基于高分卫星影像的青岛市冬小麦种植信息提取

Winter Wheat Planting Information Extraction in Qingdao:Based on High-resolution Satellite Imagery

为了准确获取青岛市主要农作物冬小麦的种植信息,以GF-1/16 m卫星影像为主要数据源,将高程、土地利用和田间调查数据作为辅助数据源,根据冬小麦主要发育期与其他地物在GF-1/16 m卫星影像上的光谱差异,计算得到4月份为青岛市冬小麦遥感面积提取的最佳时相。在最佳时相内,采用决策树分类法,通过分区解译方式,提取出青岛市2017年冬小麦种植面积和分布区域,并利用GF-2融合后1 m卫星影像、地面调查数据和统计局公布数据对分类结果进行精度验证。结果表明:利用GF-1/16 m卫星影像在幅宽、时间和空间分辨率的优势,将土地利用和高程等引入决策树分类模型,进行区域尺度的冬小麦种植面积遥感估算的方法是可行的。经精度验证,2017年青岛市冬小麦遥感解译总精度为94.3%,Kappa系数为0.857。遥感提取面积略小于统计局公布数据,面积总量提取精度为93.6%。本研究为基于高分卫星影像的区域尺度作物种植面积提取提供参考。

To accurately obtain the planting information of winter wheat,which is the main crop in Qingdao,taking the GF-1/16 m satellite imagery as the main data source,combined with the auxiliary data source including the elevation,land utilization and field survey data,we calculated and got the results that the optimum time phase for extracting the remote sensing area of winter wheat in Qingdao was April through the spectral differences between the main development period of winter wheat and other features on the satellite imagery of GF-1/16 m.During its optimum time phase in 2017,we extracted the planting area and distribution area of winter wheat in Qingdao by the decision tree classification and districting interpretation.Besides,the 1 m satellite imagery fused GF-2,ground survey data and the data released by the statistics agency were adopted to verify the accuracy of the classification results.The results showed that:the remote sensing estimation method for winter wheat planting area on regional scale was feasible by using the advantages of GF-1/16 m satellite imagery in width,time and spatial resolution to introduce land utilization and elevation into the decision tree classification model.By verification of accuracy,the total accuracy of remote sensing interpretation for winter wheat in Qingdao in 2017 was 94.3%,and the Kappa coefficient was 0.857.The remote sensing extraction area was slightly smaller than the data released by the statistics agency,and the extraction accuracy of total area was 93.6%.This study provides reference for the extraction of crops planting area on regional scale based on high-resolution satellite imagery.