CN115855836A - Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system - Google Patents
Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system Download PDFInfo
- Publication number
- CN115855836A CN115855836A CN202211593947.9A CN202211593947A CN115855836A CN 115855836 A CN115855836 A CN 115855836A CN 202211593947 A CN202211593947 A CN 202211593947A CN 115855836 A CN115855836 A CN 115855836A
- Authority
- CN
- China
- Prior art keywords
- withering
- data
- fermentation
- tea
- hyperspectral
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a method and a system for judging withering and fermentation degrees of fresh tea leaves based on hyperspectral imaging. The evaluation method comprises the following steps: collecting hyperspectral data in the tea withering and fermentation processes, and determining the contents of Tea Polyphenols (TPs), free Amino Acids (FAA) and Caffeine (CAF) in each sample; a continuous projection algorithm (SPA), a competitive adaptive re-weighting (CARS) and a non-information variable elimination (UVE) method are adopted, a characteristic waveband is selected, and a monitoring model of TPs, FAA and CAF content is established by combining a Support Vector Machine (SVM), a Random Forest (RF) and a partial least square method (PLS) and is used for quantitatively judging the withering degree and the fermentation degree. The method can improve the online monitoring efficiency of biochemical components in the tea withering and fermentation processes, and provides a basis for intelligent judgment of the withering and fermentation degrees in the black tea processing process.
Description
Technical Field
The invention belongs to the field of hyperspectral imaging estimation, and particularly relates to a method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging.
Background
Black tea originates from china. Because it has the characteristics of red leaves, red soup and sweet and mellow taste and contains rich nutrient components, it is popular with people all over the world. The black tea is processed from fresh tea leaves, the withering is the first process of black tea processing, when the withering degree is moderate, the activity of an enzyme system in the fresh leaves can be effectively improved, and a good foundation is laid for subsequent processing and the quality of finished tea. Fermentation is a key process for processing black tea, and a series of biochemical reactions centered on polyphenol enzyme oxidation can be promoted through proper fermentation, so that the contents of biochemical components such as FAA, CAF, TPs and the like are changed, and the specific flavor quality of the black tea is finally formed. Therefore, the judgment of the withering and fermentation degree is crucial to improving the quality of black tea.
Traditionally, tea makers have observed changes in tea color and aroma through subjective and empirical methods to determine the degree of withering and fermentation of tea leaves. However, this method requires a lot of time, manpower, and expertise. The evaluation result is easily influenced by factors such as the specialty of tea makers, mood and the like, and has no strict standard. In addition, biochemical analysis to determine the quality component content can be used for judging the withering and fermentation conditions. However, these chemical analysis methods are sample and time consuming and do not meet the requirements of modern production and monitoring systems. Therefore, a method and a system for rapidly and accurately judging the withering and fermentation degrees of tea leaves based on hyperspectral imaging are developed.
Disclosure of Invention
The invention provides a method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging. The hyperspectral imaging technology is used for acquiring hyperspectral data in the process of withering and fermentation of the fresh tea leaves, and TPs, FAA and CAF contents of all samples are measured. And (3) preprocessing the hyperspectral data by adopting a convolution smoothing method (S-G), a Multivariate Scattering Correction (MSC) and a 1-order derivative (1D) algorithm. A monitoring model of TPs, FAA and CAF content is established through machine learning and various algorithms, quantitative prediction of quality components in the withering and fermentation processes of fresh tea leaves can be achieved, and effective judgment of withering and fermentation degrees is achieved. The research lays a good foundation for nondestructive online detection of quality components in the tea withering and fermentation processes, and provides a new method for intelligently judging the tea withering and fermentation degrees.
In order to realize the purpose of the invention, the invention adopts the following technical scheme to realize:
the invention provides a method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging, which comprises the following steps:
s1: hyperspectral data in the tea withering and fermentation processes are collected;
s2: extracting the spectral reflectivity of the tea hyperspectral data obtained in the step S1;
s3: preprocessing the spectrum by combining the extracted spectral reflectivity;
s4: screening spectral characteristic wave bands of the preprocessed spectrum;
s5: and combining the processing results of the steps S2, S3 and S4, performing data modeling by using SVM, RF and PLS, and further verifying the data modeling.
Further, the step of collecting the tea withering and fermentation process data in the step S1 is as follows:
s11: measuring the content of Tea Polyphenols (TPs), free Amino Acids (FAA) and Caffeine (CAF) in tea;
s12: collecting hyperspectral data;
s13: black and white correction and normalization processing.
Further, in step S12, the hyperspectral camera has pixels of: 1101 × 960 (spatial × spectral) pixels.
Further, in the step S2, the spectral reflectance of the tea hyperspectral data is extracted by using the ENVI software.
Further, the algorithms used in the spectral preprocessing of step S3 include MSC algorithm, S-G algorithm and first derivative (1-D).
Further, the algorithm utilized in the step S4 of screening the spectral characteristic bands includes SPA, CARS and UVE.
Further, the step S5 is to perform data modeling by using SVM, RF, PLS, and further verify the data modeling, and the specific steps are as follows:
s51: dividing the tea hyperspectral data set into 5 parts by adopting 5-fold cross validation, taking 4 parts as training data and 1 part as test data in turn, repeating for 5 times, and then averaging the results;
s52: establishing a regression model for tea hyperspectral data and contents of TPs, FAA and CAF by using SVM, RF and PLS algorithms;
s53: the performance of the model was evaluated using the determination coefficient (R2), root Mean Square Error (RMSE), normalized Root Mean Square Error (NRMSE) and relative analytical error (RPD).
The discovery also provides a method and a system for judging the withering and fermentation degree of the fresh tea leaves based on hyperspectral imaging, and the method comprises the following steps:
the acquisition system comprises an imaging spectrum camera, a halogen lamp line light source, a computer and the like. The method is used for collecting data in the tea withering and fermentation processes;
a processing system to perform the following operations: inputting the collected data into a machine learning network and the like, and performing spectral reflectivity extraction, spectral preprocessing and spectral characteristic band screening;
and the analysis system is used for modeling and verifying the data according to the processing system, quantitatively judging the withering and fermentation degrees and providing a basis for intelligently judging the withering and fermentation degrees in the black tea processing process.
Compared with the prior art, the invention has the advantages and beneficial effects that:
according to the invention, a monitoring model of TPs, FAA and CAF content is constructed through machine learning and various algorithms, so that quantitative prediction of quality components in the withering and fermentation processes of fresh tea leaves can be realized, and effective judgment of the withering and fermentation degrees is realized. The research lays a good foundation for nondestructive online detection of quality components in the tea withering and fermentation processes, and provides a new method for intelligently judging the tea withering and fermentation degrees.
Drawings
FIG. 1 is a flow chart of a method for determining withering and fermentation degree of fresh tea leaves based on hyperspectral imaging.
Fig. 2 illustrates the noise reduction process of the built-in environment visualization program on the average hyperspectral data of all samples.
FIG. 3 is a high spectrum data characteristic wave band screening diagram of tea polyphenol, free amino acid and caffeine.
FIG. 4 is a correlation coefficient diagram of tea polyphenol, free amino acid and caffeine with a hyperspectral data model.
Detailed Description
The technical solution of the present invention will be further described in detail with reference to the following specific examples.
Example 1
A method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging are disclosed, a flow chart is shown in figure 1, and the method comprises the following steps:
step one, collecting data in the tea withering and fermentation processes:
the test was carried out at 2021, 9 and 30 days at 35 ℃ N, 119 ℃ 33' E, institute for tea science, sunshine, shandong province, china. During withering of fresh tea leaves, samples were taken 1 time per hour for a total of 19 times. And (3) taking a large amount of withered leaves for rolling when the withered leaves are withered for 16 hours, and then entering a fermentation process. During the fermentation, samples were taken 1 time every 0.5 hour for a total of 10 times. For each sampling, a hyperspectral camera is used for collecting spectral data, the sample is put into an oven to be dried to be dry enough, and then the sample is sealed and stored at the temperature of-4 ℃ in the dark. A total of 87 samples were collected for this test, each sample being repeated 3 times.
Step two, extracting the spectral reflectivity:
we first normalized the hyperspectral image after black and white correction. The hyperspectral image was opened in the image processing software Spec View (Dualix Spectral Imaging, china) and corrected using the analysis tool lens calibration and reflectance calibration.
Next, the spectral variables were extracted by ENVI5.3 (Research System Inc, america). Opening the corrected hyperspectral image in ENVI5.3, selecting the image Of the whole tea sample as a Region Of Interest (ROI), extracting the average reflection spectrum value Of the sample, and then obtaining the sample spectrum reflection curve. In total, a spectral matrix of 87 × 360 (number of samples × number of variables) is obtained.
Step three, preprocessing the spectrum by combining the extracted spectral reflectivity:
due to the influence of a hyperspectral acquisition instrument or environmental factors, the problems of scattering effect, random noise, system noise and the like exist in the original spectrum of the tea, the spectrum signal of the content of the tea can be weakened, and the establishment of a regression model is not facilitated. For this reason, we combined MSC, S-G and 1D 3 preprocessing algorithms to preprocess the raw spectral data of tea before modeling (fig. 2).
To eliminate artifacts or defective spectra in the data matrix, we use the MSC algorithm to make each spectrum closer to some "ideal" spectra; in order to obtain the optimal estimation value of a spectrum data point and effectively reduce the random noise of an average reflection spectrum, averaging or fitting is carried out on each point in a window range with a certain width of single-point spectrum data through S-G; to remove the process of baseline shift and separate the overlapping spectral peaks, we enhance a small amount of information in the spectrum by 1D, estimating the difference between two subsequent spectral data points. The arithmetic formula of the differential method-1D is shown in the formula (2).
Step four, screening spectral characteristic wave bands of the preprocessed spectrum:
in this study, we extracted spectral data for 360 bands. In order to improve the later modeling efficiency, representative bands in all spectral data are selected as 'characteristic bands' by using SPA, CARS and UVE3 algorithms, and bands which are not useful for the research are removed, so that the data operation amount is reduced (figure 3). Table 1 screening results for SPA, CARS and UVE algorithms.
TABLE 1 screening results for SPA, CARS and UVE algorithms
Step five, carrying out data modeling by using the CNN-GRU, and further verifying the data modeling:
we constructed a regression model between the spectral data of the tea sample and its quality components using 3 machine learning methods SVM, PLS and RF.
In the evaluation system of the model, we use a decision coefficient (R2), a Root Mean Square Error (RMSE), and a relative analytical error (RPD) to represent the effect of the prediction model. Wherein, the higher the R2 value is, the closer to 1 is, the higher the accuracy of the established model is; conversely, lower RMSE values, closer to 0, indicate higher accuracy of the model built; the RPD value is less than 1.4, which indicates poor prediction performance, and the RPD value is more than 1.4, which indicates that the RPD can be used for model analysis, and the larger the value, the more reliable the established model is. Table 2 results of regression models established for SVM, PLS and RF. In order to evaluate the inversion accuracy of each model, the measured values of each quality index in the test set were compared with the predicted values of the model, and the model stability of tea polyphenols, free amino acids and caffeine was verified (fig. 4).
TABLE 2 results of regression modeling of SVM, PLS and RF
The invention relates to a method and a system for judging the withering and fermentation degrees of fresh tea leaves based on hyperspectral imaging, which can quantitatively predict quality components in the withering and fermentation processes of the fresh tea leaves and realize effective judgment of the withering and fermentation degrees. The traditional withering and fermentation degree measurement depends on manual means and empirical judgment, so that misjudgment is easy to occur, and the detection efficiency is low. Therefore, the method combines the hyperspectral imaging technology with machine learning and is applied to withering and fermentation degree judgment and quality component content measurement.
The above examples are only intended to illustrate the technical solution of the present invention, and not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be apparent to those skilled in the art that various changes may be made and equivalents may be substituted for elements thereof; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions.
Claims (8)
1. A method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging are characterized by comprising the following steps:
s1: hyperspectral data in the tea withering and fermentation processes are collected;
s2: extracting the spectral reflectivity of the tea hyperspectral data obtained in the step S1;
s3: preprocessing the spectrum by combining the extracted spectral reflectivity;
s4: screening spectral characteristic wave bands of the preprocessed spectrum;
s5: and combining the processing results of the steps S2, S3 and S4, performing data modeling by using SVM, RF and PLS, and further verifying the data modeling.
2. The method and system for determining the withering and fermentation degree of fresh tea leaves based on hyperspectral imaging according to claim 1 are characterized in that the step of collecting hyperspectral data of the withering and fermentation processes of the tea leaves in the step S1 comprises the following steps:
s11: measuring the content of Tea Polyphenols (TPs), free Amino Acids (FAA) and Caffeine (CAF) in tea;
s12: collecting hyperspectral data;
s13: black and white correction and normalization processing.
3. The hyperspectral imaging based method and system for determining the withering and fermentation degree of fresh tea leaves according to claim 1, wherein in the step S12, the hyperspectral camera has pixels as follows: 1101 × 960 (spatial × spectral) pixels.
4. The method and system for determining the withering and fermentation degree of fresh tea leaves based on hyperspectral imaging according to claim 1 are characterized in that the step S2 is to extract the spectral reflectance of hyperspectral data of the tea leaves by utilizing ENVI software.
5. The method and system for determining the withering and fermentation degree of fresh tea leaves based on hyperspectral imaging as claimed in claim 1, wherein the algorithms used in the spectral preprocessing of step S3 comprise MSC algorithm, S-G algorithm and first derivative (1-D).
6. The hyperspectral imaging-based method and system for determining the withering and fermentation degree of fresh tea leaves according to claim 1 are characterized in that the algorithm for screening the spectral characteristic bands in the step S4 comprises SPA, CARS and UVE.
7. The method and system for determining the withering and fermentation degree of fresh tea leaves based on hyperspectral imaging as claimed in claim 1, wherein the step S5 is to use SVM, RF, PLS to perform data modeling and further verify, and the specific steps are as follows:
s51: adopting 5-fold cross validation, dividing the tea hyperspectral data set into 5 parts, taking 4 parts as training data in turn, 1 part is taken as test data, the test data is repeated for 5 times, and then the results are averaged;
s52: establishing a regression model for tea hyperspectral data and contents of TPs, FAA and CAF by using SVM, RF and PLS algorithms;
s53: using a determined coefficient (R) 2 ) The performance of the model was evaluated as Root Mean Square Error (RMSE), normalized Root Mean Square Error (NRMSE) and relative analytical error (RPD).
8. A method and a system for judging withering and fermentation degree of fresh tea leaves based on hyperspectral imaging are characterized by comprising the following steps:
an acquisition system: the device comprises an imaging spectrum camera, a halogen lamp line light source, a computer and the like, and is used for collecting data of the tea withering and fermentation processes;
the processing system comprises: the following operations are performed: inputting the collected data into a machine learning network and the like, and performing spectral reflectivity extraction, spectral preprocessing and spectral characteristic band screening;
an analysis system: according to the processing system, data are modeled and verified, and the data are used for quantitatively judging the withering and fermentation degrees, so that a basis is provided for intelligently judging the withering and fermentation degrees in the black tea processing process.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211593947.9A CN115855836A (en) | 2022-12-13 | 2022-12-13 | Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211593947.9A CN115855836A (en) | 2022-12-13 | 2022-12-13 | Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115855836A true CN115855836A (en) | 2023-03-28 |
Family
ID=85672294
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211593947.9A Pending CN115855836A (en) | 2022-12-13 | 2022-12-13 | Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115855836A (en) |
-
2022
- 2022-12-13 CN CN202211593947.9A patent/CN115855836A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110763698B (en) | Hyperspectral citrus leaf disease identification method based on characteristic wavelength | |
| Yang et al. | Quantitative prediction and visualization of key physical and chemical components in black tea fermentation using hyperspectral imaging | |
| Huang et al. | Rapid and nondestructive prediction of amylose and amylopectin contents in sorghum based on hyperspectral imaging | |
| CN113237836A (en) | Flue-cured tobacco leaf moisture content estimation method based on hyperspectral image | |
| CN111443043B (en) | Hyperspectral image-based walnut kernel quality detection method | |
| CN108169165B (en) | Maltose mixture quantitative analysis method based on terahertz spectrum and image information fusion | |
| CN115575405B (en) | Fruit appearance quality detection method based on multispectral image feature quantity | |
| CN111795943A (en) | Method for nondestructive detection of exogenous doped sucrose in tea based on near infrared spectrum technology | |
| CN110779875A (en) | Method for detecting moisture content of winter wheat ear based on hyperspectral technology | |
| CN114527082A (en) | Sweet corn seed germination prediction method and device | |
| CN110672578A (en) | Model universality and stability verification method for polar component detection of frying oil | |
| Wen et al. | Visualization of chlorophyll content distribution in apple leaves based on hyperspectral imaging technology | |
| CN113627709A (en) | System and method for evaluating gorgeous quality based on infrared spectrum detection | |
| CN113076692A (en) | Method for inverting nitrogen content of leaf | |
| Qian et al. | Development and application of crop monitoring system for detecting chlorophyll content of tomato seedlings | |
| CN116994126A (en) | Crop leaf area index obtaining method and device based on canopy reflectivity spectrum | |
| CN115855836A (en) | Hyperspectral imaging-based fresh tea leaf withering and fermentation degree judgment method and system | |
| Zhou et al. | Hyperspectral imaging technology for detection of moisture content of tomato leaves | |
| CN113916824B (en) | Alpha in milk s1 Intermediate infrared rapid batch detection method of casein | |
| CN113049526B (en) | Corn seed moisture content determination method based on terahertz attenuated total reflection | |
| Chen et al. | Quality detection and variety classification of pecan seeds using hyperspectral imaging technology combined with machine learning | |
| CN111289516B (en) | Method and device for detecting amino acid content of plant leaves | |
| CN113310933A (en) | Spectrum identification method for number of days for storing raw buffalo milk | |
| CN117788960B (en) | Tobacco leaf turning judging method, medium and system based on double models | |
| CN116385784A (en) | Method and system for measuring and calculating chlorophyll content of rice under cadmium stress |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |