Non-woven fiber mesh gram weight online detection method and device and storage medium

Abstract

The invention discloses a non-woven fiber web gram weight on-line detection method and device and a storage medium. The quality of subsequent series products such as non-woven fabric can be higher through detection of the previous procedure. The method comprises the following steps: acquiring an image of a non-woven fiber web; preprocessing the image; performing wavelet transformation on the preprocessed image, and converting the image into a low-frequency image and a high-frequency image; a neural network model is constructed, the input of the neural network is the low-frequency image and the high-frequency image after wavelet transformation, and the output of the neural network is the gram weight of the non-woven fiber mesh; the trained neural network model is used for online estimation of the gram weight of the non-woven fiber mesh. The method provided by the invention is convenient in sampling and high in detection efficiency; the wavelet transform is combined with the neural network, so that the detection accuracy is greatly improved; moreover, the neural network model can be continuously learned and updated, meets the detection requirements of non-woven fiber webs of different batches and different qualities, has better adaptability and robustness, and is wide in application range.

Description

Technical Field

[0001] The present invention relates to the technical field of non-woven fiber web quality detection, and particularly to an on-line detection method, device and computer-readable storage medium for the grammage of a non-woven fiber web. Background Art

[0002] In the quality detection of high-end non-woven products in the textile industry, not only is it required that the grammage of the produced non-woven fabric meets the requirements, but also there are requirements for the fiber web carded by a non-woven carding machine. According to the relevant regulations of non-woven carding machines in the textile industry, the traditional method usually uses the sampling weighing method for determination. During normal production, at least 10 samples are taken at intervals of 100 mm from the edge horizontally, 60 mm apart, and 500 mm from the edge vertically, 60 mm apart. The size of each sample is: 100 mm × 100 mm or 200 mm × 200 mm, and the samples are weighed separately using a balance. Then, the weight CV value is obtained based on the weights of a series of fiber webs to determine whether the quality of the fiber web meets the standard.

[0003] The above method has the following significant defects:

[0004] (1) The sampling operation is cumbersome: It is necessary to manually use cutting tools to cut the non-woven fiber web sample into the specified size. For batch detection, this process is time-consuming and laborious. For example, in a non-woven fabric factory with large-scale production, the non-woven fiber web detection requires the production to stop for inspection, and the inspectors need to spend a lot of time cutting samples, which seriously affects the inspection progress and production efficiency.

[0005] (2) The detection efficiency is low: The cut samples need to be weighed one by one on an electronic balance. If the number of samples is large, operations such as waiting for the balance to stably display the weight and recording data will further lengthen the detection time.

[0006] (3) The detection accuracy is limited: Random sampling has limitations. If the sampling is not representative, it may miss the quality problem areas, resulting in a large deviation between the detection result and the actual quality situation. For example, in a batch of non-woven fiber webs, there is a situation of uneven grammage locally, but the random sampling does not draw this area, so the problem cannot be found. Even if the number of samples is increased, it is difficult to completely cover all parts and production links of the product. And according to the statistical principle, to achieve a high detection accuracy, the number of samples needs to reach a certain scale, which is often difficult to achieve in actual production due to cost and time limitations.

[0007] (4) Detection has hysteresis: The weighing detection is carried out after the production of the non-woven fiber web is completed, and the product is sent to the laboratory or the detection point, and it is impossible to monitor the quality change in real time on the production line. When quality problems are detected, a large number of unqualified products may have been produced, resulting in waste of raw materials, manpower and time. Due to the inability to timely feedback quality information, quality problems caused by parameter fluctuations, equipment failures, etc. during the production process cannot be discovered and corrected in time, affecting the overall quality and production efficiency of the product. Summary of the Invention

[0008] The technical problem to be solved by the embodiments of the present application is: how to conveniently and accurately perform on-line detection of the gram weight of the non-woven fiber web.

[0009] To solve the above technical problems, in a first aspect, the embodiments of the present application provide a method for on-line detection of the gram weight of a non-woven fiber web, and the method includes the following steps:

[0010] Obtain an image of the non-woven fiber web;

[0011] Preprocess the image;

[0012] Perform wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image;

[0013] Construct a neural network model, the input of the neural network is the low-frequency image and the high-frequency image after wavelet transform, and the output is the gram weight of the non-woven fiber web; the trained neural network model is used for on-line estimation of the gram weight of the non-woven fiber web.

[0014] Preferably, the obtaining of the image of the non-woven fiber web is specifically: obtaining an image of the non-woven fiber web on the conveying device between the output system of the carding machine and the cross-laying system.

[0015] Preferably, the preprocessing method includes a light balance method and / or an image enhancement method.

[0016] Preferably, the performing of wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image specifically includes:

[0017] Step S31: Perform low-pass and high-pass filtering on each row of the preprocessed image to obtain the low-frequency component and the high-frequency component of the row;

[0018] Step S32: Downsample the low-frequency component and the high-frequency component of the row respectively to halve the data volume;

[0019] Step S33: Perform low-pass filtering and high-pass filtering on each column of the low-frequency component and the high-frequency component of the row obtained in Step S31 to obtain the low-frequency component and the high-frequency component of the column;

[0020] Step S34: Downsample the low-frequency components and high-frequency components of the column respectively to halve the data volume;

[0021] Finally, four sub-band images are obtained, namely LL, LH, HL, and HH sub-bands; among them, the LL sub-band is the wavelet coefficients generated by convolving with a low-pass wavelet filter in two directions; the HL sub-band represents the singular characteristics of the image in the horizontal direction; the LH sub-band represents the singular characteristics of the image in the vertical direction; the HH sub-band represents the diagonal edge characteristics of the image;

[0022] Use the LL sub-band as the low-frequency image, and combine the remaining sub-bands as the high-frequency image.

[0023] Preferably, the neural network is a convolutional neural network.

[0024] Furthermore, the convolutional neural network includes two convolutional layers and two pooling layers for extracting deep features of the nonwoven fabric web image; after the low-frequency image and high-frequency image of the input layer are processed by the convolutional layer and pooling layer to extract multi-level features, they are mapped to a single-node output layer through a fully connected layer to predict the grammage value of the nonwoven fabric web.

[0025] In a second aspect, an embodiment of the present application provides an on-line nonwoven fabric web grammage detection device, and the device includes:

[0026] An image acquisition module for acquiring an image of the nonwoven fabric web;

[0027] An image preprocessing module for preprocessing the image;

[0028] A wavelet transform module for performing wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image;

[0029] A neural network model construction module for constructing a neural network model, where the input of the neural network is the low-frequency image and high-frequency image after wavelet transform, and the output is the grammage of the nonwoven fabric web; the trained convolutional neural network model is used for on-line estimation of the grammage of the nonwoven fabric web.

[0030] Preferably, the wavelet transform module includes:

[0031] A row filtering module for performing low-pass and high-pass filtering on each row of the preprocessed image to obtain the low-frequency component and high-frequency component of the row;

[0032] A row downsampling module for downsampling the low-frequency component and high-frequency component of the row respectively to halve the data volume;

[0033] A column filtering module for performing low-pass filtering and high-pass filtering on each column of the low-frequency component and high-frequency component of the row obtained by the row filtering module to obtain the low-frequency component and high-frequency component of the column;

[0034] A column downsampling module is used to downsample the low-frequency component and the high-frequency component of the column respectively, so that the data volume is halved.

[0035] Finally, four sub-band images are obtained, namely LL, LH, HL, and HH sub-bands. Among them, the LL sub-band is the wavelet coefficients generated after convolution using low-pass wavelet filters in two directions; the HL sub-band represents the singular characteristics of the horizontal direction of the image; the LH sub-band represents the singular characteristics of the vertical direction of the image; the HH sub-band represents the diagonal edge characteristics of the image.

[0036] Taking the LL sub-band as the low-frequency image, and combining the remaining sub-bands as the high-frequency image.

[0037] In a third aspect, an embodiment of the present application further provides a non-woven fiber web grammage on-line detection device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. The characteristic is that when the processor executes the program, the steps of the above-mentioned non-woven fiber web grammage on-line detection method are realized.

[0038] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored. The characteristic is that when the program is executed by the processor, the steps of the above-mentioned non-woven fiber web grammage on-line detection method are realized.

[0039] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0040] (1) Convenient sampling and high detection efficiency: There is no need for manual use of cutting tools to perform cumbersome cutting and sampling on the non-woven fiber web. By directly obtaining the image of the non-woven fiber web on the production line, compared with the traditional cutting and weighing method, the steps of cutting and weighing one by one are omitted, greatly shortening the detection time, avoiding the manpower and time consumption caused by cutting. Especially in non-woven fabric factories with large-scale production, it can greatly save the time and energy of detection personnel and significantly improve the detection efficiency.

[0041] (2) High detection accuracy: Using wavelet transform to process the image, converting the image into a low-frequency image and a high-frequency image, as the input data of the neural network. Compared with directly feeding the original picture, it can greatly improve the prediction accuracy of the neural network and improve the detection accuracy.

[0042] (2) It can be detected in real time online: It can obtain the image of the non-woven fiber web on the conveying device between the carding machine output system and the cross-laying system for detection, realizing real-time monitoring on the production line, being able to promptly detect quality problems, promptly adjust parameter fluctuations, equipment failures, etc. during the production process, avoiding the output of a large number of unqualified products, reducing the waste of raw materials, labor and time, and effectively improving production efficiency and product quality.

[0043] (4) It can adaptively adjust: As the production process progresses, the neural network model can continuously learn and update to adapt to the detection requirements of non-woven fiber webs of different batches and different qualities, having good self-adaptability and robustness, and a wide range of applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0045] Figure 1 It is a flowchart of the online detection method for the grammage of the non-woven fiber web provided in Embodiment 1 of the present application;

[0046] Figure 2 It is a schematic diagram of the preprocessed image in Embodiment 1 of the present application;

[0047] Figure 3 It is a schematic diagram of the high-frequency image after wavelet transform in Embodiment 1 of the present application;

[0048] Figure 4 It is a schematic diagram of the low-frequency image after wavelet transform in Embodiment 1 of the present application;

[0049] Figure 5 It is a comparison chart of the prediction results of the conventional convolutional neural network, the optimized convolutional neural network in this embodiment, and the actual grammage of the non-woven fiber web in Embodiment 1 of the present application;

[0050] Figure 6 It is a comparison chart of the training losses of the conventional convolutional neural network and the optimized convolutional neural network in this embodiment in Embodiment 1 of the present application;

[0051] Figure 7 It is a comparison chart of the mean absolute errors of the conventional convolutional neural network and the optimized convolutional neural network in this embodiment in Embodiment 1 of the present application;

[0052] Figure 8 It is a schematic diagram of the structure of the online detection device for the grammage of the non-woven fiber web provided in Embodiment 2 of the present application.

[0053] Through the above-mentioned accompanying drawings, specific embodiments of the present application have been shown, and will be described in more detail hereinafter. These drawings and the written description are not intended to limit the scope of the concept of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. Detailed Description of the Embodiments

[0054] To better understand the above technical solution, exemplary embodiments will be described in detail herein, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numerals in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices consistent with some aspects of the present application as detailed in the appended claims.

[0055] Embodiment 1

[0056] Figure 1 It is a flowchart of an on-line detection method for the grammage of a non-woven fiber web provided in Embodiment 1 of the present application. The on-line detection method for the grammage of the non-woven fiber web includes the following steps:

[0057] Step S1: Obtain an image of the non-woven fiber web;

[0058] Specifically, the non-woven fiber web processing system includes a carding machine, a conveying device, and a cross-laying system. The non-woven fiber web formed by carding in the carding machine is conveyed to the cross-laying system through the conveying device. A plurality of CCD cameras are arranged above the conveying device, and the CCD cameras are evenly arranged along the width of the non-woven fiber web to ensure that the entire width of the non-woven fiber web is covered, so as to comprehensively detect the surface condition of the non-woven fiber web.

[0059] With the assistance of a light source, the CCD camera continuously acquires images of the non-woven fiber web on the conveying device. The light source provides sufficient light for shooting to ensure the image quality. Adjust the sampling frequency according to the running speed, exposure time, etc. of the non-woven fiber web. When the running speed is fast, increase the sampling frequency to avoid missing detections; when the exposure time is short, it is also necessary to increase the sampling frequency accordingly to ensure that enough image information is collected.

[0060] Step S2: Preprocess the image to achieve light balance and / or image enhancement;

[0061] To ensure the accuracy of subsequent analysis, preprocess the collected original image. The preprocessing methods include a light balance method and an image enhancement method.

[0062] The described illumination balance method is Contrast Limited Adaptive Histogram Equalization (CLAHE). Specifically: First, convert the image to the LAB color space, then use the Gaussian filter function to extract the illumination component in the image. According to the distribution characteristics of the image illumination component, adjust the parameters of the CLAHE function to achieve adaptive correction of the unevenly illuminated image.

[0063] The described image enhancement method is the contrast stretching method. Specifically: By changing the gray range of the image, map the gray values of the image to a wider or more appropriate interval to enhance the contrast of the image. For example, make the darker parts of the image darker and the brighter parts brighter, making the details in the image more obvious.

[0064] Step S3: Perform wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image;

[0065] Performing wavelet transform on the preprocessed image, its main purpose is to enable the subsequent convolutional neural network to more effectively extract the information of the picture, thereby improving the accuracy of the prediction results of the convolutional neural network.

[0066] The specific operation of wavelet transform is as follows:

[0067] Step S31: Perform low-pass and high-pass filtering on each row of the preprocessed image to obtain the low-frequency component and high-frequency component of that row;

[0068] Step S32: Downsample the low-frequency component and high-frequency component of the row respectively to halve the data volume;

[0069] Step S33: Perform low-pass filtering and high-pass filtering on each column of the low-frequency component and high-frequency component of the row obtained in Step S31 to obtain the low-frequency component and high-frequency component of that column;

[0070] Step S32: Downsample the low-frequency component and high-frequency component of the column respectively to halve the data volume;

[0071] Finally, four sub-band images are obtained, namely LL, LH, HL, and HH sub-bands. Among them, the LL sub-band is the wavelet coefficients generated by convolving with a low-pass wavelet filter in two directions, and it is an approximate representation of the image; the HL sub-band represents the horizontal direction singularity characteristics of the image; the LH sub-band represents the vertical direction singularity characteristics of the image; the HH sub-band represents the diagonal edge characteristics of the image.

[0072] Finally, use the obtained LL sub-band as the low-frequency image of the image, and combine the remaining sub-bands into the high-frequency image.

[0073] In a specific implementation manner, Figure 2 is a schematic diagram of the preprocessed image, Figure 3Schematic diagram of the high-frequency image after wavelet transform, Figure 4 Schematic diagram of the low-frequency image after wavelet transform.

[0074] Step S4: Construct a convolutional neural network model. The input of the convolutional neural network is the low-frequency image and the high-frequency image after wavelet transform, and the output is the grammage of the nonwoven fiber web. The trained convolutional neural network model is used for online estimation of the grammage of the nonwoven fiber web.

[0075] In a specific embodiment, the convolutional neural network model includes two convolutional layers and two pooling layers. By training on a large number of low-frequency images and high-frequency images of nonwoven fiber webs, it can automatically extract deep features of the low-frequency images and high-frequency images of nonwoven fiber webs, such as the distribution of fibers, the size and shape of pores, etc. After multiple convolutional and pooling operations, the feature map is flattened and input into the fully connected layer. The last layer of the fully connected layer is a single-node output, and a linear activation function is used to directly return the predicted value of the grammage of the nonwoven fiber web. During the training process, the mean square error (MSE) is used as the loss function, and the network parameters are optimized by the backpropagation algorithm; the optimization method selects the Adam optimizer, and the initial learning rate is set to 0.001.

[0076] Specifically, the settings of the convolutional layer and the pooling layer are as follows:

[0077] The first convolutional layer: The convolutional kernel size is 3×3, the number of channels is 16, and the activation function is ReLU;

[0078] The first pooling layer: Max pooling operation is adopted, and the window size is 2×2;

[0079] The second convolutional layer: The convolutional kernel size is 3×3, the number of channels is 32, and the activation function is ReLU;

[0080] The second pooling layer: Max pooling operation is adopted, and the window size is 2×2.

[0081] Specifically, the fully connected layer contains two layers:

[0082] The number of nodes in the first fully connected layer is 64, and the activation function is ReLU;

[0083] The second fully connected layer is a single-node output, and the activation function is a linear function.

[0084] The convolutional neural network model is implemented using the PyTorch deep learning framework. After the input layer low-frequency image and high-frequency image are processed by the convolutional layer and the pooling layer to extract multi-level features, they are mapped to the single-node output layer through the fully connected layer to predict the grammage value of the nonwoven fiber web.

[0085] The training process is completed through a training file. The mean squared error (MSE) is used as the loss function, and the performance of the model on the validation set is recorded, resulting in a trained weight file. Model prediction calls the weight file through a prediction file, enabling the direct input of low-frequency images and high-frequency images and the output of predicted grammage values.

[0086] In a specific embodiment, Figure 5 It is a comparison chart of the prediction results of a conventional convolutional neural network, the prediction results of the optimized convolutional neural network in this embodiment, and the actual grammage of the nonwoven fiber web. Among them, the horizontal axis pic_name represents the picture number, the vertical axis weight represents the grammage, true weight represents the actual grammage of the nonwoven fiber web, cnn predictweight represents the prediction results of the conventional convolutional neural network, and optimize cnn predict weight represents the prediction results of the convolutional neural network after wavelet transform optimization in this embodiment. From Figure 5 It can be seen that the prediction results of the convolutional neural network after wavelet transform optimization in this embodiment are very close to the actual grammage, and its accuracy is greater than that of the conventional convolutional neural network.

[0087] Figure 6 It is a comparison chart of the training losses of a conventional convolutional neural network and the optimized convolutional neural network in this embodiment; among them, the horizontal axis epochs represents the number of training rounds, the vertical axis Train_Loss represents the training loss, cnn_Train Loss represents the training loss of the conventional convolutional neural network, and optimize_cnnTrain Loss represents the training loss of the convolutional neural network after wavelet transform optimization in this embodiment. From Figure 6 It can be seen that the training loss of the convolutional neural network after wavelet transform optimization in this embodiment is less than that of the conventional convolutional neural network.

[0088] Figure 7 It is a comparison chart of the mean absolute errors of a conventional convolutional neural network and the optimized convolutional neural network in this embodiment; among them, the horizontal axis epochs represents the number of training rounds, the vertical axis MAE represents the mean absolute error, cnn_MAE represents the mean absolute error of the conventional convolutional neural network, and optimize_cnn MAE represents the mean absolute error of the convolutional neural network after wavelet transform optimization in this embodiment. From Figure 7 It can be seen that the mean absolute error of the convolutional neural network after wavelet transform optimization in this embodiment is less than that of the conventional convolutional neural network.

[0089] And through calculation, it can be known that when the convolutional neural network optimized by wavelet transform in this embodiment is used to estimate the grammage, the average error between the actual grammage and the estimated grammage of 26 groups of samples is about 3%, and the correlation coefficient is 0.94. While the average error of the conventional convolutional neural network without using wavelet transform is about 7%, and the correlation coefficient is 0.87. This not only shows that there is a strong correlation between the grammage estimated by the two neural network models and the actual grammage, but also the prediction accuracy of the convolutional neural network optimized by wavelet transform in this embodiment has been greatly improved compared with the conventional convolutional neural network.

[0090] Embodiment 2

[0091] Based on the same inventive concept as the non-woven fiber web grammage online detection method in the foregoing Embodiment 1, this embodiment also provides a non-woven fiber web grammage online detection device, as Figure 8 shown, the device includes:

[0092] An image acquisition module 10, configured to acquire an image of the non-woven fiber web;

[0093] An image preprocessing module 20, configured to preprocess the image to achieve illumination balance and / or image enhancement;

[0094] A wavelet transform module 30, configured to perform wavelet transform on the preprocessed image to convert the image into a low-frequency image and a high-frequency image;

[0095] A neural network model construction module 40, configured to construct a convolutional neural network model, the input of the convolutional neural network being the low-frequency image and the high-frequency image after wavelet transform, and the output being the grammage of the non-woven fiber web; the trained convolutional neural network model is used for online estimation of the grammage of the non-woven fiber web.

[0096] Further, the image preprocessing module 20 includes:

[0097] An illumination balance module, configured to adopt contrast-limited adaptive histogram equalization (CLAHE). First, convert the image into the LAB color space, then use the Gaussian filtering function to extract the illumination component in the image, and adjust the parameters of the CLAHE function according to the distribution characteristics of the image illumination component to achieve adaptive correction of the image with uneven illumination.

[0098] An image enhancement module, configured to adopt the contrast stretching method. By changing the gray range of the image, map the gray value of the image to a wider or more appropriate interval to enhance the contrast of the image. For example, make the darker part of the image darker and the brighter part brighter, so that the details in the image are more obvious.

[0099] Further, the wavelet transform module 30 includes:

[0100] A row filtering module, which is used to perform low-pass and high-pass filtering on each row of the preprocessed image to obtain the low-frequency component and the high-frequency component of that row;

[0101] A row downsampling module, which is used to perform downsampling on the low-frequency component and the high-frequency component of the row respectively to halve the amount of data;

[0102] A column filtering module, which is used to perform low-pass filtering and high-pass filtering on each column of the low-frequency component and the high-frequency component of the row obtained by the row filtering module to obtain the low-frequency component and the high-frequency component of that column;

[0103] A column downsampling module, which is used to perform downsampling on the low-frequency component and the high-frequency component of the column respectively to halve the amount of data;

[0104] Finally, four sub-band images are obtained, namely LL, LH, HL, and HH sub-bands. Among them, the LL sub-band is the wavelet coefficients generated after convolving with a low-pass wavelet filter in two directions, and it is an approximate representation of the image; the HL sub-band represents the singular characteristics of the horizontal direction of the image; the LH sub-band represents the singular characteristics of the vertical direction of the image; the HH sub-band represents the diagonal edge characteristics of the image.

[0105] Finally, the obtained LL sub-band is used as the low-frequency image of the image, and the remaining sub-bands are combined into a high-frequency image.

[0106] The various specific processes and specific examples of the nonwoven fabric web grammage online detection method in the foregoing Embodiment 1 are equally applicable to the nonwoven fabric web grammage online detection device in this embodiment. Through the detailed description of Embodiment 1, those skilled in the art can clearly know the implementation method of the nonwoven fabric web grammage online detection device in this embodiment. Therefore, for the sake of brevity of the specification, it will not be elaborated here.

[0107] Embodiment 3

[0108] Based on the same inventive concept as the nonwoven fabric web grammage online detection method in the foregoing Embodiment 1, this embodiment also provides a nonwoven fabric web grammage online detection device, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the nonwoven fabric web grammage online detection method described in Embodiment 1.

[0109] Embodiment 4

[0110] Based on the same inventive concept as the nonwoven fabric web grammage online detection method in the foregoing Embodiment 1, this embodiment also provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the steps of the nonwoven fabric web grammage online detection method described in Embodiment 1.

[0111] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0112] The present application is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0113] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that realize the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0114] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0115] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to these embodiments once they learn the basic creative concepts. Therefore, the appended claims are intended to be construed as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

[0116] Obviously, those skilled in the art can make various changes and modifications to this application without departing from the spirit and scope of this application. Thus, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, this application is also intended to cover these changes and modifications.

Claims

1. A method for online detection of grammage of a nonwoven fiber web, characterized in that: The method comprises the following steps: acquiring an image of the nonwoven web; Preprocessing the image; Perform wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image; A neural network model is constructed, wherein the input of the neural network is a low-frequency image and a high-frequency image after wavelet transformation, and the output is the grammage of the nonwoven fiber web; the trained neural network model is used for online estimation of the grammage of the nonwoven fiber web.

2. The method for online detection of grammage of a nonwoven web according to claim 1, characterized in that: The acquiring of the image of the nonwoven web specifically includes acquiring the image of the nonwoven web on a conveying device between the carding machine output system and the cross-lapping system.

3. The method for online detection of grammage of a nonwoven web according to claim 1, characterized in that: The preprocessing method includes an illumination balancing method and / or an image enhancement method.

4. The method for online detection of grammage of a nonwoven web according to claim 1, characterized in that: The performing of wavelet transform on the preprocessed image to convert the image into a low-frequency image and a high-frequency image specifically includes: Step S31: performing low-pass and high-pass filtering on each row of the preprocessed image to obtain a low-frequency component and a high-frequency component of the row; Step S32: down-sampling the low-frequency component and the high-frequency component of the row respectively to reduce the data volume by half; Step S33: performing low-pass filtering and high-pass filtering on each column of the low-frequency components and high-frequency components of the row obtained in step S31 to obtain the low-frequency components and high-frequency components of the column; Step S34: down-sampling the low-frequency component and the high-frequency component of the column respectively to reduce the data volume by half; Finally, four sub-band images are obtained, namely LL, LH, HL and HH sub-bands; among them, the LL sub-band is the wavelet coefficient generated by convolution of low-pass wavelet filters in two directions; the HL sub-band represents the horizontal singular characteristics of the image; the LH sub-band represents the vertical singular characteristics of the image; the HH sub-band represents the diagonal edge characteristics of the image; The LL sub-band is used as a low-frequency image, and the remaining sub-bands are combined into a high-frequency image.

5. The method for online detection of grammage of a nonwoven web according to claim 1, characterized in that: The neural network is a convolutional neural network.

6. The method for online detection of grammage of a nonwoven web according to claim 5, characterized in that: The convolutional neural network includes two convolutional layers and two pooling layers, which are used to extract deep features of the nonwoven fiber web image; The low-frequency and high-frequency images in the input layer are processed by convolutional layers and pooling layers to extract multi-level features, and then mapped to a single-node output layer through a fully connected layer to estimate the grammage of the nonwoven fiber web.

7. An online detection device for nonwoven fiber web weight, characterized in that: The device comprises: An image acquisition module, used for acquiring an image of the nonwoven fiber web; An image preprocessing module, used for preprocessing the image; A wavelet transform module is used to perform wavelet transform on the preprocessed image to transform the image into a low-frequency image and a high-frequency image; The neural network model building module is used to build a neural network model, the input of the neural network is a low-frequency image and a high-frequency image after wavelet transformation, and the output is the grammage of the non-woven fiber web; the trained convolutional neural network model is used for online estimation of the grammage of the non-woven fiber web.

8. The nonwoven web weight online detection device according to claim 7, characterized in that: The wavelet transform module comprises: A row filtering module is used to perform low-pass and high-pass filtering on each row of the preprocessed image to obtain a low-frequency component and a high-frequency component of the row; A row downsampling module, used to downsample the low-frequency component and the high-frequency component of the row respectively, so as to reduce the data volume by half; A column filtering module, used for performing low-pass filtering and high-pass filtering on each column of the low-frequency components and high-frequency components of the row obtained by the row filtering module to obtain the low-frequency components and high-frequency components of the column; A column downsampling module, used to downsample the low-frequency component and the high-frequency component of the column respectively, so as to reduce the data volume by half; Finally, four sub-band images are obtained, namely LL, LH, HL and HH sub-bands; among them, the LL sub-band is the wavelet coefficient generated by convolution of low-pass wavelet filters in two directions; the HL sub-band represents the horizontal singular characteristics of the image; the LH sub-band represents the vertical singular characteristics of the image; the HH sub-band represents the diagonal edge characteristics of the image; The LL sub-band is used as a low-frequency image, and the remaining sub-bands are combined into a high-frequency image.

9. An online detection device for the grammage of a nonwoven web, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the method for online detection of the grammage of a nonwoven fiber web as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method for online detection of grammage of a nonwoven fiber web as described in any one of claims 1 to 7 are implemented.

Images