A fabric material identification method based on near-infrared spectrum

Abstract

The application discloses a fabric material identification method based on near-infrared spectroscopy, and the method comprises the following steps: acquiring multiple near-infrared spectrum data of a fabric to be measured, performing first-order derivation processing on the data, performing normalization processing on part of wave bands of one of the first-order derivation data, performing PCA dimension reduction processing on all the first-order derivation data, combining the characteristic data obtained through the normalization processing and the dimension reduction processing, and then performing MLP classification processing to realize identification of the fabric material. The application has the characteristics of high identification accuracy, high speed, simplicity, convenience, higher intelligence, and no damage to the fabric.

Description

Technical Field

[0001] This invention belongs to the field of fabric material identification technology, specifically, it relates to a fabric material identification method based on near-infrared spectroscopy. Background Technology

[0002] Generally, clothes that need washing are made of various materials, including pure cotton, linen, synthetic fibers, silk, wool, down, woolen fabrics, leather, blends, mercerized cotton, chiffon, and polyester, etc. Each material requires different washing time, washing intensity, washing volume, and rinsing times. Therefore, the selection of a washing mode depends on the material and quantity of the clothes to be washed. Currently, users typically identify the material themselves or by mixing the clothes with water and heating it, then determining the material based on a temperature change table of the mixture over a certain time and the corresponding material, or by using a material sensor. However, these methods are not always accurate in identifying the material, especially since the clothes may be contaminated with other substances, leading to incorrect material identification, incorrect washing mode selection, and unnecessary damage to the clothes.

[0003] Qualitative and quantitative analysis of textile fibers contained in clothing is the key to the intelligent upgrading of clothing washing and care technology. With the advent of the AI ​​era, people have higher and higher expectations for the intelligence of washing and care equipment. Washing and care equipment can automatically and accurately obtain the material information of the clothes being washed and adjust the washing and care program accordingly, which can bring users a more intelligent washing and care experience. The main methods for identifying materials on the market are RFID system and water absorption rate judgment. (1) RFID system: Due to the popularity of clothing with built-in RFID electronic tags, the application scope of this technology is limited; at the same time, the high cost of tags and the high complexity of the construction of the clothing network have become the difficulties restricting the development of this technology; (2) Water absorption rate judgment of clothing material is the most widely used method: based on the principle of different water absorption rates of clothing, the clothing material is approximately and vaguely distinguished, but it cannot accurately obtain the type and content of clothing fibers, and it belongs to the destructive identification method in advance.

[0004] Chinese patent application CN201910389465.3 discloses a material identification device, method, washing machine, and control method thereof. The device includes a housing, a monochromatic light source, a spectral sensor, a filter, a diffuse reflector, a diffuse reflector driving module, and an identification module. The filter is mounted on the housing. The diffuse reflector is initially positioned between the monochromatic light source and the filter. Light emitted from the monochromatic light source is reflected by the diffuse reflector, and the spectral sensor obtains a reference spectrum. The diffuse reflector driving module drives the diffuse reflector to move away from its initial position, so that light emitted from the monochromatic light source is reflected by the object to be identified after passing through the filter, and the spectral sensor obtains a sample spectrum. The identification module identifies the material of the object to be identified based on the reference spectrum and the sample spectrum. The reference spectrum is based on ambient light. Processing the sample spectrum with its corresponding reference spectrum yields a standard spectrum that excludes the influence of ambient light, resulting in higher accuracy in material identification. However, clothing materials are diverse and their content varies greatly, making accurate identification and characterization difficult when spectral data processing is inadequate.

[0005] In view of this, the present invention is proposed. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a fabric material identification method based on near-infrared spectroscopy. By merging the feature data obtained by normalization and dimensionality reduction processing, the fabric material is identified. The spectral data is optimized, which improves the accuracy of fabric material identification and does not damage the fabric.

[0007] To solve the above-mentioned technical problems, the basic concept of the technical solution adopted by the present invention is: a method for identifying fabric material based on near-infrared spectroscopy, including acquiring multiple near-infrared spectral data of the fabric to be tested, performing first-order derivative processing on the data, and performing normalization and dimensionality reduction processing on the first-order derivative data respectively, merging the feature data obtained by normalization and dimensionality reduction processing, and then performing fabric material identification.

[0008] Further steps include the following:

[0009] S1: Acquire multiple near-infrared spectral data of the fabric to be tested, and perform first-order derivative processing on the data;

[0010] S2: Normalize a portion of the first-order derivative data to generate the first feature data;

[0011] S3: Perform PCA dimensionality reduction on all first-order derivative data to generate second feature data;

[0012] S4: Merge the first feature data and the second feature data into new feature data, perform MLP classification on the new feature data, and realize the identification of fabric material.

[0013] In step S1, the near-infrared spectral data is processed by first-order derivative. The main purpose is to remove the influence of amplitude on the spectral curve, preserve the overall characteristics of the curve shape, improve the validity of the data, and facilitate subsequent processing.

[0014] The normalization and PCA dimensionality reduction processes performed on the first-order derivative data in steps S2 and S3 are to simplify the data and facilitate subsequent processing.

[0015] In step S4, the first feature data and the second feature data are merged into new feature data. The new feature data contains all feature points after different processing of the near-infrared spectral data. The processing simplifies the data, eliminates interference from other factors, and facilitates rapid and accurate identification of fabric materials.

[0016] Furthermore, step S2 includes:

[0017] S21: Select one of the first-order derivative data, obtain the wavelength value of the zero point or the closest point to the zero point of the first-order derivative data, and perform segmentation processing on the bands of the near-infrared spectral data to obtain several wavelength sequences.

[0018] S22: Normalize wavelength sequences containing zero or wavelength values ​​closest to zero, and calculate the normalized values; record wavelength sequences without zero or wavelength values ​​closest to zero as 0.

[0019] S23: Arrange the normalized values ​​and 0 in the order of the wavelength sequence to generate the first feature data.

[0020] Wavelength sequences containing zero or closest to zero wavelength values ​​are sensitive bands. Normalizing these wavelength sequences can help improve the accuracy of fabric material identification.

[0021] Because near-infrared spectral data are discrete points, there may be cases where there is no corresponding wavelength when the first derivative is zero. Therefore, the wavelength value closest to zero can also be selected.

[0022] This encoding method for generating the first feature data includes information on the sensitive bands and simplifies the data to facilitate subsequent processing and improve the accuracy of fabric material identification.

[0023] Furthermore, in step S22, the normalization calculation formula is:

[0024]

[0025] Where d is the normalized value, d x d represents the wavelength value at or closest to the zero of the first-order derivative data. maxd represents the maximum wavelength value in the wavelength sequence. min It is the minimum wavelength value in the wavelength sequence.

[0026] Furthermore, step S3 includes:

[0027] S31: Arrange all the first-order derivative data into an n-row, m-column first matrix;

[0028] S32: Zero-mean normalize each row of the first matrix to obtain the second matrix;

[0029] S33: Find the covariance matrix of the second matrix, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0030] S34: Arrange the corresponding eigenvectors into a matrix from top to bottom in descending order of eigenvalues, and take the first B rows to form a transformation matrix;

[0031] S35: Multiply the transformation matrix by the first matrix to obtain the second feature data.

[0032] Where m is the number of near-infrared spectral data acquired, n is the feature dimension of the first-order derivative data, and B is the feature dimension of the second feature data.

[0033] Furthermore, in step S32, when each row of the first matrix is ​​zero-mean normalized, the formula for calculating the average value of each row is:

[0034]

[0035] Among them, X ij μ represents the data corresponding to each row in the first matrix. i Let i be the average value of each row of data, i is (1,2,……,n) and j is (1,2,……,m).

[0036] Furthermore, the covariance matrix in step S33 is:

[0037]

[0038] Where C is the covariance matrix, X' is the second matrix, and X' T This is the transpose of the second matrix.

[0039] Furthermore, in step S4, when merging the first feature data and the second feature data into new feature data, the first feature data is arranged first and the second feature data is arranged last.

[0040] Alternatively, the first feature data can be arranged last, and the second feature data can be arranged first.

[0041] The feature dimension of the new feature data is A+B, where A is the feature dimension of the first feature data.

[0042] The feature dimensions A of the first feature data, B of the second feature data, and A+B of the new feature data are all smaller than the feature dimensions of the original near-infrared spectral data. This simplifies the data to a certain extent, making it easier for subsequent processing and improving the accuracy of fabric material identification.

[0043] Furthermore, in step S4, near-infrared spectral data of different fabric materials are obtained, and the MLP is trained to obtain the MLP model.

[0044] The feature dimensions A+B of the new feature data are input into the MLP model. Through the classification processing of the MLP model, the similarity percentage between the fabric to be identified and each fabric material is finally obtained. The fabric with the highest similarity percentage is the material of the fabric to be identified.

[0045] Furthermore, the near-infrared spectral data is obtained by irradiating a point on the fabric material with a near-infrared light source to obtain the absorption data of different wavelengths of light at that point.

[0046] There are generally two types of near-infrared spectroscopy detection: point detection and spectral imaging detection. Point detection has less background information, less data volume, less computational load during recognition calculation, and lower algorithm implementation difficulty compared to spectral imaging detection. This is conducive to faster and more accurate identification of fabric materials. Therefore, this application adopts point detection.

[0047] By adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0048] This invention provides a fabric material identification method based on near-infrared spectroscopy. By performing first-order derivative processing on near-infrared spectral data, and then normalizing and reducing the dimensions of the first-order derivative data, the feature data obtained from the normalization and dimensionality reduction processes are merged before fabric material identification. This optimization of spectral data improves the accuracy of fabric material identification without damaging the fabric.

[0049] The main purpose of performing first-order derivative processing on near-infrared spectral data is to remove the influence of amplitude on the spectral curve, preserve the overall characteristics of the curve shape, improve the validity of the data, and facilitate subsequent processing.

[0050] Normalizing and performing PCA dimensionality reduction on the first-order derivative data simplifies the data and facilitates subsequent processing.

[0051] Wavelength sequences containing zero or closest to zero wavelength values ​​are sensitive bands. Normalizing these wavelength sequences can help improve the accuracy of fabric material identification.

[0052] The first feature data and the second feature data are merged into new feature data. The new feature data contains all feature points after different processing of the near-infrared spectral data. The processing simplifies the data and eliminates interference from other factors, making it easier to quickly and accurately identify the fabric material.

[0053] This invention provides a fabric material identification method based on near-infrared spectroscopy, which features high identification accuracy, fast speed, simplicity and convenience, greater intelligence, and no damage to the fabric.

[0054] The specific embodiments of the present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0055] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention, but do not constitute an undue limitation of the invention. Obviously, the drawings described below are merely some embodiments, and those skilled in the art can obtain other drawings based on these drawings without creative effort. In the drawings:

[0056] Figure 1 This is a flowchart illustrating the fabric material identification method of the present invention;

[0057] Figure 2 This is a schematic diagram of the steps of the fabric material identification method of the present invention.

[0058] It should be noted that these accompanying drawings and textual descriptions are not intended to limit the scope of the invention in any way, but rather to illustrate the concept of the invention to those skilled in the art by referring to specific embodiments. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention are described clearly and completely below with reference to some embodiments. Those skilled in the art will understand that the following embodiments are only used to explain the technical solutions of this invention and are not intended to limit the scope of protection of this invention. For example, although this application describes the steps of the method of this invention in a specific order, these orders are not restrictive. Those skilled in the art can perform the steps in different orders without departing from the basic principles of this invention.

[0060] This invention provides a fabric material identification method based on near-infrared spectroscopy, including acquiring multiple near-infrared spectral data of the fabric to be tested, performing first-order derivative processing on the data, and performing normalization and dimensionality reduction processing on the first-order derivative data respectively, merging the feature data obtained by normalization and dimensionality reduction processing, and then performing fabric material identification.

[0061] The near-infrared spectral data is obtained by irradiating a point on the fabric material with a near-infrared light source to obtain the absorption data of different wavelengths of light at that point.

[0062] like Figure 1 As shown, the present invention provides a method for identifying fabric material based on near-infrared spectroscopy, comprising the following steps:

[0063] S1: Acquire multiple near-infrared spectral data of the fabric to be tested, and perform first-order derivative processing on the data;

[0064] S2: Normalize a portion of the first-order derivative data to generate the first feature data;

[0065] S3: Perform PCA dimensionality reduction on all first-order derivative data to generate second feature data;

[0066] S4: Merge the first feature data and the second feature data into new feature data, perform MLP classification on the new feature data, and realize the identification of fabric material.

[0067] In the above scheme, the first-order derivative processing of the near-infrared spectral data in step S1 mainly removes the influence of amplitude on the spectral curve, preserves the overall characteristics of the curve shape, which helps to improve the validity of the data and facilitates subsequent processing.

[0068] The normalization and PCA dimensionality reduction processes performed on the first-order derivative data in steps S2 and S3 are to simplify the data and facilitate subsequent processing.

[0069] In step S4, the first feature data and the second feature data are merged into new feature data. The new feature data contains all feature points after different processing of the near-infrared spectral data. The processing simplifies the data, eliminates interference from other factors, and facilitates rapid and accurate identification of fabric materials.

[0070] In the above scheme, step S2 includes:

[0071] S21: Select one of the first-order derivative data, obtain the wavelength value of the zero point or the closest point to the zero point of the first-order derivative data, and perform segmentation processing on the bands of the near-infrared spectral data to obtain several wavelength sequences.

[0072] S22: Normalize wavelength sequences containing zero or wavelength values ​​closest to zero, and calculate the normalized values; record wavelength sequences without zero or wavelength values ​​closest to zero as 0.

[0073] S23: Arrange the normalized values ​​and 0 in the order of the wavelength sequence to generate the first feature data.

[0074] In the above scheme, since near-infrared spectral data are discrete points, there may be cases where there is no corresponding wavelength when the first derivative is zero. Therefore, the wavelength value closest to zero can also be selected.

[0075] In the above scheme, the normalization calculation formula in step S22 is:

[0076]

[0077] Where d is the normalized value, d x d represents the wavelength value at or closest to the zero of the first-order derivative data. max d represents the maximum wavelength value in the wavelength sequence. min It is the minimum wavelength value in the wavelength sequence.

[0078] For example: the acquired near-infrared spectral data is Z = {x} 750 ,x 751 ,x 752 ,…,x 1049 ,x 1050}, where x 750 The light absorption intensity is 750nm, and 750nm~1050nm is the band of near-infrared spectral data.

[0079] After taking the first derivative of the near-infrared spectral data Z, take the wavelength value corresponding to the zero point of the first derivative data. For example, the first derivative is zero at 909 nm and 1049 nm.

[0080] The near-infrared spectral data is segmented into bands, for example, each 50nm is a wavelength sequence, namely 750nm~800nm, 800nm~850nm, 850nm~900nm, 900nm~950nm, 950nm~1000nm, and 1000nm~1050nm.

[0081] 909nm falls within the wavelength range of 900nm to 950nm. Normalization calculations are performed on this wavelength range.

[0082]

[0083] 1049nm falls within the wavelength range of 1000nm to 1050nm. Normalization calculations are performed on this wavelength range.

[0084]

[0085] Other wavelength sequences that do not contain zero-point wavelength values ​​are denoted as 0, and are arranged sequentially with the normalized values ​​calculated above according to the order of the wavelength sequences to generate the first feature data [0,0,0,0.18,0,0.98]. The feature dimension A of the first feature data is 6.

[0086] In the above scheme, step S3 includes:

[0087] S31: Arrange all the first-order derivative data into an n-row, m-column first matrix;

[0088] S32: Zero-mean normalize each row of the first matrix to obtain the second matrix;

[0089] S33: Find the covariance matrix of the second matrix, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0090] S34: Arrange the corresponding eigenvectors into a matrix from top to bottom in descending order of eigenvalues, and take the first B rows to form a transformation matrix;

[0091] S35: Multiply the transformation matrix by the first matrix to obtain the second feature data.

[0092] Where m is the number of near-infrared spectral data acquired, n is the feature dimension of the first-order derivative data, and B is the feature dimension of the second feature data.

[0093] In the above scheme, when zero-meaning is applied to each row of the first matrix in step S32, the formula for calculating the average value of each row is as follows:

[0094]

[0095] Among them, X ij μ represents the data corresponding to each row in the first matrix. i Let i be the average value of each row of data, i is (1,2,……,n) and j is (1,2,……,m).

[0096] For example: m near-infrared spectral data were acquired, and the feature dimension of each near-infrared spectral data is n, that is, it contains the absorption intensity corresponding to n wavelength values.

[0097] The first matrix is:

[0098] The second matrix is:

[0099] Furthermore, the covariance matrix in step S33 is:

[0100]

[0101] Where C is the covariance matrix, X' T This is the transpose of the second matrix.

[0102] In the above scheme, in step S4, when merging the first feature data and the second feature data into new feature data, the first feature data is arranged first and the second feature data is arranged last.

[0103] Alternatively, the first feature data can be arranged last, and the second feature data can be arranged first.

[0104] The feature dimension of the new feature data is A+B, where the values ​​of A and B can be the same or different.

[0105] The feature dimensions A of the first feature data, B of the second feature data, and A+B of the new feature data are all smaller than the feature dimensions of the original near-infrared spectral data. This simplifies the data to a certain extent, making it easier for subsequent processing and improving the accuracy of fabric material identification.

[0106] In the above scheme, in step S4, near-infrared spectral data of different fabric materials are obtained, and the MLP is trained to obtain the MLP model.

[0107] The feature dimensions A+B of the new feature data are input into the MLP model. Through the classification processing of the MLP model, the similarity percentage between the fabric to be identified and each fabric material is finally obtained. The fabric with the highest similarity percentage is the material of the fabric to be identified.

[0108] As one implementation method, such as Figure 2 As shown, this invention provides a method for identifying fabric material based on near-infrared spectroscopy, specifically including the following steps:

[0109] S1: Acquire multiple near-infrared spectral data of the fabric to be tested, and perform first-order derivative processing on the data;

[0110] S21: Select one of the first-order derivative data, obtain the wavelength value of the zero point or the closest point to the zero point of the first-order derivative data, and perform segmentation processing on the bands of the near-infrared spectral data to obtain several wavelength sequences.

[0111] S22: Normalize wavelength sequences containing zero or wavelength values ​​closest to zero, and calculate the normalized values; record wavelength sequences without zero or wavelength values ​​closest to zero as 0.

[0112] S23: Arrange the normalized values ​​and 0 in the order of the wavelength sequence to generate the first feature data;

[0113] S31: Arrange all the first-order derivative data into an n-row, m-column first matrix;

[0114] S32: Zero-mean normalize each row of the first matrix to obtain the second matrix;

[0115] S33: Find the covariance matrix of the second matrix, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix;

[0116] S34: Arrange the corresponding eigenvectors into a matrix from top to bottom in descending order of eigenvalues, and take the first B rows to form a transformation matrix;

[0117] S35: Multiply the transformation matrix by the first matrix to obtain the second feature data;

[0118] S4: Merge the first feature data and the second feature data into new feature data, and obtain the feature dimensions of the new feature data;

[0119] S5: Obtain near-infrared spectral data of different fabric materials, train the MLP, and obtain the MLP model;

[0120] S6: Input the feature dimensions of the new feature data into the MLP model. Through the classification processing of the MLP model, the similarity percentage between the fabric to be identified and each fabric material is finally obtained. The fabric with the highest similarity percentage is the material of the fabric to be identified.

[0121] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-described technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for identifying fabric material based on near-infrared spectroscopy, characterized in that, Includes the following steps: S1: Acquire multiple near-infrared spectral data of the fabric to be tested, and perform first-order derivative processing on the data; S21: Select one of the first-order derivative data, obtain the wavelength value of the zero point or the closest point to the zero point of the first-order derivative data, and perform segmentation processing on the bands of the near-infrared spectral data to obtain several wavelength sequences. S22: Normalize wavelength sequences containing zero or wavelength values ​​closest to zero, and calculate the normalized values; record wavelength sequences without zero or wavelength values ​​closest to zero as 0. S23: Arrange the normalized values ​​and 0 in the order of the wavelength sequence to generate the first feature data; S3: Perform PCA dimensionality reduction on all first-order derivative data to generate second feature data; S4: Merge the first feature data and the second feature data into new feature data, perform MLP classification on the new feature data, and realize the identification of fabric material.

2. The fabric material identification method based on near-infrared spectroscopy according to claim 1, characterized in that, In step S22, the normalization calculation formula is: ； Where d is the normalized value, d x d represents the wavelength value at or closest to the zero of the first-order derivative data. max d represents the maximum wavelength value in the wavelength sequence. min It is the minimum wavelength value in the wavelength sequence.

3. The fabric material identification method based on near-infrared spectroscopy according to claim 1, characterized in that, Step S3 includes: S31: Arrange all the first-order derivative data into an n-row, m-column first matrix; S32: Zero-mean normalize each row of the first matrix to obtain the second matrix; S33: Find the covariance matrix of the second matrix, and calculate the eigenvalues ​​and corresponding eigenvectors of the covariance matrix; S34: Arrange the corresponding eigenvectors into a matrix from top to bottom in descending order of eigenvalues, and take the first B rows to form a transformation matrix; S35: Multiply the transformation matrix by the first matrix to obtain the second feature data; Where m is the number of near-infrared spectral data acquired, n is the feature dimension of the first-order derivative data, and B is the feature dimension of the second feature data.

4. The fabric material identification method based on near-infrared spectroscopy according to claim 3, characterized in that, In step S32, when zero-mean is applied to each row of the first matrix, the formula for calculating the average value of each row is: ； Among them, X ij µ represents the data corresponding to each row in the first matrix. i Let i be the average value of each row of data, where i is 1, 2, ..., n, and j is 1, 2, ..., m.

5. The fabric material identification method based on near-infrared spectroscopy according to claim 3, characterized in that, The covariance matrix in step S33 is: ； Where C is the covariance matrix, X' is the second matrix, and X' T This is the transpose of the second matrix.

6. The fabric material identification method based on near-infrared spectroscopy according to claim 3, characterized in that, In step S4, when merging the first feature data and the second feature data into new feature data, the first feature data is arranged first and the second feature data is arranged last. Alternatively, the first feature data can be arranged last, and the second feature data can be arranged first. The feature dimension of the new feature data is A+B, where A is the feature dimension of the first feature data.

7. The fabric material identification method based on near-infrared spectroscopy according to claim 6, characterized in that, In step S4, near-infrared spectral data of different fabric materials are obtained, and the MLP is trained to obtain the MLP model. The feature dimensions A+B of the new feature data are input into the MLP model. Through the classification processing of the MLP model, the similarity percentage between the fabric to be identified and each fabric material is finally obtained. The fabric with the highest similarity percentage is the material of the fabric to be identified.

8. A fabric material identification method based on near-infrared spectroscopy according to any one of claims 1-7, characterized in that, The near-infrared spectral data is obtained by irradiating a point on the fabric material with a near-infrared light source to obtain the absorption data of different wavelengths of light at that point.

Images