XAS Distortion Correction Algorithm of Illicit Drug by ResU-Net Model

The increasing use of illegal or recreational drugs has continually posed a threat to public health and safety. As a result, there is an urgent need for an effective technique to detect these substances, particularly in field settings for rapid on-site detection. X-ray Absorption Spectroscopy (XAS) is a spectral technique employed to qualitatively or quantitatively analyze the differences in X-ray absorption among substances. Its brief detection time and high accuracy render it well-suited for on-site detection of illicit drugs. However, traditional XAS detection algorithms analyze the feature envelope on each channel in its entirety, neglecting the temporal information and physical structural characteristics on different photon energy channels. Earlier methods for XAS substance detection have not utilized deep learning methods to correct spectral distortions. Expanding on this context, this paper introduces a novel approach for the detection of illicit drugs using X-ray Absorption Spectroscopy, proposing a new deep learning method for substance identification and spectral correction within XAS. This paper details a Residual Network-based U-Net spectral correction model, ResU-Net, that corrects distorted spectral data via deep learning, and it compares this model with conventional alternatives. Empirical testing and comparative evaluation, utilizing real-world data and performance metrics, demonstrate that the proposed ResU-Net spectral correction model surpasses alternative methods.