珠海一号高光谱影像在黄土高原大型煤矿区分类中的应用

The Application of Hyperspectral Images from Zhuhai-1 in Classification for the Large-scale Coal Fields on Loess Plateau

本文选取珠海一号影像数据,以哨兵二号数据为参考,将平朔矿区按权属划分五个子研究区,采用面向对象结合最邻近特征的分类方法,融合光谱特征、纹理特征、植被指数等特征因子,对珠海一号矿区土地利用分类精度进行评价。研究结果表明:(1)珠海一号分类总体精度为78.87%,Kappa系数为0.7285;哨兵二号分类总体精度为78.38%,Kappa系数为0.7203,珠海一号分类精度略高于哨兵二号;(2)部分耕地、草地、裸地产生误分,可能由于选取影像成像时间过晚地物特征相似导致。总体来说,珠海一号高光谱数据光谱及空间分辨率高,具有应用于矿区生态监测、复垦区植被演替特征等研究的潜力。

In this paper,the Zhuhai-1 image data is selected,and the Sentinel-2 data is used as a reference to divide the Pingshuo mining area into five sub-study areas according to ownership.The object-oriented classification method combined with the nearest features is adopted,and the spectral features,texture features,vegetation index,etc.are integrated.The characteristic factor is used to evaluate the accuracy of land use classification in Zhuhai-1 mining area.The research results show that:(1)The overall classification accuracy of Zhuhai-1 is 78.87%,and the Kappa coefficient is 0.7285;the overall classification accuracy of Sentinel-2 is 78.38%,and the Kappa coefficient is 0.7203.The classification accuracy of Zhuhai-1 is slightly higher than that of Sentinel-2.(2)The misclassification of some cultivated land,grassland and bare land may be caused by the similarity of the features of the ground objects when the imaging time is too late.In general,Zhuhai-1 hyperspectral data has high spectral and spatial resolution,and has the potential to be applied to research on ecological monitoring of mining areas and vegetation succession characteristics in reclamation areas.