An online detection method for polyester fiber content in textiles based on hyperspectral imaging
Through hyperspectral imaging technology, intelligently selecting ROI regions and building neural network models, the accuracy and efficiency of textile polyester fiber content detection are solved, and online accurate detection and automated operation of textile polyester fiber content is realized.
Patent Information
- Application Number
- CN202311632307.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-11-30
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2043-11-30
AI Technical Summary
The existing textile content detection methods have problems such as inaccurate detection, low efficiency, and inability to achieve online inspection and intelligent ROI area selection, which is difficult to meet the quality and trade needs.
Using the online detection method of textile polyester fiber content based on hyperspectral imaging, the infrared spectral image is intermittently collected, the ROI region is intelligently selected, the spectral reflectivity average value and second-order derivative are extracted, and a neural network model is constructed to predict the polyester fiber content.
It realizes online accurate detection of textile polyester fiber content, improves detection accuracy and efficiency, simplifies the process, reduces labor costs, and achieves fully automated operations.
Smart Images

Figure CN117705754B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of textile component content detection, and in particular to an online detection method for textile polyester fiber content based on hyperspectral imaging. Background Art
[0002] In textile testing, the testing methods and instruments are required to be accurate, fast and environmentally friendly to meet the needs of the market. Some existing methods for testing the content of textiles are to perform qualitative testing on blended products and then use strong acids, organic solvents and other reagents to dissolve the tested fabrics, destroying the samples. This is time-consuming and inefficient, and cannot be used for large-scale online testing. The testing is inaccurate and has large errors, making it difficult to meet the testing needs of quality and trade.
[0003] In the actual recognition process, the existing detection system does not cut and separate the invalid areas in the textile spectral image, resulting in the system being unable to accurately recognize online and unable to perform intelligent ROI area selection.
[0004] Therefore, how to provide an online detection method for polyester fiber content in textiles based on hyperspectral imaging that can intelligently select ROI areas with high detection accuracy and efficiency is an urgent problem that technicians in this field need to solve. Summary of the invention
[0005] In view of this, the present invention provides an online detection method for polyester fiber content in textiles based on hyperspectral imaging to solve the problems in existing infrared recognition technology that the ROI area cannot be intelligently selected for spectral feature extraction and the polyester fiber content cannot be accurately detected.
[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:
[0007] An online detection method for polyester fiber content in textiles based on hyperspectral imaging, comprising:
[0008] Intermittently collect infrared spectrum images of textiles in the 1000-1700nm infrared band;
[0009] Perform feature interception on the infrared spectrum image and determine the ROI area in the infrared spectrum image;
[0010] Scan and analyze all pixels in the ROI area, extract the spectral reflectance value of each pixel within the 1000-1700nm band, and determine the average spectral reflectance R of the ROI area;
[0011] The average value R of the spectral reflectance is differentiated with respect to the wavelength to obtain the second-order derivative R2 of the spectral reflectance with respect to the wavelength;
[0012] With R2 as input and polyester fiber content Q as output, an intelligent prediction model of polyester fiber content based on neural network is constructed;
[0013] The trained prediction model is imported into the detection system, and the second-order derivative R2 of the textile spectral reflectance is used as input to detect the polyester fiber content of the tested textile.
[0014] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method of intermittently collecting infrared spectral images of textiles in the infrared band of 1000-1700nm includes:
[0015] The preset spectral resolution is 9nm, and the spectral range is 1000nm-1700nm;
[0016] The infrared spectrum information of the textiles on the conveyor belt is collected by a hyperspectral camera and saved every 1 second to obtain several infrared spectrum images.
[0017] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the feature interception of the infrared spectrum image and the determination of the ROI area in the infrared spectrum image include:
[0018] Analyzing a plurality of infrared spectral images to obtain a plurality of first spectral image regions, wherein the first spectral image regions include regions occupied by each textile in the infrared spectral image;
[0019] Feature extraction is performed on a plurality of first spectral image regions to determine ROI regions in the first spectral image regions.
[0020] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method of analyzing a plurality of infrared spectral images to obtain a plurality of first spectral image regions comprises:
[0021] Perform edge detection on a number of infrared spectral images to obtain the first area occupied by each textile in the infrared spectral image within 1 second;
[0022] The first area occupied by each textile in the infrared spectrum image is cut to obtain a plurality of first spectrum image areas.
[0023] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method for determining the ROI area in the first spectral image area includes:
[0024] Scanning and analyzing the first spectral image region, detecting the grayscale values of the pixels in the first spectral image region, and integrating the grayscale values of the pixels in the same spectral image region to generate a grayscale value record set;
[0025] Compare the grayscale values in the grayscale value record set to determine the largest grayscale value in the grayscale value record set;
[0026] Using a first ratio value smaller than the maximum grayscale value as a screening condition, screening corresponding pixel points in the first spectral image area, and identifying the screened pixel points as pixel points to be removed;
[0027] If a plurality of pixels to be removed meet the preset clustering screening rule, the area mapped by the pixels to be removed in the first spectral image area is identified as an invalid area, and the invalid area is removed in the first spectral image area to obtain the ROI area.
[0028] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method of eliminating invalid areas in the first spectral image area to obtain the ROI area includes:
[0029] Evenly dividing the first spectral image region into a plurality of first blocks, and determining the pixel points in the plurality of first blocks according to the overall pixel points in the first spectral image region;
[0030] Scan and analyze the pixel points in the first blocks to obtain a number of first pixel points whose grayscale values are greater than 60% of the maximum grayscale value, calculate the ratio of all the first pixel points in each first block to the total pixel points in the first spectral image area to obtain a second ratio value, and limit the second ratio value to be less than a preset ratio value;
[0031] Scanning a plurality of pixels in the first blocks and performing grayscale value detection, and integrating the grayscale values of the pixels in the same first block to generate a plurality of grayscale value data sets, and associating the data sets with the first blocks;
[0032] Calculate the grayscale average value of all pixels in the first block, record it as the block grayscale expression value, if the block grayscale expression value is less than 60% of the maximum grayscale value, and the difference between the two is greater than the preset difference value, then the corresponding first block mark needs to be removed;
[0033] The first block marked to be removed in the first spectral image region is removed to obtain the ROI region.
[0034] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method for determining the average spectral reflectance R of the ROI area includes:
[0035] Obtain the ROI area of each textile in the infrared spectrum image;
[0036] Scan all pixel points in the ROI area to obtain a corresponding relationship between a spectral reflectance value and a wavelength, wherein the corresponding relationship between a spectral reflectance value and a wavelength includes a plurality of spectral reflectance values in the ROI area and their corresponding wavelengths;
[0037] The spectral reflectance values of the pixel points within the wavelength range of 1000-1700nm are extracted, and the average value of the spectral reflectance values within the wavelength range of 1000-1700nm is calculated to obtain the spectral average reflectance R of the ROI area.
[0038] Preferably, in the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging, the method of differentiating the average value R of spectral reflectance with respect to wavelength to obtain the second-order derivative R2 of reflectance with respect to wavelength includes:
[0039] The spectral average reflectance R is subjected to SG smoothing to obtain the smoothed spectral reflectance R1;
[0040] The second-order derivative of the smoothed spectral reflectance R1 with respect to wavelength is obtained to obtain the second-order derivative of the reflectance with respect to wavelength R2.
[0041] Preferably, in the above-mentioned online detection method of polyester fiber content in textiles based on hyperspectral imaging, the method for detecting the polyester fiber content of the tested textile comprises:
[0042] 600 textiles with different polyester fiber contents were used as training samples, and their spectral average reflectance R was collected. After SG smoothing, the second-order derivative of the smoothed spectral average reflectance R1 was calculated to obtain R2.
[0043] Taking R2 obtained from 600 training samples as input, the Levenberg-Marquardt algorithm was used, a 10-layer neural network was selected, and the polyester fiber content Q was output. The neural network-based intelligent prediction model of polyester fiber content was obtained based on the second-order derivative of the spectral average reflectance R2 and the polyester fiber content Q.
[0044] The trained neural network-based intelligent prediction model for polyester fiber content is imported into the recognition system, and R2 calculated by infrared spectrum image extraction is input, and the polyester fiber content Q is output and obtained.
[0045] The present invention provides an online detection method for polyester fiber content in textiles based on hyperspectral imaging. Compared with the prior art, the method has the following beneficial effects:
[0046] (1) The online detection method of polyester fiber content in textiles based on hyperspectral imaging of the present invention realizes intelligent selection of ROI area of textiles in infrared spectrum images based on hyperspectral imaging, and predicts the polyester fiber content Q based on the polyester fiber content intelligent prediction model of neural network, thereby realizing online accurate detection of polyester fiber content in textiles, and the detection accuracy is high and the efficiency is fast, which solves the problem that the existing infrared recognition technology cannot intelligently select ROI area and cannot accurately detect polyester fiber content;
[0047] (2) The online detection method for polyester fiber content in textiles based on hyperspectral imaging of the present invention is simple, highly accurate, and has high processing efficiency, thereby ensuring the accuracy and reliability of the detection of polyester fiber content in textiles. At the same time, the entire process does not require human intervention and can achieve fully automated operation, thereby reducing labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0049] Figure 1 It is a flow chart of an online detection method for polyester fiber content in textiles based on hyperspectral imaging provided by the present invention;
[0050] Figure 2 It is a principle diagram of an online detection method for polyester fiber content in textiles based on hyperspectral imaging provided by the present invention;
[0051] Figure 3 is a spectral image of textiles collected by the hyperspectral camera provided by the present invention;
[0052] Figure 4 yes Figure 3 Line graph of spectral reflectance and wavelength extracted from textiles in different areas. DETAILED DESCRIPTION
[0053] The specific implementation of the present invention is further described in detail below in conjunction with the accompanying drawings and examples. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.
[0054] In the description of the present application, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.
[0055] The terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of this application, unless otherwise specified, "plurality" means two or more.
[0056] In the description of this application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.
[0057] like Figure 1 As shown in FIG. 4 , the present invention provides an online detection method for polyester fiber content in textiles based on hyperspectral imaging, comprising:
[0058] S100: Intermittently collect infrared spectral images of textiles in the infrared band of 1000-1700nm. The infrared spectral images are collected by a hyperspectral camera, which is fast and efficient, and easy to collect information about each textile in the infrared spectral image.
[0059] S200: Perform feature extraction on the infrared spectrum image and determine the ROI area in the infrared spectrum image. The infrared spectrum image often contains a large amount of information. Feature extraction can highlight the key parts of the image. Selecting the ROI area of interest can reduce the complexity of data processing. Intelligent selection of the ROI area can reduce processing time and calculation amount, greatly improving processing efficiency.
[0060] S300: Scan and analyze all pixels in the ROI area, extract the spectral reflectance value of each pixel in the 1000-1700nm band, and determine the average spectral reflectance R of the ROI area. The average spectral reflectance R of the ROI can reduce the influence of noise and improve the accuracy of the processing results.
[0061] S400: The average value R of the spectral reflectance is derived with respect to the wavelength to obtain the second-order derivative R2 of the spectral reflectance with respect to the wavelength. The second-order derivative method can eliminate the baseline offset of the spectral reflectance data, increase the resolution of the spectral reflectance data, and effectively reduce and suppress noise.
[0062] S500: Taking R2 as input and polyester fiber content Q as output, a polyester fiber content intelligent prediction model based on neural network is constructed, which can improve the accuracy of the prediction of polyester fiber content Q and has strong adaptability.
[0063] S600: The trained neural network-based intelligent prediction model for polyester fiber content is imported into the recognition system, and the second-order derivative R2 of the textile spectral reflectance is used as input to detect the polyester fiber content Q of the tested textile. The composition of the tested textile is 97% polyester fiber and 3% spandex, which greatly improves the processing efficiency and accuracy. The whole process does not require human intervention and can achieve fully automated operation, reducing labor costs.
[0064] In some embodiments of the present invention, the method of discontinuously collecting infrared spectral images of textiles in the infrared band of 1000-1700 nm in step S100 includes:
[0065] S110: The preset spectral resolution is 9nm, and the spectral range is 1000nm-1700nm;
[0066] S120: The infrared spectrum information of the textile on the conveyor belt is collected by a hyperspectral camera and saved every 1 second to obtain a number of infrared spectrum images.
[0067] In some embodiments of the present invention, the method of performing feature extraction on the infrared spectrum image and determining the ROI region in the infrared spectrum image in step S200 includes:
[0068] S210: Analyze a plurality of infrared spectrum images to obtain a plurality of first spectrum image regions, wherein the first spectrum image regions include a region occupied by each textile in the infrared spectrum image;
[0069] S211: performing edge detection on a plurality of infrared spectral images to obtain a first area occupied by each textile in the infrared spectral image within 1 second, wherein the edge detection method is based on a gradient algorithm. Since the image information obtained by the hyperspectral camera is gray value information, the gradient can be directly calculated, a threshold range is set, and a portion exceeding the threshold is regarded as an edge and then cut to obtain a spectral image of the first area;
[0070] S212: cutting the first area occupied by each textile in the infrared spectrum image to obtain a plurality of first spectrum image areas;
[0071] S220: performing feature extraction on a plurality of first spectral image regions to determine ROI regions in the first spectral image regions;
[0072] S221: Scan and analyze the first spectral image region, detect the grayscale values of the pixels in the first spectral image region, and integrate the grayscale values of the pixels in the same spectral image region to generate a grayscale value record set;
[0073] S222: Compare the grayscale values in the grayscale value record set to determine the largest grayscale value in the grayscale value record set;
[0074] S223: Using a first ratio value that is smaller than the maximum grayscale value as a screening condition, screening corresponding pixel points in the first spectral image area, and identifying the screened pixel points as pixel points to be removed;
[0075] S224: If a plurality of pixels to be removed satisfy a preset clustering screening rule, the region mapped with the pixels to be removed in the first spectral image region is identified as an invalid region, and the invalid region is removed in the first spectral image region to obtain a ROI region.
[0076] In some embodiments of the present invention, the method of removing the invalid area in the first spectral image area in step S224 to obtain the ROI area includes:
[0077] S225: Evenly divide the first spectral image region into a plurality of first blocks, and determine the pixel points in the plurality of first blocks according to the overall pixel points in the first spectral image region, which is beneficial to reducing the calculation process and improving the processing efficiency;
[0078] S226: Scan and analyze the pixel points in the first blocks to obtain the first pixel points whose grayscale value is greater than 60% of the maximum grayscale value, calculate the ratio of all the first pixel points in each first block to the total pixel points in the first spectral image area, obtain a second ratio value, and limit the second ratio value to be less than a preset ratio value, so as to limit the size of each first block, so that each first block will not be too large, which is convenient for subsequent calculations;
[0079] S227: Scanning the pixels in the first blocks and performing grayscale value detection, integrating the grayscale values of the pixels in the same first block, generating a plurality of grayscale value data sets, and associating the data sets with the first blocks, so as to intelligently extract the first blocks corresponding to the grayscale value data sets and improve processing efficiency;
[0080] S228: Calculate the grayscale average value of all pixels in the first block, record it as the block grayscale expression value, if the block grayscale expression value is less than 60% of the maximum grayscale value, and the difference between the two is greater than a preset difference value, then the corresponding first block is marked to be removed, and the first block corresponding to the difference between the block grayscale expression value and 60% of the maximum grayscale value is less than the preset difference value can be retained, effectively marking the block to be removed;
[0081] S229: Eliminate the first block marked to be eliminated in the first spectral image region to obtain the ROI region, that is, intelligently select the ROI region to improve processing efficiency.
[0082] In some embodiments of the present invention, the method for determining the average spectral reflectance R of the ROI region in step S300 includes:
[0083] S310: Obtaining the ROI region of each textile in the infrared spectrum image;
[0084] S320: Scan all pixel points in the ROI area to obtain a correspondence between a spectral reflectance value and a wavelength, wherein the correspondence between a spectral reflectance value and a wavelength includes a plurality of spectral reflectance values in the ROI area and their corresponding wavelengths;
[0085] S330: extracting the spectral reflectance values of the pixel points within the wavelength range of 1000-1700nm, and calculating the average value of the spectral reflectance values within the wavelength range of 1000-1700nm to obtain the spectral average reflectance R of the ROI area.
[0086] In some embodiments of the present invention, the method of derivatizing the average value R of the spectral reflectance with respect to the wavelength in step S400 to obtain the second-order derivative R2 of the reflectance with respect to the wavelength includes:
[0087] S410: performing SG smoothing on the spectral average reflectance R to obtain a smoothed spectral reflectance R1, wherein the SG smoothing can keep the shape and width of the signal unchanged while filtering out noise. By using the moving average method, the SG smoothing can suppress noise, thereby improving the accuracy of information extraction. At the same time, it can also retain most of the subtle features of the infrared spectrum image, and can effectively improve the accuracy and stability of the spectral average reflectance R, so as to be better applied to spectral-based analysis and applications;
[0088] S420: The second-order derivative of the smoothed spectral reflectance R1 with respect to wavelength is obtained to obtain the second-order derivative of the reflectance with respect to wavelength R2, which can highlight the subtle changes in the spectral curve and better extract the characteristics of the textile. It can also effectively extract useful information and reduce the impact of noise, thereby gaining a deeper understanding of the intrinsic characteristics of the spectral data and providing valuable reference for subsequent analysis.
[0089] In some embodiments of the present invention, constructing a polyester fiber content intelligent prediction model based on a neural network in step S500 includes:
[0090] S510: 600 textiles with different polyester fiber contents are used as training samples, and their spectral average reflectance R is collected. After SG smoothing, the second-order derivative of the smoothed spectral average reflectance R1 is calculated to obtain R2;
[0091] S520: Taking R2 obtained from 600 training samples as input, the Levenberg-Marquardt algorithm is adopted, a 10-layer neural network is selected, and the polyester fiber content Q is output. The second-order derivative of the spectral average reflectance R2 and the polyester fiber content Q are obtained to obtain an intelligent prediction model for polyester fiber content based on a neural network. Through training with a large amount of data, the model can more accurately predict the polyester fiber content and has strong applicability.
[0092] The working principle of the above-mentioned online detection method for polyester fiber content in textiles based on hyperspectral imaging is as follows: first, intermittently collect infrared spectral images of textiles in the infrared band of 1000-1700nm, and determine the first spectral image area based on the infrared spectral image, scan and analyze all pixels in each first spectral image area, obtain a gray value record set, determine the maximum gray value, and evenly divide each first spectral image area into a number of first blocks, scan and analyze all pixels in each first block, and obtain the gray average value of all pixels in each first block, which is recorded as the block gray expression value. If the block gray expression value is less than 60% of the maximum gray value, and the difference between the two is greater than the preset difference value, then The corresponding first block mark needs to be removed, and the first block marked to be removed is removed to obtain the ROI area, and all pixel points in the ROI area are scanned to obtain several spectral reflectance values and their corresponding wavelengths within the wavelength range of 1000-1700nm, and the spectral reflectance values are averaged to obtain the spectral average reflectance R of the ROI area, and the spectral average reflectance R is differentiated with respect to the wavelength to obtain the second-order derivative R2 of the reflectance with respect to the wavelength, and R2 is input into the trained neural network-based intelligent prediction model of polyester fiber content to obtain the polyester fiber content Q, which effectively solves the problem that the existing infrared recognition technology cannot intelligently select the ROI area for spectral feature extraction and cannot accurately detect the polyester fiber content.
[0093] It can be seen from the above embodiments that the present invention provides an online detection method for polyester fiber content in textiles based on hyperspectral imaging, which can intelligently select ROI areas, has high polyester fiber content detection accuracy, and high processing efficiency.
[0094] In summary, the online detection method of polyester fiber content in textiles based on hyperspectral imaging of the present invention can not only realize fast and efficient intelligent selection of ROI areas of interest, reduce processing time and calculation amount, and greatly improve processing efficiency, but also can predict the polyester fiber content Q based on the polyester fiber content intelligent prediction model of neural network, thereby realizing online and accurate detection of polyester fiber content in textiles, reducing the influence of noise, and improving the accuracy of processing results. The whole process does not require human intervention, and can realize fully automated operation, thereby reducing labor costs.
[0095] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0096] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention without departing from the principles and intent of the present invention.
Claims
1. An online detection method for polyester fiber content in textiles based on hyperspectral imaging, characterized in that: include: Intermittently collect infrared spectral images of textiles in the 1000-1700nm infrared band: The preset spectral resolution is 9nm, and the spectral range is 1000nm-1700nm; The infrared spectrum information of the textiles on the conveyor belt is collected by a hyperspectral camera and saved every 1 second to obtain several infrared spectrum images; The method of analyzing a plurality of infrared spectrum images to obtain a plurality of first spectrum image regions includes: Perform edge detection on a number of infrared spectral images to obtain the first area occupied by each textile in the infrared spectral image within 1 second; Cutting the first area occupied by each textile in the infrared spectrum image to obtain a plurality of first spectrum image areas; Scanning and analyzing the first spectral image region, detecting the grayscale values of the pixels in the first spectral image region, and integrating the grayscale values of the pixels in the same spectral image region to generate a grayscale value record set; Compare the grayscale values in the grayscale value record set to determine the largest grayscale value in the grayscale value record set; Using a first ratio value smaller than the maximum grayscale value as a screening condition, screening corresponding pixel points in the first spectral image area, and identifying the screened pixel points as pixel points to be removed; Evenly dividing the first spectral image region into a plurality of first blocks, and determining the pixel points in the plurality of first blocks according to the overall pixel points in the first spectral image region; Scan and analyze the pixel points in the first blocks to obtain a number of first pixel points whose grayscale values are greater than 60% of the maximum grayscale value, calculate the ratio of all the first pixel points in each first block to the total pixel points in the first spectral image area to obtain a second ratio value, and limit the second ratio value to be less than a preset ratio value; Scanning a plurality of pixels in the first blocks and performing grayscale value detection, and integrating the grayscale values of the pixels in the same first block to generate a plurality of grayscale value data sets, and associating the data sets with the first blocks; Calculate the grayscale average value of all pixels in the first block, record it as the block grayscale expression value, if the block grayscale expression value is less than 60% of the maximum grayscale value, and the difference between the two is greater than the preset difference value, then the corresponding first block mark needs to be removed; Eliminate the first block marked to be eliminated in the first spectral image region to obtain a ROI region; Determine the average spectral reflectance R of the ROI area: Obtain the ROI area of each textile in the infrared spectrum image; Scan all pixel points in the ROI area to obtain a corresponding relationship between a spectral reflectance value and a wavelength, wherein the corresponding relationship between a spectral reflectance value and a wavelength includes a plurality of spectral reflectance values in the ROI area and their corresponding wavelengths; Extract the spectral reflectance value of the pixel point within the wavelength range of 1000-1700nm, and calculate the average value of the spectral reflectance value within the wavelength range of 1000-1700nm to obtain the spectral average reflectance R of the ROI area; The average value of the spectral reflectance R is differentiated with respect to the wavelength to obtain the second-order derivative of the reflectance with respect to the wavelength R2: The spectral average reflectance R is subjected to SG smoothing to obtain the smoothed spectral reflectance R1; The second-order derivative of the smoothed spectral reflectance R1 with respect to the wavelength is obtained to obtain the second-order derivative of the reflectance with respect to the wavelength R2; The methods for testing the polyester fiber content of the tested textiles include: 600 textiles with different polyester fiber contents were used as training samples, and their spectral average reflectance R was collected. After SG smoothing, the second-order derivative of the smoothed spectral average reflectance R1 was calculated to obtain R2. Taking R2 obtained from 600 training samples as input, the Levenberg-Marquardt algorithm was used, a 10-layer neural network was selected, and the polyester fiber content Q was output. The neural network-based intelligent prediction model of polyester fiber content was obtained based on the second-order derivative of the spectral average reflectance R2 and the polyester fiber content Q. The trained neural network-based intelligent prediction model for polyester fiber content is imported into the recognition system, and R2 calculated by infrared spectrum image extraction is input, and the polyester fiber content Q is output and obtained.
Citation Information
Patent Citations
Method and apparatus for three-dimensional visualization of sequence image
CN101256670A
Fish water content distribution detection method based on hyperspectral imaging technology
CN103674864A
Pulmonary nodule detection device and method based on shape template matching and combining classifier
CN104751178A
Hyperspectral imaging-based measuring equipment of chlorophyll content of whole rice plant and method thereof
CN105021617A
Cell detection method based on hyperspectral microimaging technology
CN106226247A
