Discrimination of white automotive paint samples using ATR-FTIR and PLS-DA for forensic purposes

The consequences of a hit-and-run car crash are significant and may include serious injuries to the victims, health system overload and even victim's death. The vehicle and driver identification are often challenging for local law enforcement. The aim of this study was to develop a methodology to discriminate between automotive paint samples according to the make of the vehicle and its color shade. 143 white samples (collected at traffic accident scenes) were analyzed in situ by Fourier transform infrared spectroscopy with attenuated total reflectance (ATR-FTIR) and coupled microscopy. Principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were performed for data analysis. The samples were split into three groups: calibration set, validation set and external test set. The figures of merit were calculated to assess the quality of the model. Sensitivity, specificity, and efficiency rates were, respectively, 98,9%, 98.4% and 98.6%, for the calibration set. For the validation group, the classification accuracy was 100%. Correct classification rates for the internal validation set and external test set were 100% and 79.1% respectively. The technique is clean, fast, relatively low-cost, and non-destructive. Damaged regions of the samples were avoided by using the attached microscope. Limiting the age of the samples to a maximum of 10 years was enough to avoid misclassifications due to the natural degradation and weathering of the sample. Since the external test group is formed by underrepresented classes, its correct classification rate (79.1%) can be potentially improved at any time, by including and analyzing more samples.