CN114414523A - Textile fiber component qualitative method based on automatic waveband selection - Google Patents
Textile fiber component qualitative method based on automatic waveband selection Download PDFInfo
- Publication number
- CN114414523A CN114414523A CN202111242181.5A CN202111242181A CN114414523A CN 114414523 A CN114414523 A CN 114414523A CN 202111242181 A CN202111242181 A CN 202111242181A CN 114414523 A CN114414523 A CN 114414523A
- Authority
- CN
- China
- Prior art keywords
- qualitative
- automatic
- band selection
- textile fiber
- model
- 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
- 238000000034 method Methods 0.000 title claims abstract description 89
- 239000004753 textile Substances 0.000 title claims abstract description 46
- 239000000835 fiber Substances 0.000 title claims abstract description 40
- 230000008569 process Effects 0.000 claims abstract description 57
- 238000004451 qualitative analysis Methods 0.000 claims abstract description 27
- 230000004913 activation Effects 0.000 claims abstract description 24
- 238000002329 infrared spectrum Methods 0.000 claims abstract description 21
- 230000003595 spectral effect Effects 0.000 claims abstract description 19
- 238000000605 extraction Methods 0.000 claims abstract description 17
- 238000004458 analytical method Methods 0.000 claims abstract description 13
- 238000013528 artificial neural network Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims description 17
- 238000001228 spectrum Methods 0.000 claims description 16
- 238000010586 diagram Methods 0.000 claims description 13
- 238000001914 filtration Methods 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012512 characterization method Methods 0.000 claims description 6
- 239000000203 mixture Substances 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 5
- 230000009467 reduction Effects 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 2
- 239000004744 fabric Substances 0.000 claims description 2
- 238000012545 processing Methods 0.000 claims description 2
- 238000002310 reflectometry Methods 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 238000009941 weaving Methods 0.000 claims description 2
- 238000013145 classification model Methods 0.000 abstract description 5
- 230000002596 correlated effect Effects 0.000 abstract 1
- 230000002269 spontaneous effect Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 8
- 229920000742 Cotton Polymers 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 238000004445 quantitative analysis Methods 0.000 description 4
- 240000008564 Boehmeria nivea Species 0.000 description 3
- 238000004497 NIR spectroscopy Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000010187 selection method Methods 0.000 description 2
- 239000000126 substance Substances 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000000862 absorption spectrum Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 238000012372 quality testing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000004513 sizing Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
- G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
- G01N21/31—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry
- G01N21/35—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light
- G01N21/359—Investigating relative effect of material at wavelengths characteristic of specific elements or molecules, e.g. atomic absorption spectrometry using infrared light using near infrared light
Landscapes
- Physics & Mathematics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a textile fiber component qualitative method based on automatic waveband selection, and designs a near infrared spectrum automatic waveband selection and qualitative analysis model for textile fiber component qualitative analysis by combining a deep neural network characteristic extraction technology. The automatic waveband selection and qualitative model based on the category activation information can realize the efficient qualitative analysis process of textile fibers. In the invention, the wave band selection process is self-adaptive, and the spontaneous wave band interval is selected based on the near infrared spectrum data characteristics to further complete qualitative analysis. The whole process comprises a spectral data preprocessing part, a waveband selecting part and a qualitative classification model. The band selection part is correlated with the qualitative classification model part, band selection is completed in the process of establishing the qualitative model, a final qualitative result is obtained through an end-to-end analysis process, manual intervention is less, the band selection process is more effective, and the accuracy of qualitative analysis is high.
Description
Technical Field
The invention relates to the technical field of near-infrared textile fiber component analysis, in particular to a qualitative model based on automatic waveband selection for the near-infrared textile fiber component qualitative analysis.
Background
The qualitative analysis of textile fiber components is an important task in textile fiber detection, which is used for identifying and identifying the textile fiber component groups, and therefore, is also an important task in the quality testing unit for textile quality determination. When the components are measured by the traditional chemical method and physical method, the time consumption is high, the cost is high, and the components are damaged to the measured object, and the near infrared spectrum analysis technology has the advantages of low cost, high efficiency, safety, environmental protection, no damage to the measured object and the like, so the near infrared spectrum analysis technology is used for qualitative components of textile fibers as a new technology.
The near infrared spectrum analysis is divided into two parts: qualitative analysis and quantitative analysis. Wherein the qualitative analysis identifies the component groups and the quantitative analysis requires the content of the components to be given. Qualitative analysis can be used as an auxiliary for quantitative analysis tasks, because the potential process of qualitative problem is to mine distinguishing features between absorption spectrum sequence data of different substances, which can provide an effective spectrum data feature extraction process for the quantitative analysis tasks, thereby further completing the determination of component content.
The textile fiber qualitative analysis method based on the near infrared spectrum comprises the following three steps: 1) preprocessing the spectral data; 2) feature selection (wavelength selection); 3) and (5) qualitatively classifying. The spectral data preprocessing part mainly aims at reducing noise of data; the main purpose of feature selection is the process of compressing and reducing dimensions of original data. For spectral data, data features that reflect sample characteristics are usually located in specific spectral sequence regions, so wavelength selection is the main feature selection method; after pretreatment and wavelength selection, the spectrum qualitative process is finally realized through a classification model.
Disclosure of Invention
The invention aims to overcome the defects and shortcomings of the existing near-infrared textile fiber component qualitative method, and provides a textile fiber component qualitative method based on automatic waveband selection by combining a deep learning method technology.
The band selection is performed by screening the spectral regions, filtering the sample data to analyze irrelevant regions, and reserving the spectral regions reflecting the characteristics of the sample. Compared with the method of directly analyzing based on full spectrum information, the method has the advantages that redundancy and noise in the spectrum are filtered through wavelength selection, data dimensionality is simplified, and difficulty and complexity of subsequent analysis tasks are reduced. Therefore, it is an important step in near infrared spectroscopy, and in the related art, the conventional band selection algorithm includes: sliding window based, statistical based, regression analysis based, and genetic algorithms based.
The method completes the wave band selection based on the class activation information, performs component qualitative analysis while selecting the wave band, and compared with the traditional textile fiber component qualitative method, the method has the advantages that the wave band selection process is self-adaptive, and is a wave band interval selected based on the spontaneity of the textile near infrared spectrum data characteristics, so the manual intervention is less, the wave band selection result is effective, and the qualitative analysis accuracy is high. The method comprises a spectrum data preprocessing part, a wave band selection part and a qualitative classification model. The band selection part is mutually associated with the qualitative classification model part, band selection is completed in the qualitative model construction process, manual intervention is not needed, the band selection result obtained by utilizing the strong characteristic extraction capability of the deep neural network is very effective, and the final qualitative result is obtained through an end-to-end process.
In order to achieve the purpose, the invention adopts the following scheme: and identifying the salient feature interval on the original waveband by utilizing the category activation graph information, and taking the waveband interval corresponding to the part exceeding the threshold value in the category activation graph as a selection interval as a standard for selecting the waveband. The features extracted by the convolutional network are classified to complete qualitative analysis, and meanwhile, the method is also used for calculating a category activation information graph. Through an end-to-end model training mode, a multi-label classification loss function is adopted to learn parameters in a feature extraction and classification module based on a convolutional network, an automatic waveband selection process is realized, and finally a qualitative analysis model is obtained for testing. The textile fiber component qualitative method based on automatic waveband selection can effectively detect textiles with different components. The scheme specifically comprises the following steps:
s1: constructing and preprocessing a textile near infrared spectrum data set;
s11: and constructing sample data required by model training. Textile fibres of known composition containing all possible components of the fibre to be measured are collected and then the known class of textile fibres is scanned using an active Near Infrared (NIR) spectrometer for the construction of a correction data set (model training set). The spectral data information obtained by scanning is X, wherein the spectral data information comprises n samples from fiber samples with different components, and the sample marking information is Y corresponding to n sample component categories.
X∈RN*M*L,Y∈RN*C
The size of the spectrum data of each sample in the sample spectrum data X is M X L, the size of L represents each wavelength sampling position determined by the resolution of near infrared equipment, M represents which kinds of original data (reflectivity, absorptivity and intensity) are adopted at each wavelength position during near infrared data processing and analysis, C represents all possible component categories, and each textile fiber sample component consists of one or more of C component categories;
s12: and preprocessing the near infrared spectrum data on the basis of S11. Respectively adopting standardization, Fourier filtering and wiener filtering for pretreatment, realizing noise reduction on spectral data, eliminating noise generated by temperature, humidity and visible light intensity factors in a collection environment on a finished spectrum, eliminating influence on the finished spectrum caused by intrinsic factors of a dye on a fabric and a measured object in a weaving mode and intrinsic errors in spectral collection equipment, and obtaining the finished spectrum by X' epsilon R after pretreatmentN*M′*L′。
S2: an automatic band selection and qualitative model based on category activation information;
s21: and (5) a near infrared spectrum feature extraction model. Adopting a convolution neural network based on Resnet as a feature extraction structure of input spectral data in a model and simultaneously as a feature extraction structure in a final definite nature analysis process;
s22: and (5) a classification module. On the basis of S21, the structure of the full-connection network + softmax classification layer is used as a classification module. The classification module receives the characteristic diagram output by the convolution network in S21, and outputs a one-dimensional vector with the size of 1 × C as C;
s23: a category activation information map is computed. On the basis of S22, the automatic waveband selection process corresponds to the clipping weight calculation process of the characteristic diagram, and the calculation result forms a category activation information diagram;
s24: and an automatic band selection module. On the basis of S23, filtering an insignificant interval by using a category activation information graph through an adjustable threshold, and discarding spectral band data corresponding to the insignificant interval to realize a band selection process.
S3: an automatic band selection process based on end-to-end training;
s31: the automatic band selection process is completed in the end-to-end training process of the neural network. In the S22 classification module, the classification module,
represents the weight of class c on the kth node of the full connectivity layer,
the larger the value of (c) is, the higher the predicted value for class c is, and thus
Reflecting the degree of contribution to the prediction result on the k-th node on the full-connection layer associated with the node. Category activation graph
Lc∈RLA graph of activation information for class c components is shown. The category significant information may be calculated once when the training of the feature extraction module is completed at S21, and the band selection is completed through the process of S24. The whole process is iterated for a plurality of times, so that the automatic waveband selection process is completed in the end-to-end training process of the neural network.
S4: qualitative model learning based on end-to-end training;
s41: the qualitative model learning process is synchronized with the automatic band selection process. In the near infrared spectrum feature extraction model of S21, the output feature diagram of the convolution network is subjected to qualitative process by an S22 classification module on one hand, and is subjected to category activation information diagram calculation by an S23 on the other hand. The qualitative model learning process is synchronized with the automatic band selection process through end-to-end learning.
S42: an end-to-end model is trained based on a multi-label classification binary cross entropy loss function (binary cross entropy loss). And (4) regarding the qualitative analysis process as a multi-label classification process, and adopting a multi-label loss function to constrain the end-to-end learning process.
S5: and (3) online real-time qualitative prediction of near infrared spectrum components of the textile fibers.
S51: deploying the model in the S2 as a cloud reasoning service after the learning processes of S3 and S4;
s52: scanning near infrared spectrum data information of the textile needing to be subjected to component qualitative determination;
s53: and calling the cloud inference service of S51, and taking the scanning data of S52 as input to finish online real-time qualitative prediction.
Compared with the prior art, the invention has the following advantages: the invention innovatively combines the automatic band selection process with the spectrum qualitative analysis process, adaptively extracts the band in the characteristic extraction process and completes the qualitative analysis, overcomes the defects of low efficiency of the wavelength selection process, low promotion of the subsequent qualitative analysis effect and the like in the textile fiber qualitative analysis based on near infrared spectrum analysis at present, provides an end-to-end near infrared textile fiber qualitative method with less manual intervention and high efficiency, and realizes the nondestructive rapid qualitative detection of textile components.
Drawings
The present application will be described in further detail with reference to the following drawings and detailed description.
FIG. 1 is a flow chart of near infrared spectroscopy analysis in accordance with the present invention.
Fig. 2 is a technical framework diagram of the present invention.
FIG. 3 is a comparison of the effect of different models on the cotton sizing task at different wavelength ranges without automatic band selection according to the present invention.
FIG. 4 is a graph showing the comparison of the effects of the method for qualitative task of ramie cotton based on the original data and based on the automatic band selection when different feature extraction methods are adopted in the present invention
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
As shown in fig. 1, the overall analysis process of the method for textile fiber component characterization based on automatic band selection disclosed by the present invention comprises the following steps:
1. inputting near infrared spectrum data;
2. initializing data and carrying out noise reduction pretreatment;
3. training a convolution-based neural network model;
4. testing the accuracy of the model and judging whether the accuracy is improved;
5. when the accuracy in step 4 meets a certain condition, setting the current model as a final classification network;
6. when the accuracy rate in the step 4 does not meet the condition, calculating the information of the category activation graph;
7. selecting wave bands based on the category activation map information calculated in the step 6, and completing feature selection and data compression dimensionality reduction of current data;
8. and taking the data in the step 7 as new data and restarting the step 3. Until the final classification network is obtained.
As shown in fig. 2, the automatic band selection and certainty model framework based on the category activation information in the present invention is composed of three parts: 1) data input and preprocessing; 2) feature extraction and automatic wavelength selection; 3) and (5) performing qualitative analysis. The core part of the automatic wavelength selection is the calculation of the class activation map, the second part of which is shown in fig. 2.
Wherein A iskRepresenting the characteristics on the kth channel in the convolutional network output characteristic diagram A, connecting the A to the full connection after global pooling, and finally finishing classification through a softmax layer,
indicating that k channels in a are assigned weights when weighted. In the framework of the present invention,
representing the weight of class c on the kth node of the full connectivity layer.
The larger the value of (c) is, the higher the predicted value for class c is, and thus
The degree of contribution to the prediction result at the kth node on the fully connected layer associated with it is reflected. Global pooling layer in network architecture AkPooled as input on the kth node of the full connectivity layer, thus establishing AkConnecting with the kth node on the full connection layer, and finally obtaining a result L through weighting of different channels in the feature diagramcThe contribution of features to the c-class prediction can be represented. And can therefore also serve as a basis for automatic band selection.
Table 1 shows the effect comparison of different models in the qualitative task of ramie cotton under different wavelength ranges when the automatic band selection is not performed. The table contains three models Resnet9, Resnet18 and long-short term memory network (LSTM) which are Reset with two different network depths and a representative LSTM of a Recurrent Neural Network (RNN) which is commonly used for sequence modeling, and the table presents experimental results of the three models in different wave band ranges on cotton-flax qualitative analysis. In addition, the experiment in Table 2 also reflects that selecting the correct band for analysis is crucial to improving the effect
Table 2 shows the comparison of the effects of the methods for extracting different features on the qualitative task of ramie cotton based on the original data and based on the automatic band selection method. The table comprises four models of Resnet, LSTM, FCN and LSTM-FCN, and compared with the qualitative analysis by using original data, the table improves the effect when the automatic band selection based on the category activation information and the qualitative analysis by using the qualitative model provided by the invention are used for qualitative analysis, and obtains better qualitative effects on four different models.
The foregoing is only a preferred embodiment of the present application, and it should be noted that, for those skilled in the art, several modifications and substitutions can be made without departing from the technical principle of the present application, and these modifications and substitutions should also be regarded as the protection scope of the present application.
Claims (6)
1. A textile fiber component qualitative method based on automatic waveband selection is characterized in that the automatic waveband selection and qualitative method comprises the following steps:
s1: constructing and preprocessing a textile near infrared spectrum data set;
s2: an automatic band selection and qualitative model based on category activation information;
s3: an automatic band selection process based on end-to-end training;
s4: qualitative model learning based on end-to-end training;
s5: and (3) online real-time qualitative prediction of near infrared spectrum components of the textile fibers.
2. The method of claim 1 for textile fiber composition characterization based on automatic band selection, wherein: the step S1 of the method for analyzing textile fiber components based on depth codec further includes:
s11: constructing sample data required by model training; collecting textile fibers with known components containing all possible components of the fibers to be detected, and scanning the textile fibers with known classes by using an active Near Infrared (NIR) spectrometer to construct a correction data set (model training set); the spectral data information obtained by scanning is X, wherein the spectral data information comprises n samples from fiber samples with different components, and the sample marking information is Y corresponding to n sample component categories;
X∈RN*M*L，Y∈RN*C
the size of the spectrum data of each sample in the sample spectrum data X is M X L, the size of L represents each wavelength sampling position determined by the resolution of near infrared equipment, M represents which kinds of original data (reflectivity, absorptivity and intensity) are adopted by each wavelength position during the processing and analysis of the near infrared data, C represents all possible component classification numbers, and each textile fiber sample component is composed of one or more of C component types;
s12: preprocessing the near infrared spectrum data on the basis of S11; respectively adopting standardization, Fourier filtering and wiener filtering for pretreatment, realizing noise reduction on spectral data, eliminating noise generated by temperature, humidity and visible light intensity factors in a collection environment on a finished spectrum, eliminating influence on the finished spectrum caused by intrinsic factors of a dye on a fabric and a measured object in a weaving mode and intrinsic errors in spectral collection equipment, and obtaining the finished spectrum by X' epsilon R after pretreatmentN*M′*L′。
3. The method of claim 1 for textile fiber composition characterization based on automatic band selection, wherein: the step S2 of the method for analyzing textile fiber components based on depth codec further includes:
s21: a near infrared spectrum feature extraction model; a convolution neural network based on a depth residual error neural network (Resnet) is adopted as a feature extraction structure of input spectral data in a model and is also used as a feature extraction structure in the final qualitative analysis process;
s22: a classification module; on the basis of S21, using the structure of the full-connection network + softmax classification layer as a classification module; the classification module receives the characteristic diagram output by the convolution network in S21, and outputs a one-dimensional vector with the size of 1 × C as C;
s23: calculating a category activation information graph; on the basis of S22, the automatic wave band selection process corresponds to the clipping weight calculation process of the characteristic diagram, and the calculation result forms a category activation information diagram;
s24: an automatic band selection module; on the basis of S23, filtering an insignificant interval by using a category activation information graph through an adjustable threshold, and discarding spectral band data corresponding to the insignificant interval to realize a band selection process.
4. The method of claim 1 for textile fiber composition characterization based on automatic band selection, wherein: the step S3 of the method for analyzing textile fiber components based on depth codec further includes:
s31: the automatic band selection process is completed during the end-to-end training process of the neural network, and in the S22 classification module,
represents the weight of class c on the kth node of the full connectivity layer,
the larger the value of (c) is, the higher the predicted value for class c is, and thus
Reflecting the degree of contribution to the prediction result on the kth node on the full-connection layer associated with the kth node; category activation graph
Lc∈RLA graph representing activation information for class c components; when the training of the S21 feature extraction module is completed, the category significant information can be calculated once, and the waveband selection is completed through the S24 process; the whole process is iterated for a plurality of times, so that the automatic wave band selection process is completed in the end-to-end training process of the neural network.
5. The method of claim 1 for textile fiber composition characterization based on automatic band selection, wherein: the step S4 of the method for analyzing textile fiber components based on depth codec further includes:
s41: the qualitative model learning process and the automatic waveband selection process are carried out synchronously; in the S21 near infrared spectrum feature extraction model, on one hand, the qualitative process of the convolution network output feature diagram is completed through an S22 classification module, and on the other hand, a category activation information diagram is calculated through S23; the qualitative model learning process and the automatic waveband selection process are synchronously carried out through end-to-end learning;
s42: training an end-to-end model based on a multi-label classification binary cross entropy loss function (binary cross entropy loss); and (4) regarding the qualitative analysis process as a multi-label classification process, and adopting a multi-label loss function to constrain the end-to-end learning process.
6. The method of claim 1 for textile fiber composition characterization based on automatic band selection, wherein: the step S5 of the method for analyzing textile fiber components based on depth codec further includes:
s51: deploying the model in the S2 as a cloud reasoning service after the learning processes of S3 and S4;
s52: scanning near infrared spectrum data information of the textile needing to be subjected to component qualitative determination;
s53: and calling the cloud inference service of S51, and taking the S52 scanning data as input to finish online real-time qualitative prediction.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111242181.5A CN114414523A (en) | 2021-10-25 | 2021-10-25 | Textile fiber component qualitative method based on automatic waveband selection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111242181.5A CN114414523A (en) | 2021-10-25 | 2021-10-25 | Textile fiber component qualitative method based on automatic waveband selection |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114414523A true CN114414523A (en) | 2022-04-29 |
Family
ID=81265259
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111242181.5A Pending CN114414523A (en) | 2021-10-25 | 2021-10-25 | Textile fiber component qualitative method based on automatic waveband selection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114414523A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786617A (en) * | 2024-02-27 | 2024-03-29 | 南京信息工程大学 | Cloth component analysis method and system based on GA-LSTM hyperspectral quantitative inversion |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19920592A1 (en) * | 1999-05-04 | 2000-11-09 | Cetex Chemnitzer Textilmaschin | Method to automatically recognise fibrous material or mixtures; involves using near infrared spectroscopy to study unmodified material sample, and using neural network to evaluate results |
| US20100036795A1 (en) * | 2005-10-13 | 2010-02-11 | Busch Kenneth W | Classification of Fabrics by Near-Infrared Spectroscopy |
| JP2011047759A (en) * | 2009-08-26 | 2011-03-10 | Shinshu Univ | Method of inspecting fiber product |
| CN107219188A (en) * | 2017-06-02 | 2017-09-29 | 中国计量大学 | A kind of method based on the near-infrared spectrum analysis textile cotton content for improving DBN |
| CN110717368A (en) * | 2018-07-13 | 2020-01-21 | 北京服装学院 | Qualitative classification method for textiles |
| WO2020234466A1 (en) * | 2019-05-23 | 2020-11-26 | Valvan Baling Systems Nv | Improved determination of textile fiber composition |
| CN113418919A (en) * | 2021-04-30 | 2021-09-21 | 深圳市菲雀兰博科技研究中心有限公司 | Textile fiber component qualitative and quantitative online analysis system and method |
-
2021
- 2021-10-25 CN CN202111242181.5A patent/CN114414523A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE19920592A1 (en) * | 1999-05-04 | 2000-11-09 | Cetex Chemnitzer Textilmaschin | Method to automatically recognise fibrous material or mixtures; involves using near infrared spectroscopy to study unmodified material sample, and using neural network to evaluate results |
| US20100036795A1 (en) * | 2005-10-13 | 2010-02-11 | Busch Kenneth W | Classification of Fabrics by Near-Infrared Spectroscopy |
| JP2011047759A (en) * | 2009-08-26 | 2011-03-10 | Shinshu Univ | Method of inspecting fiber product |
| CN107219188A (en) * | 2017-06-02 | 2017-09-29 | 中国计量大学 | A kind of method based on the near-infrared spectrum analysis textile cotton content for improving DBN |
| CN110717368A (en) * | 2018-07-13 | 2020-01-21 | 北京服装学院 | Qualitative classification method for textiles |
| WO2020234466A1 (en) * | 2019-05-23 | 2020-11-26 | Valvan Baling Systems Nv | Improved determination of textile fiber composition |
| CN113418919A (en) * | 2021-04-30 | 2021-09-21 | 深圳市菲雀兰博科技研究中心有限公司 | Textile fiber component qualitative and quantitative online analysis system and method |
Non-Patent Citations (3)
| Title |
|---|
| 谢小红 等: "深度残差网络综述", 《信息与电脑(理论版)》 * |
| 高敬鹏: "《深度学习 卷积神经网络技术与实践》", 31 July 2020, 机械工业出版社 * |
| 黄兴阳 等: "近红外光谱法用于纺织品定量分析的研究", 《中国纤检》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786617A (en) * | 2024-02-27 | 2024-03-29 | 南京信息工程大学 | Cloth component analysis method and system based on GA-LSTM hyperspectral quantitative inversion |
| CN117786617B (en) * | 2024-02-27 | 2024-04-30 | 南京信息工程大学 | Cloth component analysis method and system based on GA-LSTM hyperspectral quantitative inversion |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109493287B (en) | Deep learning-based quantitative spectral data analysis processing method | |
| CN112766075B (en) | Hyperspectral remote sensing black and odorous water body grading method based on semi-supervised learning strategy | |
| CN110717368A (en) | Qualitative classification method for textiles | |
| CN107219188B (en) | A method of based on the near-infrared spectrum analysis textile cotton content for improving DBN | |
| Millan et al. | Image analysis‐based modelling for flower number estimation in grapevine | |
| CN109460471B (en) | Method for establishing fiber category map library based on self-learning mode | |
| CN105630743A (en) | Spectrum wave number selection method | |
| CN113686804A (en) | Textile fiber component nondestructive cleaning analysis method based on deep regression network | |
| CN110736709A (en) | blueberry maturity nondestructive testing method based on deep convolutional neural network | |
| CN113392931A (en) | Hyperspectral open set classification method based on self-supervision learning and multitask learning | |
| CN111693487A (en) | Fruit sugar degree detection method and system based on genetic algorithm and extreme learning machine | |
| CN113970532B (en) | Fabric fiber component detection system and prediction method based on near infrared spectrum | |
| CN111562235A (en) | Method for rapidly identifying black-leaf outbreak disease and infection degree of tobacco leaves based on near infrared spectrum | |
| CN114414523A (en) | Textile fiber component qualitative method based on automatic waveband selection | |
| CN116754511A (en) | Indocyanine green detection method and system based on spectrum technology | |
| CN116030310A (en) | Sample classification method and system based on laser-induced breakdown spectroscopy technology | |
| CN116026795A (en) | Rice grain quality character nondestructive prediction method based on reflection and transmission spectrum | |
| CN115575404B (en) | Fruit appearance quality detection method based on ratio multispectral image | |
| CN114264626A (en) | Fabric nondestructive quantitative analysis method based on time series residual error network | |
| CN115406852A (en) | Fabric fiber component qualitative method based on multi-label convolutional neural network | |
| CN111275131A (en) | Chemical image classification and identification method based on infrared spectrum | |
| CN116429724A (en) | Method for detecting pollution degree of composite insulator material based on infrared hyperspectral analysis | |
| CN111104876A (en) | Infrared spectrum deconvolution method based on neural network | |
| CN112964719B (en) | Hyperspectrum-based food fructose detection method and device | |
| CN112697745A (en) | Method for measuring alcohol content of white spirit |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20220429 |
|
| RJ01 | Rejection of invention patent application after publication |