Identification and classification of explosives using semi-supervised learning and laser-induced breakdown spectroscopy

Public places are often under threat from explosion events, which pose health and safety risks to the public. Therefore, the detection of explosive materials has become an important concern in the fields of antiterrorism and security. Laser-induced breakdown spectroscopy (LIBS) has been demonstrated to be useful in identifying explosives but has limitations. This study focuses on using semi-supervised learning combined with LIBS for explosive identification. Labeled data were utilized for the construction of a semi-supervised model for distinguishing explosive clusters and improving the accuracy of the K-nearest neighbor algorithm. The method requires only minimal prior information, and the time for obtaining a large amount of labeled data can be saved. The results of our investigation demonstrated that semi-supervised learning with LIBS can be used to discriminate explosives from interfering substances (plastics) containing similar components. The algorithm exhibits good robustness and practicability.