利用最大似然法识别呼和浩特市绿化树种

Maximum Likelihood in Identifying Huhhot Greening Tree Species

以呼和浩特市的WorldView-Ⅱ为数据源,通过影像光谱特征分析,筛选出有利于树种识别的波段,利用最大似然法对影像进行分类。结果表明:11个树种及草类在海岸、蓝、绿、黄色波段中可分性不大,在红、红色边缘、近红外1、近红外2波段中可分性较大;11个树种及草类的整体分类精度可达62.93%,Kappa系数为0.5842,表明分类结果与实际情况具有中等的一致性;树种组分类中,分类的总体精度为72.59%,Kappa系数为0.6571,表明分类结果与影像真实情况具有高度的一致性。

With Huhhot WorldView- Ⅱ images as the data source, we analyzed the image spectrum feature and screened the band conducive to classification to classify the image by maximum likelihood. The separability of all the trees and grasses in coast, blue, green and yellow bands is not very obviously. In red, red edge, near infrared 1 and 2 bands, the separability is very obviously. The overall accuracy of tree species classification is about 62.93%, Kappa coefficient is 0.584 2, and the classification with the actual situation has medium consistency. The overall accuracy of tree species groups classification is 72.59%, Kaopa coefficient is 0.657 1, and the classification with the actual situation has high consistency.