Rapid and Non-destructive Identification of Blue Ballpoint Pen Inks Based on Hyperspectral Imaging

This study integrates hyperspectral imaging technology with chemometric methods to establish a quick and non-destructive technique for identifying brands and models of blue ballpoint pen inks. Hyperspectral images were acquired from ink traces of nine different brands and models of blue ballpoint pens. Sixty regions of interest (ROIs) were selected per sample, extracting a total of 540 averaged spectral curves. Initially, three preprocessing methods-savitzky-golay (SG) smoothing, standard normal variate (SNV) transformation, and their combination-were compared to build classification models using support vector machine (SVM) and light gradient boosting machine (LightGBM). The optimal preprocessing method was determined based on model performance. Following this, feature wavelengths were extracted using successive projections algorithm (SPA), competitive adaptive reweighted sampling (CARS), uninformative variable elimination (UVE), and their combinations to establish the classification models. The results demonstrated that the (CARS + SPA)-SVM model achieved the highest classification accuracy of 92.66% for the nine different brands and models of ballpoint pen inks. This research highlights the potential of combining hyperspectral imaging technology with machine learning as an efficient and non-destructive approach to identifying types of ballpoint pen inks.