Ramie fiber maturity detection method and system
By processing spectral torsional torque and fractional derivatives, combined with Fourier transform and support vector regression models, the problem of low accuracy in ramie fiber maturity detection was solved, achieving higher detection accuracy and model stability.
Patent Information
- Application Number
- CN202511733056.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-24
- Publication Date
- 2026-02-10
AI Technical Summary
In the current technology for detecting the maturity of ramie fibers, the fixed integer-order derivative cannot adapt to the spectral background and noise levels of different samples, resulting in low detection accuracy.
A method for detecting the maturity of ramie fibers was established by calculating the fractional derivative order using spectral torsional torque and combining it with Fourier transform and support vector regression models, based on the characteristics of spectral torsional torque and the high-low frequency energy ratio.
This improved the accuracy and stability of ramie fiber maturity detection, overcame the limitations of fixed frequency band division, and enhanced the reliability of detection results.
Smart Images
Figure CN121499418A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of maturity detection, and in particular to a ramie fiber maturity detection method and system. BACKGROUND
[0002] Ramie is an important natural cellulose fiber, and its maturity is a key physical index for evaluating quality and determining textile processing performance. In actual production, the use of ramie fibers with different maturity levels will cause a series of problems, such as spinning difficulty, increased breakage rate, fabric defects, and uneven dyeing, which seriously affect the overall quality and economic benefits of textiles.
[0003] Currently, there are three main traditional methods for detecting the maturity of ramie fibers: one is the artificial sensory evaluation method, which relies too much on the experience of inspectors and has poor repeatability; the second is the chemical method, which is relatively accurate but time-consuming and tedious; and the third is the microscope method, which requires complex sample preparation and results in low detection efficiency. In order to overcome the above-mentioned defects, existing technologies have begun to use spectral information for detection. This type of method usually first collects the spectrum of the fiber, then uses fixed pretreatment methods such as smoothing and differentiation to denoise and enhance features, and finally establishes a maturity prediction model combined with chemometrics methods.
[0004] However, this method still has limitations in practical application, especially the traditional integer order differentiation, which is a fixed pretreatment method and cannot adapt to different spectral backgrounds and noise levels of samples with different maturity or batches. This "one-size-fits-all" approach cannot effectively suppress noise while enhancing effective signals, which may result in the loss or drowning of key feature information. In addition, it also cannot adapt to the differences in frequency energy distribution caused by changes in sample spectral morphology, ultimately affecting the detection accuracy of ramie fiber maturity. SUMMARY
[0005] The present application provides a ramie fiber maturity detection method and system to solve the above-mentioned problem of existing technologies that cannot effectively suppress noise while enhancing effective signals, resulting in low detection accuracy of ramie fiber maturity.
[0006] In a first aspect, the ramie fiber maturity detection method of the present application comprises the following steps: Obtaining spectral data of the ramie fiber sample to be tested; calculating the spectral torsion moment of the spectral data, which is the integral of the product of the intensity value at each wavelength point in the spectral data and the distance from the wavelength point to the preset center wavelength along the wavelength axis. determine a fractional derivative order based on the spectral torsion moment through a preset nonlinear mapping function; process the spectral data using the fractional derivative order to obtain a differential spectral curve; perform Fourier transform on the differential spectral curve to obtain a frequency domain signal; determine a cumulative energy percentile threshold based on the spectral torsion moment, and divide the frequency domain signal into a low frequency band and a high frequency band by taking a frequency at which a cumulative integral value of energy of the frequency domain signal from low frequency to high frequency reaches the cumulative energy percentile threshold as a segmentation frequency; calculate energy integrals of the low frequency band and the high frequency band respectively, and calculate a ratio of the two to obtain a high-low frequency energy ratio; input the high-low frequency energy ratio and the spectral torsion moment as features, and determine the maturity of the ramie fiber sample through a preset support vector regression model.
[0007] Preferably, the spectral data of the to-be-tested ramie fiber sample is obtained by: scanning the ramie fiber sample in a wavelength range of 400 nm to 2500 nm with a sampling interval of 2 nm using a near-infrared spectrometer to collect a diffuse reflection spectrum of the ramie fiber sample as the spectral data.
[0008] Preferably, the spectral torsion moment of the spectral data is calculated by: setting the center wavelength 1450 nm of the spectral data collection range as the preset center wavelength.
[0009] Preferably, the nonlinear mapping function is: ; wherein, the fractional derivative order is, the spectral torsion moment is.
[0010] Preferably, the spectral data is processed using the fractional derivative order to obtain a differential spectral curve, comprising: calculating the fractional derivative of the spectral data using the Grunwald-Letnikov definition.
[0011] Preferably, the cumulative energy percentile threshold is determined based on the spectral torsion moment, comprising: determining the cumulative energy percentile threshold using a piecewise linear function , the piecewise linear function is: ; wherein, the spectral torsion moment is.
[0012] Preferably, the maturity of the ramie fiber sample is determined by the preset support vector regression model, and the method comprises the following steps of: The kernel function of the support vector regression model is set as a Gaussian radial basis kernel function, and the penalty parameter C of the support vector regression model is set as 10, and the kernel function parameter is 0.1.
[0013] Preferably, the frequency domain signal is divided into a low frequency band and a high frequency band, and the method comprises the following steps of: The frequency domain signal below the division frequency is divided into a low frequency band, and the frequency domain signal above the division frequency is divided into a high frequency band.
[0014] Preferably, the energy integration of the low frequency band is the energy integration from the lowest frequency of the low frequency band to the division frequency, and the energy integration of the high frequency band is the energy integration from the division frequency to the highest frequency of the high frequency band.
[0015] In a second aspect, the ramie fiber maturity detection system provided by the present application comprises a memory and a processor, and the memory stores computer instructions, and the processor executes the computer instructions to realize the ramie fiber maturity detection method.
[0016] The present application has the following advantages: the present application uses the spectral torsion moment as a characteristic parameter to represent the overall shape of the spectrum, and determines the order of the fractional differential and the division threshold of the frequency energy according to the spectral torsion moment. By using a specific order of fractional differential determined for different sample spectra, the spectrum is finely processed, which can enhance the weak features related to maturity compared with the traditional fixed integer order differential, and suppress noise interference. By using a division frequency determined according to the energy distribution of the spectrum to calculate the high-low frequency energy ratio, the feature is more representative, and the limitation of fixed frequency band division is overcome. The spectral torsion moment representing the global shape of the spectrum and the energy ratio feature representing the local details are fused, which provides more comprehensive and complementary information for the prediction model, thereby improving the reliability of the ramie fiber maturity detection result and the stability of the model. BRIEF DESCRIPTION OF DRAWINGS
[0017] Figure 1 A flowchart of the ramie fiber maturity detection method provided by the embodiment of the present application is shown. DETAILED DESCRIPTION
[0018] The embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings. The embodiments described below by referring to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as a limitation of the present application.
[0019] As Figure 1 shown, the embodiment of the ramie fiber maturity detection method provided by the present application comprises the following steps: S1, obtaining spectral data of a sample of ramie fiber to be tested; calculating a spectral torsion moment of the spectral data, the spectral torsion moment being an integral of a product of an intensity value at each wavelength point in the spectral data and a distance from the wavelength point to a preset central wavelength along a wavelength axis.
[0020] Specifically, the sample of ramie fiber to be tested is placed on a sample stage of a near-infrared spectrometer, scanned in an integrating sphere diffuse reflection mode, and near-infrared spectra in a wavelength range of 4000 to 10000 wave numbers are collected, with a spectral resolution of 8 wave numbers and a scanning number of 64, and an average spectrum is taken as the original spectral data. The preset central wavelength is the center point of the spectral acquisition range, for example, 7000 wave numbers. The calculation of the spectral torsion moment is as follows: first, the distance of each wavelength point from the central wavelength is calculated, then the distance is multiplied by the spectral intensity value corresponding to the wavelength point, and finally the product results at all wavelength points are accumulated and summed to obtain the spectral torsion moment.
[0021] S2, determining a fractional derivative order based on the spectral torsion moment through a preset nonlinear mapping function; processing the spectral data using the fractional derivative order to obtain a differential spectral curve; and performing Fourier transform on the differential spectral curve to obtain a frequency domain signal.
[0022] Specifically, the preset nonlinear mapping function is a function relationship established by training a large number of samples of known maturity. The function relationship takes the spectral torsion moment as input and outputs the optimal fractional derivative order. For example, when the spectral torsion moment is small, a smaller fractional derivative order corresponds, and as the spectral torsion moment increases, the order non-linearly increases and tends to an upper limit value. The specific parameters of the function relationship are determined by grid search in the order range of 0.1 to 1.9 and with the highest final model prediction accuracy as the target.
[0023] The Grunwald-Letnikov definition is used to calculate the fractional order derivative of the spectral data. The Grunwald-Letnikov definition, referred to as G-L definition for short, is a numerical calculation method for generalizing integer order derivative to fractional order. It represents the fractional order derivative of a function at a point as a weighted sum of the function values at the point and all previous points. In this embodiment, the function to be processed is the spectral data of ramie fibers, and the independent variable is the wavelength λ. Assuming that the determined fractional order derivative order v is 0.8, to calculate the 0.8 order derivative value of the spectral data at the wavelength point λ, the spectral intensity values at the point and a series of previous wavelength points, such as the intensity values at the positions of λ, λ-2 nm, λ-4 nm, etc., are needed. The G-L definition gives a summation formula containing binomial coefficients, which are determined by the determined fractional order derivative order v. By substituting the spectral data sequence and the corresponding G-L coefficients into the formula, the derivative result at the wavelength λ can be calculated. Repeating this calculation process for each wavelength point in the spectral range results in a series of new data points, which together constitute the fractional order derivative spectrum curve. The curve can amplify the subtle feature information in the original spectrum and suppress the baseline drift.
[0024] Then, the generated fractional order derivative spectrum curve is subjected to fast Fourier transform to convert the data from the wavelength domain to the frequency domain, obtaining a complex sequence containing amplitude and phase, and calculating the energy spectral density from the complex sequence, which is the frequency domain signal.
[0025] S3, based on the spectral torsion moment, determining a cumulative energy percentile threshold, and taking the frequency at which the cumulative integral value of the energy of the frequency domain signal from low frequency to high frequency reaches the cumulative energy percentile threshold as the segmentation frequency, dividing the frequency domain signal into a low frequency band and a high frequency band; calculating the energy integrals of the low frequency band and the high frequency band respectively, and calculating the ratio of the two to obtain a high-low frequency energy ratio.
[0026] Specifically, the cumulative energy percentile threshold is determined by a preset function (for example, a linear relationship) with the spectral torsion moment as the input. In the calculation, the total energy of the frequency domain signal is first calculated. Then, the energy is cumulatively integrated from the lowest frequency, and when the ratio of the cumulative energy to the total energy is equal to the cumulative energy percentile threshold, the corresponding frequency is the segmentation frequency. The frequency domain signal below the segmentation frequency is divided into a low frequency band, and the frequency domain signal above the segmentation frequency is divided into a high frequency band. The energy integral of the low frequency band is the energy integral from the lowest frequency of the low frequency band to the segmentation frequency, and the energy integral of the high frequency band is the energy integral from the segmentation frequency to the highest frequency of the high frequency band. The calculation formula of the high-low frequency energy ratio is the energy integral of the high frequency band divided by the energy integral of the low frequency band.
[0027] S4, inputting the high-low frequency energy ratio and the spectral torsion moment as features into a preset support vector regression model to determine the maturity of the ramie fiber sample.
[0028] Specifically, the preset support vector regression model is obtained by training a large number of ramie fiber samples with known maturity. During the training, a two-dimensional feature vector composed of the calculated spectral torsion moment and high-low frequency energy ratio of each sample is used as the input of the model; at the same time, the actual maturity value corresponding to the sample is used as the output label. The model adopts a radial basis kernel function, and the optimal penalty parameter and kernel parameter are determined by grid search and cross-validation, so as to establish a regression model capable of predicting the maturity of ramie fiber according to the input feature vector.
[0029] In an optional embodiment, the acquisition of the spectral data of the to-be-tested ramie fiber sample comprises: The near-infrared spectrometer is used to scan the ramie fiber sample in the wavelength range of 400 nm to 2500 nm with a sampling interval of 2 nm, and the diffuse reflectance spectrum of the ramie fiber sample is collected as the spectral data.
[0030] For example, first, prepare the to-be-tested ramie fiber sample, for example, take 5 g of cleaned and dried ramie fiber, and place the ramie fiber flatly in the sample cup of the near-infrared spectrometer, ensuring that the sample surface is flat and uniformly filled. Then, set the acquisition parameters on the operation software of the spectrometer, set the starting wavelength of the spectral scan to 400 nm, the ending wavelength to 2500 nm, and the sampling interval to 2 nm. The instrument will measure one by one at the wavelength points of 400 nm, 402 nm, 404 nm, etc., until 2500 nm, a total of 1051 data points are collected.
[0031] After starting the scan, the near-infrared light emitted by the instrument light source irradiates the surface of the ramie fiber sample, and the sample selectively absorbs and scatters light of different wavelengths according to the chemical composition and physical structure. The detector is responsible for receiving the light signal diffusely reflected from the sample surface and measuring the intensity of the light signal at each wavelength point. For example, the instrument records that the diffuse reflectance at 1200 nm is 0.6 and the diffuse reflectance at 1202 nm is 0.62. All wavelength points and corresponding diffuse reflectance data are combined to form the original diffuse reflectance spectrum of the ramie fiber sample, providing basic data for subsequent analysis.
[0032] In an optional embodiment, the calculation of the spectral torsion moment of the spectral data comprises: The center wavelength 1450 nm of the spectral data acquisition range is set as the preset center wavelength.
[0033] For example, the spectral torsion is a physical quantity used to represent the asymmetry of the spectral curve. When calculating this value, a reference center point, i.e., the center wavelength, needs to be set. In the present embodiment, since the collection range of the spectral data is from 400 nm to 2500 nm, the arithmetic center point of the collection range is 1450 nm. Therefore, 1450 nm is set as the preset center wavelength. In the calculation process, with 1450 nm as the reference center point, each data point on the spectral curve is assigned a weight equal to the difference between the wavelength value and the center wavelength 1450 nm. The spectral intensity of all data points on the left side of the center wavelength is multiplied by the corresponding weight and summed to obtain a torsion value; similarly, the torsion value of all data points on the right side of the center wavelength is calculated. The total torsion on the right side is subtracted from the total torsion on the left side, and the difference obtained is the spectral torsion of the entire spectrum. If the spectral energy is mainly distributed in the long-wave band, i.e., greater than 1450 nm, the torsion is positive, and vice versa.
[0034] In an optional embodiment, the nonlinear mapping function is: ; wherein, is the fractional order derivative order, is the spectral torsion.
[0035] In an optional embodiment, the determining the cumulative energy percentile threshold based on the spectral torsion comprises: determining the cumulative energy percentile threshold using a piecewise linear function , the piecewise linear function is: ; wherein, is the spectral torsion.
[0036] In an optional embodiment, the determining the maturity of the ramie fiber sample by the preset support vector regression model comprises: setting the kernel function of the support vector regression model as a Gaussian radial basis kernel function, and setting the penalty parameter C of the support vector regression model as 10, and the kernel function parameter is 0.1.
[0037] Specifically, the spectrum torsion moment calculated in the previous step and the high-low frequency energy ratio are combined as a feature, which is input into a support vector regression (SVR) model that has been trained. The model is used to establish a nonlinear regression relationship between the combined feature and the maturity of ramie fibers. The key configuration parameters of the model are explicitly specified in this embodiment to ensure the stability and reproducibility of the prediction. The kernel function of the model is set to a Gaussian radial basis kernel function, which is good at handling complex nonlinear problems. The penalty parameter C of the model is set to 10, which controls the degree of punishment for samples that exceed the tolerance boundary. The larger the value, the lower the tolerance of the model to training errors. The kernel function parameter is set to 0.1, which determines the size of the influence range of a single training sample. When the feature spectrum data of the sample to be tested is input into this support vector regression (SVR) model, the model will calculate according to the internal learned decision function and output a specific value, such as 88.5, which is the predicted maturity value of the ramie fiber.
[0038] The implementation principle of the ramie fiber maturity detection method of the embodiment of the present application is as follows: the present application uses the spectrum torsion moment as a feature parameter to represent the overall shape of the spectrum, and determines the order of fractional order differentiation and the segmentation threshold of frequency energy based on this. The present application processes the spectrum by a specific fractional order differentiation order determined for different sample spectra, which can enhance the weak features related to maturity and suppress noise interference compared to traditional fixed integer order differentiation. Moreover, the present application calculates the high-low frequency energy ratio by a segmentation frequency determined according to the energy distribution of the spectrum, making the feature more representative and overcoming the limitations of fixed frequency band division. Moreover, the present application fuses the spectrum torsion moment representing the global shape of the spectrum and the energy ratio feature representing the local details to provide more comprehensive and complementary information for the prediction model, thereby improving the reliability of the ramie fiber maturity detection result and the stability of the model.
[0039] The embodiment of the ramie fiber maturity detection system provided by the present application includes a memory and a processor, and the memory stores computer instructions. When the processor executes the computer instructions, the ramie fiber maturity detection method in the above embodiment is realized.
[0040] Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application. Those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.
Claims
1. A method for detecting the maturity of ramie fiber, characterized in that, Includes the following steps: Acquire the spectral data of the ramie fiber sample to be tested; calculate the spectral torsional torque of the spectral data, wherein the spectral torsional torque is the integral of the product of the intensity value at each wavelength point in the spectral data and the distance from the wavelength point to the preset center wavelength along the wavelength axis; Based on the spectral torsional torque, the fractional derivative order is determined through a preset nonlinear mapping function; the spectral data is processed using the fractional derivative order to obtain a differential spectral curve; and a Fourier transform is performed on the differential spectral curve to obtain a frequency domain signal. Based on the spectral torsional torque, the cumulative energy percentile threshold is determined, and the frequency at which the cumulative integral value of the energy of the frequency domain signal from low frequency to high frequency reaches the cumulative energy percentile threshold is used as the segmentation frequency to divide the frequency domain signal into a low-frequency band and a high-frequency band; the energy integrals of the low-frequency band and the high-frequency band are calculated respectively, and the ratio between the two is calculated to obtain the high-low frequency energy ratio. Using the high-low frequency energy ratio and the spectral torsional torque as feature inputs, the maturity of ramie fiber samples is determined through a preset support vector regression model.
2. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The acquisition of spectral data of the ramie fiber sample to be tested includes: The ramie fiber sample was scanned using a near-infrared spectrometer in the wavelength range of 400 nm to 2500 nm at a sampling interval of 2 nm, and the diffuse reflectance spectrum of the ramie fiber sample was collected as spectral data.
3. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The calculation of the spectral torsional torque of the spectral data includes: Set the center wavelength of the spectral data acquisition range, 1450nm, as the preset center wavelength.
4. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The nonlinear mapping function is: ； in, For the fractional derivative order, The spectral torsional torque is given.
5. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The process of processing the spectral data using the fractional derivative order to obtain the differential spectral curve includes: The fractional derivative of the spectral data was calculated using the Grunwald-Letnikov definition.
6. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The determination of the cumulative energy percentile threshold based on the spectral torsional torque includes: The cumulative energy percentile threshold is determined using a piecewise linear function. The piecewise linear function is: ； in, The spectral torsional torque is given.
7. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The process of determining the maturity of ramie fiber samples using a pre-defined support vector regression model includes: The kernel function of the support vector regression model is set to the Gaussian radial basis function kernel function, and the penalty parameter C of the support vector regression model is set to 10. The kernel function parameters are... It is 0.
1.
8. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The step of dividing the frequency domain signal into a low-frequency band and a high-frequency band includes: In the frequency domain, frequencies below the cutoff frequency are classified as low-frequency bands, and frequencies above the cutoff frequency are classified as high-frequency bands.
9. The method for detecting the maturity of ramie fiber according to claim 1, characterized in that, The energy integral of the low-frequency band is the energy integral from the lowest frequency of the low-frequency band to the segmented frequency, and the energy integral of the high-frequency band is the energy integral from the segmented frequency to the highest frequency of the high-frequency band.
10. A ramie fiber maturity detection system, characterized in that, It includes a memory and a processor. The memory stores computer instructions, and when the processor executes the computer instructions, it implements the ramie fiber maturity detection method as described in any one of claims 1-9.