A weighted prediction method for reconstructed spectra based on component absorption peaks

Abstract

The present invention mainly relates to the field of spectrum prediction technology. In order to solve the problem that the prediction ability of near-infrared spectroscopy equipment for multi-component samples to be tested is unstable when measuring the contents of different components, and the error of the mismeasured results is large, the present invention provides a weighted prediction method for reconstructed spectra based on component absorption peaks, firstly using near-infrared equipment to collect sample spectral data, then reconstructing the sample spectral data according to the wavelength range and number of component absorption peaks, calculating the weight coefficients of different absorption peaks of various components of the sample reconstructed spectral data, combining the reconstructed spectral data with the known component values ​​to establish a component spectral model, and finally combining various component spectral models to perform weighted prediction on the sample to be tested to obtain the final prediction value. The method described in the present invention is mainly suitable for predicting the spectral data of samples to be tested with multi-component contents.

Description

Technical Field

[0001] The present invention mainly relates to the technical field of spectrum prediction, and in particular to a weighted prediction method for reconstructed spectrum based on component absorption peaks. Background Art

[0002] With the development of micro-electromechanical technology, in recent years, miniaturized near-infrared spectroscopy equipment has been widely used in various fields such as food, beverages, medicine, coal, etc. Compared with the traditional large-scale Fourier transform spectroscopy analysis system, it has the advantages of simple structure, low cost, easy to carry, fast detection, and real-time visible results. It has become a popular product in the field of spectroscopy. In particular, the miniaturized near-infrared spectrometer also has the ability to predict multi-component indicators in parallel. It only needs to collect the spectral data of the sample to be tested once, and through specific algorithm processing, it can realize the parallel prediction of the multi-component content of the sample to be tested. The content of different components of the sample can be predicted at the same time in a very short time. For example, if the sample to be tested is pepper, the content of pepper moisture, volatile oil, hemp and other components can be obtained at the same time. If the sample to be tested is wheat, the content of wheat moisture, protein, starch and other components can be obtained at the same time. The application scenarios of miniaturized spectral detection technology have been greatly enhanced through the detection characteristics of miniaturized near-infrared spectroscopy.

[0003] Compared with large-scale near-infrared spectroscopy equipment in laboratories, miniaturized near-infrared spectroscopy detection equipment is more susceptible to the influence of light sources, detectors, manufacturing processes, usage methods, environmental conditions, etc. The signal-to-noise ratio of the spectral data it collects is relatively low, and its predictive ability is seriously affected, especially for samples with multi-component contents. For samples of the same type with different component contents, the predictive ability of miniaturized near-infrared spectroscopy equipment often varies greatly, and the predictive ability is extremely unstable. Processing spectral data through traditional preprocessing methods cannot solve this problem well. Summary of the invention

[0004] The technical problems to be solved by the present invention are:

[0005] A weighted prediction method for reconstructed spectra based on component absorption peaks is provided to solve the problem that the prediction ability of near-infrared spectroscopy equipment is unstable when measuring the contents of different components in multi-component samples and the error of mismeasured results is large.

[0006] The technical solution adopted by the present invention to solve the above technical problems is as follows:

[0007] A weighted prediction method for reconstructed spectrum based on component absorption peaks comprises the following steps:

[0008] Step 1: Use near-infrared spectroscopy equipment to collect sample spectral data;

[0009] Step 2: Reconstruct the sample spectrum data according to the wavelength range and number of component absorption peaks;

[0010] Step 3: Calculate the weight coefficient of the reconstructed spectrum;

[0011] Step 4: Combine the reconstructed spectral data with the sample component values ​​to establish a component spectral model;

[0012] Step 5: Use the component spectrum model to predict the sample, and perform weighted averaging of the prediction results combined with the weight coefficient of the reconstructed spectrum to obtain the final prediction value.

[0013] Furthermore, the near-infrared spectroscopy equipment used in step 1 is an equally divided and spliced ​​multi-sensor spectroscopy equipment. The sample spectrum is divided into two parts, one part is a sample with known components, and the other part is a sample with unknown components. The samples with known component content are used to establish a spectral model, and the samples with unknown component content are used to predict the component content.

[0014] Furthermore, the spliced ​​multi-sensor spectral data is reconstructed into multiple segments of single-sensor spectral data, and the number of wavelength points X1 included in the spectral device is calculated according to the band range and resolution collected by the equally divided spliced ​​multi-sensor spectral device, and the number of wavelength points X2 included in the single-sensor device is calculated according to the band range and resolution R of the single sensor; N absorption peaks are selected for spectral reconstruction, and when the number of absorption peaks does not exceed 3, spectral models of the corresponding number are established according to the number of specific spectral absorption peaks; when the number of absorption peaks exceeds 3, the three absorption peaks with the largest absorbance values ​​are selected for spectral reconstruction, and three spectral models are established.

[0015] Further, if there is only one absorption peak in the spectrum, that is, when N=1, the wavelength value of the wavelength point corresponding to the absorption peak is P1nm, if P1%10=0, where % is the remainder symbol, X3 wavelength points are taken to the left and right with the wavelength P1nm as the center to reconstruct the spectrum, and the calculation method of X3 is: X3=(X2-1) / 2, and the wavelength of each wavelength point of the reconstructed spectrum is [(P1-X3*R)nm, ..., (P1-R)nm, P1nm, (P1+R)nm, ..., (P1+X3*R)nm]; if P1%10≠0, the wavelength P1nm is rounded off to obtain a new P1 value, the original P1 value is replaced, and X3 wavelength points are taken to the left and right with the new P1nm as the center to reconstruct the spectrum in the same way;

[0016] If there are two absorption peaks in the spectrum, that is, when N=2, the wavelength points corresponding to the peak points of the absorption peaks are P2nm and P3nm respectively, then the two absorption peaks are processed by the spectrum data reconstruction method of a single absorption peak using P2nm and P3nm as reference points respectively, and the reconstructed spectrum contains two small band spectrum data;

[0017] If there are three absorption peaks in the spectral data or three absorption peaks with the largest absorbance values ​​are selected to reconstruct the spectral data, that is, when N=3, if the wavelength points corresponding to the peak points of the absorption peaks are P4nm, P5nm and P6nm, the three absorption peaks are processed using the spectral data reconstruction method of a single absorption peak with P4nm, P5nm and P6nm as reference points respectively, and the reconstructed spectrum contains 3 small-band spectral data.

[0018] Furthermore, if the boundary value of the spectrum after reconstruction exceeds the boundary value of the spectrum before reconstruction, the boundary value of the spectrum before reconstruction is used to replace the boundary value of the reconstructed spectrum.

[0019] Furthermore, the specific method for calculating the weight coefficient of the reconstructed spectrum in step 3 is:

[0020] When N = 1, the sample has a unique absorption peak, and the weight of the reconstructed spectrum based on the unique absorption peak is α1 = 100%;

[0021] When N = 2, first calculate the average absorbance of the two small band spectral wavelength points, recorded as and The corresponding weight coefficients α2 and α3 are:

[0022]

[0023] When N=3, the sample has three absorption peaks. The reconstructed spectrum data based on the three absorption peaks includes three small-band spectrum data. First, the average absorbance of the three small-band spectrum data is calculated respectively, which is recorded as and The corresponding weight coefficients are α4, α5, and α6 respectively:

[0024]

[0025] Further, the method for establishing the spectral model in step 4 is:

[0026] When N=1, the unique reconstructed spectrum is matched one-to-one with the spectrum data of the known component sample to perform spectrum modeling and obtain the component spectrum model M1;

[0027] When N=2, the two reconstructed small-band spectra are matched one by one with the spectral data of the known component samples to perform spectral modeling, and component spectral models M2 and M3 are obtained;

[0028] When N=3, the three reconstructed small-band spectra are respectively matched one by one with the spectral data of the known component samples to perform spectral modeling to obtain component spectral models M4, M5 and M6.

[0029] Furthermore, the specific method of obtaining the final prediction result in step 5 is:

[0030] Reconstruct the spectrum of the sample with unknown components and select the spectrum model according to the number of spectrum absorption peaks;

[0031] When the number of absorption peaks of the sample to be tested is 1, the spectral model M1 is used to predict the component content of the sample to be tested, and the component content prediction value S1 is obtained; at this time, S1 is the final spectral data of the unknown component sample to be tested;

[0032] When the number of absorption peaks of the sample to be tested is 2, the spectral models M2 and M3 are used to predict the component content of the sample to be tested, and the predicted value of the component content S is obtained. 02 and S 03 ; Combine the weight coefficients of the spectral models M2 and M3 to obtain the final spectral data S2 = α2*S of the unknown component to be tested. 02 +α3*S 03 ;

[0033] When the number of absorption peaks of the sample to be tested is 3 or the 3 absorption peaks with the largest absorbance values ​​are selected as the reference prediction points, the spectral models M4, M5 and M6 are used to predict the component content of the sample to be tested, and the component content prediction value S is obtained. 04 , S 05 and S 06 , combined with the weight coefficients of the spectral models M4, M5 and M6, the final spectral data of the unknown component to be tested is obtained by weighted average S3 = α4 * S 04 +α5*S 05 +α6*S 06 .

[0034] Beneficial effects of the present invention:

[0035] The present invention discloses a weighted prediction method for reconstructed spectra based on component absorption peaks, which reconstructs spectra based on component absorption peaks for samples of the same type with different component contents, reduces the amount of spectral data while retaining component characterization capabilities, and improves spectral analysis efficiency. At the same time, the absorbance of different absorption peaks of various components is combined to calculate the absorption peak weight coefficient, and perform weighted prediction, which effectively improves the prediction accuracy and stability of near-infrared spectral analysis technology for the contents of different components in the samples to be tested. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure is a schematic flow chart of a weighted prediction method for reconstructed spectra based on component absorption peaks described in the present invention. DETAILED DESCRIPTION

[0037] like Figure 1As shown, the weighted prediction method for reconstructed spectrum based on component absorption peaks described in the present invention first uses a near-infrared device to collect sample spectral data, then reconstructs the sample spectral data according to the wavelength range and number of component absorption peaks, calculates the weight coefficients of different absorption peaks of various components in the sample reconstructed spectral data, combines the reconstructed spectral data with the known component values ​​to establish a component spectral model, and finally combines the spectral models of various components to perform weighted prediction on the sample to be tested to obtain the final prediction value.

[0038] Example:

[0039] In this embodiment, a miniaturized, equally divided, spliced ​​multi-sensor near-infrared spectrometer with a spectral band coverage range of 800nm ​​to 2510nm is selected as a spectral acquisition device. The band range of 800nm ​​to 2510nm almost covers the entire near-infrared band range. The sample spectrum in this embodiment includes two parts, one of which is a sample with known component content, which is used to build a spectral model, and the other is a sample with unknown component content, which is used to predict the component content. The band range covered by a single sensor is limited. The commonly used single-sensor miniaturized near-infrared spectrometer covers a band range of 420 nanometers. The miniaturized near-infrared spectrometer used in this embodiment is a multi-sensor splicing type that includes 4 single sensors, and the band ranges are 800nm ​​to 1220nm, 1230nm to 1650nm, 1660nm to 2080nm, and 2090nm to 2510nm, respectively. The single sensor device that initially only covers a band range of 420 nanometers is expanded to cover 1710 nanometers, which can cover the absorption peaks of various components of the sample as comprehensively as possible.

[0040] Different samples have different component compositions. For a certain component of the sample, there may be more than one absorption peak in the near-infrared band. The number of near-infrared spectral absorption peaks corresponding to a certain component of the sample is set to N. In actual miniaturized near-infrared spectral analysis applications, commonly used sample components, such as water, starch, protein, etc., have less than 3 spectral absorption peaks. For some specific components of some samples, their spectral absorption peaks exceed 3, but selecting too many absorption peaks to reconstruct the spectrum will affect the efficiency of spectral analysis. The absorption peaks are screened, and the absorption peaks with smaller absorbance values ​​have weak representation capabilities, so they are eliminated, and the three absorption peaks with the largest absorbance values ​​are taken for spectral reconstruction. In summary, the value range of N is 1≤N≤3.

[0041] The resolution of the miniaturized splicing multi-sensor spectral device used in this embodiment is 10nm, and the band range is 800nm ​​to 2510nm. It can be calculated that the wavelength points contained in each multi-sensor spectral data are X1=1+(2510-800) / 10=172, and the band range of a single sensor is 420nm. It can be calculated that the wavelength points contained in each single sensor spectral data are X2=1+420 / 10=43. At the same time, since 1≤N≤3 and the number of near-infrared spectral absorption peaks is different, the spectral data reconstruction method is different, and the specific reconstruction method is as follows:

[0042] When N=1, the sample component has only a single absorption peak in the 800nm~2510nm band. The reconstructed spectral data is based on the single sensor spectral data. The band range and the number of wavelength points contained in the reconstructed spectral data are the same as those of the single sensor spectral data. The reconstructed spectral data based on the component absorption peak is defined as single peak data, and the wavelength range corresponding to the peak point of the absorption peak is set to P1nm. If P1%10≠0, where % is the remainder symbol, it is rounded off to obtain a new P1 value, and the original P1 value is replaced. With the wavelength range P1nm as the center point, X3 wavelength points are taken to the left and right for spectral reconstruction, where X3 is calculated as follows:

[0043] X3=(X2-1) / 2=21

[0044] The reconstructed spectral data is actually represented as a set of absorbance values ​​at 43 wavelength points, where the wavelength ranges of each wavelength point are [(P1-210)nm, (P1-200)nm, ... P1nm, ... (P1+200)nm, (P1+210)nm].

[0045] When (P1-210)≤800, the wavelength points of the reconstructed spectrum overflow to the left and the spectrum reconstruction cannot be performed. When the above situation occurs, the leftmost wavelength range of the reconstructed spectrum is selected as 800nm. It can be further known that in this case, the wavelength range of the reconstructed spectrum is 800nm~1220nm.

[0046] When (P1+210)≥2510, the wavelength points of the reconstructed spectrum overflow to the right and spectrum reconstruction cannot be performed. When the above situation occurs, the rightmost wavelength range of the reconstructed spectrum is selected as 2510nm. It can be further known that in this case, the wavelength range of the reconstructed spectrum is 2090nm~2510nm.

[0047] When N=2, there are only two absorption peaks in the sample component in the 800nm~2510nm band range. The wavelength ranges corresponding to the peak points of the two absorption peaks are set to P2nm and P3nm. The two absorption peaks are processed respectively by the spectral data reconstruction method of a single absorption peak. The reconstructed spectral data obtained in this case are actually represented by two small-band spectral data. The wavelength ranges of each wavelength point of the small-band spectral data 2 are [(P2-210)nm, (P2-200)nm, ... P2nm, ... (P2+200)nm, (P2+210)nm], and the wavelength ranges of each wavelength point of the small-band spectral data 3 are [(P3-210)nm, (P3-200)nm, ... P3nm, ... (P3+200)nm, (P3+210)nm]. Similarly, when the boundary wavelength range overflows, the spectral reconstruction method of a single absorption peak is adopted for processing.

[0048] When N=3, the sample component has three absorption peaks in the 800nm~2510nm band. The wavelength ranges corresponding to the peak points of the three absorption peaks are set to P4nm, P5nm and P6nm. The three absorption peaks are processed respectively by the spectrum data reconstruction method of a single absorption peak. The reconstructed spectrum data obtained in this case is actually represented by three small band spectrum data. The wavelength ranges of each wavelength point of the small band spectrum data 4 are [(P4-210)nm, (P4-200)nm, ... P4nm, ... (P4+200)nm, (P4-210)nm, ... P4-200)nm, ... P4+200)nm, (P4-210)nm, ... P4-200)nm, ... P4+200)nm, (P4-200)nm, ... P4+200)nm, (P4-200)nm, ... P4+200)nm, (P4-200)nm, ... P4+200)nm, (P4+ ... ... P4+200)nm, ... P4+200)nm, ... P4+200)nm, ... P4+200)nm, ... P4+200)nm, ... P4+200) 4+210)nm], the wavelength ranges of each wavelength point of small-band spectral data 5 are [(P5-210)nm, (P5-200)nm, … P5nm, … (P5+200)nm, (P5+210)nm], and the wavelength ranges of each wavelength point of small-band spectral data 6 are [(P6-210)nm, (P6-200)nm, … P6nm, … (P6+200)nm, (P6+210)nm], and similarly, when the boundary wavelength range overflows, the spectrum reconstruction method of a single absorption peak is adopted for processing.

[0049] When N=1, the reconstructed spectral data contains the same number of wavelength points as the single sensor spectral data, both of which contain 43 wavelength points; when N=2, the reconstructed spectral data contains two small band spectral data, each of which contains the same number of wavelength points as the single sensor spectral data, containing 43 wavelength points, and a total of 86 wavelength points; when N=3, the reconstructed spectral data contains three small band spectral data, each of which contains the same number of wavelength points as the single sensor spectral data, containing 43 wavelength points, and a total of 129 wavelength points. Compared with the 172 wavelength points of the original spectral data of multiple sensors, the reconstructed spectral data greatly reduces the amount of spectral data and improves the efficiency of spectral analysis while retaining the ability to characterize spectral data components.

[0050] When N=1, the sample component has only a single absorption peak in the 800nm-2510nm band, and the reconstructed spectrum based on the absorption peak can fully characterize the component, and its weight coefficient is α1=100%.

[0051] When N=2, the sample component has two absorption peaks in the 800nm-2510nm band. The reconstructed spectrum data based on the two absorption peaks includes two small-band spectrum data. The absorbance values ​​at 43 wavelength points in the small-band spectrum data 2 are set to be (A1, A2, ... A 43 ), then its absorbance mean is Assume that the absorbance values ​​at 43 wavelength points in the small band spectrum data 3 are (B1, B2, ... B 43 ), then its absorbance mean is Since the absorbance value directly reflects the absorption capacity of the sample components to near-infrared light, the larger the absorbance mean value, the stronger the ability of the absorption peak to characterize the component content. The weight coefficients α2 and α3 corresponding to the small-band spectrum data 2 and the small-band spectrum data 3 are calculated as follows:

[0052]

[0053] When N=3, the sample component has three absorption peaks in the 800nm-2510nm band. The reconstructed spectral data based on the three absorption peaks includes three small-band spectral data. Similarly, the absorbance mean of small-band spectral data 4 is set to Set the absorbance mean of small band spectral data 5 to Set the absorbance mean of small band spectral data 6 to The weight coefficients α4, α5, and α6 corresponding to the small band spectral data 4 / 5 / 6 are calculated as follows:

[0054]

[0055] When N=1, the sample component has only a single absorption peak in the 800nm~2510nm band. The reconstructed spectrum based on the absorption peak can fully characterize the component. The reconstructed spectral data based on the absorption peak of the sample component is one-to-one corresponding to the calibration value of the sample component. The conventional single-sensor spectral data modeling method is used to perform spectral modeling to obtain the component spectral model M1.

[0056] When N=2, the sample component has two absorption peaks in the 800nm~2510nm band range. The reconstructed spectral data based on the two absorption peaks includes two small-band spectral data. The two small-band spectral data are modeled separately. Taking small-band spectral data 2 as an example, the reconstructed small-band spectral data 2 is corresponded one-to-one with the sample component calibration value, and the spectral modeling is performed on it using the conventional single-sensor spectral data modeling method to obtain the component spectral model M2; similarly, the same calibration value correspondence and spectral modeling are performed on small-band spectral data 3 to obtain the spectral model M3 corresponding to small-band spectral data 3.

[0057] When N=3, the sample component has three absorption peaks in the band range of 800nm~2510nm. The reconstructed spectral data based on the three absorption peaks includes three small-band spectral data. The three small-band spectral data are modeled separately. Taking small-band spectral data 4 as an example, the reconstructed small-band spectral data 4 is corresponded one-to-one with the sample component calibration value, and the spectral modeling is performed using the conventional single-sensor spectral data modeling method to obtain the component spectral model M4; similarly, the same calibration value correspondence and spectral modeling are performed on small bands 5 / 6 to obtain the spectral models M5 and M6 corresponding to small bands 5 / 6.

[0058] When N=1, the component spectrum model M1 is used to predict the reconstructed spectrum data of the sample to be tested. The reconstructed spectrum data of the sample to be tested has the same band range and number of wavelength points as the single peak data 1, and the predicted value S is obtained. 01 , further combined with the weight coefficient of the absorption peak of the component is α1 = 100%, then when N = 1, the predicted value of the component content of the sample to be tested is calculated to be S1 = α1*S 01 =S 01 .

[0059] When N=2, the component spectrum model M2 is used to predict the small band spectrum data 02 reconstructed by the sample to be tested. The small band spectrum data 02 and the small band spectrum data 2 contain the same band range and the number of wavelength points, and the prediction value S is obtained. 02 , the component spectrum model M3 is used to predict the small band spectrum data 03 reconstructed by the sample to be tested. The small band spectrum data 03 and the small band spectrum data 3 contain the same band range and number of wavelength points, and the prediction value S is obtained.03 , further combined with the weight coefficients α2 and α3 corresponding to the small band spectrum data 2 and small band spectrum data 3, it is calculated that when N = 2, the predicted value of the component content of the sample to be tested is S2 = α2*S 02 +α3*S 03 .

[0060] When N=3, component spectrum models M4 / M5 / M6 are used to predict the reconstructed small-band spectrum data 04 / 05 / 06 of the sample to be tested. The small-band spectrum data 04 / 05 / 06 and the small-band spectrum data 4 / 5 / 6 have the same band range and number of wavelength points, and the predicted value S is obtained. 04 / S 05 / S 06 , further combined with the weight coefficients α4, α5, and α6 corresponding to the small-band spectral data 4 / 5 / 6, it is calculated that when N=3, the predicted value of the component content of the sample to be tested is S3=α4*S 04 +α5*S 05 +α6*S 06 .

Claims

1. A weighted prediction method for reconstructed spectrum based on component absorption peaks, characterized in that: The following steps are involved: Step 1: Use an average-split splicing multi-sensor near-infrared spectrometer to collect sample spectral data; Step 2: Reconstruct the sample spectral data collected by the splicing multi-sensor optical near-infrared spectral device into multiple segments of single-sensor spectral data, calculate the number of wavelength points X1 included in the spectral device according to the band range and resolution collected by the equally divided splicing multi-sensor spectral device, and calculate the number of wavelength points X2 included in the single sensor device according to the band range and resolution R of the single sensor; select N absorption peaks for spectral reconstruction, when the number of absorption peaks does not exceed 3, establish a spectral model of the corresponding number according to the number of specific spectral absorption peaks, when the number of absorption peaks exceeds 3, select the three absorption peaks with the largest absorbance values ​​for spectral reconstruction, and establish three spectral models; Step 3: Calculate the weight coefficient of the reconstructed spectrum; the specific method of calculating the weight coefficient of the reconstructed spectrum in step 3 is: When N = 1, the sample has a unique absorption peak, and the weight of the reconstructed spectrum based on the unique absorption peak is α1 = 100%; When N = 2, first calculate the average absorbance of the two small band spectral wavelength points, recorded as and The corresponding weight coefficients α2 and α3 are: When N=3, the sample has three absorption peaks. The reconstructed spectral data based on the three absorption peaks includes three small-band spectral data. First, the average absorbance of the three small-band spectral data is calculated respectively, which is recorded as and The corresponding weight coefficients are α4, α5, and α6 respectively: Step 4: Combine the reconstructed spectral data with the sample component values ​​to establish a component spectral model; When N=1, the unique reconstructed spectrum is matched one-to-one with the spectrum data of the known component sample to perform spectrum modeling and obtain the component spectrum model M1; When N=2, the two reconstructed small-band spectra are matched one by one with the spectral data of the known component samples to perform spectral modeling, and component spectral models M2 and M3 are obtained; When N=3, the three reconstructed small-band spectra are matched one by one with the spectral data of the known component samples to perform spectral modeling, and component spectral models M4, M5 and M6 are obtained; Step 5: Use the component spectrum model to predict the sample, and perform weighted averaging of the prediction results combined with the weight coefficient of the reconstructed spectrum to obtain the final prediction value.

2. The method for weighted prediction of reconstructed spectrum based on component absorption peak according to claim 1, characterized in that: In step 1, the sample spectrum is divided into two parts, one part is a sample with known components, and the other part is a sample with unknown components. The sample with known component content is used to establish a spectral model, and the sample with unknown component content is used to predict the component content.

3. The method for weighted prediction of reconstructed spectrum based on component absorption peak according to claim 1, characterized in that: If there is only one absorption peak in the spectrum, that is, when N=1, the wavelength value of the wavelength point corresponding to the absorption peak is P1nm. If P1%10=0, where % is the remainder symbol, X3 wavelength points are taken to the left and right with the wavelength P1nm as the center to reconstruct the spectrum. The calculation method of X3 is: X3=(X2-1) / 2, and the wavelength of each wavelength point of the reconstructed spectrum is [(P1-X3*R)nm, ..., (P1-R)nm, P1nm, (P1+R)nm, ..., (P1+X3*R)nm]; if P1%10≠0, the wavelength P1nm is rounded off to obtain a new P1 value, and the original P1 value is replaced. With the new P1nm as the center, X3 wavelength points are taken to the left and right to reconstruct the spectrum in the same way; If there are two absorption peaks in the spectrum, that is, when N=2, the wavelength points corresponding to the peak points of the absorption peaks are P2nm and P3nm respectively, then the two absorption peaks are processed by the spectrum data reconstruction method of a single absorption peak using P2nm and P3nm as reference points respectively, and the reconstructed spectrum contains two small band spectrum data; If there are three absorption peaks in the spectral data or three absorption peaks with the largest absorbance values ​​are selected to reconstruct the spectral data, that is, when N=3, if the wavelength points corresponding to the peak points of the absorption peaks are P4nm, P5nm and P6nm, the three absorption peaks are processed using the spectral data reconstruction method of a single absorption peak with P4nm, P5nm and P6nm as reference points respectively, and the reconstructed spectrum contains 3 small-band spectral data.

4. The method for weighted prediction of reconstructed spectrum based on component absorption peaks according to claim 3, characterized in that: If the boundary value of the spectrum after reconstruction exceeds the boundary value of the spectrum before reconstruction, the boundary value of the spectrum before reconstruction is used to replace the boundary value of the reconstructed spectrum.

5. The method for weighted prediction of reconstructed spectrum based on component absorption peak according to claim 1, characterized in that: The specific method of obtaining the final prediction result in step 5 is: Reconstruct the spectrum of the sample with unknown components and select the spectrum model according to the number of spectrum absorption peaks; When the number of absorption peaks of the sample to be tested is 1, the spectral model M1 is used to predict the component content of the sample to be tested, and the component content prediction value S1 is obtained; at this time, S1 is the final spectral data of the unknown component sample to be tested; When the number of absorption peaks of the sample to be tested is 2, the spectral models M2 and M3 are used to predict the component content of the sample to be tested, and the predicted value of the component content S is obtained. 02 and S 03 ; Combine the weight coefficients of the spectral models M2 and M3 to obtain the final spectral data S2 = α2*S of the unknown component to be tested. 02 +α3*S 03 ; When the number of absorption peaks of the sample to be tested is 3 or the 3 absorption peaks with the largest absorbance values ​​are selected as the reference prediction points, the spectral models M4, M5 and M6 are used to predict the component content of the sample to be tested, and the component content prediction value S is obtained. 04 , S 05 and S 06 , combined with the weight coefficients of the spectral models M4, M5 and M6, the final spectral data of the unknown component to be tested is obtained by weighted average S3 = α4 * S 04 +α5*S 05 +α6*S 06 .

Images