Polyester Filament Hairiness Detection Method and System Based on Image Processing Technology

Abstract

The present invention provides a method and system for detecting hairiness of polyester filaments based on image processing technology, which relates to the technical field of chemical fiber quality detection. The present invention respectively collects images of polyester filaments under the irradiation of two light sources, visible light and near-infrared light, preprocesses the images and then generates a composite image by weighted average, and extracts edge and texture features by using edge detection and gray-level co-occurrence matrix, and further constructs a composite feature vector; trains samples with known hairiness distribution through a deep learning network to generate a hairiness detection model; finally, collects images of samples to be detected in real time and generates corresponding composite feature vectors, and inputs them into the trained hairiness detection model to judge the distribution of hairiness. This method combines image processing and deep learning technologies to achieve accurate detection of hairiness of polyester filaments.

Description

Technical Field

[0001] The present invention relates to the technical field of chemical fiber quality detection, and specifically provides a method and system for detecting wool filaments of polyester filaments based on image processing technology. Background Art

[0002] During the production process of polyester filaments, wool filaments have a significant impact on product quality. These wool filaments not only affect the appearance quality of polyester filaments but may also affect the equipment on the production line during subsequent processing. Traditional methods for detecting wool filaments mostly rely on manual inspection, often suffering from low detection efficiency and high false alarm and missed detection rates, making it difficult to meet the quality control requirements of large-scale production and resulting in production losses and resource waste. To overcome these deficiencies, automated detection solutions based on image processing technology have gradually attracted attention. However, existing image processing technologies are still insufficient in feature extraction, especially in the case of changing lighting conditions and background interference, making it difficult to effectively extract the features of wool filaments on polyester filaments. Therefore, there is an urgent need for a new method to improve the accuracy and efficiency of wool filament detection.

[0003] In the prior art, the publication number CN111415349B discloses a method for detecting wool filaments of polyester filaments based on image processing technology, which obtains a clear image of the polyester filaments to be measured in a dark box, calibrates the correspondence between the pixels of the polyester filaments image to be measured and the actual size, and preprocesses the calibrated polyester filaments image through a multi-directional span filtering combination method; uses a double-image double-threshold filament and wool filament segmentation method to process the polyester filaments image after preprocessing, segment the filament stem, and at the same time use the double-image double-threshold filament and wool filament segmentation method to segment the wool filaments; determines the background position from the non-filament-stem positions and non-wool-filament positions, calculates the length of each wool filament using the view calibration method, and performs statistical processing to achieve the detection and recognition of wool filaments.

[0004] The main problems of the above solution are: it depends on obtaining clear images in a dark box, requires uniform light and low background noise, is difficult to operate in complex production environments, and has weak adaptability; and the double-image double-threshold segmentation method has high requirements for computing resources and long processing time, which will affect the efficiency of real-time detection.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and thus it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and system for detecting wool filaments of polyester filaments based on image processing technology to solve the problems raised in the above background art.

[0007] To achieve the above purpose, the present invention provides the following technical solutions:

[0008] A method for detecting hairiness of polyester filaments based on image processing technology, and the specific steps include:

[0009] Step 1: Install a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensure that it is directly facing the running track of the polyester filaments, and keep the distance from the polyester filament production line at 50 - 100 cm;

[0010] Step 2: Pass a polyester filament sample with known hairiness distribution data through the irradiation areas of the visible light LED lamp and the near-infrared LED lamp, collect visible light images and near-infrared images, scale the collected images to a size of 224×224, convert the images into grayscale images after denoising, and set them as the first recognition image and the second recognition image respectively. Establish a plane rectangular coordinate system in the first recognition image, and map the second recognition image one by one with the first recognition image in the coordinate system;

[0011] Step 3: Perform weighted averaging on the gray values of the pixel points of the first recognition image and the second recognition image, and generate a composite image through image fusion;

[0012] Step 4: Extract the edge pixel points of the composite image through canny edge detection to generate an edge pixel map of the composite image, extract the texture features of the composite image based on the gray level co-occurrence matrix, generate an edge feature vector based on the edge pixel map, generate a texture feature vector based on the texture features, and combine the edge feature vector and the texture feature vector to generate a composite feature vector;

[0013] Step 5: Generate a deep learning network, use the composite feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, input them into the deep learning network for training, and obtain a hairiness detection model with the composite feature vector as the input value and the hairiness distribution data as the output value;

[0014] Step 6: Real-time collect the visible light image and the near-infrared image of the polyester filament to be detected. After generating the composite feature vector according to the above steps, input the composite feature vector into the trained hairiness detection model to output the hairiness distribution data of the polyester filament to be detected.

[0015] Further, the visible light LED lamp is a white light LED lamp, and the near-infrared LED lamp is a near-infrared LED lamp with a wavelength of 850 nm.

[0016] Further, fix the white light LED lamp and the near-infrared LED lamp on both sides of the image acquisition device respectively, and the included angle between the irradiation area of the LED lamp and the polyester filament production line is between 30° and 45°;

[0017] In the first recognition image, a plane rectangular coordinate system is established with the column where the leftmost pixel point of the first recognition image is located as the y-axis and the row where the bottommost pixel point is located as the x-axis. All pixel points are in the first quadrant, and each pixel point has a unique coordinate.

[0018] Furthermore, the formula for the weighted average of pixel points is:

[0019] I fus I(x,y) = α·I vis (x,y) + (1 - α)·I NIR (x,y)

[0020] where I fus (x,y) represents the gray value of the pixel point with coordinates (x,y) in the composite image, I vis (x,y) represents the gray value of the pixel point with coordinates (x,y) in the first recognition image, I NIR (x,y) represents the gray value of the pixel point with coordinates (x,y) in the second recognition image, and α represents the proportion weight of the pixel points in the first recognition image in image fusion.

[0021] Furthermore, the principle for generating the edge intensity of the composite image is:

[0022] For each pixel point in the composite image, the pixel point and its neighboring pixel points are respectively convolved with the horizontal direction template and the vertical direction template of the Prewitt operator to generate the gray difference of the pixel point in the horizontal direction and the vertical direction. The formula is:

[0023]

[0024] where P X represents the horizontal direction template of the Prewitt operator, P Y represents the vertical direction template of the Prewitt operator, G x represents the horizontal direction difference of the pixel point, G y represents the vertical direction difference of the pixel point, and (x,y) represents the coordinates of the pixel point;

[0025] According to the gray differences in the horizontal direction and the vertical direction, the gradient magnitude of each pixel point is generated. The formula is:

[0026]

[0027] where G(x,y) represents the gradient magnitude of the pixel point with coordinates (x,y), G x represents the horizontal direction difference of the pixel point, G y represents the vertical direction difference of the pixel point;

[0028] A preset edge threshold is set. When the gradient amplitude of a pixel is higher than the edge threshold, the pixel is retained as an edge pixel, and all edge pixels form an edge pixel map.

[0029] Furthermore, the principle for extracting the texture features of the comprehensive image is as follows:

[0030] The formula for generating the gray-level co-occurrence matrix is:

[0031]

[0032] Where P(i,j) represents the value in the i-th row and j-th column of the gray-level co-occurrence matrix, i represents the gray value of the pixel value in the comprehensive image, j represents the gray value of another pixel value different from i in the comprehensive image, and i = 0, 1, 2,..., N, j = 0, 1, 2,..., N, N represents the total number of gray levels that appear in the image, g(x,y) represents the gray value of the pixel at coordinates (x,y) in the comprehensive image, g(x+dx,y+dy) represents the gray value of the pixel at coordinates (x+dx,y+dy) in the comprehensive image, and dx, dy represent the relative displacement between pixels;

[0033] Based on the gray-level co-occurrence matrix, the formula for extracting the contrast of the comprehensive image is:

[0034]

[0035] Where C represents the contrast of the comprehensive image;

[0036] The formula for extracting the energy of the comprehensive image is:

[0037]

[0038] Where E represents the energy of the comprehensive image;

[0039] The formula for extracting the homogeneity of the comprehensive image is:

[0040]

[0041] Where H represents the homogeneity of the comprehensive image.

[0042] Furthermore, the principle for generating the comprehensive feature vector is as follows:

[0043] The formula for generating the edge feature vector is:

[0044]

[0045] F L =[L avg ,L std ​

[0046] Among them, L avg represents the average edge intensity of edge pixels, (x0, y0) represents the coordinates of a pixel in the edge pixel map, X represents the number of pixels in the horizontal direction of the edge pixel map, Y represents the number of pixels in the vertical direction of the edge pixel map, M represents the total number of pixels in the edge pixel map, and M = X·Y. G0(x0, y0) represents the gradient amplitude of the edge pixel at the coordinates (x0, y0), and L std represents the edge standard deviation of edge pixels, and F L represents the edge feature vector; the formula for generating the texture feature vector is:

[0047] F T = [C, E, H]

[0048] Among them, F T represents the texture feature vector, and C, E, and H respectively represent the contrast, energy, and homogeneity of the comprehensive image;

[0049] The generated comprehensive feature vector is:

[0050] K = [F L , F T

[0051] Among them, K represents the comprehensive feature vector.

[0052] The present invention also provides a polyester filament hairiness detection system based on image processing technology. The system is used to implement the above-mentioned polyester filament hairiness detection method based on image processing technology, and specifically includes:

[0053] An equipment installation module, which is used to install a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensure that it is directly opposite the running track of the polyester filament, and maintain a distance of 50 - 100 cm from the polyester filament production line;

[0054] An image acquisition module, which is used to pass the polyester filament sample with known hairiness distribution data through the irradiation areas of the visible light LED lamp and the near-infrared LED lamp, acquire visible light images and near-infrared images, scale the acquired images to a size of 224×224, convert the images into grayscale images after denoising, and respectively set them as the first recognition image and the second recognition image. A plane rectangular coordinate system is established in the first recognition image, and the second recognition image is mapped one by one with the first recognition image in the coordinate system;

[0055] An image fusion module, which is used to perform weighted averaging on the grayscale values of the pixel points of the first recognition image and the second recognition image, and generate a comprehensive image through image fusion;

[0056] A feature extraction module, which is used to extract the edge pixel points of the comprehensive image through Canny edge detection to generate an edge pixel map of the comprehensive image, extract the texture features of the comprehensive image based on the gray-level co-occurrence matrix, generate an edge feature vector based on the edge pixel map, generate a texture feature vector based on the texture features, and combine the edge feature vector and the texture feature vector to generate a comprehensive feature vector;

[0057] A model training module, which is used to generate a deep learning network. Using the comprehensive feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, input them into the deep learning network for training to obtain a hairiness detection model with the comprehensive feature vector as the input value and the hairiness distribution data as the output value;

[0058] A data output module, which is used to collect the visible light image and the near-infrared image of the polyester filament to be detected in real time. After generating the comprehensive feature vector according to the above steps, input the comprehensive feature vector into the trained hairiness detection model to output the hairiness distribution data of the polyester filament to be detected.

[0059] Compared with the prior art, the beneficial effects of the present invention are:

[0060] The present invention simultaneously collects images of polyester filaments under visible light and near-infrared light. Visible light provides rich color and surface details, while near-infrared images can penetrate a certain depth of the material to capture internal features of the material. By fusing images in two bands through a weighted average method, the information in the two bands can be combined to more effectively extract and analyze target features, making the detection of polyester filaments more comprehensive and accurate.

[0061] The present invention also extracts edge features and texture features through Canny edge detection and gray-level co-occurrence matrix, and generates edge feature vectors and texture feature vectors respectively, combining the shape and surface structure of polyester filaments, making the expression of features more comprehensive and accurate. The Canny edge detection algorithm has strong robustness in a noisy environment and can effectively suppress interference, while the texture features extracted by the gray-level co-occurrence matrix can remain stable under different lighting and conditions, improving the adaptability of the overall detection scheme and the recognition performance in complex scenarios; training a deep learning network with polyester filament samples with known hairiness conditions, and then generating a hairiness detection model. The model can accurately learn the features and conditions of hairiness, adapt to diverse input features, thereby improving the versatility of the system, and through continuous detection, generating new sample data to continuously optimize and update the model to improve the detection accuracy of the model. Description of the Drawings

[0062] Figure 1 It is a schematic flowchart of the method of the embodiment of the present invention;

[0063] Figure 2Schematic diagram of the system module according to an embodiment of the present invention. Detailed implementation manner

[0064] To make the objectives, technical solutions, and advantages of the present invention more clear and understandable, the present invention will be further described in detail below in conjunction with specific embodiments.

[0065] It should be noted that unless otherwise defined, the technical terms or scientific terms used in the present invention should have the ordinary meanings understood by those with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar terms used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or objects appearing before this term cover the elements or objects listed after this term and their equivalents, without excluding other elements or objects. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right" are only used to represent relative position relationships, and when the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0066] Embodiment:

[0067] Please refer to Figure 1 , the present invention provides a technical solution:

[0068] A method for detecting hairiness of polyester filaments based on image processing technology, the specific steps include:

[0069] Step 1: Install a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensure that they are directly facing the running track of the polyester filaments, and keep the distance from the polyester filament production line at 50-100 cm;

[0070] In this embodiment, the visible light LED lamp is a white light LED lamp, and the near-infrared LED lamp is a near-infrared LED lamp with a wavelength of 850 nm.

[0071] Step 2: Pass a polyester filament sample with known hairiness distribution data through the irradiation areas of the visible light LED lamp and the near-infrared LED lamp, collect visible light images and near-infrared images, scale the collected images to a size of 224×224, convert the images into grayscale images after denoising, and respectively set them as the first recognition image and the second recognition image. Establish a plane rectangular coordinate system in the first recognition image, and map the second recognition image and the first recognition image one by one in the coordinate system;

[0072] In this embodiment, the part of the polyester filament irradiated by the LED lamp is divided into several small detectable regions. For each region, it is determined whether there are broken filaments. After the determination of the presence or absence of broken filaments in all regions of this part of the polyester filament is completed, the broken filament region of the polyester filament is generated, which is the broken filament distribution data;

[0073] In this embodiment, the white light LED lamp and the near-infrared LED lamp are respectively fixed on both sides of the image acquisition device, and the included angle between the irradiation region of the LED lamp and the polyester filament production line is between 30° and 45°;

[0074] The included angle of the light irradiation is set to be between 30° and 45°, aiming to reduce the influence of light reflection and enhance the contrast of the image. The formula for converting the denoised image into a grayscale image is:

[0075] Z = 0.299·R + 0.587·G + 0.114·B

[0076] where Z represents the grayscale value of the pixel point, and R, G, and B respectively represent the red, green, and blue channel values of the pixel point;

[0077] In the first recognition image, a plane rectangular coordinate system is established with the column where the leftmost pixel point of the first recognition image is located as the y-axis and the row where the bottommost pixel point is located as the x-axis, so that all pixel points are in the first quadrant and each pixel point has a unique coordinate.

[0078] Step 3: Perform weighted averaging on the pixel point grayscale values of the first recognition image and the second recognition image, and generate a composite image through image fusion;

[0079] In this embodiment, the formula for weighted averaging of pixel points is:

[0080] I fus (x,y) = α·I vis (x,y) + (1 - α)·I NIR (x,y)

[0081] where I fus (x,y) represents the grayscale value of the pixel point with coordinates (x,y) in the composite image, I vis (x,y) represents the grayscale value of the pixel point with coordinates (x,y) in the first recognition image, I NIR(x, y) represents the gray value of the pixel at coordinates (x, y) in the second recognition image. α represents the proportion weight of the pixel in the first recognition image in image fusion. Visible light images can usually capture rich detail information including color, texture, and contour in the image subject. Near-infrared images are mainly used to identify the defective parts in the image subject, that is, the hairiness in polyester filaments. Therefore, the value of the proportion weight is α = 0.6, while retaining the details of the visible light, making full use of the advantages of the near-infrared image.

[0082] Step 4: Extract the edge pixels of the composite image through canny edge detection to generate the edge pixel map of the composite image. Extract the texture features of the composite image based on the gray-level co-occurrence matrix. Generate the edge feature vector based on the edge pixel map, generate the texture feature vector based on the texture features, and combine the edge feature vector and the texture feature vector to generate the composite feature vector;

[0083] In this embodiment, the principle for generating the edge intensity of the composite image is as follows:

[0084] For each pixel in the composite image, convolve the pixel and its neighborhood pixels with the horizontal direction template and the vertical direction template of the Prewitt operator respectively to generate the gray difference in the horizontal direction and the vertical direction of the pixel. The formula is as follows:

[0085]

[0086] Among them, P X represents the horizontal direction template of the Prewitt operator, P Y represents the vertical direction template of the Prewitt operator, G x represents the horizontal direction difference of the pixel, G y represents the vertical direction difference of the pixel, and (x, y) represents the coordinates of the pixel;

[0087] According to the gray differences in the horizontal direction and the vertical direction, generate the gradient amplitude of each pixel. The formula is as follows:

[0088]

[0089] Among them, G(x, y) represents the gradient amplitude of the pixel at coordinates (x, y), G x represents the horizontal direction difference of the pixel, G y represents the vertical direction difference of the pixel;

[0090] Preset the edge threshold. When the gradient amplitude of the pixel is higher than the edge threshold, retain the pixel as an edge pixel. All edge pixels form the edge pixel map.

[0091] When setting the edge threshold, the initial edge threshold is set to 70% of the maximum gradient magnitude of all pixel points. After setting the initial edge threshold, edge pixel points are extracted, and the expert evaluates the extraction effect of the edge pixel points. If there are too few edge pixel points, the edge threshold is decreased; if there are too many edge pixel points, the edge threshold is increased.

[0092] The principle for extracting the texture features of the composite image is as follows:

[0093] The formula for generating the gray-level co-occurrence matrix is as follows:

[0094]

[0095] Among them, P(i,j) represents the value at the i-th row and j-th column in the gray-level co-occurrence matrix. i represents the gray value of the pixel value in the composite image, j represents another gray value different from i in the composite image, and i = 0, 1, 2, ……, N, j = 0, 1, 2, …… N. N represents the total number of gray levels that appear in the image. g(x,y) represents the gray value of the pixel point with coordinates (x,y) in the composite image, and g(x+dx,y+dy) represents the gray value of the pixel point with coordinates (x+dx,y+dy) in the composite image. dx and dy represent the relative displacements between pixels;

[0096] The gray-level co-occurrence matrix reflects the occurrence probability of pixel point pairs with different gray values at a specific direction and distance;

[0097] Based on the gray-level co-occurrence matrix, the formula for extracting the contrast of the composite image is as follows:

[0098]

[0099] Among them, C represents the contrast of the composite image;

[0100] Contrast reflects the intensity of the gray value change in the image. The higher the contrast, the more obvious the texture transformation of the image. It is usually used to identify textures with high contrast. In hairiness detection, a higher contrast can identify different types of textures and distinguish filaments from hairiness.

[0101] The formula for extracting the energy of the composite image is as follows:

[0102]

[0103] Among them, E represents the energy of the composite image;

[0104] Energy reflects the uniformity of the image texture. The higher the energy, the more uniform the texture in the image. It is usually used in regions with lower complexity and fewer changes. In hairiness detection, it can distinguish regular filament structures from irregular hairiness structures.

[0105] The homogeneity of the synthesized image is extracted based on the following formula:

[0106]

[0107] where H represents the homogeneity of the synthesized image.

[0108] Homogeneity reflects the similarity between different gray values and is used to identify regions with consistent texture in the image. In hairiness detection, it can identify the continuous region of filaments and the burst region of hairiness.

[0109] The principle for generating the synthesized feature vector is as follows:

[0110] The formula for generating the edge feature vector is:

[0111]

[0112] F L =[L avg ,L std

[0113] where L avg represents the average edge intensity of edge pixel points, X represents the number of pixels in the horizontal direction of the edge pixel map, Y represents the number of pixels in the vertical direction of the edge pixel map, M represents the total number of pixels in the edge pixel map, and M = X·Y. G0(x,y) represents the gradient magnitude of the edge pixel point, L std represents the edge standard deviation of the edge pixel point, and F L represents the edge feature vector;

[0114] The average edge intensity reflects the overall intensity level of the edges in the image. A higher average edge intensity indicates more obvious edges, suggesting a higher complexity of the image content, while a lower edge mean indicates fewer edges and the image tends to be flat or blurred. The edge standard deviation reflects the degree of dispersion or volatility of the edge intensity. A higher standard deviation means there are edges with varying strengths in the image, indicating the presence of multiple objects or a more complex structure in the image, while a lower standard deviation indicates that the edge intensities in the image tend to be consistent. When constructing the edge feature vector, the average edge intensity and the edge standard deviation jointly provide comprehensive information about the edge features. The average edge intensity reflects the overall intensity of the edges, and the edge standard deviation reflects the degree of change in the edge intensity. Combining these two features can comprehensively describe the edge features of the image.

[0115] The formula for generating the texture feature vector is:

[0116] F T =[C,E,H]

[0117] where F TIt represents the texture feature vector, and C, E, and H respectively represent the contrast, energy, and homogeneity of the comprehensive image;

[0118] The generated comprehensive feature vector is:

[0119] K = [F L , F T

[0120] Among them, K represents the comprehensive feature vector.

[0121] Step 5: Generate a deep learning network. Using the comprehensive feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, input them into the deep learning network for training to obtain a hairiness detection model with the comprehensive feature vector as the input value and the hairiness distribution data as the output value;

[0122] In this embodiment, the structure of the deep learning network is:

[0123] Input layer: It contains 1 neuron and inputs the comprehensive feature vector;

[0124] The first hidden layer: It contains 128 neurons and is activated through the ReLU function;

[0125] The second hidden layer: It contains 64 neurons and is activated through the ReLU function;

[0126] Output layer: It contains 1 neuron and outputs a judgment on whether there is hairiness.

[0127] Step 6: Real-time collect the visible light image and near-infrared image of the polyester filament to be detected. After generating the comprehensive feature vector according to the above steps, input the comprehensive feature vector into the trained hairiness detection model to output the hairiness distribution data of the polyester filament to be detected.

[0128] Divide a to-be-detected image into several regions, and respectively detect each region through the hairiness detection model. Determine the distribution of hairiness in the polyester filament according to the judgment of the presence or absence of hairiness in different regions.

[0129] Please refer to Figure 2 , the present invention also provides a polyester filament hairiness detection system based on image processing technology. The system is used to implement the above-mentioned polyester filament hairiness detection method based on image processing technology, and specifically includes:

[0130] The device installation module is used to install a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensure that it is directly facing the running track of the polyester filament, and maintain a distance of 50 - 100 cm from the polyester filament production line;

[0131] ​The image acquisition module is used to irradiate a polyester filament sample with known hairiness distribution data through the illumination areas of visible light LED lights and near-infrared LED lights, acquire visible light images and near-infrared images, scale the acquired images to a size of 224×224, convert the images into grayscale images after denoising, and respectively set them as the first recognition image and the second recognition image. A plane rectangular coordinate system is established in the first recognition image, and the second recognition image is mapped one by one with the first recognition image in the coordinate system;

[0132] The image fusion module is used to perform weighted averaging on the grayscale values of the pixel points of the first recognition image and the second recognition image, and generate a composite image through image fusion;

[0133] The feature extraction module is used to extract the edge pixel points of the composite image through canny edge detection to generate an edge pixel map of the composite image, extract the texture features of the composite image based on the gray-level co-occurrence matrix, generate an edge feature vector based on the edge pixel map, generate a texture feature vector based on the texture features, and combine the edge feature vector and the texture feature vector to generate a composite feature vector;

[0134] The model training module is used to generate a deep learning network, take the composite feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, input them into the deep learning network for training, and obtain a hairiness detection model with the input value being the composite feature vector and the output value being the hairiness distribution data;

[0135] The data output module is used to collect the visible light image and near-infrared image of the polyester filament to be detected in real time. After generating the composite feature vector according to the above steps, input the composite feature vector into the trained hairiness detection model, and output the hairiness distribution data of the polyester filament to be detected.

[0136] The above formulas are all dimensionless and take their numerical calculations. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0137] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by the combination of computer software and electronic hardware. Whether these functions are executed by hardware or software methods depends on the specific application and design constraints of the technical solution.

[0138] The unit described as a separating component may or may not be physically separated. The component shown as a unit may or may not be a physical unit, and it may be located in one place or distributed over multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0139] As mentioned above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of changes or substitutions, which should all be covered within the protection scope of the present application.

Claims

1. A method for detecting hairiness of polyester filaments based on image processing technology, characterized in that The specific steps include: Step 1: Install a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensuring that they are directly facing the running track of the polyester filament and maintaining a distance of 50 - 100 cm from the polyester filament production line; Step 2: Pass the polyester filament sample with known hairiness distribution data through the irradiation areas of the visible light LED lamp and the near-infrared LED lamp, collect visible light images and near-infrared images, scale the collected images to a size of 224×224, convert the images into grayscale images after denoising, and respectively set them as the first recognition image and the second recognition image. Establish a plane rectangular coordinate system in the first recognition image, and map the second recognition image to the first recognition image one by one in the coordinate system; Step 3: Perform weighted averaging on the gray values of the pixel points of the first recognition image and the second recognition image, and generate a composite image through image fusion; Step 4: Extract the edge pixel points of the composite image through canny edge detection to generate an edge pixel map of the composite image. Extract the texture features of the composite image based on the gray level co-occurrence matrix, generate an edge feature vector based on the edge pixel map, generate a texture feature vector based on the texture features, and combine the edge feature vector and the texture feature vector to generate a composite feature vector; Step 5: Generate a deep learning network, use the composite feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, input them into the deep learning network for training, and obtain a hairiness detection model with the input value being the composite feature vector and the output value being the hairiness distribution data; Step 6: Real-time collect the visible light image and the near-infrared image of the polyester filament to be detected. After generating the composite feature vector according to the above steps, input the composite feature vector into the trained hairiness detection model to output the hairiness distribution data of the polyester filament to be detected; The formula for weighted averaging of pixel points is: Among them, represents the gray value of the pixel at the coordinate in the comprehensive image, represents the gray value of the pixel at the coordinate in the first recognition image, represents the gray value of the pixel at the coordinate in the second recognition image, represents the proportion weight of the pixel of the first recognition image in image fusion; The principle for generating the composite feature vector is: The formula for generating the edge feature vector is: Among them, represents the average edge intensity of edge pixels, represents the coordinates of the pixels in the edge pixel map, represents the number of pixels in the horizontal direction of the edge pixel map, represents the number of pixels in the vertical direction of the edge pixel map, represents the total number of pixels in the edge pixel map, and , represents that at the coordinate the gradient amplitude of the edge pixel, represents the edge standard deviation of the edge pixel, represents the edge feature vector; The formula for generating the texture feature vector is: Among them, represents the texture feature vector, respectively represent the contrast, energy and homogeneity of the composite image; The generated composite feature vector is: Among them, represents the comprehensive feature vector.

2. The polyester filament hairiness detection method based on image processing technology according to claim 1, wherein: In Step 1, the visible light LED lamp is a white light LED lamp, and the near-infrared LED lamp is a near-infrared LED lamp with a wavelength of 850 nm.

3. The polyester filament hairiness detection method based on image processing technology according to claim 1, wherein: In Step 2, fix the white light LED lamp and the near-infrared LED lamp on both sides of the image acquisition device respectively, and the included angle between the irradiation area of the LED lamp and the polyester filament production line is between 30° and 45°; In the first recognition image, take the column where the leftmost pixel point of the first recognition image is located as the y-axis, and the row where the bottommost pixel point is located as the x-axis to establish a plane rectangular coordinate system in which all pixel points are in the first quadrant and each pixel point has a unique coordinate.

4. A method for detecting hairiness of polyester filament based on image processing technology according to claim 1, characterized in that: The principle for generating the edge pixel map of the composite image in Step 4 is: For each pixel point in the composite image, convolve the pixel point and its neighborhood pixel points with the horizontal direction template and the vertical direction template of the Prewitt operator respectively to generate the gray level difference of the pixel point in the horizontal direction and the vertical direction. The formula is: Among them, represents the horizontal direction template of the Prewitt operator, represents the vertical direction template of the Prewitt operator, represents the horizontal direction difference of the pixel point, represents the vertical direction difference of the pixel point, represents the coordinates of the pixel point; According to the gray level difference in the horizontal direction and the vertical direction, generate the gradient amplitude of each pixel point. The formula is: Among them, represents the gradient magnitude of the pixel point with coordinates , represents the horizontal direction difference of the pixel point, represents the vertical direction difference of the pixel point; A preset edge threshold is set. When the gradient amplitude of a pixel point is higher than the edge threshold, the pixel point is retained as an edge pixel point, and all the edge pixel points form an edge pixel map.

5. A method for detecting hairiness of polyester filament based on image processing technology according to claim 1, characterized in that: The principle for extracting the texture features of the composite image in step 4 is as follows: The formula for generating the gray-level co-occurrence matrix is: Among them, represents the value of the i-th row and j-th column in the gray-level co-occurrence matrix, where i represents the gray value of the pixel value in the comprehensive image, j represents another gray value different from i in the comprehensive image, and , , represents the total number of gray levels that appear in the image, represents the gray value of the pixel point with coordinates in the comprehensive image, represents the gray value of the pixel point with coordinates in the comprehensive image, represents the relative displacement between pixels; Based on the gray-level co-occurrence matrix, the formula for extracting the contrast of the composite image is: Among them, represents the contrast of the composite image; The formula for extracting the energy of the composite image is: Among them, represents the energy of the composite image; The formula for extracting the homogeneity of the composite image is: Among them, represents the homogeneity of the composite image.

6. A polyester filament hairiness detection system based on image processing technology, characterized in that: The system is used to implement the polyester filament hairiness detection method based on image processing technology according to any one of claims 1-5, and specifically includes: An equipment installation module for arranging a visible light LED lamp and a near-infrared LED lamp side by side directly above the polyester filament production line, ensuring that it is directly facing the running track of the polyester filament, and maintaining a distance of 50-100 cm from the polyester filament production line; An image acquisition module for passing a polyester filament sample with known hairiness distribution data through the illumination areas of the visible light LED lamp and the near-infrared LED lamp, acquiring a visible light image and a near-infrared image, scaling the acquired images to a size of 224×224, converting the images into gray-level images after denoising, respectively setting them as the first recognition image and the second recognition image, establishing a plane rectangular coordinate system in the first recognition image, and mapping the second recognition image one by one with the first recognition image in the coordinate system; An image fusion module for performing weighted averaging on the gray values of the pixel points of the first recognition image and the second recognition image, and generating a composite image through image fusion; A feature extraction module for extracting the edge pixel points of the composite image through canny edge detection to generate an edge pixel map of the composite image, extracting the texture features of the composite image based on the gray-level co-occurrence matrix, generating an edge feature vector based on the edge pixel map, generating a texture feature vector based on the texture features, and generating a composite feature vector by combining the edge feature vector and the texture feature vector; A model training module for generating a deep learning network, using the composite feature vector of the polyester filament sample as the input and the hairiness distribution data as the label, inputting them into the deep learning network for training, and obtaining a hairiness detection model with the input value being the composite feature vector and the output value being the hairiness distribution data; A data output module for real-time collecting the visible light image and the near-infrared image of the polyester filament to be detected, generating a composite feature vector according to the above steps, inputting the composite feature vector into the trained hairiness detection model, and outputting the hairiness distribution data of the polyester filament to be detected.

Images