李峰等

摘要：针对目前省域冬小麦种植面积遥感监测业务化运行中所存在的监测精度与成本之间相互制约的问题，本研究选取冬小麦主产区——山东省为研究区域，利用HJ-1卫星提供的CCD遥感影像，采用决策树分类法，同时以山东省土地利用类型之林地数据和野外地面调查数据作为辅助，通过分区解译提取出山东省冬小麦种植面积和分布区域。研究表明：利用HJ-1 CCD影像进行山东省冬小麦种植面积提取是可行的，位置精度和面积总量精度分别达到93.0%和96.8%，对国产卫星的推广应用具有良好的示范作用。

关键词：HJ-1 CCD影像；山东省；冬小麦；决策树分类

中图分类号：S127文献标识号：A文章编号：1001-4942（2015）05-0109-06

Study on Winter Wheat Planting Area

Extraction Based on HJ-1 CCD Images

Li Feng1， Zhao Hong1， Zhao Yujin1， Shang Lin1， Cao Zhangchi2

（1.Shandong Provincial Climate Center， Jinan 250031， China； 2.Linyi Meteorological Bureau， Linyi 276004， China）

AbstractOn account of the mutual restricting problems among the accuracy and cost in the measurement of winter wheat planting area at provincial scale， it was studied to extract its planting and distribution areas using the HJ-1 CCD images in this paper. Shandong Province as the main production area of winter wheat was selected for this research. Using the decision tree classification and partition interpretation methods， which solved the spectral difference problems caused by growth period， species and farming， the winter wheat planting and distribution areas were extracted with the forest land data and the field survey data as assist. The results showed that the position accuracy and total area accuracy were up to 93.0% and 96.8% respectively， which preferably reflected the winter wheat planting area and distribution in Shandong Province. It was also a good model for the promotion and application of domestic satellites.

Key wordsHJ-1 CCD images； Shandong Province； Winter wheat； Decision tree classification

山东是我国冬小麦生产大省，占全国总产量的20%以上。及时、准确地获取其冬小麦种植面积及空间分布状况信息，对于准确估计和预测产量，加强生产管理，优化种植空间格局，确保山东省乃至全国粮食安全都具有重要意义。

目前，中国法定发布的农作物种植面积数据一直采用两种传统的方式进行统计，一种是按行政单元逐级统计并层层上报的全面统计，另一种是抽样调查。多年来，这两种方式发挥了重要作用，但也存在着很大的局限性。层层上报受人为因素干扰严重，且费时费力，准确性差，难以达到理想精度。抽样调查方法虽然人为干扰小、准确性高，但工作量大，人力、物力、财力投入大，且只适用于省级以上大范围农作物种植面积监测，难以获得区县级别的小区域作物种植面积。……