基于珠海一号高光谱数据的农作物精细化分类研究

Crop Refinement Classification Based on ZH-1 Hyperspectral Data

利用高光谱遥感影像识别精细化农作物类型,已经成为农业遥感的热点领域。本文以绥化市北林区内的水稻、玉米、大豆、烤烟等典型农作物为分类对象,以ZH-1高光谱数据为主要数据源,结合外业采集样本,探讨利用高光谱数据对精细化农作物类型信息提取的方法。针对高光谱数据的冗余问题,对其进行PAC降维处理,采用支持向量机分类方法对农作物种植结构进行提取。最后,采用混淆矩阵方法对分类结果进行验证,分类总体精度达到87.08%,Kappa系数为0.86。分类结果表明,高光谱数据在精细化农业分类研究上得到较好的应用。

Using hyperspectral remote sensing images to identify refined crop types has become a hot field of agricultural remote sensing.This article takes typical crops such as rice,corn,soybeans and flue-cured tobacco in Beilin district,Suihua city as classification objects,and uses ZH-1 hyperspectral data as the main data source.Combined with field sample collection,this article explores the method of using hyperspectral data to extract refined crop type information.Aiming at the redundancy problem of hyperspectral data,PAC dimensionality reduction was applied to it,and the crop planting structure was extracted by support vector machine(SVM)classification method.Finally,the confusion matrix method was used to verify the classification results,and the overall accuracy of classification was 87.08%,and Kappa coefficient was 0.86.The classification results showed that hyperspectral data had been well applied in the research of refined agricultural classification.