基于全极化SAR数据散射机理的农作物分类

Crop classification based on scattering model using full-polarization SAR data

提出了利用极化相干矩阵对雷达遥感影像进行特征值分解,通过分析分解特征图达到农作物识别、分类的目的。该方法不仅能识别不同农作物,而且对水稻、玉米、大豆较好的识别精度。以江苏省睢宁地区农作物识别为例,利用2009年8月23日Radarsat-2全极化SAR影像数据,结合极化相干矩阵的方法提取不同农作物散射特征的3个参量,分析不同农作物散射机理;结合地面GPS数据进行Wishart监督分类和非监督分类。分类结果表明:城市及水体散射特征特点明显、识别清晰;水稻由于植株底部具有含水层,易与其它农作物区别,水稻分类精度达到97.92%;根据平均散射角的差异性对大豆和玉米进行了区分,最终总分类精度达到78.1%。研究结果表明,全极化雷达数据能提供更为丰富的地物散射信息,是农作物遥感监测的重要数据来源。

The results of eigenvalue decomposition based on the polarimetric coherency matrix using synthetic aperture radar（SAR） data were analyzed for the purposes of identification and classification of crops.This method can identify different crops and have better classification accuracy of rice,corn,and soybean.Taking Suining experimental area in Jiangsu Province as an example,using Radarsat-2 polarimetric SAR image data of August 23,2009,three parameters of scattering characteristics of different crops were extracted and the scattering mechanisms of these crops were analyzed.Using ground GPS data,the different crops were classified with Wishart supervised and unsupervised method.The results showed that the scattering characteristics of city and water were clear.Rice was easily distinguished from other crops for having water layer in plant bottom with the classification accuracy being 97.92%.Corn and soybean were respectively identified according to the difference of average scattering angle.The total accuracy of paddy identification was 78.1%.Overall,full-polarization SAR data can provide richer land scattering information which is an important data source of remote sensing monitoring of crops.