%0 Journal Article %T 基于多时相环境卫星的冬前油菜种植面积估算 %T Estimation of rapeseed planting area before winter based on multitemporal HJ-1A/B satellite data %A 魏传文 %A 黄敬峰※ %A 杨玲波 %A Wei Chuanwen %A Huang Jingfeng※ %A Yang Lingbo %J 中国农业信息 %J China agricultural informatics %@ 1672-0423 %V 31 %N 5 %D 2019 %P 38-48 %K 油菜;面积估算;决策树;冬前;HJ-1A/B(环境一号卫星);江陵县 %K rapeseed;area estimation;decision tree;before overwintering;HJ-1A/B;Jiangling county %X 【目的】油菜是我国重要的油料作物之一,监测油菜种植面积有助于了解油菜生长状况,为油菜病虫害、湿渍害、冻害等灾害损失评估提供数据基础。【方法】文章以湖北省荆州市江陵县为研究区,使用国产HJ-1A/B 30 m 分辨率时序多光谱数据,通过地面调查及资料分析确定油菜与其它易混淆作物的主要NDVI 时序特征,建立油菜识别决策树,估算了2009—2015 年(不包括2011—2012 年生长季)冬前油菜种植面积。将基于油菜开花期影像的最大似然法提取的油菜面积作为定性验证数据。以油菜籽面积统计数据和GoogleEarth 高分辨率影像数据对冬前油菜提取的面积和空间位置结果进行定量评价。【结果】定性评价结果:2009—2011 年生长季的决策树方法提取冬前油菜面积结果与开花期影像最大似然法提取结果基本一致,2012—2015 年生长季的油菜提取面积空间分布差异较大。定量评价结果:决策树方法提取冬前油菜面积的用户精度达到80.40%~95.56%,生产者精度达到82.56%~91.43%,相对误差低于15%。【结论】基于NDVI 时间序列特征的决策树算法估算冬前油菜面积具有可行性,但仍受到云和冬小麦的影响。 %X [Purpose]Rapeseed is the main cash crop in the Yangtze River Basin. Monitoring the planting area of rapeseed is helpful to understand the growth status of rapeseed and provide data basis for the loss assessment of rapeseed diseases and insect pests,waterlogging,freezing and other disasters.[Method]The study area is located at Jiangling County,Jingzhou City, Hubei Province in this paper. Decision tree was constructed using HJ-CCD time series data with the spatial resolution of 30 m. Planting area of rapeseed was estimated from 2009 to 2015 growing seasons(excluding 2011—2012)based on the NDVI time series characteristics of rapeseed and other easily confused crops. The rapeseed area extracted by maximum likelihood method was used as the qualitative verification data. The accuracy of area and spatial location of rapeseed were quantitatively evaluated by the statistical yearbook of rapeseed area statistics and Google Earth high resolution image data.[Result]The result of qualitative evaluation shows that the spatial distribution of flowering period of rapeseed in the growing season of 2009—2010 and 2010—2011 was basically consistent with the results obtained before overwintering. The spatial distribution of rapeseed in the growing season of 2012—2015 is quite different. The quantitative evaluation results show that accuracy of rapeseed area extracted by decision tree method was less than 15%. The accuracy ranges from 80.40% to 95.56% for user and 82.56% to 91.43% for producers. [Conclusion]The decision tree algorithm based on NDVI time series characteristics is feasible for estimating rapeseed area before overwintering. Cloud and winter wheat are the main factors affecting the estimation of rapeseed area before overwintering. %R 10.12105/j.issn.1672-0423.20190504 %U http://www.cjarrp.com/zgnyxx/ch/reader/view_abstract.aspx %1 JIS Version 3.0.0