面向对象的高光谱影像目标识别方法

Object-oriented target identification from hyperspectral remote sensing image

将面向对象影像分析和决策树方法综合应用于高光谱遥感影像的目标提取,在分析待提取地物特征的基础上,总结归纳其特性知识,构建分类规则,设计和实现决策树以完成目标识别。该方法充分应用分割所得到的均质对象的位置、形状、纹理等特征实现了决策规则提取和决策树构建,在应用OMIS和PHI影像进行目标识别的实践中取得了很好的效果。

Object-oriented image analysis is combined with decision tree to target identification from hyperspeetral remote sensing image. After analyzing the features of extracted target on hyperspectral remote sensing, those valid properties and knowledge are used to derive related rules, so the decision tree is generated for target identification based on those rules. Different features of homogenous objects such as location, spectra, shape and texture are fully used to construct decision tree and extract targets. The experimental results from OMIS and PHI images show the effectiveness of the proposed approach.