LSWI最高值数据集成特征曲线数据集;用CART决策树分类方法对5种农作物种植面积进行提取,并利用地面样方数据进行精度验证。【结果】建立了小麦、油菜、玉米、大豆、水稻的标准特征曲线,并针对不同区域构建了适合该区域的农作物信息提取模型;地面样方数据验证相对精度达到77%以上。【结论】综合考虑遥感地理特征曲线,用CART决策树分类方法大尺度快速提取主要农作物面积是可行的。

【Objective】To study the method for extracting the growing area of main crops in Hulun Buir City and Hinggan League in Inner Mongolia by using remote sensing geographic feature curve data.【Methods】NDVI calculated by using MOD09Q1 as the dataset,46th,2018;reconstructed time series dataset by using the method of NRF filter and integrate feature curve data sets by adding NDVI minimum in May,DEM data,the slope data and LSWI maximum in May.The CART decision tree classification method was used to extract the growing area of 5 main crops,and the ground quadrat data was used to verify the accuracy.【Results】The standard feature curve of wheat,oilseed rape,corn,soybean,and rice was built,and the crop information extraction models were constructed for the different areas;the relative accuracy of ground quadrat data verification reached more than 77%.【Conclusion】Considering the remote sensing geographic feature curves,it was feasible to quickly extract the area of main crops on a large scale by using the CART decision tree classification method.