A quality evaluation method and device based on non-woven fabric defect detection

Abstract

This application relates to a quality assessment method and apparatus based on nonwoven fabric defect detection. The quality assessment method includes the following steps: selecting the camera and light source settings based on the material information of the nonwoven fabric to be inspected; acquiring environmental data, nonwoven fabric image data, and corresponding label data from the nonwoven fabric production line, and performing data analysis to obtain a nonwoven fabric dataset for training; constructing a quality assessment model for defect detection, and training and optimizing the quality assessment model using the nonwoven fabric dataset to obtain a nonwoven fabric quality assessment model; acquiring images of the target nonwoven fabric to be identified on the nonwoven fabric production line, as well as real-time environmental data, and performing quality assessment and analysis using the nonwoven fabric quality assessment model. This application's quality assessment method based on nonwoven fabric defect detection achieves full automation from image acquisition and data analysis to quality assessment, reducing manual intervention and improving the accuracy and stability of detection.

Description

Technical Field

[0001] This application relates to the field of nonwoven fabric defect detection technology, specifically to a quality assessment method and apparatus based on nonwoven fabric defect detection. Background Technology

[0002] Nonwoven fabrics, as an important material, are widely used in many fields such as medical and health care, packaging, clothing, and home furnishing. With the increasing demand for nonwoven fabrics from various industries, higher requirements are being placed on their quality. The presence of defects can seriously affect the performance and quality of nonwoven fabrics. For example, defects such as holes, broken threads, stains, and wrinkles can reduce the strength, breathability, and aesthetic appearance of nonwoven fabrics. Therefore, accurate detection of defects in nonwoven fabrics and quality assessment are crucial. Traditional methods for detecting defects in nonwoven fabrics mainly rely on manual visual inspection. Manual visual inspection has many drawbacks: inspectors are prone to fatigue from long hours of work, leading to decreased accuracy; subjective judgments vary greatly among inspectors, making it difficult to guarantee consistent results; and manual inspection is inefficient and cannot meet the rapid testing needs of large-scale production.

[0003] Therefore, there is an urgent need for a quality assessment method based on nonwoven fabric defect detection that is efficient, accurate, and can integrate multiple factors. Summary of the Invention

[0004] To address the problems existing in the prior art, this application aims to provide a quality assessment method and apparatus based on nonwoven fabric defect detection. The quality assessment method based on nonwoven fabric defect detection automates the entire process from image acquisition and data analysis to quality assessment, reducing manual intervention, minimizing errors caused by human factors, and improving the accuracy and stability of detection.

[0005] The quality assessment method based on nonwoven fabric defect detection described in this application includes the following steps:

[0006] S1. Select the camera and light source settings based on the material information of the non-woven fabric for which defect detection is required;

[0007] S2. Obtain environmental data, non-woven fabric image data, and corresponding label data from the non-woven fabric production line, and perform data analysis and processing to obtain a non-woven fabric dataset for training.

[0008] S3. Construct a quality assessment model for defect detection, and train and optimize the quality assessment model using the nonwoven fabric dataset to obtain the nonwoven fabric quality assessment model.

[0009] S4. Obtain the target nonwoven fabric image to be identified on the nonwoven fabric production line, as well as real-time environmental data, and perform quality assessment and analysis in conjunction with the nonwoven fabric quality assessment model.

[0010] Preferably, step S1 specifically includes:

[0011] The camera and light source can be set in two ways: facing forward and facing backward.

[0012] The forward orientation is set such that the camera and the light source are located on the same side of the nonwoven fabric surface, and the optical axis of the camera forms a certain angle α with the light emitted from the light source.

[0013] The back-facing configuration is such that the camera and the light source are located on opposite sides of the non-woven fabric, and the camera and the light source are positioned opposite each other, so that the optical axis of the camera is collinear with the light emitted from the light source.

[0014] To obtain the chemical fiber content C of nonwoven fabric cf Natural fiber content C nf Thickness D and surface roughness R a Satisfying C cf +C nf =1, and calculate the positive setting comprehensive index S according to the following formula. f and back-to-back setting comprehensive index S b :

[0015] S f =μ1*C cf +μ2*(1-D)+μ3*(1-R a )

[0016] S b =ω1*C nf +ω2*D+ω3*R a

[0017] Where μ1, μ2, μ3, ω1, ω2 and ω3 are all weighting coefficients, satisfying: μ1+μ2+μ3=1 and ω1+ω2+ω3=1;

[0018] If S f >S b If so, then select the positive setting;

[0019] If S f <S b Then select the aforementioned back-facing setting;

[0020] If S f =S b Then select the forward setting or the reverse setting.

[0021] Preferably, step S2 specifically includes:

[0022] Acquire environmental data from the nonwoven fabric production line, including temperature data, humidity data, and tension data;

[0023] The nonwoven fabric image data and corresponding label data on the nonwoven fabric production line are obtained. The nonwoven fabric image data includes image samples with defect-free features, image samples with hole features, image samples with broken filament features, image samples with stain features, and image samples with wrinkle features. In the label data, the label corresponding to the defect-free feature is 0, the label corresponding to the hole feature is 1, the label corresponding to the broken filament feature is 2, the label corresponding to the stain feature is 3, and the label corresponding to the wrinkle feature is 4.

[0024] The environmental data is subjected to outlier processing and smoothing to obtain the processed environmental data;

[0025] The nonwoven fabric image data is processed by weighted averaging, noise is removed by median filtering, image contrast is enhanced by histogram equalization, and image edges are extracted by Canny edge detection algorithm to obtain the processed nonwoven fabric image data.

[0026] Preferably, step S2 further includes:

[0027] A convolutional neural network model based on the LeNet structure is constructed. The processed nonwoven fabric image data is input into the convolutional neural network model to perform convolution operations, extract local features of the nonwoven fabric image, and obtain feature maps.

[0028] The feature map is downsampled, and the extracted local features are integrated to obtain the spatial feature vector of the nonwoven fabric image;

[0029] A single-layer long short-term memory network model was constructed, and the continuously acquired non-woven fabric image sequence in the non-woven fabric image data was organized into a time series data format to obtain time series data.

[0030] By inputting the time series data into the single-layer long short-term memory network model and processing the time series data using a gating mechanism, the time feature vector of the nonwoven fabric image is obtained.

[0031] The processed environmental data are subjected to statistical feature calculations, including mean, variance, and rate of change, to obtain the environmental data statistical features.

[0032] The spatial feature vector, the temporal feature vector, and the environmental data statistical features are connected and fused to obtain spatiotemporal fused feature data.

[0033] The spatiotemporal fusion feature data is divided into a training set and a test set according to the ratio A1:A2. The image data of the training set and the test set are paired and integrated with the corresponding label data to construct a nonwoven fabric dataset for training.

[0034] Preferably, step S3 specifically includes:

[0035] A quality assessment model for a three-layer fully connected neural network is constructed, comprising an input layer, a hidden layer, and an output layer.

[0036] The dimension of the spatiotemporal fusion feature data is determined according to the principal component analysis method, and the number of input neurons in the input layer is set to be the same as the dimension of the spatiotemporal fusion feature data.

[0037] The number of hidden neurons in the hidden layer is set to 1.5 times the number of input neurons, and ReLU is used as the activation function of the hidden layer.

[0038] The number of output neurons in the output layer is set to 5, and the Softmax function is used as the activation function of the output layer.

[0039] Set the number of training rounds to Q and the number of training batches to e;

[0040] The parameters of the quality assessment model are initialized, and the training set is input into the quality assessment model in the training batch. The training set passes through the input layer, the hidden layer and the output layer in sequence to obtain the predicted probability distribution of each sample for each label.

[0041] The cross-entropy loss function was used to analyze the difference between the probability distribution predicted by the quality assessment model and the true label, and the average loss value for each batch was calculated.

[0042] Based on the calculated average loss value, the backpropagation algorithm is used to calculate the gradient of the loss function with respect to each connection weight and bias in the model;

[0043] Based on the gradients of each connection weight and bias obtained from the calculation, the connection weights and biases of the model are updated using the stochastic gradient descent optimization algorithm.

[0044] Repeat the training until the training rounds are completed.

[0045] Preferably, step S3 further includes:

[0046] The quality assessment model that has completed the test rounds is tested using the test set to obtain the identification results, and the accuracy E, recall R, F1 score and test loss L are calculated.

[0047] The accuracy E is calculated using the following formula:

[0048]

[0049] The recall rate R is calculated using the following formula:

[0050]

[0051] The formula for calculating the F1 value is as follows:

[0052]

[0053] The test loss value L is calculated using the cross-entropy loss function;

[0054] Where D represents the number of correctly identified image samples; A represents the total number of image samples in the test set; X i M represents the number of image samples that correctly identified the i-th type of feature; i The total number of image samples representing the i-th type of feature; i = 1, 2, 3, 4, 5; i = 1 represents the flawless feature; i = 2 represents the hole feature; i = 3 represents the broken thread feature; i = 4 represents the stain feature; i = 5 represents the wrinkle feature;

[0055] The overall coefficient K for model testing is calculated using the following formula:

[0056] K = C1*E + C2*R + C3*F1 + C4*L

[0057] Where C1, C2, C3 and C4 are all weighting coefficients, satisfying C1+C2+C3+C4=1;

[0058] If K≥0.8, the quality assessment model is deemed to meet the requirements and is used as the quality assessment model for nonwoven fabrics.

[0059] If K < 0.8, the quality assessment model is deemed unsuitable and retrained until K ≥ 0.8 is satisfied.

[0060] Preferably, step S4 specifically includes:

[0061] A target nonwoven fabric image to be identified is acquired on the nonwoven fabric production line. The target nonwoven fabric image is then identified using the nonwoven fabric quality assessment model, and the first area proportion S of the pore features is statistically obtained. K The first number of broken filaments, Z, and the second area percentage of the stain, S. W and the first length Y cm of the fold feature;

[0062] If S K <5%, Z<5, S W If both <3% and Y < 10 are met, the nonwoven fabric is judged to be of qualified quality.

[0063] If S K <5%, Z<5, S WIf either <3% or Y < 10 is not met, the nonwoven fabric is deemed to be of substandard quality, and a defect correlation analysis is performed in conjunction with the real-time environmental data.

[0064] Preferably, the defect correlation analysis includes:

[0065] Acquire real-time environmental data from the nonwoven fabric production line, including real-time temperature values ​​W. S Real-time humidity value H S and tension real-time value F S And obtain the temperature change rate dW, humidity change rate dH and tension change rate dF over the time interval ΔT;

[0066] If W S >W max If dW > 2, then a temperature anomaly is determined. If Z ≥ 5, then the temperature is determined to affect the generation of the wire breakage characteristic. W When the concentration is ≥3%, it is determined that temperature affects the formation of the stain characteristics;

[0067] If H S ≠[H min H max If dH > 5%, then the humidity is considered abnormal. K A humidity level greater than 5% indicates that humidity affects the formation of the hole feature; a humidity level greater than 10 indicates that humidity affects the formation of the wrinkle feature.

[0068] If F S ≠[F min F max If dF > 1.5, then the tension is considered abnormal. If Z ≥ 5, then the tension is considered to affect the generation of the wire breakage characteristic. K A tension greater than 5% is considered to have an impact on the formation of the aforementioned hole characteristics;

[0069] Among them, W max Indicates the theoretical maximum temperature value; H min and H max These represent the theoretical minimum humidity value and the theoretical maximum humidity value, respectively; F min and F max These represent the minimum and maximum standard tension values, respectively; temperature is expressed in °C; tension is expressed in N.

[0070] Preferably, the defect correlation analysis further includes:

[0071] When the temperature anomaly occurs, the temperature on the nonwoven fabric production line is adjusted until it returns to normal. Then, new target nonwoven fabric images from the same batch are re-identified to obtain the second number Z′ of the broken filament feature and the third area percentage S of the stain feature. W′;

[0072] If Z′<5, it is determined that the occurrence of the broken yarn feature is related to temperature; otherwise, it is determined that the occurrence of the broken yarn feature is unrelated to temperature, and the batch of nonwoven fabric has a quality defect.

[0073] If S W If the value is less than 3%, it is determined that the stain characteristics are related to temperature; otherwise, it is determined that the stain characteristics are not related to temperature, and the batch of nonwoven fabric is determined to have a quality defect.

[0074] When the humidity anomaly is detected, the humidity on the nonwoven fabric production line is processed until it returns to normal. Then, the image of a new target nonwoven fabric from the same batch is re-identified to obtain the fifth area ratio S of the pore feature. K The second length Y′cm of the fold feature;

[0075] If S K If the value is less than 5%, it is determined that the generation of the hole feature is related to humidity; otherwise, it is determined that the generation of the hole feature is not related to humidity, and the batch of nonwoven fabric is determined to have a quality defect.

[0076] If Y′ < 10, it is determined that the generation of the wrinkle feature is related to humidity; otherwise, it is determined that the generation of the wrinkle feature is not related to humidity, and the batch of nonwoven fabric has a quality defect.

[0077] When the tension anomaly exists, the tension on the nonwoven fabric production line is processed until the tension returns to normal. Then, the images of new target nonwoven fabrics from the same batch are re-identified to obtain the third quantity Z″ of the broken filament feature and the sixth area ratio S of the hole feature. K ";

[0078] If Z″ < 5, it is determined that the occurrence of the broken filament feature is related to tension; otherwise, it is determined that the occurrence of the broken filament feature is unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect.

[0079] If S K If the value is less than 5%, it is determined that the generation of the hole feature is related to tension; otherwise, it is determined that the generation of the hole feature is unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect.

[0080] This application also proposes a quality assessment device based on nonwoven fabric defect detection, comprising:

[0081] The component setting module is used to select the camera and light source settings based on the material information of the nonwoven fabric for which defect detection is required.

[0082] The data acquisition and analysis module is used to acquire environmental data, non-woven fabric image data and corresponding label data on the non-woven fabric production line, and to perform data analysis and processing to obtain a non-woven fabric dataset for training.

[0083] The model building module is used to build a quality assessment model, and to train and optimize the quality assessment model using the nonwoven fabric dataset to obtain the nonwoven fabric quality assessment model.

[0084] The quality assessment and analysis module is used to acquire images of the target nonwoven fabric to be identified on the nonwoven fabric production line, as well as real-time environmental data, and to perform quality assessment and analysis in conjunction with the nonwoven fabric quality assessment model.

[0085] The quality assessment method and apparatus based on nonwoven fabric defect detection described in this application have the following advantages:

[0086] 1. This application's quality assessment method based on nonwoven fabric defect detection ensures high-quality and accurate image acquisition by selecting camera and light source settings according to the nonwoven fabric's material information. This helps reduce problems such as uneven lighting and reflection, thereby improving defect detection accuracy. By acquiring environmental data, nonwoven fabric image data, and corresponding label data from the production line and performing data analysis, a high-quality training dataset can be generated. This helps the quality assessment model learn more accurate defect features, thus improving detection accuracy. By training and optimizing the training dataset using the constructed quality assessment model, an efficient and accurate nonwoven fabric quality assessment model can be obtained, capable of quickly identifying and analyzing defects on nonwoven fabrics, significantly improving detection efficiency. The quality assessment method based on nonwoven fabric defect detection achieves full automation from image acquisition and data analysis to quality assessment, reducing manual intervention, minimizing errors caused by human factors, and improving detection accuracy and stability.

[0087] 2. The quality assessment device based on nonwoven fabric defect detection disclosed in this application achieves a high degree of automation and intelligence in nonwoven fabric defect detection and quality assessment through the collaborative work of a component setting module, a data acquisition and analysis module, a model building module, and a quality assessment and analysis module. It can select the camera and light source settings according to the material information of the nonwoven fabric to be inspected, enabling the device to adapt to the inspection needs of different types and specifications of nonwoven fabrics. The quality assessment and analysis module can combine a nonwoven fabric quality assessment model to perform high-precision quality assessment and analysis of the target nonwoven fabric image. It can acquire data from the nonwoven fabric production line in real time and perform real-time quality assessment and analysis, enabling timely detection and resolution of problems in the production process and avoiding the generation of unqualified products. Furthermore, it can optimize the allocation of production resources, improving production efficiency and product quality. Attached Figure Description

[0088] Figure 1 This is a flowchart illustrating a quality assessment method based on nonwoven fabric defect detection as described in this application;

[0089] Figure 2 This is a schematic diagram of the forward orientation of the camera and light source in a quality assessment method based on nonwoven fabric defect detection as described in this application;

[0090] Figure 3 This is a schematic diagram showing the back-to-back arrangement of the camera and light source in a quality assessment method based on nonwoven fabric defect detection as described in this application.

[0091] Explanation of reference numerals in the attached figures:

[0092] 10-Non-woven fabric;

[0093] 20-Camera;

[0094] 30 - Light source. Detailed Implementation

[0095] like Figure 1 As shown, the quality assessment method based on nonwoven fabric defect detection described in this application includes the following steps:

[0096] S1. Based on the material information of the non-woven fabric 10 for which defect detection is required, select the settings for the camera 20 and the light source 30.

[0097] S2. Obtain environmental data, non-woven fabric image data, and corresponding label data from the non-woven fabric production line, and perform data analysis and processing to obtain a non-woven fabric dataset for training.

[0098] S3. Construct a quality assessment model for defect detection, and train and optimize the quality assessment model using a non-woven fabric dataset to obtain a non-woven fabric quality assessment model.

[0099] S4. Acquire images of the target nonwoven fabric to be identified on the nonwoven fabric production line, as well as real-time environmental data, and combine them with the nonwoven fabric quality assessment model to conduct quality assessment and analysis.

[0100] Furthermore, in this embodiment, step S1 specifically includes:

[0101] The camera 20 and the light source 30 can be set in two ways: facing forward and facing backward.

[0102] The forward orientation is configured such that camera 20 and light source 30 are located on the same side of the nonwoven fabric 10 surface, and the optical axis of camera 20 forms an angle α with the light emitted from light source 30; Figure 2The above shows the forward orientation, where both the camera 20 and the light source 30 are located above the surface of the nonwoven fabric 10. The optical axis of the camera 20 points to the surface of the nonwoven fabric 10 where the image to be acquired is to be obtained, and the light emitted from the light source 30 also points to the surface of the nonwoven fabric 10 where the image to be acquired is to be obtained. Furthermore, the optical axis of the camera 20 and the light emitted from the light source 30 form a certain angle α. The forward orientation can acquire a clearer image of the nonwoven fabric surface by increasing the reflected light from the surface of the nonwoven fabric 10.

[0103] The camera 20 and the light source 30 are positioned opposite each other on opposite sides of the nonwoven fabric 10, with the camera 20 and the light source 30 facing each other, such that the optical axis of the camera 20 is collinear with the light emitted from the light source 30; Figure 3 The image shown is a back-facing setup, where the camera 20 is positioned above the surface of the nonwoven fabric 10, and the light source 30 is positioned below the nonwoven fabric 10, forming a relative arrangement between the camera 20 and the light source 30. The optical axis of the camera 20 points towards the surface of the nonwoven fabric 10 to which the image needs to be captured, while the light emitted from the light source 30 points towards the back of the nonwoven fabric 10 and passes through the nonwoven fabric 10 towards the camera 20. The back-facing setup can improve the clarity and contrast of the captured image of the nonwoven fabric surface by increasing the transmitted light.

[0104] Obtain the chemical fiber content C of nonwoven fabric 10 cf Natural fiber content C nf Thickness D and surface roughness R a Satisfying C cf +C nf =1, and calculate the positive setting comprehensive index S according to the following formula. f and back-to-back setting comprehensive index S b :

[0105] S f =μ1*C cf +μ2*(1-D)+μ3*(1-R a )

[0106] S b =ω1*C nf +ω2*D+ω3*R a

[0107] Where μ1, μ2, μ3, ω1, ω2 and ω3 are all weighting coefficients, satisfying: μ1+μ2+μ3=1 and ω1+ω2+ω3=1;

[0108] If S f >S b If so, select the positive setting;

[0109] If S f <S b Then select the back-facing setting;

[0110] If Sf =S b If so, you can choose either forward setting or reverse setting, that is, choose either forward setting or reverse setting arbitrarily according to the actual situation;

[0111] The angle between the optical axis of camera 20 and the light rays emitted from light source 30 is selected as follows:

[0112] The gloss level G on the surface of nonwoven fabric 10 was measured using a gloss meter, and the light intensity in the nonwoven fabric production environment was measured using an illuminometer. The running speed V of nonwoven fabric 10 on the nonwoven fabric production line was obtained, and the angle selection coefficient B was used. α The formula is as follows:

[0113] B α =λ1*G+λ2*U+λ3*V

[0114] Wherein, λ1 represents the glossiness influence coefficient; λ2 represents the light intensity influence coefficient; and λ3 represents the operating speed influence coefficient.

[0115] If B α If the value is less than 6, then the angle α should be chosen as 0 < α ≤ 15°; the optimal value is 15°.

[0116] If 6≤B α If the value is less than 8, then the angle α should be chosen as 15° < α ≤ 30°; the optimal value is 30°.

[0117] If 8≤B α If the value is less than 10, then the angle α should be chosen as 30° < α ≤ 45°; the optimal value is 45°.

[0118] If 10≤B α If the value is less than 12, then the angle α should be chosen as 45° < α ≤ 60°; the optimal value is 60°.

[0119] If 12≤B α If the value is less than 14, then the angle α should be chosen as 60° < α ≤ 75°; the optimal value is 70°.

[0120] If B α If the value is ≥14, then the angle α should be chosen as 75°<α≤90°; the optimal value is 80°.

[0121] Example as follows:

[0122] The chemical fiber content (C) of nonwoven fabric 10 was measured using an infrared spectrometer. cf =0.6 and the proportion of natural fibers C nf =0.4, the thickness D of nonwoven fabric 10 was measured to be 0.3 mm using a thickness measuring instrument, and the surface roughness R of nonwoven fabric 10 was measured to be 0.3 mm using a roughness measuring instrument. a =0.2;

[0123] μ1=0.3, μ2=0.4, μ3=0.3, ω1=0.2, ω2=0.5, ω3=0.3;

[0124] S f =0.3*0.6+0.4*(1-0.3)+0.3*(1-0.2)=0.7;

[0125] S b =0.2*0.4+0.5*0.3+0.2*0.2=0.27;

[0126] Therefore S f >S b If so, select the positive setting;

[0127] The gloss of the surface of nonwoven fabric 10 was measured to be G=30 using a gloss meter, and the light intensity in the nonwoven fabric production environment was measured to be U=100Lux using an illuminometer. The running speed of nonwoven fabric 10 on the nonwoven fabric production line was obtained as V=10m / min.

[0128] λ1=0.1, λ2=0.05, λ3=0.1;

[0129] B α =0.1*30+0.05*100+0.1*10=9, so the angle α should be chosen as 30°<α≤45°, with 45° being the optimal choice.

[0130] Furthermore, in this embodiment, step S2 specifically includes:

[0131] Obtain environmental data from the nonwoven fabric production line, including temperature, humidity, and tension data;

[0132] Acquire nonwoven fabric image data and corresponding label data from the nonwoven fabric production line. The nonwoven fabric image data includes image samples with defect-free features, image samples with hole features, image samples with broken filament features, image samples with stain features, and image samples with wrinkle features. In the label data, the label corresponding to defect-free features is 0, the label corresponding to hole features is 1, the label corresponding to broken filament features is 2, the label corresponding to stain features is 3, and the label corresponding to wrinkle features is 4.

[0133] The environmental data is processed by outlier handling and smoothing to obtain the processed environmental data.

[0134] The nonwoven fabric image data was processed by weighted averaging, noise was removed by median filtering, image contrast was enhanced by histogram equalization, and image edges were extracted by Canny edge detection algorithm to obtain the processed nonwoven fabric image data.

[0135] Furthermore, in this embodiment, step S2 also includes:

[0136] A convolutional neural network model based on the LeNet structure is constructed. The processed nonwoven fabric image data is input into the convolutional neural network model for convolution operation to extract local features of the nonwoven fabric image and obtain feature maps.

[0137] The feature map is downsampled, and the extracted local features are integrated to obtain the spatial feature vector of the nonwoven fabric image;

[0138] A single-layer long short-term memory network model was constructed, and the continuously acquired non-woven fabric image sequence in the non-woven fabric image data was organized into a time series data format to obtain time series data.

[0139] By inputting time series data into a single-layer long short-term memory network model and using a gating mechanism to process the time series data, the temporal feature vector of the nonwoven fabric image is obtained.

[0140] The processed environmental data are subjected to statistical characteristic calculations, including mean, variance, and rate of change, to obtain the statistical characteristics of the environmental data.

[0141] Spatial feature vectors, temporal feature vectors, and environmental data statistical features are connected and fused to obtain spatiotemporal fused feature data.

[0142] The spatiotemporal fusion feature data is divided into training and testing sets according to the ratio A1:A2. The image data of the training and testing sets are paired and integrated with the corresponding label data to construct a nonwoven fabric dataset for training. A1:A2 = 8:2, that is, A1 = 8 and A2 = 2.

[0143] Furthermore, in this embodiment, step S3 specifically includes:

[0144] Construct a quality assessment model for a three-layer fully connected neural network, which includes an input layer, a hidden layer, and an output layer.

[0145] The dimension of the spatiotemporal fusion feature data is determined based on the principal component analysis method, and the number of input neurons in the input layer is set to be the same as the dimension of the spatiotemporal fusion feature data.

[0146] The number of hidden neurons in the hidden layer is set to 1.5 times the number of input neurons, and ReLU is used as the activation function for the hidden layer.

[0147] The number of output neurons in the output layer is set to 5, and the Softmax function is used as the activation function of the output layer.

[0148] Set the number of training rounds to Q and the number of training batches to e; Q is set to Q = 50 times and e is set to e = 32 batches.

[0149] Initialize the parameters of the quality assessment model, input the training set into the quality assessment model in training batches, and the training set passes through the input layer, hidden layer and output layer in sequence to obtain the predicted probability distribution of each sample for each label;

[0150] The cross-entropy loss function was used to analyze the difference between the probability distribution predicted by the quality assessment model and the true label, and the average loss value for each batch was calculated.

[0151] Based on the calculated average loss value, the backpropagation algorithm is used to calculate the gradient of the loss function with respect to each connection weight and bias in the model;

[0152] Based on the gradients of each connection weight and bias obtained from the calculation, the connection weights and biases of the model are updated using the stochastic gradient descent optimization algorithm.

[0153] Repeat the training until the training round is completed.

[0154] Furthermore, in this embodiment, step S3 also includes:

[0155] The quality assessment model that has completed the test rounds is tested using a test set to obtain the identification results, and the accuracy E, recall R, F1 score and test loss L are calculated.

[0156] The formula for calculating accuracy E is as follows:

[0157]

[0158] The formula for calculating recall R is as follows:

[0159]

[0160] The formula for calculating the F1 value is as follows:

[0161]

[0162] The test loss value L is calculated using the cross-entropy loss function;

[0163] Where D represents the number of correctly identified image samples; A represents the total number of image samples in the test set; X i M represents the number of image samples that correctly identified the i-th type of feature; i The total number of image samples representing the i-th feature; i = 1, 2, 3, 4, 5; i = 1 represents the flawless feature; i = 2 represents the hole feature; i = 3 represents the broken thread feature; i = 4 represents the stain feature; i = 5 represents the wrinkle feature;

[0164] The overall coefficient K for model testing is calculated using the following formula:

[0165] K = C1*E + C2*R + C3*F1 + C4*L

[0166] Where C1, C2, C3 and C4 are all weighting coefficients, satisfying C1+C2+C3+C4=1;

[0167] If K≥0.8, the quality assessment model is deemed to meet the requirements and will be used as the quality assessment model for nonwoven fabrics.

[0168] If K < 0.8, the quality assessment model is deemed unsuitable and retrained until K ≥ 0.8 is met.

[0169] Example as follows:

[0170] If the total number of image samples in the test set is A = 200, and the number of correctly identified image samples is D = 180, then E = 90%;

[0171] M1 = 40, X1 = 37, M2 = 30, X2 = 26, M3 = 30, X3 = 28, M4 = 50, X4 = 45, M5 = 50, X5 = 44, then R = 90.1%;

[0172] F1=90%, L=0.2, C1=0.4, C2=0.4, C3=0.1, C4=0.1;

[0173] Since K = 0.83 > 0.8, the quality assessment model is deemed to meet the requirements and will be used as the quality assessment model for nonwoven fabrics.

[0174] Furthermore, in this embodiment, step S4 specifically includes:

[0175] Images of the target nonwoven fabric to be identified on the nonwoven fabric production line are acquired. The images are then identified using a nonwoven fabric quality assessment model, and the first area proportion S of the pore features is statistically calculated. K The first number of broken fibers (Z) and the second area percentage of stains (S) W The first length Ycm of the fold feature;

[0176] If S K <5%, Z<5, S W If both <3% and Y < 10 are met, the nonwoven fabric is judged to be of qualified quality.

[0177] If S K <5%, Z<5, S W If either <3% or Y < 10 is not met, the nonwoven fabric is deemed to be of substandard quality, and a defect correlation analysis is performed in conjunction with real-time environmental data.

[0178] Example as follows:

[0179] The first area proportion S of the hole feature was obtained statistically. K =3%, the first number of broken fibers Z = 3, the second area percentage of stains S W =2% and the first length Y of the fold feature = 8cm;

[0180] Then S K <5%, Z<5, S W The nonwoven fabric is deemed to be of acceptable quality if both <3% and Y < 10 are met.

[0181] Furthermore, in this embodiment, the defect correlation analysis includes:

[0182] Acquire real-time environmental data from the nonwoven fabric production line, including real-time temperature values ​​(W). S Real-time humidity value H S and tension real-time value F S And obtain the temperature change rate dW, humidity change rate dH and tension change rate dF over the time interval ΔT;

[0183] If W S >W max If dW > 2, then a temperature anomaly is determined. If Z ≥ 5, then temperature is determined to affect the generation of wire breakage characteristics. W When the concentration is ≥3%, it is determined that temperature has an impact on the formation of stain characteristics; among them, high temperature or unstable temperature environment is a factor leading to yarn breakage; high temperature environment will make certain substances more likely to adhere to the surface of non-woven fabric and thus form stains;

[0184] If H S ≠[H min H max If dH > 5%, then the humidity is considered abnormal. K A humidity level greater than 5% indicates that humidity affects the formation of pores, while a humidity level greater than 10 indicates that humidity affects the formation of wrinkles. In particular, unsuitable humidity can affect the forming process of nonwoven fabrics, leading to the formation of pores. Humidity changes can cause uneven shrinkage or expansion of nonwoven fabric materials, resulting in wrinkles.

[0185] If F S ≠[F min F max If dF > 1.5, then the tension is considered abnormal. When Z ≥ 5, the tension is considered to affect the generation of the wire breakage characteristic. K >5% indicates that tension affects the generation of hole features; among them, unstable tension on nonwoven fabrics on equipment can lead to fiber breakage or damage to the nonwoven fabric structure, which can easily produce broken filament features and hole features.

[0186] Among them, W max Indicates the theoretical maximum temperature value; Hmin and H max These represent the theoretical minimum humidity value and the theoretical maximum humidity value, respectively; F min and F max These represent the minimum and maximum standard tension values, respectively; temperature is in °C; tension is in N.

[0187] Example as follows:

[0188] Real-time temperature value W S =40℃, real-time humidity value H S =55%, real-time tension value F S =35N;

[0189] ΔT=10min,W max =35℃,H min =20%, H max =50%, F min =15N,F max =30;

[0190] In the time interval ΔT, the temperature rises from 20℃ to 45℃, so the rate of temperature change dW = 2.5℃ / min. Therefore, W S >W max If dW > 2, the temperature is considered abnormal;

[0191] If the first number of broken wires Z = 6 > 5, then it is determined that temperature has an impact on the generation of broken wires.

[0192] The second area percentage of the stain feature S W If 4% > 3%, then temperature is considered to have an impact on the formation of stain characteristics.

[0193] If the humidity changes from 35% to 55% within a time interval ΔT, then the rate of change of humidity dH = 2% / min, therefore H S ≠[H min H max If dH > 5%, then the humidity is considered abnormal.

[0194] The first area percentage S of the hole feature K If 6% > 5%, then humidity is determined to have an impact on the formation of hole characteristics;

[0195] The first length of the fold feature, Y = 12cm > 10cm, indicates that humidity affects the generation of the fold feature.

[0196] During the time interval ΔT, the tension changes from 15N to 35N, so the rate of change of tension dF = 2N / min. Therefore, dF ≠ [F]. min F max If dF > 1.5, then the tension is considered abnormal;

[0197] If the first number of broken wires Z = 6 > 5, then it is determined that tension has an impact on the generation of broken wires.

[0198] The first area percentage S of the hole feature K If 6% > 5%, then tension is considered to have an impact on the generation of hole characteristics.

[0199] Furthermore, in this embodiment, the defect correlation analysis also includes:

[0200] When an abnormal temperature is detected, the temperature on the nonwoven fabric production line is controlled until it returns to normal. Then, new target nonwoven fabric images from the same batch are re-identified to obtain the second number of broken filaments, Z′, and the third area percentage of stains, S. W ′;

[0201] If Z′<5, it is determined that the occurrence of the broken yarn feature is related to temperature; otherwise, it is determined that the occurrence of the broken yarn feature is not related to temperature, and the batch of nonwoven fabric has a quality defect.

[0202] If S W If the value is less than 3%, the stain characteristics are determined to be related to temperature; otherwise, the stain characteristics are determined to be unrelated to temperature, and the batch of nonwoven fabric is determined to have quality defects.

[0203] When abnormal humidity is detected, the humidity on the nonwoven fabric production line is adjusted until it returns to normal. Then, images of new target nonwoven fabrics from the same batch are re-identified to obtain the fifth area proportion S of the pore features. K The second length Y′cm of the fold feature;

[0204] If S K If the value is less than 5%, the generation of the pore features is determined to be related to humidity; otherwise, the generation of the pore features is determined to be unrelated to humidity, and the batch of nonwoven fabric is determined to have quality defects.

[0205] If Y′<10, it is determined that the wrinkle feature is related to humidity; otherwise, it is determined that the wrinkle feature is not related to humidity, and the batch of nonwoven fabric has a quality defect.

[0206] When tension anomalies are present, the tension on the nonwoven fabric production line is adjusted until it returns to normal. Then, images of new target nonwoven fabrics from the same batch are re-identified to obtain the third quantity Z″ of broken filament features and the sixth area proportion S of void features. K ";

[0207] If Z″ < 5, it is determined that the occurrence of the broken yarn feature is related to tension; otherwise, it is determined that the occurrence of the broken yarn feature is unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect.

[0208] If S K If the value is less than 5%, the generation of the hole feature is determined to be related to tension; otherwise, the generation of the hole feature is determined to be unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect.

[0209] Example as follows:

[0210] The real-time temperature value W of the nonwoven fabric production line was detected. S =40°C, theoretical maximum temperature value W max =35℃, and the rate of temperature change dW = 2.5℃ within the time interval ΔT = 10 min, which meets the temperature anomaly judgment condition W. S >W max And dW > 2;

[0211] At this point, the temperature of the nonwoven fabric production line is adjusted to return it to a normal range, such as 25°C. Then, new target nonwoven images from the same batch of nonwoven fabric are re-identified, yielding the second number of broken filaments Z′ = 3, and the third area percentage S of the stain feature. W =2%;

[0212] Since Z′=3<5, the occurrence of the broken wire characteristic is related to temperature.

[0213] Because S W Since ' = 2% < 3%, the determination of stain characteristics is related to the temperature at which the stain was formed;

[0214] The real-time humidity value H of the nonwoven fabric production line was detected. S =55%, theoretical minimum humidity value H min =20%, theoretical maximum humidity value H max =50%, and the humidity change rate is 2% / min within the time period ΔT = 10min, which meets the humidity anomaly judgment condition H. S ≠[H min H max And dH > 5%;

[0215] At this point, the humidity of the nonwoven fabric production line is adjusted to normal, for example, to 30%. Then, new target nonwoven images in the same batch of nonwoven fabric are re-identified to obtain the fifth area proportion S of the pore features. K Y′c = 4%, and the second length of the fold feature is Y′c = 8cm;

[0216] Because S K =4% < 5%, therefore the occurrence of hole characteristics is related to humidity;

[0217] Since Y′=8<10, it is determined that the formation of wrinkle features is related to humidity;

[0218] The real-time tension value F of the nonwoven fabric production line was detected. S =35N, minimum standard tension F min =15N, standard maximum tension F max =30, and the rate of change of tension dF = 2N / min during the time interval ΔT = 10min, which satisfies the condition for judging tension anomalies: dF ≠ [F min F max And dF > 1.5;

[0219] At this point, the tension of the nonwoven fabric production line is adjusted to restore it to the normal range, such as 22N. Then, the new target nonwoven images in the same batch of nonwoven fabric are re-identified, and the third number of broken filaments Z″ = 4, and the sixth area ratio S of the hole feature is obtained. K =4%.

[0220] Since Z″=4<5, the occurrence of the broken wire characteristic is related to tension;

[0221] Because S K =4% < 5%, indicating that the formation of hole characteristics is related to tension.

[0222] This application also proposes a quality assessment device based on nonwoven fabric defect detection, comprising:

[0223] The component setting module is used to select the camera and light source settings based on the material information of the nonwoven fabric for which defect detection is required.

[0224] The data acquisition and analysis module is used to acquire environmental data, non-woven fabric image data and corresponding label data on the non-woven fabric production line, and to perform data analysis and processing to obtain a non-woven fabric dataset for training.

[0225] The model building module is used to build a quality assessment model. The quality assessment model is trained and optimized using a non-woven fabric dataset to obtain a non-woven fabric quality assessment model.

[0226] The quality assessment and analysis module is used to acquire images of the target nonwoven fabric to be identified on the nonwoven fabric production line, as well as real-time environmental data, and to perform quality assessment and analysis in conjunction with the nonwoven fabric quality assessment model.

[0227] In the description of this application, it should be understood that the orientation or positional relationship indicated by directional terms such as "front, back, up, down, left, right", "horizontal, vertical, horizontal" and "top, bottom" is usually based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this application and simplifying the description. Unless otherwise stated, these directional terms do not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the scope of protection of this application.

[0228] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this application.

Claims

1. A quality assessment method based on nonwoven fabric defect detection, characterized in that, Includes the following steps: S1. Based on the material information of the non-woven fabric (10) to be inspected for defects, select the settings of the camera (20) and the light source (30); S2. Obtain environmental data, non-woven fabric image data, and corresponding label data from the non-woven fabric production line, and perform data analysis and processing to obtain a non-woven fabric dataset for training. S3. Construct a quality assessment model for defect detection, and train and optimize the quality assessment model using the nonwoven fabric dataset to obtain the nonwoven fabric quality assessment model. S4. Obtain the target nonwoven fabric image to be identified on the nonwoven fabric production line, as well as real-time environmental data, and combine it with the nonwoven fabric quality assessment model to perform quality assessment and analysis. Step S1 specifically includes: The camera (20) and light source (30) can be set in two ways: facing forward and facing backward. The forward orientation is configured such that the camera (20) and the light source (30) are located on the same side of the nonwoven fabric (10) surface, and the optical axis of the camera (20) forms a certain angle with the light emitted from the light source (30). ; The back-facing configuration has the camera (20) and the light source (30) located on opposite sides of the non-woven fabric (10), and the camera (20) and the light source (30) are arranged opposite to each other, so that the optical axis of the camera (20) is collinear with the light emitted from the light source (30); Obtain the chemical fiber percentage of nonwoven fabric (10) Natural fiber content ,thickness and surface roughness ,satisfy And calculate the positive setting comprehensive index according to the following formula. and back-to-back setting comprehensive indicators : in, , , , , and All are weighting coefficients, satisfying: and ; like > If so, then select the positive setting; like < Then select the aforementioned back-facing setting; like = Then select the forward setting or the reverse setting.

2. The quality assessment method based on nonwoven fabric defect detection according to claim 1, characterized in that, Step S2 specifically includes: Acquire environmental data from the nonwoven fabric production line, including temperature data, humidity data, and tension data; The nonwoven fabric image data and corresponding label data on the nonwoven fabric production line are obtained. The nonwoven fabric image data includes image samples with defect-free features, image samples with hole features, image samples with broken filament features, image samples with stain features, and image samples with wrinkle features. In the label data, the label corresponding to the defect-free feature is 0, the label corresponding to the hole feature is 1, the label corresponding to the broken filament feature is 2, the label corresponding to the stain feature is 3, and the label corresponding to the wrinkle feature is 4. The environmental data is subjected to outlier processing and smoothing to obtain the processed environmental data; The nonwoven fabric image data is processed by weighted averaging, noise is removed by median filtering, image contrast is enhanced by histogram equalization, and image edges are extracted by Canny edge detection algorithm to obtain the processed nonwoven fabric image data.

3. The quality assessment method based on nonwoven fabric defect detection according to claim 2, characterized in that, Step S2 further includes: A convolutional neural network model based on the LeNet structure is constructed. The processed nonwoven fabric image data is input into the convolutional neural network model to perform convolution operations, extract local features of the nonwoven fabric image, and obtain feature maps. The feature map is downsampled, and the extracted local features are integrated to obtain the spatial feature vector of the nonwoven fabric image; A single-layer long short-term memory network model was constructed, and the continuously acquired non-woven fabric image sequence in the non-woven fabric image data was organized into a time series data format to obtain time series data. By inputting the time series data into the single-layer long short-term memory network model and processing the time series data using a gating mechanism, the time feature vector of the nonwoven fabric image is obtained. The processed environmental data are subjected to statistical feature calculations, including mean, variance, and rate of change, to obtain the environmental data statistical features. The spatial feature vector, the temporal feature vector, and the environmental data statistical features are connected and fused to obtain spatiotemporal fused feature data. The spatiotemporal fusion feature data is divided into a training set and a test set according to the ratio A1:A2. The image data of the training set and the test set are paired and integrated with the corresponding label data to construct a nonwoven fabric dataset for training.

4. The quality assessment method based on nonwoven fabric defect detection according to claim 3, characterized in that, Step S3 specifically includes: A quality assessment model for a three-layer fully connected neural network is constructed, comprising an input layer, a hidden layer, and an output layer. The dimension of the spatiotemporal fusion feature data is determined according to the principal component analysis method, and the number of input neurons in the input layer is set to be the same as the dimension of the spatiotemporal fusion feature data. The number of hidden neurons in the hidden layer is set to 1.5 times the number of input neurons, and ReLU is used as the activation function of the hidden layer. The number of output neurons in the output layer is set to 5, and the Softmax function is used as the activation function of the output layer. Set the number of training rounds to Q and the number of training batches to e; The parameters of the quality assessment model are initialized, and the training set is input into the quality assessment model in the training batch. The training set passes through the input layer, the hidden layer and the output layer in sequence to obtain the predicted probability distribution of each sample for each label. The cross-entropy loss function was used to analyze the difference between the probability distribution predicted by the quality assessment model and the true label, and the average loss value for each batch was calculated. Based on the calculated average loss value, the backpropagation algorithm is used to calculate the gradient of the loss function with respect to each connection weight and bias in the model; Based on the gradients of each connection weight and bias obtained from the calculation, the connection weights and biases of the model are updated using the stochastic gradient descent optimization algorithm. Repeat the training until the training rounds are completed.

5. The quality assessment method based on nonwoven fabric defect detection according to claim 4, characterized in that, Step S3 further includes: The quality assessment model that has completed the training rounds is tested using the test set to obtain recognition results and calculate accuracy. Recall rate , Value and test loss value ; The accuracy The calculation formula is as follows: The recall rate The calculation formula is as follows: The The formula for calculating the value is as follows: The test loss value is calculated using the cross-entropy loss function. ; Where D1 represents the number of correctly identified image samples; This represents the total number of image samples in the test set; This indicates that the first one was correctly identified. Number of image samples for class features; Indicates the first The total number of image samples with class features; =1, 2, 3, 4, 5; =1 indicates the defect-free feature; =2 indicates the hole feature; =3 indicates the broken wire feature; =4 indicates the characteristics of the stain; =5 indicates the wrinkle feature; The overall coefficient of the model test is calculated using the following formula. : in, , , and All are weighting coefficients, satisfying + + + =1; like If the value is ≥0.8, the quality assessment model is deemed to meet the requirements and will be used as the quality assessment model for nonwoven fabrics. like If the value is less than 0.8, the quality assessment model is deemed unqualified and retrained until it meets the requirements. ≥0.

8.

6. The quality assessment method based on nonwoven fabric defect detection according to claim 5, characterized in that, Step S4 specifically includes: The target nonwoven fabric image to be identified is acquired on the nonwoven fabric production line. The target nonwoven fabric image is then identified using the nonwoven fabric quality assessment model, and the first area ratio of the pore features is statistically calculated. The first number of the broken wire features The root, the second area ratio of the stain feature and the first length of the fold feature The first length The unit is cm; like <5% <5. <3% and If all conditions <10 are met, the nonwoven fabric is deemed to be of acceptable quality. like <5% <5. <3% and If any of the following conditions is not met, the nonwoven fabric is deemed to be of substandard quality, and a defect correlation analysis is performed in conjunction with real-time environmental data.

7. The quality assessment method based on nonwoven fabric defect detection according to claim 6, characterized in that, The defect correlation analysis includes: Acquire real-time environmental data from the nonwoven fabric production line, including real-time temperature values. Real-time humidity value and tension real-time value and obtain Temperature change rate over time Humidity change rate and rate of change of tension ; like > and If the value is greater than 2, then the temperature is considered abnormal. When the temperature is ≥5, it is determined that the temperature affects the generation of the broken wire characteristic. When the concentration is ≥3%, it is determined that temperature affects the formation of the stain characteristics; like ≠[ , ]and If the humidity is greater than 5%, it is considered abnormal. A humidity level greater than 5% is considered to affect the formation of the aforementioned hole characteristics. A humidity level greater than 10 indicates that humidity affects the formation of the wrinkle features; like ≠[ , ]and If the value is greater than 1.5, then the tension is considered abnormal. When the tension is ≥5, it is determined that the tension affects the generation of the wire breakage characteristic. A tension greater than 5% is considered to have an impact on the formation of the aforementioned hole characteristics; in, This represents the theoretical maximum temperature value; This represents the theoretical minimum humidity value. This represents the theoretical maximum humidity value. This represents the minimum standard tension. The unit for the maximum standard tension is °C; the unit for the maximum standard temperature is N.

8. The quality assessment method based on nonwoven fabric defect detection according to claim 7, characterized in that, The defect correlation analysis also includes: When the temperature anomaly is detected, the temperature on the nonwoven fabric production line is adjusted until it returns to normal. Then, the images of new target nonwoven fabrics from the same batch are re-identified to obtain the second number of broken filament features. The third area ratio of the root and the stain features ; like If the value is less than 5, it is determined that the occurrence of the broken filament feature is related to temperature; otherwise, it is determined that the occurrence of the broken filament feature is unrelated to temperature, and the batch of nonwoven fabric is determined to have a quality defect. like If the value is less than 3%, it is determined that the stain characteristics are related to temperature; otherwise, it is determined that the stain characteristics are not related to temperature, and the batch of nonwoven fabric is determined to have a quality defect. When the humidity anomaly is detected, the humidity on the nonwoven fabric production line is processed until it returns to normal. Then, the images of new target nonwoven fabrics from the same batch are re-identified to obtain the fifth area ratio of the pore features. and the second length of the fold feature The two lengths The unit is cm; like If the value is less than 5%, it is determined that the generation of the pore feature is related to humidity; otherwise, it is determined that the generation of the pore feature is not related to humidity, and the batch of nonwoven fabric is determined to have a quality defect. like If the value is less than 10, it is determined that the wrinkle feature is related to humidity; otherwise, it is determined that the wrinkle feature is not related to humidity, and the batch of nonwoven fabric has a quality problem. When the tension anomaly exists, the tension on the nonwoven fabric production line is adjusted until it returns to normal. Then, the images of new target nonwoven fabrics from the same batch are re-identified to obtain the third quantity of the broken filament feature. and the sixth area ratio of the hole feature ; like If the value is less than 5, it is determined that the occurrence of the broken filament feature is related to tension; otherwise, it is determined that the occurrence of the broken filament feature is unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect. like If the value is less than 5%, it is determined that the generation of the hole feature is related to tension; otherwise, it is determined that the generation of the hole feature is unrelated to tension, and the batch of nonwoven fabric is determined to have a quality defect.

9. A quality assessment device based on nonwoven fabric defect detection, characterized in that, include: The component setting module is used to select the setting method of the camera (20) and the light source (30) according to the material information of the non-woven fabric (10) to be inspected for defects; The data acquisition and analysis module is used to acquire environmental data, non-woven fabric image data and corresponding label data on the non-woven fabric production line, and to perform data analysis and processing to obtain a non-woven fabric dataset for training. The model building module is used to build a quality assessment model, and to train and optimize the quality assessment model using the nonwoven fabric dataset to obtain the nonwoven fabric quality assessment model. The quality assessment and analysis module is used to acquire images of the target nonwoven fabric to be identified on the nonwoven fabric production line, as well as real-time environmental data, and to perform quality assessment and analysis in conjunction with the nonwoven fabric quality assessment model. The camera (20) and light source (30) can be set in two ways: facing forward and facing backward. The forward orientation is configured such that the camera (20) and the light source (30) are located on the same side of the nonwoven fabric (10) surface, and the optical axis of the camera (20) forms a certain angle with the light emitted from the light source (30). ; The back-facing configuration has the camera (20) and the light source (30) located on opposite sides of the non-woven fabric (10), and the camera (20) and the light source (30) are arranged opposite to each other, so that the optical axis of the camera (20) is collinear with the light emitted from the light source (30); Obtain the chemical fiber percentage of nonwoven fabric (10) Natural fiber content ,thickness and surface roughness ,satisfy And calculate the positive setting comprehensive index according to the following formula. and back-to-back setting comprehensive indicators : in, , , , , and All are weighting coefficients, satisfying: and ; like > If so, then select the positive setting; like < Then select the aforementioned back-facing setting; like = Then select the forward setting or the reverse setting.

Images