基于导函数层序聚类的光谱成分预测方法研究(英文)

A New Color Component Prediction Method Based on First Derivative Clustering Analysis

多光谱成像技术通过增加颜色通道的维数,成功的实现了基于光谱的颜色复制。然而,由于其颜色信息维数较高,此方法在提高色度精度的同时引入了较大的计算及存储压力。为此,最常用的方法就是通过对光谱数据进行分组,并利用每组光谱数据集中的主成分向量来对各个光谱曲线进行线性表示,从而实现数据的降维处理。提出了1种新的针对光谱数据导函数曲线的聚类分析方法,并利用伪逆算法进行光谱重建;本研究采用孟塞尔光泽色卡及无光泽色卡作为实验数据集,并将提出的导函数聚类分析法与现有的主成分分析法、聚类分析法以及色相角分类法相比较,实验结果证明其颜色预测精度在色度匹配及光谱匹配方面均优于现有方法。

By extending the number of signal channels to high dimensions, the multispectral imaging technology fulfils the spectral color reproduction. However, it also brings out calculating and storing problem because of the large spectral database. The solution for this matter is to divide the database into several subgroups and find their basic components for each individual subgroup for linear representation[1] A new approach was proposed for grouping similar spectrum from spectral database. Based on the first derivative of each spectral curve, we used agglomerative hierarchical cluster analysis to divide the database into ten groups and found their basic components by principal component analysis. The pseudo inverse method was employed to reconstruct the spectral reflectance and the results are compared to three related methods. In the experiment, we grouped the 1250 matt Munsell color chips and the 1600 glossy Munsell color chips to compare the performance of different methods. The results proved that the prooosed method outperformed others and showed excellent spectral and colorimetric accuracy.