Significance  Due to its advantages of low energy, high transmission, fingerprint spectroscopy, and safety, terahertz spectroscopy technology can be used to detect and identify single substances, mixtures, and hidden explosives. It is a promising detection technology in the field of security and has broad application prospects. However, owing to various factors inlcuding instruments, environment, and samples, the signal-to-noise ratio of terahertz spectral data decreases, making it difficult to obtain effective spectral characteristic peaks, which seriously affects the effectiveness and accuracy of peak finding comparison methods for identifying explosives. In order to improve the accuracy of terahertz spectral data, the application of machine learning technology in terahertz spectral analysis technology for detecting explosives was explored. It was pointed out that machine learning can further improve the accuracy of terahertz spectral technology detection, and develop more refined data processing techniques and more effective data augmentation techniques. Combining machine learning with traditional spectral analysis techniques is an important development direction in the field of terahertz spectral technology for detecting explosives.

Progress  Firstly, spectral technology is a commonly used method for substance detection. Based on the different absorption, reflection, and scattering of electromagnetic waves by explosives in different frequency bands, the main detection methods include Raman spectroscopy, terahertz technology, millimeter wave technology, X-ray technology, etc. Among these detection methods, terahertz technology is widely used in explosive detection due to its non invasiveness, high resolution of biological tissues, strong penetration, specific fingerprint spectra, and safety advantages. Next, the principle of terahertz spectroscopy technology and its research in explosive detection are introduced in detail. This article summarizes the research on terahertz spectroscopy technology in detecting single substances and their mixed explosives, detecting hidden explosives, and analyzing the characteristic absorption peaks of explosives. It is of great significance for establishing an explosive spectrum database and also promotes the application of terahertz spectroscopy technology in the field of security inspection. However, due to the influence of factors such as samples, environment, and detection instruments on the detection results, the ability and reliability of using characteristic peak comparison method to identify substances are greatly weakened. Machine learning technology has the characteristics of adaptability, efficiency, and accuracy, which can solve these problems to a certain extent. This emphasizes the important application of machine learning in terahertz spectroscopy technology.

Conclusions and Prospects  Terahertz spectroscopy technology has great application prospects in the field of explosive detection, and plays an important technical support role in explosive risk monitoring and prevention. However, during the detection process, factors such as instruments, environment, and sample status can affect the accuracy of identifying characteristic terahertz spectral peaks. With the continuous optimization of machine learning algorithms and the gradual expansion of datasets, machine learning will become an important means of detecting explosives in future terahertz spectroscopy technology. The future application of machine learning in terahertz spectral analysis has the following trends: Firstly, it is necessary to develop more sophisticated data processing techniques and more effective data augmentation techniques to solve the problem of large amounts of data; Secondly, machine learning needs to be combined with traditional spectral analysis to meet the requirements of fine interpretation, such as anomaly detection.