基于地块特征基元与多时相遥感数据的冬小麦播种面积快速提取

Winter wheat planting area extraction using multi-temporal remote sensing data based on field parcel characteristic

冬小麦播种面积监测是农情遥感的重要研究内容之一,及时、准确地获取冬小麦的播种面积对冬小麦产量估算具有重要的意义。该文在面向对象技术的支持下,首先利用融合的高空间分辨率SPOT5遥感影像提取农田地块专题层信息;然后在专题层控制下对多时相的ETM+遥感数据统一尺度分割,得到不同时相遥感影像相同形状的地块特征基元;通过光谱特征规则集构建不同时相的冬小麦信息提取模型,实现对各个时期冬小麦播种田块对象的提取;最后通过交叉验证,确定最终的冬小麦播种面积。结果表明,该方法能够快速获得冬小麦播种面积,总体精度达到90%以上,基本上能够满足农情遥感监测的需求,为冬小麦种植面积遥感快速监测提供了一种可行的方法。

Winter wheat area estimation by mean of remote sensing is one of important research content in agriculture information monitoring field. It is very important for yield estimation to obtain planting area timely and accurately. This paper proposed a new method for extraction winter wheat area based on object oriented technology used multi-sources and multi-temporal remote sensing data. Field parcel information was extracted from SPOT5 panchromatic and multi-temporal fusion data. Multi-temporal ETM＋ images were segmented assistant by the thematic layer of field parcel at the same segmentation scale, and then, the segmentation results with the same shape feature were acquired to construct winter wheat extraction model based on dynamic spectral feature rule set. Through the cross validation, the winter wheat area was extracted. The experiment results show that the method can rapidly recognize winter wheat area. The overall accuracy of the result is over 90%,which can meet the calculation precision requirements of agriculture information remote sensing monitoring. Therefore, the method presented in the paper is practicable and can be widely used for winter wheat area quick extraction in future.