Colorimetric representation of laser-induced breakdown spectroscopy for classification based on CIE L*a*b* features

Laser-induced breakdown spectroscopy (LIBS) is a practical technique for elemental analysis, in which classification represents an important application. However, LIBS classification is often challenged by the high dimensionality and redundancy of spectral data, while the dimensionality reduction is generally based on statistical transformations with limited physical interpretability. In this study, the CIE L*a*b* color space is introduced to represent LIBS spectra from a colorimetric perspective, enabling spectral compression into low-dimensional features. Different from conventional approaches, this method establishes a direct link between plasma emission and human visual perception, thereby providing a physical interpretation for feature extraction. For evaluation, three sample sets (rare earth, Astragalus, and coal) are selected to cover typical spectral patterns of LIBS. In classification, LIBS spectra are transformed into CIE L*a*b* coordinates, and sample classification is performed using support vector machine (SVM), k-nearest neighbors (KNN), and linear discriminant analysis (LDA). The results demonstrate that the CIE L*a*b* method achieves high classification accuracy for samples with broadband spectral emissions (e.g., rare earth samples), whereas its performance degrades for spectra dominated by narrow-band emissions (e.g., coal samples). Furthermore, it is found that the CIE L*a*b*-based representation can maintain stable classification performance under varying detection conditions. This work provides a new perspective for LIBS feature extraction by bridging plasma emission with perceptual color representation, and the CIE L*a*b* method offers a physically interpretable approach for dimensionality reduction in LIBS classification.