Capsaicin compound quantitative prediction model construction method and device, and storage medium

Abstract

The invention discloses a capsaicin compound quantitative prediction model construction method and device, and a storage medium. The method comprises the following steps: S1, collecting visible-near infrared spectrum and Raman spectrum data of a chili powder sample; s2, determining the content of a taste substance capsaicin compound in the chili powder by utilizing an HPLC (High Performance Liquid Chromatography) technology; s3, directly splicing and fusing the collected visible-visible-near infrared and Raman spectrum data to form a new fusion matrix; and S4, performing variable screening according to the new fusion matrix, and establishing a quantitative model for predicting the content of the capsaicin compound as the taste substance of the chili powder. By adopting the technical scheme provided by the invention, in-situ nondestructive rapid evaluation of trace active components in a complex food matrix is realized, and pretreatment loss and aging bottleneck of traditional destructive detection are effectively avoided.

Description

Technical Field

[0001] This invention belongs to the field of food quality testing technology, and in particular relates to a method and apparatus for constructing a quantitative prediction model for capsaicin compounds, as well as a storage medium. Background Technology

[0002] Chili peppers are one of the core spices. Globally, the demand for spicy foods, diverse cultural cuisines (especially popular in Asian, Latin American, and African cuisines), condiments, and convenience foods continues to grow, making the market prospects for chili pepper products stable and broad. Chili peppers have a pungent and distinctive flavor and are commonly used in various condiments. Furthermore, a growing body of research indicates that the active ingredients in chili peppers can improve bodily functions and possess antioxidant, anti-inflammatory, and metabolism-boosting properties. Currently, China has become the world's largest chili pepper-growing country in terms of both area and yield. Fujian, Guizhou, Ningxia, Sichuan, Yunnan, Henan, Xinjiang, and Hunan are major chili pepper producing areas in China, and their chili peppers and chili pepper products are of significant reference value. Chili peppers have a high moisture content and a short shelf life; therefore, drying them into powder is a common processing method. Chili powder, as the most important primary processed product of chili peppers, contains characteristic flavor compounds (especially capsaicin compounds), which are the main contributors to its spiciness. These compounds are easily lost during processing and storage due to thermal degradation, enzymatic hydrolysis, and non-enzymatic oxidation. Flavor compounds in food have received widespread attention in recent years as important indicators of food quality and safety. Chili peppers, as a spice, contain capsaicin and dihydrocapsaicin, natural substances that give them their spiciness. Capsaicin accounts for 70% of capsaicinoids, while dihydrocapsaicin accounts for 22%. The capsaicin content usually directly reflects the spiciness of chili powder. Therefore, detecting the capsaicin content in chili powder is an effective method for assessing its quality and spiciness, providing a theoretical basis for evaluating its flavor and quality.

[0003] Current methods for detecting volatile organic gases mainly include high-performance liquid chromatography (HPLC), gas chromatography-mass spectrometry (GC-MS), and ultraviolet-visible spectrophotometry (UV-Vis). These methods are cumbersome, labor-intensive, time-consuming, susceptible to interference, and cannot be performed online, thus impacting sensitivity and accuracy. Furthermore, HPLC and GC-MS are large-scale instruments, expensive, and require specialized personnel for complex operation. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method, apparatus and storage medium for constructing a quantitative prediction model for capsaicin compounds.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A method for constructing a quantitative prediction model for capsaicin compounds includes:

[0007] Step S1: Collect visible-near-infrared and Raman spectral data of chili powder samples;

[0008] Step S2: Determine the content of capsaicin compounds, flavor compounds in chili powder, using HPLC technology;

[0009] Step S3: Directly splice and fuse the collected visible-near infrared and Raman spectral data to form a new fused matrix;

[0010] Step S4: Variable selection is performed based on the fused new matrix to establish a quantitative model for predicting the content of capsaicin compounds, which are flavor compounds in chili powder.

[0011] Preferably, step S3 is as follows: Spectral data of chili powder are collected using visible-near-infrared spectroscopy; the 350-1000 nm wavelength range of the obtained visible-near-infrared spectral data of chili powder is extracted; spectral data of chili powder are collected using micro Raman spectroscopy; the 352-1900 nm wavelength range of the obtained Raman spectral data of chili powder is extracted; the extracted wavelength ranges of the Raman spectrum and the visible-near-infrared spectrum are fused; and the extracted Raman spectral feature matrix and the visible-near-infrared spectral feature matrix are directly spliced ​​and fused to form a new fused matrix.

[0012] Preferably, in step S4, the fused new matrix is ​​preprocessed with homogenization to remove spectral noise, and the feature matrix is ​​screened by an algorithm to combine Raman and visible-near-infrared spectra into a new feature matrix. The new feature matrix is ​​used as input, and the HPLC detection of capsaicin and dihydrocapsaicin in chili powder is used as output. A quantitative model for predicting the content of capsaicin in chili powder is constructed by variable screening, and the correlation coefficients RC / RP and RMSEC / RMSEP are used as the criteria for evaluating the model's effectiveness.

[0013] This invention also provides a device for constructing a quantitative prediction model for capsaicin compounds, comprising:

[0014] The first processing module is used to collect visible-near-infrared and Raman spectral data of chili powder samples;

[0015] The second processing module is used to determine the content of capsaicin compounds, flavor compounds in chili powder, using HPLC technology.

[0016] The third processing module is used to directly splice and fuse the collected visible-near infrared and Raman spectral data to form a new fused matrix;

[0017] The fourth processing module is used to filter variables based on the fused new matrix and establish a quantitative model for predicting the content of capsaicin compounds, a flavor compound in chili powder.

[0018] Preferably, the third processing module uses visible-near-infrared spectroscopy to collect spectral data of chili powder, extracts the visible-near-infrared data of chili powder in the wavelength range of 350-1000 nm, and uses micro Raman spectroscopy to collect spectral data of chili powder, extracts the Raman data of chili powder in the wavelength range of 352-1900 nm, fuses the extracted wavelength ranges of Raman and visible-near-infrared spectroscopy, and directly splices and fuses the extracted Raman spectral feature matrix and visible-near-infrared spectral feature matrix to form a new fused matrix.

[0019] Preferably, the fourth processing module is used to perform homogenization preprocessing on the fused new matrix and then preprocessing to remove spectral noise. The feature matrix is ​​then screened by an algorithm, and the Raman spectrum and the visible-near infrared spectrum are spliced ​​into a new feature matrix. The new feature matrix is ​​used as input, and the capsaicin and dihydrocapsaicin of chili powder are detected by HPLC as output. A quantitative model for predicting the content of capsaicin in chili powder is constructed by variable screening. The correlation coefficient RC / RP and RMSEC / RMSEP are used as the criteria for evaluating the model effect.

[0020] The present invention also provides a storage medium storing a computer program, which executes a method for constructing a quantitative prediction model for capsaicin compounds when running.

[0021] Compared with the prior art, the present invention has the following technical effects:

[0022] 1. This invention utilizes multi-module spectral data fusion combined with chemometrics to achieve rapid and non-destructive detection of capsaicin-like compounds, a flavoring substance in chili powder.

[0023] 2. This invention deeply mines multiple spectral information to improve the predictive performance of the detection model for the content of capsaicin-like compounds, a flavor compound in dried chili powder.

[0024] 3. The machine learning model for predicting the content of capsaicin-like compounds in chili powder established in this invention takes the collected visible-near-infrared spectral information and Raman spectral information fused into a new matrix as input, and uses HLPC technology to detect the physicochemical values ​​of capsaicin-like compounds in chili powder as output. It constructs a quantitative model for predicting the content of capsaicin-like compounds in chili powder, which has good versatility, improves efficiency, and provides important characteristic information of capsaicin-like compound content in chili powder and a reference for evaluating the taste quality of chili powder. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0026] Figure 1 This is a flowchart of the method for constructing a quantitative prediction model for capsaicin compounds according to an embodiment of the present invention;

[0027] Figure 2 Raman and visible-near-infrared spectral image data collected from chili powder samples of the present invention; wherein, A is visible-near-infrared spectral image data; B is Raman spectral image data;

[0028] Figure 3 This is a graph showing the content of capsaicin compounds, a flavor compound in chili powder, detected using the HPLC technique of this invention; where A represents standard samples with different concentration gradients of capsaicin; and B represents chili powder samples from eight regions.

[0029] Figure 4 This is a schematic diagram of the model based on the fusion of visible-near-infrared and Raman spectral data of chili powder according to the present invention; wherein, A is a schematic diagram of the BOSS model; B is a schematic diagram of the CARS model; C is a schematic diagram of the VCPA model; D is a schematic diagram of the CARS-GA model; and E is a schematic diagram of the VCPA-IRIV model.

[0030] Figure 5The figures show various model data graphs for the present invention, with the input of separate Raman spectral and visible-near-infrared spectral data matrices and the output of HPLC technology for detecting the content of capsaicin-like compounds, a flavor compound in chili powder; where B1 is a schematic diagram of the VCPA-IRIV model, B2 is a schematic diagram of the BOSS model, B3 is a schematic diagram of the CARS model, and B4 is a schematic diagram of the VCPA model; Figure C is a schematic diagram of the model of the Raman spectral matrix of chili powder, where C1 is a schematic diagram of the VCPA-IRIV model, C2 is a schematic diagram of the BOSS model, C3 is a schematic diagram of the CARS model, and C4 is a schematic diagram of the VCPA model. Detailed Implementation

[0031] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0032] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0033] Example 1:

[0034] like Figure 1 As shown, this embodiment of the invention provides a method for constructing a quantitative prediction model for capsaicin compounds, comprising:

[0035] Step S1: Collect visible-near-infrared and Raman spectral data of chili powder samples.

[0036] Spectroscopic data were collected from chili powder samples from eight regions (Fujian, Hunan, Guizhou, Sichuan, Xinjiang, Henan, Ningxia, and Yunnan) using both visible-near-infrared (VNIIR) and micro Raman spectrometers. The data was acquired using a VNIIR / VNIIR spectral acquisition device. To eliminate environmental noise in the infrared information, such as scattering effects caused by diffuse reflection, baseline shift, overlapping peaks, and the signal-to-noise ratio in the spectral data, the spectral detection module was encapsulated in an illumination box. A computer was connected via fiber optic cable to a portable Vis-NIR spectrometer with a spectral range of 340.886-1074.36 nm. In reflectance mode, the integration time was set to 5-10 ms, the number of scans to 10-20, and the pixel smoothing to 5-10. During VNIIR spectral acquisition, the chili powder samples were placed in a small circular aluminum box to ensure consistent thickness. Raman spectral data acquisition: Raman measurements were performed using a Portman-TS9214 Raman-Vis-NIR dual-spectral analyzer equipped with a 785.025 nm laser-excited spectrometer. A 20x eyepiece was used, and the light source was manually adjusted to focus the sample at a single point. To ensure uniform thickness and minimize light scattering, the chili powder was placed on a foil plate and pressed flat with a flat plate. Spectral characteristics were observed in the range of 300–3000 cm⁻¹. - Within the specified range, Raman spectra were collected using a laser power of 150mW-200mW and an integration time of 5s. Each point was collected three times to minimize error, and the average value represents the final spectral data. Using the above spectroscopic instrument and its parameters, 120 samples were collected, yielding 120 spectra for each sample. Figure 2 As shown, Figure A contains visible-near-infrared spectral image data, and Figure B contains Raman spectral image data.

[0037] Step S2: Determine the content of capsaicin-like compounds, the flavor compounds in chili powder, using HPLC technology.

[0038] Based on the national standard GB / T21266-2007, HPLC technology was used to analyze the capsaicin and dihydrocapsaicin content of chili powder from eight different origins. Specifically, the capsaicin content in the chili powder was determined using a Tongwei high-performance liquid chromatograph. Equipped with a UV detector, the chromatographic column is C184.6×250mm 5μm. To better distinguish the analytes, the analytical instrument method was modified based on the national standard GB / T21266-2007 and the actual situation. The mobile phase A:B—methanol:water = 72:28, the flow rate was 1ml / min, the injection volume was 20μL, the detection wavelength was 280nm, and the column temperature was 30℃.

[0039] Capsaicin standard was prepared as a capsaicin-based standard stock solution: Accurately weigh 0.0526 g of capsaicin standard and 0.0548 g of dihydrocapsaicin standard, dissolve them in methanol, and dilute to 50.0 ml. Prepare a mixed standard solution of capsaicin and dihydrocapsaicin, each with a concentration of 1 mg / ml. Seal and store at 4°C for later use. Capsaicin-based standard working solutions: Pipette 0 ml, 0.5 ml, 1.0 ml, 1.5 ml, 2.0 ml, 2.5 ml, 3.0 ml, 3.5 ml, and 4.0 ml of the standard stock solution, respectively, and dilute to 25 ml with methanol. This standard series has concentrations of 0 g / ml, 20 μg / ml, 40 μg / ml, 60 μg / ml, 80 μg / ml, 100 μg / ml, 120 μg / ml, 140 μg / ml, and 160 μg / ml.

[0040] Add 5 ml of a 1:1 methanol-tetrahydrofuran mixture to the sample and extract using an ultrasonic oscillator for 30 min. Filter with filter paper, collect the filtrate, and then add 5 ml of the 1:1 methanol-tetrahydrofuran mixture back to the filter residue along with the filter paper. Extract using an ultrasonic oscillator for 10 min, repeating twice. Combine the filtrates collected from the three filtrations, concentrate to 2 mL–4 mL using a rotary evaporator at 70 °C, then bring the volume to 10 mL with a 5.4 mixture. Filter through a 0.45 μm filter membrane and perform chromatographic analysis.

[0041] High-performance liquid chromatography (HPLC) detection results are as follows Figure 3 Figure A shows that the peak elution times of standard samples at different concentration gradients are in the range of 9-11 min, and the peak area increases with increasing concentration gradient. Furthermore, a regression equation was derived from the standard curve prepared according to the national standard GB / T21266-2007 to calculate the capsaicin content of the chili powder samples. Figure B shows the peak shape and peak area of ​​the average spectra of chili powder samples from eight regions.

[0042] Step S3: Directly stitch and fuse the collected visible-near-infrared and Raman spectral data to form a new fused matrix.

[0043] Visible-near-infrared spectroscopy was used to collect spectral data of chili powder. The 350-1000 nm wavelength range of the obtained visible-near-infrared spectral data of chili powder was extracted. Similarly, micro Raman spectroscopy was used to collect spectral data of chili powder. The 352-1900 nm wavelength range of the obtained Raman spectral data of chili powder was extracted. The extracted wavelength ranges of the Raman and visible-near-infrared spectra were fused together. The extracted Raman spectral feature matrices and visible-near-infrared spectral feature matrices were directly spliced ​​and fused to form a new fused matrix.

[0044] Step S4: Filter variables based on the fused new matrix to establish a quantitative model for predicting the content of capsaicin-like compounds, a flavor compound in chili powder.

[0045] First, the new matrix is ​​preprocessed using homogenization and then smoothed using SG filtering to remove spectral noise. Variable selection of the feature matrix is ​​performed using VCPA, VCPA-IRIV, CARS, BOSS, and CARS-GA algorithms. Raman and visible-near-infrared spectra are then concatenated into a new feature matrix. Using this new feature matrix as input, and HPLC detection of capsaicin and dihydrocapsaicin in chili powder as output, a quantitative model for predicting the content of capsaicin-like compounds in chili powder is constructed. Figure 4 As shown, A is a schematic diagram of the BOSS model; B is a schematic diagram of the CARS model; C is a schematic diagram of the VCPA model; D is a schematic diagram of the CARS-GA model; and E is a schematic diagram of the VCPA-IRIV model. The VCPA-IRIV variable selection result is the best. The model performance is evaluated using correlation coefficients RC / RP and RMSEC / RMSEP. The VCPA-IRIV variable selection model has RC of 0.985, RP of 0.975, RMSEC of 0.056, and RMSEP of 0.078. Compared with the modeling performance of individual visible-near-infrared spectral feature matrices and individual Raman spectral feature matrices, the new matrix model that fuses visible-near-infrared and Raman spectral data shows improved prediction performance.

[0046] Furthermore, the modeling of individual visible-near-infrared (VNIIR) spectral feature matrices and individual Raman spectral feature matrices is based on the feature matrices constructed from individual Raman spectral data, individual VNIIR data, and physicochemical values ​​determined by HPLC. Specifically: Visible-near-infrared spectral data of chili powder is collected using visible-near-infrared spectroscopy. The obtained VNIIR spectral data of chili powder is then extracted into a 350-1000 nm wavelength range and combined with physicochemical values ​​determined by HPLC to construct a feature matrix for individual VNIIR spectral data. A predictive model is then established through variable screening using VCPA, VCPA-IRIV, CARS, and BOSS. Conversely, Raman-visible-near-infrared dual-spectral equipment is used to collect spectral data of chili powder. The obtained Raman spectral data of chili powder is then extracted into a 352-1900 nm wavelength range and combined with physicochemical values ​​determined by HPLC to construct a feature matrix for individual VNIIR spectral data. A predictive model is then established through variable screening using VCPA, VCPA-IRIV, CARS, and BOSS. Prediction models were established based on different individual spectral feature matrices. The model performance was evaluated using correlation coefficients (RC / RP) and RMSEC / RMSEP as criteria. For example,... Figure 5As shown in B1, B2, B3, and B4, the best-performing visible-near-infrared spectral feature matrix model is the prediction model constructed using VCPA-IRIV variable selection, with RC of 0.982, RP of 0.968, RMSEC of 0.061, and RMSEP of 0.089. Figure 5 As shown in C1, C2, C3, and C4, the Raman spectral feature matrix model with the best performance is the prediction model constructed by VCPA-IRIV variable screening, with RC of 0.9321, RP of 0.8746, RMSEC of 0.117, and RMSEP of 0.159.

[0047] This invention integrates high-performance liquid chromatography (HPLC) with a visible-near-infrared (NIR)-Raman dual-spectroscopy analytical framework to characterize the spatial distribution of capsaicin compounds in chili powder matrices from multiple geographical origins. By acquiring multimodal heterogeneous spectral data (NIR vibrational absorption spectra and Raman scattering spectra) and chromatographic separation quantitative calibration values, a chemometrically driven multi-source information collaborative characterization model is constructed. Based on feature fusion and variable optimization algorithms, a highly robust quantitative inversion system for capsaicin-like flavor substances is established. This invention enables in-situ, non-destructive, and rapid assessment of trace active components in complex food matrices, effectively avoiding the preprocessing losses and time bottlenecks of traditional destructive detection methods.

[0048] Example 2:

[0049] This invention also provides a device for constructing a quantitative prediction model for capsaicin compounds, comprising:

[0050] The first processing module is used to collect visible-near-infrared and Raman spectral data of chili powder samples;

[0051] The second processing module is used to determine the content of capsaicin compounds, flavor compounds in chili powder, using HPLC technology.

[0052] The third processing module is used to directly splice and fuse the collected visible-near infrared and Raman spectral data to form a new fused matrix;

[0053] The fourth processing module is used to filter variables based on the fused new matrix and establish a quantitative model for predicting the content of capsaicin compounds, a flavor compound in chili powder.

[0054] In one embodiment of the present invention, the third processing module uses visible-near-infrared spectroscopy to collect spectral data of chili powder, extracts visible-near-infrared data of chili powder in the wavelength range of 350-1000 nm from the obtained visible-near-infrared spectral data of chili powder, and uses micro Raman spectroscopy to collect spectral data of chili powder, extracts Raman data of chili powder in the wavelength range of 352-1900 nm from the obtained Raman spectral data of chili powder, fuses the extracted wavelength ranges of Raman spectrum and visible-near-infrared spectrum, and directly splices and fuses the extracted Raman spectral feature matrix and visible-near-infrared spectral feature matrix to form a new fused matrix.

[0055] As one embodiment of the present invention, the fourth processing module is used to perform homogenization preprocessing on the fused new matrix and then perform preprocessing to remove spectral noise. The feature matrix is ​​screened by an algorithm, and Raman spectrum and visible-near infrared spectrum are spliced ​​into a new feature matrix. The new feature matrix is ​​used as input, and the HPLC detection of capsaicin and dihydrocapsaicin in chili powder is used as output. A quantitative model for predicting the content of capsaicin in chili powder is constructed by variable screening. The correlation coefficient RC / RP and RMSEC / RMSEP are used as the criteria for evaluating the model effect.

[0056] Example 3:

[0057] This invention also provides a storage medium storing a computer program, which executes a method for constructing a quantitative prediction model for capsaicin compounds during runtime.

[0058] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A method for constructing a quantitative prediction model for capsaicin-like compounds, characterized in that, include: Step S1: Collect visible-near-infrared and Raman spectral data of chili powder samples; Step S2: Determine the content of capsaicin compounds, flavor compounds in chili powder, using HPLC technology; Step S3: Directly splice and fuse the collected visible-near infrared and Raman spectral data to form a new fused matrix; Step S4: Variable selection is performed based on the fused new matrix to establish a quantitative model for predicting the content of capsaicin compounds, which are flavor compounds in chili powder.

2. The method for constructing a quantitative prediction model for capsaicin compounds as described in claim 1, characterized in that, Step S3 is as follows: Spectral data of chili powder are collected using visible-near-infrared spectroscopy. The obtained visible-near-infrared spectral data of chili powder is then extracted into the 350-1000 nm wavelength range. Similarly, spectral data of chili powder are collected using micro Raman spectroscopy. The obtained Raman spectral data of chili powder is then extracted into the 352-1900 nm wavelength range. The extracted wavelength ranges from the Raman and visible-near-infrared spectroscopy are fused together. The extracted Raman spectral feature matrix and the visible-near-infrared spectral feature matrix are then directly spliced ​​and fused to form a new fused matrix.

3. The method for constructing a quantitative prediction model for capsaicin compounds as described in claim 2, characterized in that, In step S4, the fused new matrix is ​​preprocessed with homogenization to remove spectral noise. The feature matrix is ​​then filtered by an algorithm to combine Raman and visible-near-infrared spectra into a new feature matrix. The new feature matrix is ​​used as input, and the capsaicin and dihydrocapsaicin content of chili powder detected by HPLC is used as output. A quantitative model for predicting the capsaicin content of chili powder is constructed through variable filtering. The correlation coefficients RC / RP and RMSEC / RMSEP are used as the criteria for evaluating the model's effectiveness.

4. A device for constructing a quantitative prediction model for capsaicin-like compounds, characterized in that, include: The first processing module is used to collect visible-near-infrared and Raman spectral data of chili powder samples; The second processing module is used to determine the content of capsaicin compounds, flavor compounds in chili powder, using HPLC technology. The third processing module is used to directly splice and fuse the collected visible-near infrared and Raman spectral data to form a new fused matrix; The fourth processing module is used to filter variables based on the fused new matrix and establish a quantitative model for predicting the content of capsaicin compounds, a flavor compound in chili powder.

5. The apparatus for constructing a quantitative prediction model for capsaicin compounds as described in claim 4, characterized in that, The third processing module uses visible-near-infrared spectroscopy to collect spectral data of chili powder. It extracts the visible-near-infrared data of chili powder in the wavelength range of 350-1000 nm. It also uses micro Raman spectroscopy to collect spectral data of chili powder. It extracts the Raman data of chili powder in the wavelength range of 352-1900 nm. The extracted wavelength ranges of the Raman spectrum and the visible-near-infrared spectrum are fused together. The extracted Raman spectral feature matrix and the visible-near-infrared spectral feature matrix are directly spliced ​​and fused to form a new fused matrix.

6. The apparatus for constructing a quantitative prediction model for capsaicin compounds as described in claim 5, characterized in that, The fourth processing module is used to perform homogenization preprocessing on the fused new matrix and then further preprocessing to remove spectral noise. The feature matrix is ​​then filtered for variables using an algorithm. Raman and visible-near-infrared spectra are spliced ​​into a new feature matrix. The new feature matrix is ​​used as input, and the capsaicin and dihydrocapsaicin of chili powder are detected by HPLC as output. A quantitative model for predicting the capsaicin content of chili powder is constructed through variable filtering. The correlation coefficients RC / RP and RMSEC / RMSEP are used as the criteria for evaluating the model's effectiveness.

7. A storage medium, characterized in that, The storage medium stores a computer program, which, when running, executes the method for constructing a quantitative prediction model for capsaicin compounds as described in any one of claims 1-3.

Images