基于邻近集计算的光谱相似性测度方法研究

发布时间：2018-05-29 15:08

本文选题：近红外光谱 + 邻域保持投影　；参考：《光谱学与光谱分析》2017年07期

【摘要】：在近红外光谱数据相似性测度时,由于光谱数据高维、非线性、重叠等特点,会出现测度距离失效、数据信息处理困难等难题。针对传统相似性测度方法在高维空间出现的不适应性,提出了基于邻近集计算的光谱相似性测度方法。首先,采用邻域保持投影neighborhood preserving projections(NPP)算法对原始光谱数据进行降维处理,该降维方法可以很好的保留原始光谱数据非线性结构信息和数据点的邻域信息。然后,在光谱数据降维后的低维空间中,采用改进的邻近集计算方法,实现对近红外光谱数据的相似性测度。实验结果表明,基于邻近集计算的光谱相似性测度方法,有效的实现了光谱数据的相似性测度,在烟叶风格判定和品质分析方面有较好的应用前景,同时也为高维光谱数据相似性测度提供了一个良好的解决方法。
[Abstract]:In the near infrared spectral data similarity measurement, due to the spectral data high dimensional, nonlinear, overlapping and other characteristics, there will be measurement distance failure, data information processing difficulties. Aiming at the inadaptability of traditional similarity measurement methods in high dimensional space, a spectral similarity measurement method based on neighborhood set computation is proposed. Firstly, the neighborhood preserving projection neighborhood preserving projects (NPP) algorithm is used to reduce the dimension of the original spectral data, which can preserve the nonlinear structural information of the original spectral data and the neighborhood information of the data points. Then, in the reduced dimension space of the spectral data, the similarity measurement of the near infrared spectral data is realized by using the improved neighborhood set calculation method. The experimental results show that the spectral similarity measurement method based on adjacent set calculation can effectively measure the similarity of spectral data and has a good application prospect in tobacco style determination and quality analysis. At the same time, it also provides a good solution to measure the similarity of high dimensional spectral data.
【作者单位】：中国海洋大学信息科学与工程学院;中国海洋大学信息工程中心;青岛农业大学网络管理中心;
【基金】：国家自然科学基金项目(41276085)资助
【分类号】：O657.33

【相似文献】