基于决策树理论的高光谱影像目标提取技术研究

Study of the Method of Determinedly Spectral Feature Matching Based on the Theory of Scale Space

利用决策树分级匹配分类的思想，完成高光谱遥感影像分类过程中的光谱特征优化、特征参量构造和相应的匹配策略选择。该方法结合专家知识和光谱特征分析结果，能够充分利用高光谱遥感数据中丰富且细致的光谱信息，灵活采用合适的匹配参量和匹配方法。试验结果表明，该方法能够显著提高高光谱遥感影像目标提取的精度。

By utilizing the thought of matching on different grades from decision tree, the paper completed the process of optimizing spectral features, constructing feature parameters and selecting corresponding matching strategies. This method combined the expertise with result of analyzing spectral features, could make the best use of the abundant and fine spectrum information in hyperspectral remote sensing data, and adopted appropriate feature parameters and matching strategies smartly. The experiment indicated that the technology could improve the precision during extracting the targets from hyperspectral images.