超光谱技术结合化学计量法区分直液式走珠笔墨迹种类

Hyperspectral Technology Combined with Chemometrical Method for Distinction of Ink Type of Straight Liquid Ballpoint Pen

为了填补直液式走珠笔墨迹种类区分的空白,试验对市面上常见的29种直液式走珠笔墨迹样品进行了400~1000nm光谱区间的超光谱成像,将得到的光谱数据进行简单的平滑预处理后,按照其光谱走势进行了初步的鉴别分类,采用SPSS.25分析软件中的沃尔德系统聚类分析法进一步分类,并采用主成分分析法评价了聚类分析的分类结果。试验结果表明:按照650~1000nm光谱区间曲线上升趋势的差异,可将29个墨迹样品分为第Ⅰ大类和第Ⅱ大类,根据系统聚类分析的树状图和按集中计划表绘制的散点图能够进一步将这29种墨迹样品分为8小类,利用主成分分析法分别提取Ⅰ、Ⅱ大类光谱数据中7个和4个主成分,其中,前两个主成分累计方差贡献率分别达到了87%,97%,用其绘制主成分得分图,结果显示8个小类中的样品均能聚在一起,表明系统聚类效果良好。

In order to fill the gap in the distinction of the ink type of the straight liquid ballpoint pen,a hyperspectral imaging test in the spectral interval of 400-1000 nm was performed on the 29 common straight ballpoint ink samples from the market.The obtained spectral data was pretreated by simple smoothing,and a preliminary classification was carried out for the spectra according to their spectral trends.Further classification was made by using the Wald hierarchical clustering analytical method with SPSS.25 software,and its classification result was evaluated by using principal component analysis.It was found by results that the 29 ink samples were divided into the first category and the second category by the difference in the upward trend of the spectra at the spectra interval of 650-1000 nm.Based on the dendrogram and scatter charts drawn according to the centralized plan of hierarchical clustering analysis,the 29 ink samples were further divided into 8 subcategories.7 and 4 principal components of spectral data of the first and second categories were extracted by the principal component analysis,respectively,of which the cumulative variance contribution rate of the first two principal components reached 87%and 97%respectively.Score charts were drawn by values of the first two principal components,and it was shown that the samples in 8 subcategories could all be well clustered together,indicating that the effect of the hierarchinal clustering analysis was good.