Transmission Infrared Microscopy and Machine Learning Applied to the Forensic Examination of Original Automotive...

Area of Science:

• Forensic Science
• Spectroscopy
• Machine Learning

Background:

• Automotive paint analysis is crucial in forensic investigations.
• Infrared (IR) spectroscopy is a key technique for identifying paint layers.
• Existing methods can be time-consuming and may face challenges with spectral variations.

Purpose of the Study:

• To enhance the accuracy and speed of forensic automotive paint examination.
• To validate the alternate least squares (ALS) procedure for spectral reconstruction of individual paint layers.
• To apply machine learning for classifying vehicle origin based on paint spectra.

Main Methods:

• Alternate least squares (ALS) reconstruction of IR spectra from cross-sectioned automotive paint samples.
• Analysis of spectra using an in-house library and IR microscope, addressing peak shifts with a correction algorithm.
• Application of machine learning algorithms for classification of vehicle manufacturer and assembly plant.

Main Results:

• ALS procedure validated for spectral reconstruction of individual paint layers.
• A correction algorithm successfully mitigated frequency shifts between different IR spectral collection methods.
• Machine learning correctly classified all twenty-six automotive paint samples by vehicle make, model, and matched them to the correct library sample.

Conclusions:

• The combined ALS reconstruction and machine learning approach significantly improves automotive paint analysis.
• This method offers a faster and more accurate alternative for forensic identification of vehicle paint.
• The developed technique has practical applications in identifying vehicle origin from paint evidence.