Classifying paint colour using acoustic data from laser ablation

We present an approach for identifying the paint colour and tone on a surface by using the acoustic signal collected during laser ablation. We trained convolutional neural networks to classify the colour (from 8 different colours) and the tone (the percentage of black in the paint). The colour was predicted with ∼91% accuracy and the tone with an R-value of 0.95. This technique has significant potential for supporting real-time optimisation in laser-material processing, particularly for high-precision laser cleaning, as well as broader applications where direct visual observation of the sample is not feasible.