基于NDVI密度分割的冬小麦种植面积提取

为解决作物面积遥感监测中常遇的混合像元问题,选用江苏省沭阳县冬小麦扬花期HJ-1A卫星遥感影像,基于不同地物光谱信息的差异性与可分割性,提出基于归一化植被指数(NDVI)密度分割的冬小麦种植面积提取方法。在利用GPS实地取样调查和建立解译标志的基础上,对HJ-1A卫星影像进行了几何与大气校正。利用NDVI灰度影像提取混合像元训练样本的NDVI值和小麦种植面积,计算小麦面积权重,确定混合像元的NDVI阈值。利用NDVI再归一化结果对NDVI灰度影像进行密度分割,依据不同密度分割系数下像元总面积及其所对应的小麦面积权重关系,最终得到沭阳县冬小麦种植面积。结果表明,根据NDVI密度分割法提取冬小麦面积为8.37×104 hm2,面积精度为92.37%,样本精度为93.31%。基于密度分割系数(P0.5)制作沭阳县冬小麦种植分布图,获取了全县冬小麦空间分布特征信息。以上结果说明NDVI密度分割法能较准确地提取研究区内冬小麦种植面积,可有效解决农作物种植面积提取中混合像元问题。

Area of Winter Wheat Extracted on NDVI Density Slicing

With the advantages of quick, non destructive and large area to obtain, remote sensing technology has been widely used in the monitoring of crop acreage. In crop acreage monitoring, mixed pixel problem has been considered to be the key issues. Taking the Shuyang county of Jiangsu province as the study area, one HJ 1A image at the flowering stage of winter wheat was chosen. To extract the total crop acreage based on the differences and severability spectral information of different objects, this paper referenced Normalized Difference Vegetation Index ( NDVI ) density slicing method. Firstly, the GPS data and the interpretation marks were used to correct HJ 1A image. To determine the NDVI threshold of wheat mixed pixel, the image of NDVI were used to extract NDVI and the area of training quadrat pixel, and to calculate the weights of wheat area. Then, through NDVI renormalized, the gray image of NDVI was segmented. According to the relationship between area of pixel and weights of wheat area at different partition coefficient, the acreage of winter wheat of Shuyang County finally were extracted. The results showed that the wheat area in Shuyang County was 8.37×10 hm based on NDVI density slicing. The area accuracy was 92.37%, and the accuracy of samples was 93.31%. The distribution map of winter wheat planting area in Shuyang County were got by partition coefficient (P>0.5). Then, the information on the spatial distribution of winter wheat of Shuyang County was obtained. The method of NDVI density segmentation can extract the area of winter wheat accurately, and solve the problem of mixed pixel effectively in extraction. It can also provide technical support for extracting crop acreage information in the south fast and accurately.