基于“资源一号”02D数据的植被提取效果对比研究

Comparative Study on Vegetation Extraction Effect Based on ZY-1 02D Data

基于“资源一号”02D卫星高光谱影像数据在可见光波段连续成像以及光谱信息丰富的特点,通过自适应波段选择的方法选取计算可见光植被指数的最佳波段,探究在经过波段优选后不同种可见光植被指数对植被的提取效果。利用Otsu’s阈值法对可见光植被指数计算结果中的植被覆盖区域进行提取,并利用同幅影像进行基于支持向量机的监督分类得到验证数据,建立混淆矩阵,利用提取结果与监督分类结果进行植被提取精度的量化评价。评价结果表明:文章研究选取的8种可见光植被指数中,以超红指数(EXR)的提取效果最佳,其制图精度为88.23%,用户精度为90.50%,总体精度为87.71%,Kappa系数为0.7482,像元错分漏分现象都处于较低水平,在研究区内提取植被的效果优于其他7种指数,然而超红指数无法正确区分研究区内的植被与水体,后续对其改进中应着重进行增大水体与植被区分度的研究。

Based on the characteristics of continuous imaging of hyperspectral image data of “resource-1”02D satellite in visible band and rich spectral information,the best band for calculating visible vegetation index is selected through adaptive band selection method,so as to explore the extraction effect of different visible vegetation indexes on vegetation after band optimization.Otsu’s threshold method is used to extract the vegetation coverage area in the calculation results of visible vegetation index,and the same image is used for supervised classification based on support vector machine to obtain the verification data,establish the confusion matrix,and quantitatively evaluate the accuracy of vegetation extraction by using the extraction results and supervised classification results.The evaluation results show that among the eight visible vegetation indexes selected in this paper,the extraction effect of ultra red index(EXR) is the best,with mapping accuracy of 88.23%,user accuracy of 90.50%,overall accuracy of 87.71%,kappa coefficient of 0.7482,pixel misclassification and missing points are at a low level,and the effect of extracting vegetation in the study area is better than the other seven indexes.However,the ultra red index can not correctly distinguish vegetation from water in the study area,In the subsequent improvement,we should focus on increasing the differentiation between water body and vegetation.