Hazardous Chemicals Detection and Classification Through Millimeter Wave and Machine Learning

Abstract

This paper demonstrates the effectiveness of integrating computational intelligence to enhance the reliability of millimeter wave technology as a detection device for hazardous chemicals. The research explores the use of millimeter wave as an efficient and dependable alternative technology for chemical detection with the aid of machine learning to further improve its reliability and accuracy. This advancement is crucial in enabling security agencies, and authorities to remotely identify hazardous chemicals, minimizing risks to human lives and properties. The millimeter wave relies on natural non-ionizing radiation, which is of low power and considered safe for human exposure. The millimeter wave region used in this study is 77–81 GHz that offers short-pulse transmission capabilities, producing a wide spectrum of frequencies. These short pulses serve as the source for collecting the broadband spectral identity of chemicals, and the subsequent detection is post-processed with machine learning to increase the level of accuracy. The result of this study shows that by using computational intelligence models such as decision tree, k-nearest neighbor, support vector machine, and random forest, enhances the overall device reliability, and achieves higher detection accuracy based on the received reflected power. This result is comparable to an X-ray system device.