A Raman spectral image data preprocessing method based on low-rank tensor algorithm

Abstract

A Raman spectral image data preprocessing method based on a low-rank tensor algorithm, including: converting the collected Raman spectral image data into a third-order tensor form, and constructing a new tensor by using tensor decomposition and reconstruction methods Low-rank third-order tensor; obtain the low-rank approximation tensor of the original data tensor through iterative optimization algorithm and optimal approximation algorithm; build a database of noise distribution, and judge whether the noise component in the original data conforms to the noise distribution based on big data statistics; If so, the low-rank approximation tensor at this time is the best low-rank approximation tensor, and the best low-rank approximation tensor is the spectral information part in the original data, and a large amount of white noise and photon noise are removed, thereby Improved signal-to-noise ratio for Raman spectral image data. The method is simple and easy to implement, and does not need to modify the acquisition instrument. For the existing Raman spectral imaging technology, the signal acquisition time can be greatly reduced, and it will have broad prospects in the research of Raman spectral imaging technology.

Description

technical field [0001] The invention relates to the technical field of spectrum detection, in particular to a Raman spectrum image data preprocessing method based on a low-rank tensor algorithm. Background technique [0002] Raman spectroscopy imaging technology is a new development of Raman spectroscopy analysis technology, which effectively combines Raman spectroscopy technology with microscopic technology, and improves the spatial resolution of Raman measurement with the help of confocal micro Raman spectrometer and signal detection device To the micron scale, the single-point analysis method in the traditional Raman spectroscopy technology is expanded, and a comprehensive analysis is performed within a certain range, so that the spatial distribution of the physical and chemical properties of the sample is displayed in the form of an image. It has a wide range of applications in science and life sciences. [0003] At present, the noise processing methods for Raman image ...

AI Technical Summary

Problems solved by technology

[0003] At present, the noise processing method for Raman image data is relatively simple, limited to some conventional digital signal filtering methods, such as Fourier filtering and wavelet filtering

[0004] These conventional filtering methods cannot essentially improve the signal-to-noise ratio of the signal

Therefore, for signal acquisition, in order to ensure the validity of the data, the Raman acquisition process requires a long integration time, which makes the entire signal acquisition process very long, which greatly limits the application and development of Raman spectral imaging technology.

Images