基于尺度空间理论的定性光谱特征匹配方法研究

Qualitative spectrum feature matching based on the theory of scale space

本文利用尺度空间理论对高光谱遥感数据中包含的精细光谱信息进行多尺度观察,在特定的尺度层次下提取能够对地物类别属性做出判断的定性约束特征,在此基础上结合光谱相似性测度最终确定像元所属类别。试验结果表明该方法可有效减少传统匹配算法由于噪声、成像环境等因素引起的误判、错分问题,提高分类识别的精度。

Through observing the fine spectral information which is included in hyperspectral remote sensing data with the theory of scale space, the paper extracted qualitative constraint features that under specific scale. These features can distinguish the attribute of object class. Combining with spectrum similarity measure, the pixel class can be determined. Test result indicated that this method can efficiently reduce the problems of miscarriage of justice from noise and environment of imaging and distinctly increase the precision of classification and discrimination.