摘要
为了去除高光谱影像的数据冗余,提高高光谱影像处理的精度和效率,提出了一种基于线性表示的高光谱影像波段选择算法。针对每一个波段,建立与其他波段的线性表示关系,依据复相关系数确定相关程度最高的波段,将其作为冗余波段去除;对剩余波段重复上述过程,得到最小波段集;并证明了利用该波段集和全波段所选的端元是一致的,在不影响端元提取的前提下,最大程度地去除了冗余波段。通过2组实验结果证明了该波段选择算法的可行性和有效性。
In order to remove the data redundancy of hyperspectral image and improve the accuracy and efficiency of hyperspectral image processing,this paper proposes a band selection method based on linear representation of hyperspectral image. A linear relationship is established for a band with the other bands,and the most relevant band is removed as a redundant band which is determined based on the multiple correlation coefficient. The set of minimum bands is finally obtained by repeating the above process for the remaining bands. It is proved that the set of selected endmembers by using the above bands is consistent with the set selected by using all bands,and the redundancy bands are removed to the greatest extent without affecting the endmember extraction. The experimental results show that the band selection algorithm in the paper is feasible and effective.
出处
《国土资源遥感》
CSCD
北大核心
2017年第4期39-42,共4页
Remote Sensing for Land & Resources
基金
国家自然科学基金项目"高分辨率遥感影像信息提取的特征结构化多尺度分析方法研究"(编号:41571346)
"基于高维马尔可夫网结构统计方法的高光谱图像分割研究"(编号:40971217)和"无限时滞脉冲泛函微分方程及其在经济中的应用"(编号:11201038)共同资助
关键词
高光谱影像
线性表示
波段选择
复相关系数算法
hyperspectral images
linear representation
band selection
multiple correlation coefficient algorithm