压缩感知光谱重构中的字典原子选取优化方法

发布时间：2018-05-12 01:24

本文选题：光谱学 + 光谱重构　；参考：《光学学报》2016年09期

【摘要】：针对常用的迭代追踪类算法难以保证低采样下光谱重构的成功率与精度的问题,提出了一种在低采样下光谱重构中字典原子选取的优化方法。利用AVIRIS和ROSIS高光谱数据构建光谱稀疏字典并进行压缩感知光谱重构实验,分别从光谱重构精度、稀疏成分提取能力、光谱重构的成功率和光谱识别的准确率等不同角度进行了分析。实验结果表明,本文方法不仅优于传统的匹配追踪算法,同时也优于公认的精度较高的FOCUSS、MSBL等其他类型的算法。
[Abstract]:Aiming at the problem that the common iterative tracing algorithms can not guarantee the success rate and precision of spectral reconstruction under low sampling, an optimization method for selecting dictionary atoms in spectral reconstruction under low sampling is proposed. Using AVIRIS and ROSIS hyperspectral data to construct spectral sparse dictionaries and to carry out experiments of compressed sensing spectral reconstruction, respectively, from spectral reconstruction accuracy, sparse component extraction ability. The success rate of spectral reconstruction and the accuracy of spectral recognition were analyzed. The experimental results show that this method is not only superior to the traditional matching tracking algorithm, but also superior to other algorithms such as FOCUSS MSBL, which has high accuracy.
【作者单位】：中国科学院光电研究院定量遥感信息重点实验室;中国科学院大学;
【基金】：国家863计划(2013AA12904) 中国科学院/国家外国专家局创新国际团队(2013AA1229)
【分类号】：TP751

【参考文献】