CN-122016759-A - Raman-near infrared spectrum combined intelligent sensor and detection method

CN122016759ACN 122016759 ACN122016759 ACN 122016759ACN-122016759-A

Abstract

The invention discloses a Raman-near infrared spectrum combined intelligent sensor and a detection method, which relate to the technical field of spectrum intelligent sensing detection, wherein a constant current pulse driving module excites a Raman laser source to generate pulse laser, the Raman laser source is combined with a near infrared source to form Raman-near infrared mixed light, a MOEMS (metal oxide optical element) beam splitter module is used for splitting detection after wavelength dispersion, a dynamic gain adjustment method is used for dynamically amplifying two paths of electric signals, after cross-talk is removed through dual-band notch filtering, synchronous analog-to-digital conversion and wavelet transformation are performed to reduce noise, a characteristic peak space association mapping is established based on a molecular vibration-rotation coupling principle, raman spectrum comparison is combined to perform qualitative and correction partial least square quantification, trace components are identified through characteristic peak expansion, and accurate qualitative and quantitative detection of a sample is finally achieved.

Inventors

TIAN HAIBO
WANG YANG

Assignees

东虹芯光(上海)高科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260228

Claims (10)

1. The Raman-near infrared spectrum combined detection method is characterized by comprising the following steps of: acquiring a Raman signal and a near infrared signal output by an optical perception module; Carrying out dynamic amplification on two paths of electric signals by a dynamic gain adjustment method, carrying out filtering separation based on a dual-band filtering method to obtain pure Raman signals and pure near infrared signals, carrying out synchronous ADC (analog-to-digital conversion) on the pure Raman signals and the pure near infrared signals, and carrying out noise reduction by wavelet transformation to obtain Raman digital signals and near infrared digital signals; According to the principle of molecular vibration-rotation coupling, carrying out space association analysis on characteristic peaks of a Raman digital signal and a near infrared digital signal, establishing a mapping relation of space association of the Raman-near infrared characteristic peaks, determining sample component groups according to Raman characteristic peak comparison, calculating component content by adopting a partial least square method corrected based on the mapping relation, expanding the characteristic peaks through a characteristic peak half-width expansion algorithm, extracting sidelobe information to identify micro components, and iteratively confirming qualitative component groups and qualitative component content.
2. The method for combined detection of raman and near infrared spectra according to claim 1, wherein establishing a mapping relationship of spatial association of raman and near infrared characteristic peaks comprises: respectively performing baseline correction and peak shape normalization processing on the time-synchronous Raman digital signal and the near-infrared digital signal, and extracting a Raman characteristic peak parameter set and a near-infrared characteristic peak parameter set; According to the molecular vibration-rotation coupling principle, calculating a spatial position matching item and a relative intensity matching item of a Raman characteristic peak and a near infrared characteristic peak, multiplying to obtain a correlation degree, screening characteristic peak pairs meeting a correlation degree threshold, integrating correlation characteristic peak information, and establishing a Raman-near infrared characteristic peak spatial correlation mapping relation table.
3. A combined raman-near infrared spectrum detection method according to claim 1, wherein the dynamic gain adjustment amplifies two electrical signals, comprising: A double-channel self-adaptive gain amplification circuit is built, a sliding time window method is adopted for extracting pulse peak values and pulse effective values aiming at Raman signals, and a continuous sampling average algorithm is adopted for calculating direct current average values and fluctuation amplitude values aiming at near infrared signals; And setting an amplitude interval for the two paths of signals, adjusting gain according to fine steps, calibrating amplification precision, and outputting amplified Raman signals and near infrared signals.
4. A combined raman-near infrared spectrum detection method according to claim 1, wherein the filtering separation is performed by a dual band filtering method, comprising: performing frequency characteristic analysis on the amplified electric signals by adopting fast Fourier transform, and identifying characteristic frequency segments and cross-talk frequency segments of the Raman signals and the near infrared signals; The method comprises the steps of adopting double-secondary notch filtering and cascading an all-pass filter for carrying out phase compensation on a Raman signal, adopting Butterworth low-pass filtering and carrying out direct-current component compensation on a near-infrared signal, respectively carrying out low-pass smoothing filtering on two paths of signals, and outputting a pure Raman signal and a pure near-infrared signal without cross-talk.
5. A combined raman-near infrared spectrum detection method according to claim 1, wherein synchronizing ADC conversion and wavelet noise reduction comprises: After synchronous ADC conversion is carried out on the pure Raman signal and the pure near infrared signal by adopting a double-channel analog-to-digital converter, a time synchronous Raman original digital signal and a near infrared original digital signal are generated; And selecting an adaptive wavelet base and a decomposition layer number according to the two types of digital signals, calculating a noise standard deviation, processing each order of high-frequency subband coefficient through a soft threshold function, executing inverse wavelet transformation, and outputting a Raman digital signal and a near infrared digital signal after noise reduction.
6. The method for combined detection of raman and near infrared spectroscopy according to claim 1, wherein the identification of the micro-component comprises: After the composition group and the content of the sample are preliminarily determined, calculating the sum of the content of the components and comparing the sum with a component content threshold value; And expanding the half-width of the Raman and near-infrared characteristic peaks according to the multi-component spectrum superposition principle, scanning an expansion interval to extract side lobe signals, and determining the types of the micro-ingredients by comparing with a standard Raman database and combining with characteristic peak space correlation mapping verification.
7. A raman-near infrared spectrum combination detection method according to claim 1 or 2, wherein the component content calculation comprises: Extracting relevant near infrared characteristic peaks corresponding to the preliminary substances according to a Raman-near infrared characteristic peak space association mapping relation table to form a dedicated near infrared characteristic peak subset of each preliminary substance; And constructing an initial correlation model of the characteristic peak intensity and the component content by adopting a partial least square method, introducing the correlation degree as a correction coefficient to correct the initial correlation model, substituting the peak intensity of the near infrared characteristic peak subset into the corrected initial correlation model, and calculating to obtain the content of each preliminary substance.
8. The method for combined raman-near infrared spectroscopy detection according to claim 5, wherein the time synchronization of the raman raw digital signal and the near infrared raw digital signal comprises: Generating a synchronous sampling clock by adopting a double-channel analog-to-digital converter and a phase-locked loop, detecting the pulse rising edge of a Raman signal, and generating a synchronous trigger signal; And configuring sampling rates according to the characteristics of the two paths of signals, synchronously acquiring pure Raman signals and pure near infrared signals, and generating Raman original digital signals and near infrared original digital signals containing time stamps and amplitude sequences.
9. A combined raman-near infrared spectroscopy detection method according to claim 6 wherein initially determining a group of sample constituents comprises: the standard Raman database is called, the Raman characteristic peak parameter set is normalized, the position coincidence degree and the relative intensity similarity of the Raman characteristic peak and the standard Raman characteristic peak are calculated, the comprehensive matching degree is obtained through weighting, and candidate substances with the comprehensive matching degree meeting the matching degree threshold are screened; And verifying the existence of the associated near infrared characteristic peak corresponding to the candidate substance by combining with a Raman-near infrared characteristic peak space association mapping relation table, and determining the sample composition group.
10. The Raman-near infrared spectrum combined intelligent sensor is used for realizing the Raman-near infrared spectrum combined detection method according to any one of claims 1-9, and is characterized by comprising an optical sensing module, a dual-channel sensing unit, a signal processing unit and a data analysis unit: The optical perception module separates the Raman-near infrared mixed light by using the MOEMS beam splitting assembly according to wavelength dispersion, and the Raman signal and the near infrared signal are converted by the corresponding detector; the dual-channel sensing unit acquires a Raman signal and a near infrared signal output by the optical sensing module; The signal processing unit dynamically amplifies the two paths of electric signals through a dynamic gain adjustment method, performs filtering separation based on a dual-band filtering method to obtain pure Raman signals and pure near infrared signals, performs synchronous ADC (analog-to-digital conversion) on the pure Raman signals and the pure near infrared signals, and performs noise reduction through wavelet transformation to obtain Raman digital signals and near infrared digital signals; The data analysis unit performs space association analysis on characteristic peaks of the Raman digital signal and the near infrared digital signal according to a molecular vibration-rotation coupling principle, establishes a mapping relation of space association of the Raman characteristic peaks and the near infrared characteristic peaks, determines sample composition groups according to Raman characteristic peak comparison, calculates component content by adopting a partial least square method corrected based on the mapping relation, expands the characteristic peaks through a characteristic peak half-width expansion algorithm, extracts sidelobe information to identify micro components, and iteratively confirms qualitative composition groups and qualitative composition content.

Description

Raman-near infrared spectrum combined intelligent sensor and detection method Technical Field The invention relates to the technical field of spectrum intelligent sensing detection, in particular to a Raman-near infrared spectrum combined intelligent sensor and a detection method. Background In the technical field of spectrum intelligent sensing detection, although Raman spectrum and near infrared spectrum combined use can complement each other to realize qualitative and quantitative analysis of substances, an integrated intelligent sensor becomes a main development direction of spectrum detection, the technical bottleneck is that firstly, a signal processing unit of the sensor adopts a fixed gain amplifying circuit, characteristics of the transient, narrow pulse, large amplitude fluctuation and near infrared signal stability and low fluctuation of a Raman signal cannot be matched, pulse waveform distortion or unstable amplitude of a direct current signal is easily caused, a filtering module of the sensor is depended on an experience setting frequency section, cross crosstalk between two types of signals is inhibited inaccurately, crosstalk residues interfere spectral characteristic peak extraction to reduce sensor perception precision, secondly, a data analysis unit of the sensor lacks cross spectrum correlation verification logic, qualitative and quantitative analysis is carried out singly depending on Raman or near infrared spectrum comparison, characteristic peak overlapping or background interference to cause error of sample component groups, a quantitative model is not combined with correlation characteristics of the two types to correct, system error is difficult to identify characteristic peak content of the main component, and the characteristic peak is hard to mask, and the infrared spectrum sensor cannot meet the requirements of high-to meet the requirements of the combination of the traditional optical sensor, and the intelligent sensor has high-to develop the accuracy. Disclosure of Invention The invention aims to provide a Raman-near infrared spectrum combined intelligent sensor and a detection method, which achieve the purposes of restraining crosstalk and realizing efficient, accurate, qualitative and quantitative sample through MOEMS (metal oxide optical element array) accurate light splitting, differential signal optimization and cross-spectrum characteristic correlation analysis. The technical scheme for realizing the purpose of the invention is as follows: in one aspect, a raman-near infrared spectrum combined intelligent detection method comprises the following steps: acquiring a Raman signal and a near infrared signal output by an optical perception module; Carrying out dynamic amplification on two paths of electric signals by a dynamic gain adjustment method, carrying out filtering separation based on a dual-band filtering method to obtain pure Raman signals and pure near infrared signals, carrying out synchronous ADC (analog-to-digital conversion) on the pure Raman signals and the pure near infrared signals, and carrying out noise reduction by wavelet transformation to obtain Raman digital signals and near infrared digital signals; According to the principle of molecular vibration-rotation coupling, carrying out space association analysis on characteristic peaks of a Raman digital signal and a near infrared digital signal, establishing a mapping relation of space association of the Raman-near infrared characteristic peaks, determining sample composition groups according to Raman characteristic peak comparison, calculating component content by adopting a partial least square method corrected based on the mapping relation, expanding the characteristic peaks through a characteristic peak half-width expansion algorithm, extracting sidelobe information to identify micro components, and iteratively confirming qualitative composition groups and qualitative component content. Further, establishing a mapping relation of the space association of the Raman-near infrared characteristic peaks comprises the following steps: Preprocessing a time-synchronous Raman digital signal and a near-infrared digital signal, subtracting a fluorescent background by adopting a self-adaptive iterative re-weighting punishment least square method aiming at the Raman digital signal to eliminate baseline interference, smoothing by adopting a Gaussian function to weaken peak shape saw tooth distortion caused by random noise, correcting a baseline by adopting a moving average method aiming at the near-infrared digital signal, calculating regulation peak shape by combining Savitzky-Golay filtering to avoid characteristic peak baseline inclination caused by low-frequency fluctuation, respectively extracting characteristic peak parameters of two types of spectrums after preprocessing is completed, extracting wave numbers, peak intensities and half-widths of characteristic peaks from the Raman spectrums to form a Raman cha