提出了一种基于统计抽样总量控制下的中高分辨率遥感影像玉米种植面积信息提取方法, 该方法首先利用分层抽样技术对调查目标总体(玉米)进行分层抽样；然后对抽样小区进行目视解译, 反推区域总量真值；最后在总量控制下进行区域目标作物的空间分布提取。以河北省三河市中部地区的部分影像为研究区, 以该区2006-08-21的10m分辨率的SPOT 5多光谱影像为基础数据进行了试验研究。结果表明该方法基于群样本检验的总体精度达到93.8%, Kappa系数达到0.88, 均高于最大似然监督分类结果的精度。另外, 所提出的方法在大范围农作物信息提取方面具有一定的实用性和推广性。

More and more attention has been paid to the best use of medium and high resolution images and statistical data,combined with low resolution images on crop area estimation. However,information abstraction with high and medium resolution images also has many uncertainties due to factors such as spectral difference within classes, spectral similarity between classes,and the mixed pixels. This paper presents a method for crop area estimation with high and medium resolution images based on statistical sampling and amount controlling. Firstly, sample units are obtained by stratified sampling. Then sampling units are interpreted,and the estimator of crop planting acreage is extrapolated. Finally the spatial distribution mapping is classified and refined under the restriction derived from sampling estimator.Moreover,we validate the method presented above by using a SPOT-5 subset image (with resolution of 10m, August 21, 2006) of Sanhe, Hebei province. The results indicate that the overall accuracy of the new method is 93.8%, with kappa 0.88, based on cluster samples,which is higher than that of MLC method. The new method has promising practicality and popularity in large-cover measurement of crop planting acreage.