天宫一号高光谱数据尚未得到普遍应用,其数据的质量和应用潜力仍在进一步实践求证和挖掘.See5.0数据挖掘工具是一种能够找出训练样本中模式类隐含特征,并可以自动建立决策规则的分类算法,可避免人为建立分类规则的主观性.本文首先通过光谱曲线分析,选择地物光谱分离性最好的波段组合,然后利用See5.0工具生成规则集,再利用规则集对同一幅天宫一号高光谱数据在不同分类级别上进行分类,并利用相同的验证样本进行精度验证.经过光谱分析发现分类不同森林类型的最佳谱段中心波长分别为:655 nm、673 nm、802 nm、866 nm、984 nm,See5.0分类结果表明在同一树种不同生长期及不同亚种的分类级别上,分类精度在45%以下,表现出了一定局限性,但在树种分类级别上,天宫一号数据表现出了高光谱的优越性,分类精度皆在80%以上,植被类型分类级别,分类精度可达到90%以上.

Tiangong-1 hyperspectral data has not been widely used, its data quality and application potential still need to further proof and mining practice. See5.0 data mining tool is a classification algorithm that is able to find out the hidden features of model classes in the training sample and automatically establish decision rules, Avoiding the subjectivity of artificially build the decision tree. First, by analyzing the spectral curves of different objects, select the spectral separability best band combination to Hyperspectral data dimensionality reduction,then,using See5.0 tool to generate rule sets,finally,making the classification on the same hyperspectral data from Tiangong-1 in different classification level base on rule sets and validating the results respectively by the same validation samples .In result, finding that the best separability of different feature type appeared at 655 nm and 673 nm and 781 nm and 866 nm, 984 nm of Center Wavelength respectively by spectral curve analysis,result of classification show that at level of different growth period and different subspecies of same species, classification accuracy is below 45%, showing a certain limitations, But in species classification level, classification accuracy is above 80%, showing the superiority of hyperspectral data, at the level of vegetation classification, classification accuracy can reach more than 90%.