Fusion spectrum detection method and system based on near infrared spectrum and Raman spectrum

Abstract

The invention discloses a fusion spectrum detection method and system based on a near infrared spectrum and a Raman spectrum, and the method comprises the steps: obtaining a near infrared spectrogram and a Raman spectrogram of a to-be-detected sample, inputting the near infrared spectrogram and the Raman spectrogram into a preset spectrum recognition model, and determining first recognition data; when the first identification data cannot be acquired, characteristic wavelength extraction and spectrogram splicing are carried out through a plurality of preset characteristic wavelength optimization algorithms, and a first fusion characteristic spectrogram is determined; analyzing according to the first fusion characteristic spectrogram, and determining a second target component and a corresponding second characteristic spectrogram; analyzing according to the second target component and the corresponding second characteristic spectrogram, and determining second identification data; generating a mixed spectrogram processing scheme according to the second identification data and storing the mixed spectrogram processing scheme in a database; and visually displaying the first identification data and the second identification data through a preset terminal. According to the invention, characteristic wavelength extraction and spectrogram splicing are carried out through a plurality of preset characteristic wavelength optimization algorithms, so that the accuracy of spectrum detection can be effectively improved.

Description

Technical Field

[0001] The present application relates to the field of spectral detection technology, and more specifically, to a fusion spectral detection method and system based on near-infrared spectroscopy and Raman spectroscopy. Background Art

[0002] Against the backdrop of trade globalization and rapid economic development, global trade in hazardous chemicals has surged. However, traditional detection methods (such as chromatography and chemical analysis) are time-consuming, costly, and require complex pre-processing, making them inadequate for rapid customs clearance at ports of entry. Spectroscopic technology, due to its rapidity and non-destructive nature, has gradually become mainstream.

[0003] However, existing single spectral techniques have limitations. Raman spectroscopy is sensitive to fluorescence interference and has weak signal intensity, making it difficult to detect samples with low concentrations or high fluorescence backgrounds (such as some edible oils and biological tissues). Near-infrared spectroscopy is insufficiently sensitive to non-polar substances (such as certain fatty acids), and peak overlap affects quantitative accuracy.

[0004] Existing fusion technologies mostly use simple data splicing, which does not fully utilize the complementarity of the two spectra, resulting in high model redundancy and poor generalization ability.

[0005] Therefore, the prior art has defects and is in urgent need of improvement. Summary of the Invention

[0006] In view of the above problems, the purpose of the present invention is to provide a fusion spectrum detection method and system based on near-infrared spectroscopy and Raman spectroscopy, which can effectively improve the accuracy of spectrum detection.

[0007] The first aspect of the present invention provides a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy, comprising:

[0008] Obtaining near-infrared spectra and Raman spectra of the sample to be tested;

[0009] Inputting the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data;

[0010] When the first identification data cannot be obtained, characteristic wavelength extraction and spectrum splicing are performed on the near-infrared spectrum and the Raman spectrum respectively using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fused characteristic spectra;

[0011] Analyze the plurality of first fused characteristic spectra to determine a second target component and a corresponding second characteristic spectra;

[0012] Analyze the second target component and the corresponding second characteristic spectrum to determine second identification data;

[0013] generating a mixed spectrum processing solution according to the second identification data and storing the solution in a database;

[0014] The first identification data and the second identification data are visually displayed through a preset terminal.

[0015] In this solution, the inputting of the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine the first recognition data includes:

[0016] spectral noise reduction is performed on the near-infrared spectrum and the Raman spectrum respectively by a cyclic three-point zero-order SG filtering method, and a wide-area spectrum is determined by spectral data vector normalization and data parallel fusion;

[0017] Extracting key feature data of the wide-area spectrum graph by PCA-LDA feature dimensionality reduction method;

[0018] The wide-area spectrum is compared with a standard characteristic spectrum of a preset target component according to the key characteristic data to determine first identification data.

[0019] In this solution, the near-infrared spectrum and the Raman spectrum are respectively subjected to characteristic wavelength extraction and spectrum splicing by using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fused characteristic spectra, including:

[0020] Extracting characteristic wavelengths from the near-infrared spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of first spectra;

[0021] Extracting characteristic wavelengths from the Raman spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of second spectra;

[0022] The first spectrum graph and the second spectrum graph processed by different preset characteristic wavelength optimization algorithms are spectrally fused to obtain multiple first fused characteristic spectra.

[0023] In this solution, the analysis based on the multiple first fused characteristic spectra to determine the second target component and the corresponding second characteristic spectra includes:

[0024] Determine the preset target components based on the testing requirements;

[0025] When all characteristic peaks of the preset target component exist in the first fused characteristic spectrum, and the peak value of each characteristic peak is greater than the corresponding preset peak threshold, the preset target component is determined as the first target component, and the first fused characteristic spectrum is determined as the first characteristic spectrum;

[0026] Calculating a first curve similarity between the first characteristic spectrum and a standard characteristic spectrum of the first target component;

[0027] When the first curve similarity is greater than a first preset curve similarity threshold, the first target component is determined as a second target component, and the first characteristic spectrum is determined as a second characteristic spectrum.

[0028] In this solution, the analysis based on the second target component and the corresponding second characteristic spectrum to determine the second identification data includes:

[0029] Count the number of occurrences of the second target component;

[0030] If the number of occurrences of the second target component is greater than a preset occurrence threshold, determining the second target component as the second identification data;

[0031] On the contrary, from the first characteristic spectrum and the second characteristic spectrum corresponding to the second target component, the characteristic peak image having the smallest peak fluctuation value with respect to each characteristic peak in the standard characteristic spectrum of the second target component is sequentially selected for spectrum splicing to determine a spliced characteristic spectrum;

[0032] Calculating a second curve similarity between the spliced characteristic spectrum and the standard characteristic spectrum of the second target component;

[0033] When the second curve similarity is greater than a second preset curve similarity threshold, the second target component is determined as second recognition data.

[0034] In this solution, generating a mixed spectrum processing solution based on the second identification data and storing it in a database includes:

[0035] Determining, based on the second identification data, the mixing weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks, and generating a mixed spectrum processing solution;

[0036] Binding the mixed spectrum processing scheme and the key characteristic peak of the second target component corresponding to the second identification data and storing them in a database;

[0037] During the subsequent detection process, if all the key characteristic peaks of the second target component exist in the near-infrared spectrum and Raman spectrum of the detection sample, the corresponding hybrid spectrum processing scheme is called from the database to perform spectrum fusion processing on the near-infrared spectrum and Raman spectrum of the detection sample.

[0038] In this solution, determining the mixed weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks according to the second identification data to generate a mixed spectrum processing solution includes:

[0039] When the number N of second characteristic spectra corresponding to the second identification data is greater than the preset number of spectra, each characteristic peak of the second identification data is analyzed in order of wavelength interval from small to large, and the characteristic peak a is calculated based on the standard characteristic spectrum in the second characteristic spectrum F. n The peak fluctuation value x a-n , combined with the number of times y the first preset characteristic wavelength optimization algorithm i is used in all second characteristic spectra of characteristic peak a a-i Determine the mixing weight k of the first preset characteristic wavelength optimization algorithm i in the wavelength range corresponding to the characteristic peak a i-a ;

[0040]

[0041] Wherein, k1 is the influence weight of the peak fluctuation value, k2 is the influence weight of the number of times the preset characteristic wavelength optimization algorithm is used, k1+k2=1;

[0042] When the number N of second characteristic spectra corresponding to the second identification data is less than or equal to the preset number of spectra, a second preset characteristic wavelength optimization algorithm j corresponding to the minimum peak fluctuation value of the characteristic peak a is retrieved from all the second characteristic spectra;

[0043] The second preset characteristic wavelength optimization algorithm, the first preset characteristic wavelength optimization algorithm and the corresponding mixing weights of the wavelength interval corresponding to each characteristic peak are integrated to determine a mixed spectrum processing scheme.

[0044] In this solution, if all key characteristic peaks of the second target component exist in the near-infrared spectrum and Raman spectrum of the test sample, the corresponding hybrid spectrum processing solution is called from the database to perform spectrum fusion processing on the near-infrared spectrum and Raman spectrum of the test sample, including:

[0045] The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the first preset characteristic wavelength optimization algorithm i to determine the first peak value H of the characteristic peak a. 1(i-a) ;

[0046] Multiply all the first peak values of the characteristic peak a by the corresponding mixing weights, accumulate the calculation results, and determine the second peak value H of the characteristic peak a. 2(a) ;

[0047] The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the second preset characteristic wavelength optimization algorithm j to determine the third peak value H of the characteristic peak a. 3(a) ;

[0048] The second peak H of all characteristic peaks 2(a) and the third peak H 3(a) Performing integration to determine a second fusion feature spectrum;

[0049] Calculating a second curve similarity between the second fused characteristic spectrum and the standard characteristic spectrum of the second target component;

[0050] When the second curve similarity is greater than a third preset curve similarity, the second target component is determined as second recognition data.

[0051] A second aspect of the present invention provides a fusion spectrum detection system based on near-infrared spectroscopy and Raman spectroscopy, comprising:

[0052] A data acquisition module is used to obtain near-infrared spectra and Raman spectra of the sample to be tested;

[0053] a model analysis module, configured to input the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data;

[0054] a fusion spectrum splicing module, configured to, when the first identification data cannot be obtained, extract characteristic wavelengths and splice spectra of the near-infrared spectrum and the Raman spectrum using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fusion characteristic spectra;

[0055] a fusion spectrum analysis module, configured to analyze the plurality of first fusion feature spectra to determine a second target component and a corresponding second feature spectrum; and to analyze the second target component and the corresponding second feature spectrum to determine second recognition data;

[0056] a mixed spectrum processing solution generating module, configured to generate a mixed spectrum processing solution according to the second recognition data and store the solution in a database;

[0057] The visual display module is used to visually display the first identification data and the second identification data through a preset terminal.

[0058] A third aspect of the present invention provides a computer-readable storage medium, which includes a program for a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy. When the program for a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy is executed by a processor, the steps of the above-mentioned fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy are implemented.

[0059] The present invention discloses a fusion spectrum detection method and system based on near-infrared spectroscopy and Raman spectroscopy. The method includes: obtaining a near-infrared spectrum and a Raman spectrum of a sample to be tested and inputting them into a preset spectrum recognition model to determine first recognition data; when the first recognition data cannot be obtained, extracting characteristic wavelengths and splicing spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a first fused characteristic spectrum; analyzing the first fused characteristic spectrum to determine a second target component and a corresponding second characteristic spectrum; analyzing the second target component and the corresponding second characteristic spectrum to determine second recognition data; generating a mixed spectrum processing scheme based on the second recognition data and storing it in a database; and visually displaying the first recognition data and the second recognition data through a preset terminal. The present invention uses a plurality of preset characteristic wavelength optimization algorithms to extract characteristic wavelengths and splice spectra, which can effectively improve the accuracy of spectral detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 The present invention provides a flow chart of a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy;

[0061] Figure 2 A flow chart showing a first identification data determination method provided by the present invention is shown;

[0062] Figure 3 The flowchart of the first fusion feature spectrum fusion method provided by the present invention is shown;

[0063] Figure 4 The block diagram of a fusion spectrum detection system based on near-infrared spectroscopy and Raman spectroscopy provided by the present invention is shown. DETAILED DESCRIPTION

[0064] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that, in the absence of conflict, the embodiments of the present application and the features therein can be combined with each other.

[0065] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0066] Figure 1 The flowchart of the present invention shows a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy.

[0067] like Figure 1 As shown, the present invention discloses a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy, comprising:

[0068] S102, obtaining a near-infrared spectrum and a Raman spectrum of the sample to be tested;

[0069] S104, inputting the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data;

[0070] S106, when the first identification data cannot be obtained, extracting characteristic wavelengths and splicing spectra of the near-infrared spectrum and the Raman spectrum using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fused characteristic spectra;

[0071] S108, analyzing the plurality of first fused characteristic spectra to determine a second target component and a corresponding second characteristic spectra;

[0072] S110, analyzing the second target component and the corresponding second characteristic spectrum to determine second identification data;

[0073] S112, generating a mixed spectrum processing solution based on the second recognition data and storing it in a database;

[0074] S114: Visually display the first identification data and the second identification data through a preset terminal.

[0075] According to an embodiment of the present invention, the same spectrometer is preferably used to collect signals from both near-infrared and Raman spectra. Based on the characteristic that the laser used for near-infrared spectroscopy can generate multi-wavelength laser light required for Raman spectroscopy excitation, a wavelength-tunable pulsed Nd:YAG laser is selected as a shared light source for both near-infrared and Raman spectroscopy. Based on the temporal characteristics of Raman and near-infrared spectroscopy excitation, an enhanced charge-coupled device (ICCD) with time-resolved function is selected as a shared detection and collection optical path. Zemax optical path simulation and analysis software is used to study and establish a spectral response model for near-infrared and Raman spectroscopy under the same spectrometer, and an echelle grating spectrometer system is designed. By broadening the response range of the spectrometer, signals from both near-infrared and Raman spectra can be collected in the same spectrometer. By sharing the light source, the collection and detection optical paths, and the broadening of the spectrometer's response range, the equipment can be miniaturized and portable.

[0076] The optical path for collecting near-infrared and Raman spectra is designed to be located at the same location, with near-infrared and Raman spectra emitted and collected in separate time sequences. Raman excitation light is irradiated onto the sample surface via a dichroic mirror. Raman scattered light is focused by a collection lens group, passes through a pinhole, and is focused by a coupling lens and a long-pass filter onto a slit before entering the Raman spectrometer. A fixed grating is used for spectrometry to improve vibration resistance and stability, while a CT symmetrical structure design achieves compactness and high resolution. Near-infrared diffuse reflected light is focused onto a slit by a collection lens group, a perforated reflector, and a coupling lens group before entering the near-infrared spectrometer. A linear gradient filter is coupled to an InGaAs linear array detector to improve stability.

[0077] Among them, the preset spectral recognition model is obtained by training historical detection data, and the historical detection data includes historical near-infrared spectra and historical Raman spectra of each sample, as well as corresponding detection and recognition data.

[0078] After the infrared spectrum and Raman spectrum are input into the preset spectral recognition model, the model performs spectral denoising on the near-infrared spectrum and Raman spectrum through the cyclic three-point zero-order SG filtering method, extracts key feature data through the PCA-LDA feature dimensionality reduction method, and compares it with the standard feature spectrum of the preset target component stored in the database to determine the first recognition data.

[0079] However, due to the influence of data errors, it is impossible to obtain all the characteristic data of infrared spectra and Raman spectra only through the cyclic three-point zero-order SG filtering method and PCA-LDA feature dimensionality reduction method, and it may not be possible to accurately identify the target components in the sample to be tested. That is, it may not be possible to accurately extract the characteristic wavelengths of all components through any single characteristic wavelength extraction method.

[0080] The system uses multiple preset characteristic wavelength optimization algorithms to extract characteristic wavelengths from infrared spectra and Raman spectra, respectively. The first and second spectra processed by different preset characteristic wavelength optimization algorithms are spectrally fused to obtain multiple first fused characteristic spectra. Each first fused characteristic spectrum is analyzed in turn. When all characteristic peaks of a preset target component are present in the spectrum, the preset target component is determined to be the first target component. The similarity between the first curve of the standard characteristic spectrum of the first target component and the first fused characteristic spectrum is calculated. When the first curve similarity is greater than the first preset curve similarity threshold, the first target component is determined to be the second target component, and the first characteristic spectrum is determined to be the second characteristic spectrum. The second target component is verified based on the number of occurrences of the second target component to determine the second identification data. The first identification data and the second identification data are visualized through a preset terminal such as a display or mobile phone.

[0081] At the same time, the second identification data can be analyzed to determine the preset characteristic wavelength optimization algorithm and the corresponding mixing weight used in the wavelength range corresponding to each characteristic peak of the second target component. By using one or more preset characteristic wavelength optimization algorithms, the infrared spectrum and Raman spectrum of the detection sample are processed one by one, so that the peak values of all characteristic peaks can be obtained from the processed spectrum, and all target components that may exist in the sample to be tested can be determined.

[0082] Figure 2 A flow chart of a first identification data determination method provided by the present invention is shown.

[0083] like Figure 2 As shown, according to an embodiment of the present invention, the near-infrared spectrum and the Raman spectrum are input into a preset spectrum recognition model to determine the first recognition data, including:

[0084] S202, performing spectral noise reduction on the near-infrared spectrum and the Raman spectrum respectively by a cyclic three-point zero-order SG filtering method, and determining a wide-area spectrum by spectral data vector normalization and data parallel fusion;

[0085] S204, extracting key feature data of the wide-area spectrum using a PCA-LDA feature dimensionality reduction method;

[0086] S206 , comparing the wide-area spectrum with a standard characteristic spectrum of a preset target component according to the key characteristic data to determine first recognition data.

[0087] It should be noted that the cyclic three-point zero-order SG filtering method is used for dual-spectrum noise reduction, and the spectral baseline is corrected by wavelet analysis technology to solve the problem that the dual-spectrum absorption intensity is weak and the spectral signal is easily weakened by noise and baseline interference. The wide-area spectrum is obtained by using spectral data vector normalization and data parallel fusion method, and then the PCA-LDA feature dimensionality reduction method is used to extract key feature data to solve the problem that the signal intensity of Raman and near-infrared spectral data is quite different and there are too many feature data. The preset spectral recognition model is a back propagation neural network algorithm. By adjusting and optimizing parameters such as the input layer, hidden layer and error threshold, the wide-area spectrum is compared with the standard feature spectrum of the preset target component to determine the component of the sample to be tested and output the first recognition data.

[0088] Among them, the types of preset target components are set by technical personnel in this field according to actual needs, and the standard characteristic spectrum corresponding to each preset target component is obtained through sample testing, network query, etc.

[0089] Figure 3 A flow chart of the first fusion feature spectrum fusion method provided by the present invention is shown.

[0090] like Figure 3 As shown, according to an embodiment of the present invention, characteristic wavelength extraction and spectrum splicing are performed on the near-infrared spectrum and the Raman spectrum respectively through multiple preset characteristic wavelength optimization algorithms to determine multiple first fused characteristic spectra, including:

[0091] S302, extracting characteristic wavelengths from the near-infrared spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of first spectra;

[0092] S304, extracting characteristic wavelengths from the Raman spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of second spectra;

[0093] S306 , performing spectrum fusion on the first spectrum and the second spectrum processed by different preset characteristic wavelength optimization algorithms to obtain a plurality of first fused characteristic spectra.

[0094] It should be noted that the preset characteristic wavelength optimization algorithms are set by the system, including BPSO, GA, CARS, and SPA. The system uses a plurality of preset characteristic wavelength optimization algorithms to extract characteristic wavelengths from the acquired near-infrared spectrum and Raman spectrum, and performs spectrum fusion on the extracted first and second spectrum according to wavelength, thereby obtaining multiple sets of fused spectra of near-infrared spectra and Raman spectra processed by different preset characteristic wavelength optimization algorithms.

[0095] According to an embodiment of the present invention, analyzing the plurality of first fused characteristic spectra to determine the second target component and the corresponding second characteristic spectra includes:

[0096] Determine the preset target components based on the testing requirements;

[0097] When all characteristic peaks of the preset target component exist in the first fused characteristic spectrum, and the peak value of each characteristic peak is greater than the corresponding preset peak threshold, the preset target component is determined as the first target component, and the first fused characteristic spectrum is determined as the first characteristic spectrum;

[0098] Calculating a first curve similarity between the first characteristic spectrum and a standard characteristic spectrum of the first target component;

[0099] When the first curve similarity is greater than a first preset curve similarity threshold, the first target component is determined as the second target component, and the first characteristic spectrum is determined as the second characteristic spectrum.

[0100] It should be noted that, according to the user's detection requirements for the sample to be tested, the corresponding preset target component is selected, the wavelength range corresponding to each characteristic peak in the standard sample spectrum of the preset target component is determined, and the first fusion characteristic spectrum is analyzed based on the wavelength range. When there are peaks in all wavelength ranges, the peak values of each characteristic peak are compared with the corresponding preset peak threshold values to determine whether the preset target component exists in the sample to be tested, and the preset target component present in the sample to be tested is determined as the first target component. By comparing the first curve similarity between the first characteristic spectrum and the standard characteristic spectrum of the first target component, it can be determined whether there are other target components in addition to the first target component in the sample to be tested. When the first curve similarity is greater than the first preset curve similarity threshold, it is determined that the first target component in the sample to be tested is the main component and is determined as the second target component; otherwise, it is determined that the sample to be tested is composed of multiple target components.

[0101] Among them, the preset peak threshold of each characteristic peak and the first preset curve similarity threshold are set by technical personnel in this field according to actual needs. The preset peak threshold of each characteristic peak can be set with reference to the peak size of each characteristic peak in the preset target component standard characteristic spectrum. The larger the peak value of the characteristic peak in the standard characteristic spectrum, the larger the corresponding preset peak threshold.

[0102] According to an embodiment of the present invention, analyzing the second target component and the corresponding second characteristic spectrum to determine the second identification data includes:

[0103] Count the number of occurrences of the second target component;

[0104] If the number of occurrences of the second target component is greater than a preset occurrence threshold, determining the second target component as the second identification data;

[0105] On the contrary, the characteristic peak images having the smallest peak fluctuation values with respect to each characteristic peak in the standard characteristic spectrum of the second target component are sequentially selected from the first characteristic spectrum and the second characteristic spectrum corresponding to the second target component to perform spectrum splicing to determine a spliced characteristic spectrum;

[0106] Calculate the second curve similarity between the spliced characteristic spectrum and the standard characteristic spectrum of the second target component;

[0107] When the second curve similarity is greater than a second preset curve similarity threshold, the second target component is determined as the second recognition data.

[0108] It should be noted that after the analysis of all first fused characteristic spectra is completed, the number of occurrences of each second target component is counted, and those with a number of occurrences greater than a preset occurrence threshold are directly determined as the second identification data. Conversely, the peak fluctuation value of each characteristic peak in the corresponding first characteristic spectrum and the second characteristic spectrum is calculated, and the characteristic peak image with the smallest peak fluctuation value of each characteristic peak is selected for spectrum splicing to determine the spliced characteristic spectrum. By calculating the second curve similarity between the spliced characteristic spectrum and the standard characteristic spectrum of the second target component, it is determined whether the second target component is determined as the second identification data.

[0109] The preset occurrence threshold and the second preset curve similarity threshold are both set by those skilled in the art according to actual needs.

[0110] According to an embodiment of the present invention, generating a mixed spectrum processing solution based on the second identification data and storing it in a database includes:

[0111] Determining, based on the second identification data, the mixing weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks, and generating a mixed spectrum processing solution;

[0112] Binding the mixed spectrum processing solution and the key characteristic peak of the second target component corresponding to the second identification data and storing them in a database;

[0113] During the subsequent detection process, if all the key characteristic peaks of the second target component exist in the near-infrared spectrum and Raman spectrum of the test sample, the corresponding hybrid spectrum processing scheme is called from the database to perform spectrum fusion processing on the near-infrared spectrum and Raman spectrum of the test sample.

[0114] It should be noted that after the second identification data is determined, a corresponding mixed spectrum processing scheme can be generated based on the second identification data, and the mixed spectrum processing scheme can be stored in the database. When all the key characteristic peaks of the second target component are detected in the near-infrared spectrum and Raman spectrum of the test sample, the corresponding mixed spectrum processing scheme can be directly called to perform spectrum fusion, and there is no need to perform spectrum fusion on the dual spectrum through multiple preset characteristic wavelength optimization algorithms.

[0115] According to an embodiment of the present invention, determining the mixing weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks based on the second identification data, and generating a mixed spectrum processing solution, includes:

[0116] When the number N of second characteristic spectra corresponding to the second identification data is greater than the preset number of spectra, each characteristic peak of the second identification data is analyzed in order from small to large wavelength range, and the characteristic peak a is calculated based on the standard characteristic spectrum in the second characteristic spectrum F. n The peak fluctuation value x a-n, combined with the number of times y the first preset characteristic wavelength optimization algorithm i is used in all second characteristic spectra of characteristic peak a a-i Determine the mixing weight k of the first preset characteristic wavelength optimization algorithm i in the wavelength range corresponding to the characteristic peak a i-a ;

[0117]

[0118] Wherein, k1 is the influence weight of the peak fluctuation value, k2 is the influence weight of the number of times the preset characteristic wavelength optimization algorithm is used, k1+k2=1;

[0119] When the number N of second characteristic spectra corresponding to the second identification data is less than or equal to the preset number of spectra, a second preset characteristic wavelength optimization algorithm j corresponding to the minimum peak fluctuation value of the characteristic peak a is retrieved from all the second characteristic spectra;

[0120] The second preset characteristic wavelength optimization algorithm, the first preset characteristic wavelength optimization algorithm and the corresponding mixing weights of the wavelength interval corresponding to each characteristic peak are integrated to determine a mixed spectrum processing scheme.

[0121] It should be noted that the characteristic peak a in the second characteristic spectrum F n The peak fluctuation value x a-n From the characteristic peak a in the second characteristic spectrum F n The ratio of the peak value of the characteristic peak a to the peak value of the corresponding standard characteristic spectrum is determined by combining the number of times y that the first preset characteristic wavelength optimization algorithm i is used in all the second characteristic spectra. a-i , calculate the mixing weight k of the first preset characteristic wavelength optimization algorithm i for the wavelength interval corresponding to the characteristic peak a by the mixing weight calculation formula of the preset characteristic wavelength optimization algorithm i-a .

[0122] The number of preset spectra is set by those skilled in the art according to actual needs.

[0123] According to an embodiment of the present invention, if all key characteristic peaks of the second target component exist in the near-infrared spectrum and Raman spectrum of the test sample, a corresponding hybrid spectrum processing scheme is called from the database to perform spectrum fusion processing on the near-infrared spectrum and Raman spectrum of the test sample, including:

[0124] The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the first preset characteristic wavelength optimization algorithm i to determine the first peak value H of the characteristic peak a. 1(i-a) ;

[0125] Multiply all the first peak values of the characteristic peak a by the corresponding mixing weights, accumulate the calculation results, and determine the second peak value H of the characteristic peak a. 2(a) ;

[0126] The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the second preset characteristic wavelength optimization algorithm j to determine the third peak value H of the characteristic peak a. 3(a) ;

[0127] The second peak H of all characteristic peaks 2(a) and the third peak H 3(a) Performing integration to determine a second fusion feature spectrum;

[0128] Calculating a second curve similarity between the second fused characteristic spectrum and the standard characteristic spectrum of the second target component;

[0129] When the second curve similarity is greater than the third preset curve similarity, the second target component is determined as the second recognition data.

[0130] It should be noted that when all the key characteristic peaks of the second target component are present in the near-infrared spectrum and Raman spectrum of the test sample, the second peak value or the third peak value of each characteristic peak is calculated respectively by the first preset characteristic wavelength optimization algorithm i and the second preset characteristic wavelength optimization algorithm j in the hybrid spectrum processing scheme, and the spectral curve corresponding to each characteristic peak is adjusted to determine the second fused characteristic spectrum. When the similarity of the second curve is greater than the similarity of the third preset curve, the second target component is directly determined as the second identification data. There is no need to perform spectrum fusion on the dual spectrum through multiple preset characteristic wavelength optimization algorithms to identify whether the second target component is present in the sample to be tested.

[0131] The third preset curve similarity is set by those skilled in the art according to actual needs.

[0132] Figure 4 The block diagram of a fusion spectrum detection system based on near-infrared spectroscopy and Raman spectroscopy provided by the present invention is shown.

[0133] like Figure 4 As shown, the second aspect of the present invention provides a fusion spectrum detection system based on near-infrared spectroscopy and Raman spectroscopy, comprising:

[0134] A data acquisition module is used to obtain near-infrared spectra and Raman spectra of the sample to be tested;

[0135] A model analysis module is used to input the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data;

[0136] A fusion spectrum splicing module is used to extract characteristic wavelengths and splice spectra of the near-infrared spectrum and the Raman spectrum respectively using a plurality of preset characteristic wavelength optimization algorithms when the first identification data cannot be obtained, so as to determine a plurality of first fusion characteristic spectra;

[0137] a fusion spectrum analysis module, configured to analyze the plurality of first fusion characteristic spectra to determine a second target component and a corresponding second characteristic spectrum; and to analyze the second target component and the corresponding second characteristic spectrum to determine second recognition data;

[0138] A mixed spectrum processing solution generating module, configured to generate a mixed spectrum processing solution according to the second recognition data and store the solution in a database;

[0139] The visual display module is used to visually display the first identification data and the second identification data through a preset terminal.

[0140] A third aspect of the present invention provides a computer-readable storage medium, which includes a program for a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy. When the program for a fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy is executed by a processor, the steps of the above-mentioned fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy are implemented.

[0141] The information involved in this application (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals (including but not limited to signals transmitted between user terminals and other devices, etc.) are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions. For example, the "near-infrared spectrum and Raman spectrum of the sample to be tested" and the "preset target components" involved in this disclosure are all obtained with full authorization.

[0142] The present invention discloses a fusion spectrum detection method and system based on near-infrared spectroscopy and Raman spectroscopy. The method includes: obtaining a near-infrared spectrum and a Raman spectrum of a sample to be tested and inputting them into a preset spectrum recognition model to determine first recognition data; when the first recognition data cannot be obtained, extracting characteristic wavelengths and splicing spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a first fused characteristic spectrum; analyzing the first fused characteristic spectrum to determine a second target component and a corresponding second characteristic spectrum; analyzing the second target component and the corresponding second characteristic spectrum to determine second recognition data; generating a mixed spectrum processing scheme based on the second recognition data and storing it in a database; and visually displaying the first recognition data and the second recognition data through a preset terminal. The present invention uses a plurality of preset characteristic wavelength optimization algorithms to extract characteristic wavelengths and splice spectra, which can effectively improve the accuracy of spectral detection.

[0143] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0144] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed across multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0145] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0146] Those skilled in the art will appreciate that all or part of the steps of the above-mentioned method embodiments may be implemented by hardware associated with program instructions, and the aforementioned program may be stored in a computer-readable storage medium. When the program is executed, the program executes the steps of the above-mentioned method embodiments. The aforementioned storage medium includes various media that can store program codes, such as mobile storage devices, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical disks.

[0147] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

Claims

1. A fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy, characterized in that: include: Obtaining near-infrared spectra and Raman spectra of the sample to be tested; Inputting the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data; When the first identification data cannot be obtained, characteristic wavelength extraction and spectrum splicing are performed on the near-infrared spectrum and the Raman spectrum respectively using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fused characteristic spectra; Analyze the plurality of first fused characteristic spectra to determine a second target component and a corresponding second characteristic spectra; Analyze the second target component and the corresponding second characteristic spectrum to determine second identification data; generating a mixed spectrum processing solution according to the second identification data and storing the solution in a database; The first identification data and the second identification data are visually displayed through a preset terminal.

2. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 1 is characterized in that: The step of inputting the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data includes: spectral noise reduction is performed on the near-infrared spectrum and the Raman spectrum respectively by a cyclic three-point zero-order SG filtering method, and a wide-area spectrum is determined by spectral data vector normalization and data parallel fusion; Extracting key feature data of the wide-area spectrum graph by PCA-LDA feature dimensionality reduction method; The wide-area spectrum is compared with a standard characteristic spectrum of a preset target component according to the key characteristic data to determine first identification data.

3. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 1 is characterized in that: The method of extracting characteristic wavelengths and splicing the spectra of the near-infrared spectrum and the Raman spectrum respectively by using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fused characteristic spectra includes: Extracting characteristic wavelengths from the near-infrared spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of first spectra; Extracting characteristic wavelengths from the Raman spectra using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of groups of second spectra; The first spectrum graph and the second spectrum graph processed by different preset characteristic wavelength optimization algorithms are spectrally fused to obtain multiple first fused characteristic spectra.

4. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 1 is characterized in that: The step of analyzing the plurality of first fused characteristic spectra to determine a second target component and a corresponding second characteristic spectra includes: Determine the preset target components based on the testing requirements; When all characteristic peaks of the preset target component exist in the first fused characteristic spectrum, and the peak value of each characteristic peak is greater than the corresponding preset peak threshold, the preset target component is determined as the first target component, and the first fused characteristic spectrum is determined as the first characteristic spectrum; Calculating a first curve similarity between the first characteristic spectrum and a standard characteristic spectrum of the first target component; When the first curve similarity is greater than a first preset curve similarity threshold, the first target component is determined as a second target component, and the first characteristic spectrum is determined as a second characteristic spectrum.

5. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 1 is characterized in that: The step of analyzing the second target component and the corresponding second characteristic spectrum to determine the second identification data includes: Count the number of occurrences of the second target component; If the number of occurrences of the second target component is greater than a preset occurrence threshold, determining the second target component as the second identification data; On the contrary, from the first characteristic spectrum and the second characteristic spectrum corresponding to the second target component, the characteristic peak image having the smallest peak fluctuation value with respect to each characteristic peak in the standard characteristic spectrum of the second target component is sequentially selected for spectrum splicing to determine a spliced characteristic spectrum; Calculating a second curve similarity between the spliced characteristic spectrum and the standard characteristic spectrum of the second target component; When the second curve similarity is greater than a second preset curve similarity threshold, the second target component is determined as second recognition data.

6. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 1 is characterized in that: The generating of a mixed spectrum processing scheme according to the second identification data and storing the scheme in a database comprises: Determining, based on the second identification data, the mixing weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks, and generating a mixed spectrum processing solution; Binding the mixed spectrum processing scheme and the key characteristic peak of the second target component corresponding to the second identification data and storing them in a database; During the subsequent detection process, if all the key characteristic peaks of the second target component exist in the near-infrared spectrum and Raman spectrum of the detection sample, the corresponding hybrid spectrum processing scheme is called from the database to perform spectrum fusion processing on the near-infrared spectrum and Raman spectrum of the detection sample.

7. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 6, characterized in that: The step of determining, based on the second identification data, the mixed weights of the spectrum processing methods for the wavelength intervals corresponding to the characteristic peaks to generate a mixed spectrum processing solution includes: When the number N of second characteristic spectra corresponding to the second identification data is greater than the preset number of spectra, each characteristic peak of the second identification data is analyzed in order of wavelength interval from small to large, and the characteristic peak a is calculated based on the standard characteristic spectrum in the second characteristic spectrum F. n The peak fluctuation value x a-n , combined with the number of times y the first preset characteristic wavelength optimization algorithm i is used in all second characteristic spectra of characteristic peak a a-i Determine the mixing weight k of the first preset characteristic wavelength optimization algorithm i in the wavelength range corresponding to the characteristic peak a i-a ; Wherein, k1 is the influence weight of the peak fluctuation value, k2 is the influence weight of the number of times the preset characteristic wavelength optimization algorithm is used, k1+k2=1; When the number N of second characteristic spectra corresponding to the second identification data is less than or equal to the preset number of spectra, a second preset characteristic wavelength optimization algorithm j corresponding to the minimum peak fluctuation value of the characteristic peak a is retrieved from all the second characteristic spectra; The second preset characteristic wavelength optimization algorithm, the first preset characteristic wavelength optimization algorithm and the corresponding mixing weights of the wavelength interval corresponding to each characteristic peak are integrated to determine a mixed spectrum processing scheme.

8. The fusion spectrum detection method based on near infrared spectroscopy and Raman spectroscopy according to claim 7, characterized in that: If all key characteristic peaks of the second target component exist in the near-infrared spectrum and the Raman spectrum of the test sample, a corresponding hybrid spectrum processing scheme is called from the database to perform spectrum fusion processing on the near-infrared spectrum and the Raman spectrum of the test sample, including: The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the first preset characteristic wavelength optimization algorithm i to determine the first peak value H of the characteristic peak a. 1(i-a) ; Multiply all the first peak values of the characteristic peak a by the corresponding mixing weights, accumulate the calculation results, and determine the second peak value H of the characteristic peak a. 2(a) ; The spectrum of the wavelength range corresponding to the characteristic peak a is processed by the second preset characteristic wavelength optimization algorithm j to determine the third peak value H of the characteristic peak a. 3(a) ; The second peak H of all characteristic peaks 2(a) and the third peak H 3(a) Performing integration to determine a second fusion feature spectrum; Calculating a second curve similarity between the second fused characteristic spectrum and the standard characteristic spectrum of the second target component; When the second curve similarity is greater than a third preset curve similarity, the second target component is determined as second recognition data.

9. A fusion spectrum detection system based on near-infrared spectroscopy and Raman spectroscopy, for implementing the fusion spectrum detection method based on near-infrared spectroscopy and Raman spectroscopy according to any one of claims 1 to 8, characterized in that: include: A data acquisition module is used to obtain near-infrared spectra and Raman spectra of the sample to be tested; a model analysis module, configured to input the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data; a fusion spectrum splicing module, configured to, when the first identification data cannot be obtained, extract characteristic wavelengths and splice spectra of the near-infrared spectrum and the Raman spectrum using a plurality of preset characteristic wavelength optimization algorithms to determine a plurality of first fusion characteristic spectra; a fusion spectrum analysis module, configured to analyze the plurality of first fusion characteristic spectra to determine a second target component and a corresponding second characteristic spectrum; Analyze the second target component and the corresponding second characteristic spectrum to determine second identification data; a mixed spectrum processing solution generating module, configured to generate a mixed spectrum processing solution according to the second recognition data and store the solution in a database; The visual display module is used to visually display the first identification data and the second identification data through a preset terminal.

10. The fusion spectrum detection system based on near infrared spectroscopy and Raman spectroscopy according to claim 9, characterized in that: The step of inputting the near-infrared spectrum and the Raman spectrum into a preset spectrum recognition model to determine first recognition data includes: spectral noise reduction is performed on the near-infrared spectrum and the Raman spectrum respectively by a cyclic three-point zero-order SG filtering method, and a wide-area spectrum is determined by spectral data vector normalization and data parallel fusion; Extracting key feature data of the wide-area spectrum graph by PCA-LDA feature dimensionality reduction method; The wide-area spectrum is compared with a standard characteristic spectrum of a preset target component according to the key characteristic data to determine first identification data.

Images