Polyimide fabric dyeing uniformity online detection method, system and device

Abstract

The application discloses a polyimide fabric dyeing uniformity online detection method, system and equipment, relates to the related technical field of intelligent detection, and the method comprises the steps of constructing a multispectral imaging module, obtaining a visible light fabric image and a near-infrared fabric image of a target polyimide fabric; building a dyeing uniformity detection double channel; performing uniformity evaluation through the visible light dyeing uniformity detection channel; if the preset fabric dyeing uniformity standard is reached, activating the near-infrared dyeing uniformity detection channel to perform uniformity detection and obtain the internal fabric dyeing uniformity; and weighted fusion evaluation is used to determine the fabric dyeing uniformity detection result. The technical problems that the surface visible light analysis cannot evaluate the penetration of the dye in the fiber, resulting in poor detection accuracy and product quality of the polyimide fabric are solved, the technical effects of realizing the synchronous detection of the surface and internal dyeing uniformity of the polyimide fabric and improving the detection accuracy and product quality are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent detection, in particular to an online detection method, system and device for dyeing uniformity of a polyimide fabric. BACKGROUND

[0002] Polyimide is a high-performance polymer material widely used in aerospace, electronic packaging, protective clothing and other fields. In the textile industry, polyimide fibers are used to manufacture special protective fabrics due to their excellent heat resistance and flame retardant properties. However, the dense molecular structure and inert surface of polyimide fibers result in poor dyeing performance, making it prone to uneven dyeing, which affects the quality and appearance of the final product. Traditional dyeing uniformity detection can only evaluate the color distribution on the surface of the fabric, and cannot detect the penetration of the dye inside the fiber, making it difficult to comprehensively evaluate the dyeing uniformity of the polyimide fabric. Since the dyeing process of the polyimide fabric involves the diffusion and fixation of the dye inside the fiber, surface detection alone cannot fully reflect the dyeing quality.

[0003] Therefore, in the related art, the surface visible light analysis cannot evaluate the penetration of the dye inside the fiber, resulting in poor detection accuracy and product quality of the polyimide fabric. SUMMARY

[0004] The present application provides an online detection method, system and device for dyeing uniformity of a polyimide fabric, which solves the technical problem of poor detection accuracy and product quality of the polyimide fabric caused by the inability of surface visible light analysis to evaluate the penetration of the dye inside the fiber in the prior art, and achieves the technical effect of simultaneous detection of surface and internal dyeing uniformity of the polyimide fabric, improving detection accuracy and product quality.

[0005] The present application provides an online detection method for dyeing uniformity of a polyimide fabric, which comprises: constructing a multi-spectral imaging module, the multi-spectral imaging module integrates a visible light perception component and a near-infrared perception component, and synchronously acquires and pre-processes a visible light fabric image and a near-infrared fabric image of the target polyimide fabric through the multi-spectral imaging module; building a dyeing uniformity detection double channel, the dyeing uniformity detection double channel includes a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel; evaluating the uniformity of the visible light fabric image through the visible light dyeing uniformity detection channel to obtain the surface dyeing uniformity of the fabric; if the surface dyeing uniformity of the fabric meets the preset fabric dyeing uniformity standard, activating the near-infrared dyeing uniformity detection channel to detect the uniformity of the near-infrared fabric image, and obtaining the internal dyeing uniformity of the fabric; and weighting and fusing the surface dyeing uniformity of the fabric and the internal dyeing uniformity of the fabric to evaluate the dyeing uniformity of the fabric and determine the detection result.

[0006] In a possible implementation, the online detection method for dyeing uniformity of the polyimide fabric further performs the following processing: synchronously collecting an initial visible light image and an initial near-infrared image of the target polyimide fabric by the multispectral imaging module; initializing a visible light filter and a near-infrared filter according to noise characteristics of the initial visible light image and the initial near-infrared image; filtering and denoising the initial visible light image and the initial near-infrared image by using the visible light filter and the near-infrared filter to obtain a usable visible light image and a usable near-infrared image; performing camera parameter calibration and image coordinate registration on the usable visible light image and the usable near-infrared image to obtain the visible light fabric image and the near-infrared fabric image.

[0007] In a possible implementation, the online detection method for dyeing uniformity of the polyimide fabric further performs the following processing: collecting a fabric standard visible light image and a fabric defect visible light image set with dyeing uniformity identification, and a fabric standard near-infrared image and a fabric defect near-infrared image set with dyeing uniformity identification; performing dyeing uniformity evaluation training on the fabric defect visible light image set with dyeing uniformity identification according to the fabric standard visible light image to construct a visible light dyeing uniformity detection channel; performing dyeing uniformity evaluation training on the fabric defect near-infrared image set with dyeing uniformity identification based on the fabric standard near-infrared image to obtain a near-infrared dyeing uniformity detection channel; and coupling the visible light dyeing uniformity detection channel and the near-infrared dyeing uniformity detection channel in parallel to build the dyeing uniformity detection double channel.

[0008] In a possible implementation, the online detection method for dyeing uniformity of the polyimide fabric further performs the following processing: constructing a visible light detection channel architecture, the visible light detection channel architecture including a feature extraction layer, a defect classification layer, and a dyeing uniformity evaluation layer, wherein the feature extraction layer adopts a ResNet-50 network; performing feature extraction on the fabric standard visible light image and the fabric defect visible light image set with dyeing uniformity identification based on the feature extraction layer to obtain a standard fabric dyeing feature set and a defect fabric dyeing feature set; performing loss comparison and defect identification on the defect fabric dyeing feature set according to the standard fabric dyeing feature set by the defect classification layer to generate a defect fabric loss distribution feature set; performing color difference weighted calculation and dyeing uniformity evaluation on the defect fabric loss distribution feature set by the dyeing uniformity evaluation layer to obtain a defect fabric surface dyeing sample set; and performing feedback training and optimization on the visible light detection channel architecture based on the defect fabric surface dyeing sample set to construct the visible light dyeing uniformity detection channel.

[0009] In a possible implementation, the online detection method for the dyeing uniformity of the polyimide fabric further performs the following processing: a near-infrared detection channel architecture is constructed, the near-infrared detection channel architecture including a three-dimensional reconstruction layer, a penetration defect identification layer, and a dyeing uniformity analysis layer; feature extraction is performed on the fabric standard near-infrared image and the fabric defect near-infrared image set with the dyeing uniformity identifier based on the three-dimensional reconstruction layer, to obtain a standard fabric dyeing reconstruction model and a defect fabric dyeing reconstruction model set; the penetration defect identification layer performs dye concentration distribution volume loss calculation on the defect fabric dyeing reconstruction model set according to the standard fabric dyeing reconstruction model, to generate a defect fabric loss volume feature set; the dyeing uniformity analysis layer performs internal dyeing uniformity evaluation on the defect fabric loss volume feature set, to obtain a defect fabric internal dyeing sample set; the near-infrared detection channel architecture is fed back and trained and optimized based on the defect fabric internal dyeing sample set, to obtain the near-infrared dyeing uniformity detection channel.

[0010] In a possible implementation, the online detection method for the dyeing uniformity of the polyimide fabric further performs the following processing: the penetration defect color block identification is performed on the defect fabric loss volume feature set by the dyeing uniformity analysis layer, to obtain a fabric penetration defect color block set; distribution color difference calculation is sequentially performed on the fabric penetration defect color block set, to obtain a fabric defect color block distribution color difference set; internal dyeing uniformity evaluation is performed based on the fabric defect color block distribution color difference set, to obtain the defect fabric internal dyeing sample set.

[0011] In a possible implementation, the online detection method for the dyeing uniformity of the polyimide fabric further performs the following processing: according to the polyimide fabric characteristics, the dyeing process, and the dyeing uniformity detection target, a surface dyeing weight factor and an internal dyeing weight factor are determined; the fabric surface dyeing uniformity and the fabric internal dyeing uniformity are weighted and fused for evaluation based on the surface dyeing weight factor and the internal dyeing weight factor, to determine the fabric dyeing uniformity detection result.

[0012] In a possible implementation, the online detection method for the dyeing uniformity of the polyimide fabric further performs the following processing: a laser displacement sensor is used to monitor the fabric tension fluctuation in real time; a fabric tension-color difference compensation model is established through testing, the fabric tension fluctuation is compensated and analyzed based on the fabric tension-color difference compensation model, a fabric color difference compensation parameter is determined, and the fabric dyeing uniformity detection result is dynamically corrected through the fabric color difference compensation parameter.

[0013] The application also provides an online detection system for dyeing uniformity of a polyimide fabric, comprising: a fabric image acquisition unit, configured to construct a multispectral imaging module, the multispectral imaging module integrating a visible light sensing component and a near-infrared sensing component, and configured to synchronously acquire and pre-process a visible light fabric image and a near-infrared fabric image of a target polyimide fabric; a detection channel building unit, configured to build a dyeing uniformity detection double channel, the dyeing uniformity detection double channel comprising a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel; a uniformity evaluation unit, configured to evaluate the uniformity of the visible light fabric image through the visible light dyeing uniformity detection channel to obtain a fabric surface dyeing uniformity; a uniformity detection unit, configured to activate the near-infrared dyeing uniformity detection channel to detect the uniformity of the near-infrared fabric image to obtain a fabric internal dyeing uniformity if the fabric surface dyeing uniformity reaches a preset fabric dyeing uniformity standard; and a detection result determination unit, configured to perform weighted fusion evaluation on the fabric surface dyeing uniformity and the fabric internal dyeing uniformity to determine a fabric dyeing uniformity detection result.

[0014] The application also provides an electronic device, comprising: a memory configured to store executable instructions; and a processor configured to execute the executable instructions stored in the memory to implement the online detection method for dyeing uniformity of a polyimide fabric.

[0015] The online detection method, system and device for dyeing uniformity of a polyimide fabric provided by the application construct a multispectral imaging module to obtain a visible light fabric image and a near-infrared fabric image of a target polyimide fabric, build a dyeing uniformity detection double channel, perform uniformity evaluation through a visible light dyeing uniformity detection channel, activate a near-infrared dyeing uniformity detection channel to perform uniformity detection to obtain a fabric internal dyeing uniformity if a preset fabric dyeing uniformity standard is reached, and perform weighted fusion evaluation to determine a fabric dyeing uniformity detection result. The technical problems of the prior art that the surface visible light analysis cannot evaluate the penetration of dyes in the fiber, resulting in poor detection accuracy and product quality of the polyimide fabric are solved, and the technical effects of realizing synchronous detection of the surface and internal dyeing uniformity of the polyimide fabric and improving the detection accuracy and product quality are achieved. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings of the embodiments of the present disclosure will be briefly introduced below. The flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the foregoing or the following operations are not necessarily performed in sequence. On the contrary, various steps can be processed in reverse order or simultaneously as needed. Meanwhile, other operations can be added to these processes, or a step or several steps can be removed from these processes.

[0017] Figure 1 The flow chart of the online detection method of the dyeing uniformity of the polyimide fabric provided by the embodiment of the application is shown.

[0018] Figure 2 The structure diagram of the online detection system of the dyeing uniformity of the polyimide fabric provided by the embodiment of the application is shown.

[0019] Figure 3 The structure diagram of the electronic device provided by the embodiment of the application is shown.

[0020] The label explanation: fabric image acquisition unit 10, detection channel building unit 20, uniformity evaluation unit 30, uniformity detection unit 40, detection result determination unit 50, input device 401, processor 402, memory 403, output device 404. DETAILED DESCRIPTION

[0021] The above description is only a summary of the technical scheme of the application, in order to more clearly understand the technical means of the application, which can be implemented according to the content of the specification, and in order to make the above and other purposes, characteristics and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described.

[0022] In order to make the purpose, technical scheme and advantage of the application more clear, the following will combine the drawings to make further detailed description of the application, the described embodiment should not be regarded as the limitation of the application, all other embodiments obtained by the person skilled in the art without creative labor belong to the protection scope of the application.

[0023] In the following description, "some embodiments" are related to a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subset of all possible embodiments, and can be combined with each other without conflict, the term "first\second" is only to distinguish similar objects, and does not represent the specific order of the object. The terms "include" and "have" and any variants, are intended to cover non-exclusive inclusion, for example, the process, method, system, product or server including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or modules not clearly listed or inherent to these processes, methods, products or devices, unless otherwise defined, all technical and scientific terms used in this paper are the same as the meanings understood by the person skilled in the art belonging to the technical field of the application. The terms used in this paper are only for the purpose of describing the embodiments of the application.

[0024] The embodiment of the application provides an online detection method of dyeing uniformity of polyimide fabric,Figure 1 The method comprises the following steps:

[0025] In step S100, a multispectral imaging module is constructed, which integrates a visible light sensing component and a near-infrared sensing component, and through the multispectral imaging module, a visible light fabric image and a near-infrared fabric image of the target polyimide fabric are synchronously acquired and preprocessed.

[0026] Preferably, a high-resolution industrial camera and a standard D65 white light LED array are selected to form the visible light sensing component, which captures the optical image of the fabric in the visible light band (400-700 nm) for analyzing the surface color distribution and appearance uniformity to detect color spots, color differences, stripes, etc. The image acquisition is realized through spectral filtering and image processing, and then the coverage uniformity of the dye on the fabric surface is evaluated, such as whether there is uneven dyeing, dyeing omission, etc. A short-wave infrared camera is selected to form the near-infrared sensing component in combination with a near-infrared light source, such as a halogen lamp or an LED, which captures the spectral image of the fabric in the near-infrared band (700-2500 nm) for analyzing the penetration depth and distribution uniformity of the dye in the fiber by using the penetration of near-infrared. The image acquisition is realized based on spectral analysis and three-dimensional reconstruction. The visible light sensing component, the near-infrared sensing component and the supporting optical components are integrated to construct the multispectral imaging module, which realizes the synchronous detection and capture of the surface color of the polyimide fabric and the internal dye penetration.

[0027] Preferably, the image acquisition is performed by the multispectral imaging module, that is, a synchronization signal is sent by the FPGA or PLC to trigger the visible light camera and the near-infrared camera to expose at the same time, and the optical images of the polyimide fabric in the visible light band (400-700 nm) and the near-infrared band (700-2500 nm) are acquired at the same time. The synchronization signal is sent by the FPGA or PLC to trigger the visible light camera and the near-infrared camera to expose at the same time. The optical common path includes a dichroic prism and two cameras placed side by side, that is, the two cameras are installed side by side. The parallax is eliminated by calibration. The incident light is divided into two paths by the dichroic mirror and enters the visible light sensing component and the near-infrared sensing component, respectively, and then the initial visible light fabric image and the initial near-infrared fabric image are acquired. Then, the initial visible light fabric image and the initial near-infrared fabric image are preprocessed, including image denoising, radiation correction, multispectral registration and image enhancement. Specifically, the initial visible light fabric image is denoised by non-local mean to retain texture details, and the initial near-infrared fabric image is denoised by wavelet transform to suppress noise. A standard white board is then shot, and the correction coefficient of each pixel is calculated to eliminate the non-uniformity of the light source. Then, multispectral registration is performed, that is, the SIFT feature points of the initial visible light fabric image and the initial near-infrared fabric image are extracted, the affine transformation matrix is calculated by the RANSAC algorithm, and the double-camera external parameter is calibrated simultaneously using a chessboard calibration board with a near-infrared reflective coating. Finally, image enhancement is performed, that is, histogram equalization is performed to enhance the color difference in low-contrast areas, and band ratio operation is performed based on the dye absorption peak to highlight the penetration difference, and finally the visible light fabric image and the near-infrared fabric image of the target polyimide fabric are obtained.

[0028] Further, step S100 further includes step S110 of synchronously acquiring an initial visible light image and an initial near-infrared image of the target polyimide fabric by the multispectral imaging module; step S120 of initializing a visible light filter and a near-infrared filter according to the noise characteristics of the initial visible light image and the initial near-infrared image; step S130 of filtering and denoising the initial visible light image and the initial near-infrared image by the visible light filter and the near-infrared filter to obtain a usable visible light image and a usable near-infrared image; and step S140 of calibrating camera parameters and registering image coordinates of the usable visible light image and the usable near-infrared image to obtain the visible light fabric image and the near-infrared fabric image.

[0029] Preferably, the pulse signal is sent by the FPGA or high-precision timer to trigger the visible light camera and the near-infrared camera to expose at the same time, and the time error is controlled within 1 millisecond, that is, when the fabric passes through the production line photoelectric sensor, the dual-camera shooting is started synchronously, the visible light sensing component collects and captures the 400-700nm waveband light reflected by the fabric to generate a color image, which is the initial visible light image of the target polyimide fabric, and may contain noise such as uneven lighting and sensor noise; the near-infrared sensing component collects and captures the absorption / scattering characteristics of the fabric to the 700-2500nm waveband light, which is recorded by the InGaAs sensor to generate a grayscale image, which is the initial near-infrared image of the target polyimide fabric, and may be disturbed by thermal noise and dark current. Analyzing the initial visible light image and the initial near-infrared image to determine the noise characteristics, the main noise of the initial visible light image is Gaussian noise (sensor noise), salt and pepper noise (transmission interference), and high-frequency random noise, which may cover the subtle color difference; the main noise of the initial near-infrared image is thermal noise and stripe noise (circuit interference), and low-frequency band noise, which may distort the dye penetration distribution. Then, a non-local mean filter is selected to initialize the visible light filter, and a wavelet threshold denoising is selected to initialize the near-infrared filter, so as to effectively suppress the noise.

[0030] Preferably, the initial visible light image and the initial near-infrared image are filtered and denoised by the visible light filter and the near-infrared filter. Specifically, for the initial visible light image, for each pixel, search for similar local blocks (such as a 7x7 window) in the image, and generate a denoised pixel value by weighted averaging, and then perform contrast enhancement, that is, limit the contrast adaptive histogram equalization, and improve the color difference recognition. For the initial near-infrared image, wavelet denoising is performed, including using a Daubechies wavelet basis function to decompose the image, performing soft thresholding on the high-frequency coefficients, and removing the noise after reconstruction to obtain a usable visible light image and a usable near-infrared image. Then, camera parameter calibration is performed on the usable visible light image and the usable near-infrared image, that is, a chessboard calibration plate with a near-infrared reflective coating is used to calculate the intrinsic parameters (focal length, distortion) and extrinsic parameters (relative position) of the visible light and near-infrared cameras, respectively, and then image registration is performed, that is, the SIFT / SURF feature points of the usable visible light image and the usable near-infrared image are extracted, and the affine transformation matrix is calculated by the RANSAC algorithm, for example, the near-infrared image is transformed to the coordinate system of the visible light image, and finally the visible light fabric image and the near-infrared fabric image are determined.

[0031] Step S200, a dyeing uniformity detection double channel is built, the dyeing uniformity detection double channel includes a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel.

[0032] The step S200 further comprises the steps of: S210, collecting a fabric standard visible light image and a fabric defect visible light image set with dye uniformity identification, and a fabric standard near-infrared image and a fabric defect near-infrared image set with dye uniformity identification; S220, performing dye uniformity evaluation training on the fabric defect visible light image set with dye uniformity identification according to the fabric standard visible light image, to construct a visible light dye uniformity detection channel; S230, performing dye uniformity evaluation training on the fabric defect near-infrared image set with dye uniformity identification based on the fabric standard near-infrared image, to obtain a near-infrared dye uniformity detection channel; and S240, coupling the visible light dye uniformity detection channel and the near-infrared dye uniformity detection channel in parallel, to build the dye uniformity detection double channel.

[0033] Preferably, the standard visible light images covering different colors and textures are photographed, and the polyimide fabric with dye uniformity identification is collected to determine a fabric defect visible light image set, wherein the defect types include color difference, stripe, insufficient penetration, dye aggregation, etc. For example, the color difference area of the visible light image is marked, and the penetration uneven area of the near-infrared image is verified by a microscopic infrared spectrometer and marked as an internal defect. A deep learning model is constructed, and dye uniformity evaluation training is performed on the fabric standard visible light image and the fabric defect visible light image set with dye uniformity identification, including using a ResNet-50 network (pre-trained weight) to extract multi-scale features, outputting a multi-dimensional feature vector, mapping the defect features to preset categories such as color difference and stain through a cross-entropy loss function, calculating the color difference between the defect area and the standard image, outputting the uniformity score and the corresponding defect position distribution, and then obtaining a visible light detection channel for evaluating the surface color difference of the target polyimide fabric.

[0034] Preferably, the fabric defect near-infrared image set marked with dyeing uniformity is similarly modeled trained based on the fabric standard near-infrared image to establish an internal dye penetration uniformity analysis model. Specifically, a U-Net network is used to convert a 2D near-infrared image into a 3D dye concentration volume (voxel resolution 0.1 mm³), and then the fabric standard near-infrared image and the fabric defect near-infrared image set are input for dyeing uniformity evaluation training, and a 128x128x64 voxel grid is output, wherein the Z-axis is the penetration depth. Then, penetration defect recognition is performed, including comparing the concentration distribution of the defect voxel and the standard body, calculating the KL divergence as the penetration deviation index, and marking the defect area according to the deviation threshold, outputting the internal uniformity score. Finally, a 3D convolution kernel is used to capture the spatial penetration features, and finally the near-infrared dyeing uniformity detection channel is determined, which is used to evaluate the internal dye penetration uniformity of the target polyimide fabric. Then, the visible light dyeing uniformity detection channel and the near-infrared dyeing uniformity detection channel are coupled in parallel, that is, the GPU server deploys a double model, and the visible light dyeing uniformity detection channel and the near-infrared dyeing uniformity detection channel run independently to determine the dyeing uniformity detection double channel, which covers the defects in the whole dyeing process through surface and internal collaborative detection.

[0035] Further, step S220 further includes step S221 of constructing a visible light detection channel architecture including a feature extraction layer, a defect classification layer, and a dyeing uniformity evaluation layer, wherein the feature extraction layer adopts a ResNet-50 network; step S222 of performing feature extraction on the fabric standard visible light image and the fabric defect visible light image set marked with dyeing uniformity based on the feature extraction layer to obtain a standard fabric dyeing feature set and a defect fabric dyeing feature set; step S223 of performing loss comparison and defect identification on the defect fabric dyeing feature set according to the standard fabric dyeing feature set by the defect classification layer to generate a defect fabric loss distribution feature set; step S224 of performing color difference weighting calculation and dyeing uniformity evaluation on the defect fabric loss distribution feature set by the dyeing uniformity evaluation layer to obtain a defect fabric surface dyeing sample set; and step S225 of performing feedback training and optimization on the visible light detection channel architecture based on the defect fabric surface dyeing sample set to construct the visible light dyeing uniformity detection channel.

[0036] Preferably, a three-level linkage visible light detection channel architecture is constructed based on a deep learning model, including a feature extraction layer, a defect classification layer, and a dyeing uniformity evaluation layer. The feature extraction layer adopts a ResNet-50 network, uses ImageNet pre-training weights to initialize the ResNet-50, and is used to convert fabric images into high-dimensional feature vectors to capture key information such as color and texture. Then, the feature extraction layer is used to extract features from the standard fabric visible light image and the fabric defect visible light image set with dyeing uniformity identification, including using the last layer of the initialized ResNet-50 network to extract a 1024-dimensional feature vector, which constitutes a standard fabric dyeing feature set. Similarly, a 1024-dimensional feature of the defect image is extracted to constitute a defect fabric dyeing feature set. The defect classification layer is used to compare the differences between the defect features and the standard features, locate and classify the defect areas, and the loss comparison refers to calculating the cosine similarity of the defect features and the standard feature set. The area with a difference greater than a threshold (such as <0.8) is marked as a potential defect. Then, the difference features are mapped to the preset defect categories, such as color difference and uneven penetration, through a fully connected layer and a Softmax, to generate a defect fabric loss distribution feature set containing defect type, location and confidence.

[0037] Preferably, the dyeing uniformity evaluation layer is used to convert the defect information into a quantifiable uniformity score, i.e., the dyeing uniformity evaluation layer is used to calculate the color difference of the defect area and the standard area and weighted fusion, and the weight is dynamically adjusted according to the defect area ratio and severity. Then, the dyeing uniformity is evaluated, i.e., the overall color difference is mapped to the 0~100 interval, and the defect fabric surface dyeing sample set is output, which contains the score and defect position distribution of each image. Finally, the visible light detection channel architecture is trained and optimized based on the defect fabric surface dyeing sample set, i.e., the model parameters are iteratively optimized to improve the detection accuracy. Specifically, the model output is calculated using the labeled data in the sample set, such as defect classification results and uniformity scores. Then, the true difference is determined by the loss function and compared. Based on the comparison result, the weights of the ResNet-50 feature extraction layer and the subsequent classification and evaluation layers are adjusted through the back propagation algorithm, focusing on optimizing the recognition sensitivity of the defect area and the color difference quantification accuracy. Finally, the visible light dyeing uniformity detection channel is constructed.

[0038] Further, step S230 further comprises step S231 of constructing a near-infrared detection channel architecture, the near-infrared detection channel architecture comprising a three-dimensional reconstruction layer, a penetration defect identification layer, and a dye uniformity analysis layer; step S232 of performing feature extraction on the fabric standard near-infrared image and the set of fabric defect near-infrared images with dye uniformity identification based on the three-dimensional reconstruction layer, to obtain a standard fabric dyeing reconstruction model and a set of defect fabric dyeing reconstruction models; step S233 of performing dye concentration distribution volume loss calculation on the set of defect fabric dyeing reconstruction models according to the standard fabric dyeing reconstruction model by the penetration defect identification layer, to generate a set of defect fabric loss volume features; step S234 of performing internal dye uniformity evaluation on the set of defect fabric loss volume features by the dye uniformity analysis layer, to obtain a set of defect fabric internal dyeing samples; and step S235 of performing feedback training and optimization on the near-infrared detection channel architecture based on the set of defect fabric internal dyeing samples, to obtain the near-infrared dye uniformity detection channel.

[0039] Preferably, the three-level linkage near-infrared detection channel architecture based on the deep learning model comprises a three-dimensional reconstruction layer, a penetration defect identification layer, and a dye uniformity analysis layer, and the penetration distribution of the dye in the polyimide fiber is analyzed through the near-infrared image, so as to realize end-to-end detection from 3D reconstruction to uniformity scoring. Specifically, the three-dimensional reconstruction layer is used to convert a 2D near-infrared image into a 3D dye concentration distribution model, feature extraction is performed on the fabric standard near-infrared image and the set of fabric defect near-infrared images with dye uniformity identification through the three-dimensional reconstruction layer, U-Net++ network is used for extraction, multi-scale feature fusion is supported, and the network is suitable for small defect detection. The absorption rate of near-infrared light is positively correlated with the dye concentration, depth information is inverted through a diffuse reflection model, and model reconstruction is performed, so as to obtain a standard fabric dyeing reconstruction model and a set of defect fabric dyeing reconstruction models. The penetration defect identification layer is used to compare the volume difference between the defect model and the standard model, identify the penetration uneven area, extract spatial features using a 3D convolutional neural network, calculate the KL divergence between the defect voxel and the standard voxel to measure the distribution difference, and mark KL divergence > 0.3 as a defect, so as to output a set of defect fabric loss volume features, including defect type, position coordinates, and volume proportion.

[0040] Preferably, the dyeing uniformity evaluation layer is used to convert the volume defects into quantifiable internal uniformity scores. The internal dyeing uniformity of the defect fabric loss volume feature set is evaluated by the dyeing uniformity analysis layer. Specifically, according to the defect volume proportion and the KL divergence value, the local unevenness index is calculated, and the weights are set to 0.6 and 0.4 respectively. The overall defect area is integrated and mapped to 0-100 points. Then the internal dyeing sample set of the defect fabric is obtained. Finally, the near-infrared detection channel architecture is trained and optimized based on the internal dyeing sample set of the defect fabric, that is, the model parameters are iteratively optimized to improve the detection accuracy. Specifically, the labeled data in the sample set is used to calculate the model output, such as defect classification results and uniformity scores. Then the real difference is determined by the loss function and compared. Based on the comparison result, the weights of each layer are adjusted by the back propagation algorithm, and the accuracy of 3D reconstruction and defect recognition in the defect area is optimized. Finally, the near-infrared dyeing uniformity detection channel is constructed, and the defect detection rate is ensured.

[0041] Further, step S340 further includes step S341, identifying the penetration defect color block set of the defect fabric loss volume feature set by the dyeing uniformity analysis layer; step S342, calculating the distribution color difference of the fabric penetration defect color block set in sequence to obtain the fabric defect color block distribution color difference set; and step S343, evaluating the internal dyeing uniformity based on the fabric defect color block distribution color difference set to obtain the internal dyeing sample set of the defect fabric.

[0042] Preferably, the dyeing uniformity analysis layer is used to identify the penetration defect color block set of the defect fabric loss volume feature set, and extract discrete dye penetration abnormal areas (i.e. color blocks). Specifically, the 3D connected domain analysis algorithm is used to cluster continuous high difference voxels into independent color blocks, and then obtain the fabric penetration defect color block set. Each fabric penetration defect color block contains spatial position, volume size and depth deviation. The distribution color difference of the fabric penetration defect color block set is calculated in sequence to quantify the dye distribution difference between the color blocks, reflect the spatial pattern of internal unevenness, and avoid small color blocks interfering with overall evaluation. The standard deviation of the color block center point is calculated to evaluate the defect aggregation degree as the spatial distribution dispersion, and then the fabric defect color block distribution color difference set is obtained. Then the internal dyeing uniformity is evaluated based on the fabric defect color block distribution color difference set, and the color block distribution features are converted into operable uniformity scores. Specifically, a scoring model is constructed based on multivariate linear regression, the input parameters include global color difference, Z-axis dispersion and maximum color block volume proportion, and the output is a uniformity score of 0-100. Finally, the internal dyeing sample set of the defect fabric is obtained.

[0043] Step S300, evaluating the uniformity of the fabric surface dyeing by the visible light dyeing uniformity detection channel.

[0044] Preferably, the visible light fabric image is input into the visible light dyeing uniformity detection channel for uniformity evaluation to quantify the color distribution uniformity of the polyimide fabric surface, such as color difference, stripes, stains, and other appearance defects, and output a quantifiable uniformity score (such as 0-100 points) and defect positioning results. Specifically, the visible light fabric image is input into the feature extraction layer for feature extraction, i.e., extracting a 1024-dimensional feature vector through the last convolutional layer of ResNet-50, calculating the similarity with the pre-established standard fabric feature set to determine the feature difference distribution, i.e., marking the high difference area as a potential defect. Then, based on the feature difference distribution, the defect classification layer is used for defect classification and positioning, including identifying common defects based on the fully connected layer and Softmax, and generating a defect attention map based on Grad-CAM to locate the abnormal area. Then, through the dyeing uniformity evaluation layer, the color difference between the defect area and the standard area is calculated to obtain the uniformity score and adjust it according to the defect area ratio, and finally output the dyeing uniformity of the fabric surface.

[0045] Step S400, if the fabric surface dyeing uniformity reaches the preset fabric dyeing uniformity standard, activate the near-infrared dyeing uniformity detection channel to detect the near-infrared fabric image for uniformity, and obtain the internal dyeing uniformity of the fabric.

[0046] Preferably, when the surface dyeing uniformity score output by the visible light dyeing uniformity detection channel reaches the preset fabric dyeing uniformity standard, the near-infrared detection channel is automatically activated to avoid redundant internal detection of obviously surface unqualified fabric. Specifically, the fabric image and its spatial coordinates detected by the surface are transmitted to the near-infrared dyeing uniformity detection channel, the visible light source is turned off, the near-infrared LED array (such as 1050nm wavelength) is started, and the near-infrared camera captures the image in global shutter mode and checks the registration error between the near-infrared and visible light images (which should be <1 pixel). If the deviation is too large, trigger the mechanical arm to adjust the fabric position; then perform 3D dye penetration modeling, i.e., use the U-Net++ network to generate a 128x128x64 voxel grid with a Z-axis resolution of 0.1mm, reconstruct a 3D dye concentration volume model, and mark abnormal areas as red voxels; then calculate the KL divergence of the defect area by comparing with the standard model to determine the penetration depth deviation and the maximum defect volume, and then calculate the uniformity score to output the internal dyeing uniformity of the fabric. If the internal score is <60, trigger the mechanical arm to remove the fabric and record the defect type.

[0047] Step S500, weight and fuse the fabric surface dyeing uniformity and the fabric internal dyeing uniformity for evaluation to determine the fabric dyeing uniformity detection result.

[0048] The step S500 further comprises a step S510 of determining a surface dyeing weight factor and an internal dyeing weight factor according to the polyimide fabric characteristics, the dyeing process and the dyeing uniformity detection target; a step S520 of performing weighted fusion evaluation on the fabric surface dyeing uniformity and the fabric internal dyeing uniformity based on the surface dyeing weight factor and the internal dyeing weight factor to determine the fabric dyeing uniformity detection result.

[0049] Preferably, the surface dyeing weight factor and the internal dyeing weight factor are dynamically allocated according to the polyimide fabric characteristics, the dyeing process and the dyeing uniformity detection target. For the polyimide fabric characteristics, the surface weight of a fabric with high fiber density and difficult penetration is 0.3, and the internal weight is 0.7; the surface weight of a fabric with loose fibers and easy dyeing is 0.7, and the internal weight is 0.3. For the dyeing process, the surface weight of a high-pressure high-temperature penetration process is 0.4, and the internal weight is 0.6; the surface weight of a normal temperature surface coating process is 0.8, and the internal weight is 0.2. For the dyeing uniformity detection target, the surface weight of a fabric with appearance dominance (such as decorative cloth) is 0.9, and the internal weight is 0.1; the surface weight of a fabric with function dominance (such as flame-retardant cloth) is 0.5, and the internal weight is 0.5. Then, the fabric surface dyeing uniformity and the fabric internal dyeing uniformity are weighted and fused based on the surface dyeing weight factor and the internal dyeing weight factor to determine the fabric dyeing uniformity detection result. If the pass level is 70 points, the fabric with internal penetration unevenness below 70 points is determined as unqualified, and the fabric dyeing uniformity detection result is determined. If the excellent level is 85 points, it is directly released. If the good level is 70-84 points, the data is recorded and the process is adjusted. Thus, the synchronous detection of the polyimide fabric surface and internal dyeing uniformity is realized, and the defect detection rate and detection accuracy in high requirement scenarios are improved.

[0050] Further, the step S500 further comprises a step S530 of monitoring the fabric tension fluctuation in real time by using a laser displacement sensor; and a step S540 of testing and establishing a fabric tension-color difference compensation model, performing compensation analysis on the fabric tension fluctuation based on the fabric tension-color difference compensation model, determining fabric color difference compensation parameters, and dynamically correcting the fabric dyeing uniformity detection result by using the fabric color difference compensation parameters.

[0051] Preferably, the polyimide fabric is stretched or wrinkled during dyeing and testing, the tension fluctuation of the conveying belt changes the local fiber density, affects the dye color development, and deviates the detection value of the color difference meter or multispectral imaging from the true dyeing effect. By real-time monitoring of the tension change through the laser displacement sensor, a tension-color difference mathematical model is established, and the detection result is dynamically corrected to eliminate mechanical interference. Specifically, the laser displacement sensor is installed above the fabric conveying path, vertically irradiates and measures the fabric surface height change, records the fabric position fluctuation in real time and calculates the fabric tension fluctuation, then adjusts the tension in a controllable environment, synchronously collects the tension data, true color difference value and original color difference value by using the laser displacement sensor, offline spectrophotometer and multispectral imaging, then establishes a relationship based on multiple linear regression and fits by least squares method to obtain a fabric tension-color difference compensation model, compensates the fabric tension fluctuation based on the fabric tension-color difference compensation model, determines the fabric color difference compensation parameter, and then dynamically corrects the fabric dyeing uniformity test result by the fabric color difference compensation parameter, that is, the compensation parameter is superimposed on the original fabric dyeing uniformity test result, and finally the corrected fabric dyeing uniformity test result is output, avoiding misjudgment caused by mechanical tension interference, eliminating detection errors introduced by line speed changes or mechanical vibration, and improving the reliability of the detection result.

[0052] In the foregoing, with reference to Figure 1 The polyimide fabric dyeing uniformity online detection method according to the embodiment of the present application is described in detail. Next, with reference to Figure 2 The polyimide fabric dyeing uniformity online detection system according to the embodiment of the present application will be described.

[0053] The polyimide fabric dyeing uniformity online detection system according to the embodiment of the present application is used to solve the technical problem that the surface visible light analysis cannot evaluate the penetration of the dye inside the fiber in the prior art, resulting in poor detection accuracy and product quality of the polyimide fabric, and achieves the technical effects of realizing synchronous detection of the surface and internal dyeing uniformity of the polyimide fabric and improving the detection accuracy and product quality. As Figure 2 shown, the polyimide fabric dyeing uniformity online detection system includes a fabric image acquisition unit 10, a detection channel building unit 20, a uniformity evaluation unit 30, a uniformity detection unit 40, and a detection result determination unit 50.

[0054] The fabric image acquisition unit 10 is used to construct a multispectral imaging module integrating a visible light sensing component and a near-infrared sensing component, and the visible light fabric image and the near-infrared fabric image of the target polyimide fabric are obtained by synchronously acquiring and preprocessing the multispectral imaging module; the detection channel building unit 20 is used to build a dyeing uniformity detection double channel, and the dyeing uniformity detection double channel includes a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel; the uniformity evaluation unit 30 is used to evaluate the uniformity of the visible light fabric image through the visible light dyeing uniformity detection channel to obtain the surface dyeing uniformity of the fabric; the uniformity detection unit 40 is used to activate the near-infrared dyeing uniformity detection channel to detect the uniformity of the near-infrared fabric image if the surface dyeing uniformity of the fabric reaches the preset fabric dyeing uniformity standard, and the internal dyeing uniformity of the fabric is obtained; and the detection result determination unit 50 is used to evaluate the surface dyeing uniformity and the internal dyeing uniformity of the fabric by weighted fusion, and determine the fabric dyeing uniformity detection result.

[0055] In the following, the specific configuration of the fabric image acquisition unit 10 will be described in detail. The fabric image acquisition unit 10 further includes: synchronously acquiring the initial visible light image and the initial near-infrared image of the target polyimide fabric through the multispectral imaging module; initializing the visible light filter and the near-infrared filter according to the noise characteristics of the initial visible light image and the initial near-infrared image; filtering and denoising the initial visible light image and the initial near-infrared image by using the visible light filter and the near-infrared filter to obtain the available visible light image and the available near-infrared image; and performing camera parameter calibration and image coordinate registration on the available visible light image and the available near-infrared image to obtain the visible light fabric image and the near-infrared fabric image.

[0056] In the following, the specific configuration of the detection channel building unit 20 will be described in detail. The detection channel building unit 20 further includes: acquiring a fabric standard visible light image and a fabric defect visible light image set with a dyeing uniformity identifier, and a fabric standard near-infrared image and a fabric defect near-infrared image set with a dyeing uniformity identifier; constructing a visible light dyeing uniformity detection channel by performing dyeing uniformity evaluation training on the fabric defect visible light image set with the dyeing uniformity identifier according to the fabric standard visible light image; obtaining a near-infrared dyeing uniformity detection channel by performing dyeing uniformity evaluation training on the fabric defect near-infrared image set with the dyeing uniformity identifier based on the fabric standard near-infrared image; and coupling the visible light dyeing uniformity detection channel and the near-infrared dyeing uniformity detection channel in parallel to build the dyeing uniformity detection double channel.

[0057] Next, the specific configuration of the detection channel building unit 20 will be described in detail. The detection channel building unit 20 further comprises: constructing a visible light detection channel architecture, which comprises a feature extraction layer, a defect classification layer and a dye uniformity evaluation layer, wherein the feature extraction layer adopts a ResNet-50 network; based on the feature extraction layer, feature extraction is performed on the fabric standard visible light image and the fabric defect visible light image set with dye uniformity identification, to obtain a standard fabric dyeing feature set and a defect fabric dyeing feature set; the defect classification layer performs loss comparison and defect identification on the defect fabric dyeing feature set according to the standard fabric dyeing feature set, to generate a defect fabric loss distribution feature set; the dye uniformity evaluation layer is used to perform color difference weighted calculation and dye uniformity evaluation on the defect fabric loss distribution feature set, to obtain a defect fabric surface dyeing sample set; based on the defect fabric surface dyeing sample set, feedback training optimization is performed on the visible light detection channel architecture, to construct the visible light dye uniformity detection channel.

[0058] Next, the specific configuration of the detection channel building unit 20 will be described in detail. The detection channel building unit 20 further comprises: constructing a near-infrared detection channel architecture, which comprises a three-dimensional reconstruction layer, a penetration defect identification layer and a dye uniformity analysis layer; based on the three-dimensional reconstruction layer, feature extraction is performed on the fabric standard near-infrared image and the fabric defect near-infrared image set with dye uniformity identification, to obtain a standard fabric dyeing reconstruction model and a defect fabric dyeing reconstruction model set; the penetration defect identification layer performs dye concentration distribution volume loss calculation on the defect fabric dyeing reconstruction model set according to the standard fabric dyeing reconstruction model, to generate a defect fabric loss volume feature set; the dye uniformity analysis layer is used to perform internal dye uniformity evaluation on the defect fabric loss volume feature set, to obtain a defect fabric internal dyeing sample set; based on the defect fabric internal dyeing sample set, feedback training optimization is performed on the near-infrared detection channel architecture, to obtain the near-infrared dye uniformity detection channel.

[0059] Next, the specific configuration of the detection channel building unit 20 will be described in detail. The detection channel building unit 20 further comprises: performing penetration defect color block identification on the defect fabric loss volume feature set through the dye uniformity analysis layer, to obtain a fabric penetration defect color block set; sequentially performing distribution color difference calculation on the fabric penetration defect color block set, to obtain a fabric defect color block distribution color difference set; based on the fabric defect color block distribution color difference set, internal dye uniformity evaluation is performed, to obtain the defect fabric internal dyeing sample set.

[0060] The specific configuration of the detection result determination unit 50 will be described in detail below. The detection result determination unit 50 further comprises: determining the surface dyeing weight factor and the internal dyeing weight factor according to the polyimide fabric characteristics, the dyeing process and the dyeing uniformity detection target; and determining the fabric dyeing uniformity detection result by weighted fusion evaluation of the fabric surface dyeing uniformity and the fabric internal dyeing uniformity based on the surface dyeing weight factor and the internal dyeing weight factor.

[0061] The specific configuration of the detection result determination unit 50 will be described in detail below. The detection result determination unit 50 further comprises: determining the surface dyeing weight factor and the internal dyeing weight factor according to the polyimide fabric characteristics, the dyeing process and the dyeing uniformity detection target; and determining the fabric dyeing uniformity detection result by weighted fusion evaluation of the fabric surface dyeing uniformity and the fabric internal dyeing uniformity based on the surface dyeing weight factor and the internal dyeing weight factor.

[0062] The polyimide fabric dyeing uniformity online detection system provided by the embodiment of the present application can perform the polyimide fabric dyeing uniformity online detection method provided by any embodiment of the present application, and has the corresponding function modules and beneficial effects of the execution method.

[0063] Figure 3 Figure 1 is a structural schematic diagram of an electronic device provided by an embodiment of the present application, which shows a block diagram of an exemplary electronic device suitable for implementing the embodiments of the present application. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functions and use range of the embodiments of the present application. The electronic device is in the form of a general computing device, and its components can include but are not limited to an input device 401, a processor 402, a memory 403 and an output device 404. The processor 402 can be one or more; the memory 403 can include a computer readable medium and at least one program product, which has a set of (at least one) program modules configured to perform the functions of the embodiments of the present application.

[0064] The memory 403 shown in the embodiments of the present application can adopt any combination of one or more computer readable media; the computer readable storage medium can be but is not limited to an infrared ray, a semiconductor system, a device or a component, or a combination of any of the above, for storing software programs, computer executable programs and modules, such as the program instructions / modules corresponding to the polyimide fabric dyeing uniformity online detection method in the embodiments of the present application. The processor 402 performs various functional applications and data processing of the computer device by running the software programs, instructions and modules stored in the memory 403, that is, implements the above-mentioned polyimide fabric dyeing uniformity online detection method.

[0065] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or the server, the various units and modules included are only divided according to the functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific name of each functional unit is only for the convenience of mutual differentiation, and is not used to limit the protection scope of the present application.

[0066] The above detailed description does not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for on-line detection of dyeing uniformity of polyimide fabric, characterized in that, The method comprises: constructing a multispectral imaging module integrating a visible light sensing component and a near-infrared sensing component, synchronously collecting and preprocessing a visible light fabric image and a near-infrared fabric image of a target polyimide fabric through the multispectral imaging module; building a dyeing uniformity detection double channel, the dyeing uniformity detection double channel comprising a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel; evaluating the uniformity of the visible light fabric image through the visible light dyeing uniformity detection channel to obtain fabric surface dyeing uniformity; if the fabric surface dyeing uniformity reaches a preset fabric dyeing uniformity standard, activating the near-infrared dyeing uniformity detection channel to detect the uniformity of the near-infrared fabric image to obtain fabric internal dyeing uniformity; weighting and fusing the fabric surface dyeing uniformity and the fabric internal dyeing uniformity to evaluate and determine a fabric dyeing uniformity detection result; the building of the dyeing uniformity detection double channel comprises: collecting a fabric standard visible light image set and a fabric defect visible light image set with dyeing uniformity identification, and a fabric standard near-infrared image set and a fabric defect near-infrared image set with dyeing uniformity identification; performing dyeing uniformity evaluation training on the fabric defect visible light image set with dyeing uniformity identification according to the fabric standard visible light image to construct a visible light dyeing uniformity detection channel; performing dyeing uniformity evaluation training on the fabric defect near-infrared image set with dyeing uniformity identification based on the fabric standard near-infrared image to obtain a near-infrared dyeing uniformity detection channel; parallel coupling the visible light dyeing uniformity detection channel and the near-infrared dyeing uniformity detection channel to build the dyeing uniformity detection double channel; the obtaining of the near-infrared dyeing uniformity detection channel comprises: constructing a near-infrared detection channel architecture comprising a three-dimensional reconstruction layer, a penetration defect identification layer and a dyeing uniformity analysis layer; performing feature extraction on the fabric standard near-infrared image set and the fabric defect near-infrared image set with dyeing uniformity identification based on the three-dimensional reconstruction layer to obtain a standard fabric dyeing reconstruction model and a defect fabric dyeing reconstruction model set; the penetration defect identification layer performs dye concentration distribution volume loss calculation on the defect fabric dyeing reconstruction model set according to the standard fabric dyeing reconstruction model to generate a defect fabric loss volume feature set; performing internal dyeing uniformity evaluation on the defect fabric loss volume feature set through the dyeing uniformity analysis layer to obtain a defect fabric internal dyeing sample set; performing feedback training and optimization on the near-infrared detection channel architecture based on the defect fabric internal dyeing sample set to obtain the near-infrared dyeing uniformity detection channel.

2. The method for on-line detection of polyimide fabric dyeing uniformity according to claim 1, characterized in that, the synchronous collection and preprocessing of a visible light fabric image and a near-infrared fabric image of a target polyimide fabric through the multispectral imaging module comprises: synchronously collecting an initial visible light image and an initial near-infrared image of a target polyimide fabric through the multispectral imaging module; According to the noise characteristics of the initial visible light image and the initial near-infrared image, a visible light filter and a near-infrared filter are initialized; The initial visible light image and the initial near-infrared image are filtered and denoised by using the visible light filter and the near-infrared filter to obtain a usable visible light image and a usable near-infrared image; The camera parameter calibration and image coordinate registration are performed on the usable visible light image and the usable near-infrared image to obtain the visible light fabric image and the near-infrared fabric image.

3. The method of on-line detection of polyimide fabric dyeing uniformity according to claim 1, wherein, The visible light dyeing uniformity detection channel is constructed, including: The visible light detection channel architecture is constructed, including a feature extraction layer, a defect classification layer, and a dyeing uniformity evaluation layer, wherein the feature extraction layer adopts a ResNet-50 network; Based on the feature extraction layer, feature extraction is performed on the fabric standard visible light image and the fabric defect visible light image set with dyeing uniformity identification to obtain a standard fabric dyeing feature set and a defect fabric dyeing feature set; The defect classification layer performs loss comparison and defect identification on the defect fabric dyeing feature set according to the standard fabric dyeing feature set to generate a defect fabric loss distribution feature set; The dyeing uniformity evaluation layer performs color difference weighted calculation and dyeing uniformity evaluation on the defect fabric loss distribution feature set to obtain a defect fabric surface dyeing sample set; Based on the defect fabric surface dyeing sample set, feedback training optimization is performed on the visible light detection channel architecture to construct the visible light dyeing uniformity detection channel.

4. The method of on-line detection of polyimide fabric dyeing uniformity according to claim 1, wherein, The defect fabric internal dyeing sample set is obtained, including: The dyeing uniformity analysis layer performs penetration defect color block identification on the defect fabric loss volume feature set to obtain a fabric penetration defect color block set; The fabric penetration defect color block set is sequentially subjected to distribution color difference calculation to obtain a fabric defect color block distribution color difference set; Based on the fabric defect color block distribution color difference set, internal dyeing uniformity evaluation is performed to obtain the defect fabric internal dyeing sample set.

5. The method of on-line detection of polyimide fabric dyeing uniformity according to claim 1, wherein, The fabric dyeing uniformity detection result is determined, including: According to the polyimide fabric characteristics, dyeing process, and dyeing uniformity detection target, surface dyeing weight factor and internal dyeing weight factor are determined; Based on the surface dyeing weight factor and the internal dyeing weight factor, the fabric surface dyeing uniformity and the fabric internal dyeing uniformity are weighted and fused to evaluate the fabric dyeing uniformity detection result.

6. The method of on-line detection of polyimide fabric dyeing uniformity according to claim 1, wherein, The method further includes: A laser displacement sensor is used to monitor the fabric tension fluctuation in real time; A fabric tension-color difference compensation model is established, and based on the fabric tension-color difference compensation model, compensation analysis is performed on the fabric tension fluctuation to determine fabric color difference compensation parameters, and the fabric dyeing uniformity detection result is dynamically corrected by the fabric color difference compensation parameters.

7. An on-line system for detecting the dyeing uniformity of polyimide fabrics, characterized by, The system is used to implement the online detection method of the polyimide fabric dyeing uniformity according to any one of claims 1 to 6, and the system includes: The fabric image acquisition unit is used for constructing a multispectral imaging module integrated with a visible light sensing component and a near-infrared sensing component, and the visible light fabric image and the near-infrared fabric image of the target polyimide fabric are synchronously acquired and preprocessed through the multispectral imaging module. The detection channel building unit is used for building a dyeing uniformity detection double channel, and the dyeing uniformity detection double channel includes a visible light dyeing uniformity detection channel and a near-infrared dyeing uniformity detection channel. The uniformity evaluation unit is used for evaluating the uniformity of the visible light fabric image through the visible light dyeing uniformity detection channel to obtain the surface dyeing uniformity of the fabric. The uniformity detection unit is used for activating the near-infrared dyeing uniformity detection channel to detect the uniformity of the near-infrared fabric image to obtain the internal dyeing uniformity of the fabric if the surface dyeing uniformity of the fabric reaches the preset fabric dyeing uniformity standard. The detection result determination unit is used for performing weighted fusion evaluation on the surface dyeing uniformity of the fabric and the internal dyeing uniformity of the fabric to determine the fabric dyeing uniformity detection result.

8. An electronic device, comprising: The electronic device includes: a memory for storing executable instructions; a processor for executing the executable instructions stored in the memory, realizing the online detection method of the polyimide fabric dyeing uniformity according to any one of claims 1-6.

Images