Study on Background Subtraction of X-ray Spectrum by Improved Fourier Transform Based on Adaptive Threshold Judgment

In the field of nuclear analysis, background superposition affects the accuracy of qualitative and quantitative analysis. The research on background deduction of X-ray spectrum has always been an important step to optimize the spectral line data. Fourier transform [6] is one of the important methods of background deduction, but the threshold judgment of background deduction is involved in the calculation process, which needs to be calculated separately and the steps are cumbersome. Therefore, an adaptive hierarchical threshold judgment method based on a neural network model is proposed. Firstly, the data of experimental spectrum are regularized and preprocessed, data with low sparse characteristics are screened, seven obvious characteristic peaks are optimized, and 809 data with small weight are screened, and the accuracy is relatively improved by 79.00%. After preprocessing, the neural network model adaptive layered breakpoint algorithm [1] is used to judge the threshold. Finally, the iterative Fourier transform method is used to deduct the background of the X-ray spectrum, and the stability of the model is studied. It is concluded that the model can accurately deduct the spectral background of X-ray, and has strong stability and accuracy under the allowable error.