Glass fiber reinforced plastic surface cleaning quality monitoring method and system and electronic equipment

Abstract

The invention relates to the technical field of glass fiber reinforced plastic surface cleaning quality monitoring scheme design, in particular to a glass fiber reinforced plastic surface cleaning quality monitoring method and system and electronic equipment. The method comprises the following steps: firstly, acquiring an original image of the surface of the glass fiber reinforced plastic, then preprocessing the image, and denoising by adopting median filtering of different filtering windows for different texture areas through graying; extracting texture, surface roughness and glossiness characteristics, and carrying out normalization processing; then, constructing a cleaning quality evaluation model containing a three-layer neural network, and training the cleaning quality evaluation model; and finally, inputting the normalized features into the model to predict the cleaning quality grade. According to the method, the glass fiber reinforced plastic surface cleaning quality is evaluated more accurately and comprehensively, the problems of low efficiency, poor accuracy and the like of a traditional monitoring method are solved, and the actual requirements of glass fiber reinforced plastic surface cleaning quality monitoring are met.

Description

Technical Field

[0001] The present invention relates to the technical field of design of a glass fiber reinforced plastic surface cleaning quality monitoring scheme, and in particular to a glass fiber reinforced plastic surface cleaning quality monitoring method and system, and electronic equipment. Background Art

[0002] In the existing technology, the monitoring of FRP surface cleaning quality mainly faces the following technical problems: Inadequate monitoring accuracy: Existing monitoring methods often rely on simple visual image comparison, making it difficult to accurately identify minor stains, scratches, or uneven cleaning on FRP surfaces. For example, simply comparing the grayscale values of images before and after cleaning cannot accurately distinguish between normal surface texture changes and residual stains caused by incomplete cleaning.

[0003] Lack of comprehensive assessment: Failure to comprehensively consider the impact of multiple factors on cleaning quality. The cleaning quality of FRP surfaces is not only related to the presence of stains, but also to multiple factors such as surface roughness and gloss. However, existing technologies lack a comprehensive assessment solution.

[0004] Poor real-time performance: Some monitoring systems have slow processing speeds when inspecting the surface of FRP and cannot meet the real-time monitoring needs of the production line.

[0005] Therefore, the existing technology needs to be further developed. Summary of the Invention

[0006] The purpose of the present invention is to overcome the above technical deficiencies and provide a method and system for monitoring the surface cleanliness quality of glass fiber reinforced plastics, as well as electronic equipment, to solve the problems existing in the prior art.

[0007] To achieve the above technical objectives, according to a first aspect of the present invention, the present invention provides a method for monitoring the surface cleanliness quality of glass fiber reinforced plastics, comprising: S100, image acquisition: acquiring the original image of the FRP surface; S200, image preprocessing: grayscale the collected original image, and then use the median filter algorithm to denoise the grayscale image; S300, feature extraction: extracting texture features, surface roughness features and gloss features from the preprocessed image; S400, feature normalization: normalizing the extracted texture features, roughness features, and glossiness features respectively so that their value ranges are within a preset range; S500, constructing a cleaning quality assessment model: using a machine learning algorithm to construct a cleaning quality assessment model, and using the texture features, roughness features, and gloss features of existing FRP surface samples with known cleaning quality levels as training data to train the model; S600, cleaning quality prediction: input the normalized texture features, roughness features, and glossiness features into the trained cleaning quality assessment model to obtain the predicted cleaning quality level of the FRP surface.

[0008] Specifically, in the image preprocessing of step S200, the grayscale processing adopts the weighted average method, and the calculation formula is Igray=0.299R+0.587G+0.114B, where R, G, and B are the pixel values of the image in the red, green, and blue channels respectively.

[0009] Specifically, before performing median filtering denoising, texture feature analysis is performed on the image of the fiberglass surface to determine the region type to which each pixel belongs. The region types include regions with finer textures, regions with coarser textures, or regions with more noise. A gray-level co-occurrence matrix method is used to calculate the texture feature parameters of the image. The texture feature parameters include contrast, energy, and entropy. Based on the calculated texture feature parameters, threshold conditions are set to determine the region type of each pixel. For pixels that are determined to be regions with finer textures by the texture feature analysis results, a 3×3 filter window is used for median filtering. For pixels that are determined to be regions with coarser textures or regions with more noise by the texture feature analysis results, a 5×5 filter window is used for median filtering.

[0010] Specifically, the setting of a threshold condition to determine the area type of each pixel includes: When the contrast is less than the first preset threshold, the energy is greater than the second preset threshold, and the entropy is less than the third preset threshold, the pixel is determined to belong to an area with finer texture; otherwise, the pixel is determined to belong to an area with coarser texture or more noise.

[0011] Specifically, the method further includes: The middle value of the grayscale values of the sorted pixels is taken as the value of the pixel after filtering; if the number of pixels is even, the average of the two middle values is taken as the filtered value.

[0012] Specifically, the method includes: Laser triangulation is used to extract surface roughness features. A specific area is selected on the FRP surface. The laser transmitter and receiver are installed at a preset height from the FRP surface at a preset angle. The laser transmitter emits a laser beam to irradiate the FRP surface, and the reflected light is received by the receiver. Based on the principle of laser triangulation, the height information h (x, y) of each point on the FRP surface is calculated according to the position change of the laser spot received by the receiver. (x, y) is the coordinate of the point in the surface area. The height data is then converted into a roughness value, including calculating the standard deviation of the height using the following formula, which is then recorded as the roughness data: Among them, M and N are the number of rows and columns of points in the selected area, is the average height of the area.

[0013] Specifically, the method further includes: Place the measuring head of the gloss meter vertically on the FRP surface, keep the contact pressure between the measuring head and the surface stable, measure the gloss at the selected measuring point, and obtain the gloss value G.

[0014] Specifically, the construction of the cleaning quality assessment model includes: A three-layer neural network is constructed with 4 input layer nodes, corresponding to the three features of texture, roughness, and glossiness plus a bias node. The number of hidden layer nodes is 15, and the number of output layer nodes is 11. The 11 output layer nodes correspond to the cleaning quality levels 0-10. Back propagation algorithm is used for training, and the learning rate η is set to 0.01. The loss function is the cross entropy loss function: Among them, y i For a true clean quality grade label, is the probability of the cleaning quality grade predicted by the model. During the training process, when the training error on the validation set is less than 0.01, the training is stopped and the trained neural network cleaning quality assessment model is obtained.

[0015] According to a second aspect of the present invention, there is provided a glass fiber reinforced plastic surface cleaning quality monitoring system, comprising: Image acquisition module, used to acquire images of the fiberglass surface; Image preprocessing module, used to perform grayscale processing and median filtering denoising on the collected images; A feature extraction module is used to extract texture features, surface roughness features and glossiness features from the preprocessed image; The feature normalization module is used to normalize the extracted features so that their value range is within a preset range; The cleaning quality assessment model module contains a trained cleaning quality assessment model, which is used to predict the cleaning quality of the normalized features.

[0016] According to a third aspect of the present invention, an electronic device is provided, comprising: a memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the above-mentioned method for monitoring the surface cleaning quality of glass fiber reinforced plastics is implemented.

[0017] Beneficial effects: This method significantly improves the accuracy of cleaning quality monitoring by comprehensively considering the texture, roughness, and gloss characteristics of the FRP surface and employing a suitable machine learning algorithm to construct an evaluation model. The overall solution can rapidly process images and generate monitoring results, meeting the real-time requirements of production lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 1 is a flow chart of a method for monitoring the surface cleanliness quality of glass fiber reinforced plastics provided in a specific embodiment of the present invention; Figure 2 It is a schematic diagram of the system composition of the glass fiber reinforced plastic surface cleaning quality monitoring system provided in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0019] In order to enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention is clearly and completely described below in conjunction with the drawings of the present invention. Based on the embodiments in this application, other similar embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application. In addition, the directional words mentioned in the following embodiments, such as "up", "down", "left", "right", etc., are only reference to the directions of the drawings. Therefore, the directional words used are used to illustrate rather than limit the invention.

[0020] The present invention will be further described below with reference to the accompanying drawings and preferred embodiments.

[0021] See also Figure 1 The present invention provides a method for monitoring the surface cleanliness quality of glass fiber reinforced plastics, comprising: S100, image acquisition: acquiring the original image of the fiberglass surface.

[0022] S200, image preprocessing: grayscale the collected original image, and then use the median filter algorithm to denoise the grayscale image.

[0023] Specifically, in the image preprocessing of step S200, the grayscale processing adopts the weighted average method, and the calculation formula is Igray=0.299R+0.587G+0.114B, where R, G, and B are the pixel values of the image in the red, green, and blue channels respectively.

[0024] Specifically, before performing median filtering denoising, texture feature analysis is performed on the image of the fiberglass surface to determine the region type to which each pixel belongs. The region types include regions with finer textures, regions with coarser textures, or regions with more noise. A gray-level co-occurrence matrix method is used to calculate the texture feature parameters of the image. The texture feature parameters include contrast, energy, and entropy. Based on the calculated texture feature parameters, threshold conditions are set to determine the region type of each pixel. For pixels that are determined to be regions with finer textures by the texture feature analysis results, a 3×3 filter window is used for median filtering. For pixels that are determined to be regions with coarser textures or regions with more noise by the texture feature analysis results, a 5×5 filter window is used for median filtering.

[0025] As you can understand, for pixels in areas with fine texture determined by texture feature analysis, a 3×3 filter window is used for median filtering. The 3×3 filter window is a moderate size, effectively removing noise while preserving fine texture details. For areas with coarse texture or high noise determined by texture feature analysis, a 5×5 filter window is used for median filtering. Larger filter windows can better handle noise, but may blur texture details, and are therefore more suitable for areas with coarse texture or high noise levels.

[0026] Specifically, the calculation formula for contrast is as follows: Among them, P (i, j) is the gray level co-occurrence matrix element, which represents the joint probability of pixel pairs with gray level values ​​i and j appearing in the image.

[0027] Specifically, the calculation formula for energy is as follows: Among them, P (i, j) is the gray level co-occurrence matrix element, which represents the joint probability of pixel pairs with gray level values ​​i and j appearing in the image.

[0028] Specifically, the calculation formula for entropy is as follows: Among them, P (i, j) is the gray level co-occurrence matrix element, which represents the joint probability of pixel pairs with gray level values ​​i and j appearing in the image.

[0029] Specifically, the setting of a threshold condition to determine the area type of each pixel includes: When the contrast is less than the first preset threshold, the energy is greater than the second preset threshold, and the entropy is less than the third preset threshold, the pixel is determined to belong to an area with finer texture; otherwise, the pixel is determined to belong to an area with coarser texture or more noise.

[0030] Preferably, when the contrast is less than 30, the energy is greater than 100, and the entropy is less than 5, the pixel is determined to belong to a region with finer texture; otherwise, the pixel is determined to belong to a region with coarser texture or more noise.

[0031] Specifically, the method further includes: The middle value of the grayscale values of the sorted pixels is taken as the value of the pixel after filtering; if the number of pixels is even, the average of the two middle values is taken as the filtered value.

[0032] Specifically, the method includes: Input: Original image I of the FRP surface, image size is M×N (M and N are the number of rows and columns of the image, respectively).

[0033] Output: denoised FRP surface image I′.

[0034] S1. Grayscale the original image I to obtain a grayscale image G with a grayscale range of 0-255.

[0035] S2. Initialize the texture feature matrix T, which has the same size as the grayscale image G and is used to store the texture feature type of each pixel (area with fine texture or area with coarse texture or more noise).

[0036] S3, for each pixel (i, j) in the grayscale image G (0≤i <M,0≤j<N): ① Taking pixel point (i, j) as the center, calculate the gray-level co-occurrence matrix P in its 5×5 neighborhood.

[0037] ② Calculate contrast C, energy E and entropy H based on the gray-level co-occurrence matrix P.

[0038] ③ Determine the region type to which the pixel point (i, j) belongs based on the set threshold, and store the result in the corresponding position T (i, j) of the texture feature matrix T.

[0039] S4. Initialize the denoised image I′ to have the same size as the original image I.

[0040] S5, for each pixel (i, j) in the grayscale image G (0≤i <M,0≤j<N): ① If T(i, j) represents a region with finer texture: With pixel point (i, j) as the center, select a 3×3 filter window W1.

[0041] The pixel points in the filter window W1 are sorted in ascending order according to the grayscale value to obtain a sorted pixel value sequence S1.

[0042] Calculate the middle value v1 of the sorted pixel value sequence S1 (if the number of pixels is even, take the average of the two middle values).

[0043] Assign v1 to the corresponding pixel (i, j) of the denoised image I′.

[0044] ② If T(i, j) represents a region with coarse texture or more noise: With pixel point (i, j) as the center, select a 5×5 filter window W2.

[0045] The pixel points in the filter window W2 are sorted in ascending order according to the grayscale value to obtain a sorted pixel value sequence S2.

[0046] Calculate the middle value v2 of the sorted pixel value sequence S2 (if the number of pixels is even, take the average of the two middle values).

[0047] Assign v2 to the corresponding pixel (i, j) of the denoised image I′.

[0048] ③ When selecting the filter window and calculating the intermediate value, if the filter window exceeds the image boundary, the boundary pixels are processed using the mirror filling method.

[0049] S6. Output the denoised FRP surface image I′.

[0050] Specifically, the method includes: Laser triangulation is used to extract surface roughness features. A specific area is selected on the FRP surface. The laser transmitter and receiver are installed at a preset angle (preferably 45 degrees) and a preset height (preferably 50 mm) from the FRP surface. The laser transmitter emits a laser beam to irradiate the FRP surface, and the reflected light is received by the receiver. Based on the principle of laser triangulation, the height information h(x, y) of each point on the FRP surface is calculated according to the position change of the laser spot received by the receiver. (x, y) is the coordinate of the point in the surface area. The height data is then converted into a roughness value, including calculating the standard deviation of the height using the following formula, which is then recorded as the roughness data: Among them, M and N are the number of rows and columns of points in the selected area, is the average height of the area.

[0051] Specifically, the method further includes: Place the measuring head of the gloss meter vertically on the FRP surface, keep the contact pressure between the measuring head and the surface stable, measure the gloss at the selected measuring point, and obtain the gloss value G.

[0052] Specifically, place the measuring head of the gloss meter vertically on the FRP surface (the contact pressure with the surface is maintained at 0.1N. In actual application, the appropriate pressure can be determined according to the instructions of the gloss meter and actual operation), and measure the gloss at selected measuring points (for example, 10 measuring points are evenly selected) to obtain the gloss value G.

[0053] S300, feature extraction: extracting texture features, surface roughness features and glossiness features from the preprocessed image.

[0054] S400, feature normalization: normalizing the extracted texture features, roughness features, and glossiness features respectively so that their value ranges are within a preset range.

[0055] The extracted texture features T, roughness features R, and glossiness features G are normalized to the range [0, 1]. Let the original feature value be x and the normalized feature value be y. The normalization formula is y = (x − min(x)) / (max(x) − min(x)), where min(x) and max(x) are the minimum and maximum values of the feature value x, respectively.

[0056] It can be understood that the formula for texture feature normalization is as follows: The formula for normalizing roughness features is as follows: The formula for normalizing glossiness features is as follows: in, are the minimum and maximum values in the texture feature dataset, respectively. are the minimum and maximum values in the roughness feature data set, are the minimum and maximum values in the glossiness feature dataset, respectively. These minimum and maximum values can be obtained based on the feature values of all samples in the training dataset.

[0057] S500. Construct a cleaning quality assessment model: Use a machine learning algorithm to construct a cleaning quality assessment model, and use the texture features, roughness features, and gloss features of existing FRP surface samples with known cleaning quality levels as training data to train the model.

[0058] Specifically, the construction of the cleaning quality assessment model includes: Collect at least 100 FRP surface samples with known cleanliness quality levels, use the texture features, roughness features, and glossiness features of each sample as input, and the cleanliness quality level (e.g., cleanliness level is defined as 0-10, 0 means extremely unclean and 10 means very clean) as output.

[0059] A three-layer neural network is constructed with 4 input layer nodes, corresponding to the three features of texture, roughness, and glossiness plus a bias node. The number of hidden layer nodes is 15, and the number of output layer nodes is 11. The 11 output layer nodes correspond to the cleaning quality levels 0-10. Back propagation algorithm is used for training, and the learning rate η is set to 0.01. The loss function is the cross entropy loss function: Where yi is the true cleaning quality level label (for example, if the true cleaning quality level of a sample is 5, then y5=1, and the rest yi=0, i≠5), To predict the cleanliness quality grade probability, the model is trained using the texture, roughness, and gloss characteristics (normalized) of existing FRP surface samples with known cleanliness quality grades. During each training iteration, the model inputs the training samples, calculates the output through forward propagation, and then calculates the error between the predicted result and the true label using a loss function. Backpropagation is then used to update the model's weights and bias parameters to minimize this error. Training is terminated when the training error (e.g., the average cross-entropy loss function) on the validation set is less than 0.01, resulting in a trained neural network cleanliness quality assessment model.

[0060] S600, cleaning quality prediction: input the normalized texture features, roughness features, and glossiness features into the trained cleaning quality assessment model to obtain the predicted cleaning quality level of the FRP surface.

[0061] Specifically, the normalized texture feature T, roughness feature R, and glossiness feature G are input into the trained cleaning quality assessment model M to obtain the predicted cleaning quality grade Q of the FRP surface.

[0062] Specifically, the normalized texture features, roughness features, and glossiness features obtained in the previous steps are input into the trained cleaning quality assessment model. After forward propagation calculation, the model obtains 11 output values, corresponding to cleaning quality levels 0-10. The cleaning quality level corresponding to the node with the largest output value is selected as the predicted cleaning quality level of the FRP surface. For example, if the output value is =0.3, =0.8, =0.2, and other values are less than 0.8, then the predicted cleaning quality level of the FRP surface is judged to be 5.

[0063] As can be appreciated, the present invention significantly improves the accuracy of cleaning quality monitoring by comprehensively considering the texture, roughness, and gloss characteristics of the FRP surface and employing a suitable machine learning algorithm to construct an evaluation model. The overall solution of the present invention enables rapid image processing and monitoring results, meeting the real-time requirements of production lines.

[0064] See also Figure 2 The present invention provides another embodiment, which provides a glass fiber reinforced plastic surface cleaning quality monitoring system, the glass fiber reinforced plastic surface cleaning quality monitoring system comprising: Image acquisition module 100, used to acquire images of the fiberglass surface; The image preprocessing module 200 is used to perform grayscale processing and median filtering denoising on the collected image; A feature extraction module 300 is used to extract texture features, surface roughness features and gloss features from the pre-processed image; The feature normalization module 400 is used to normalize the extracted features so that their numerical values fall within a preset range; The cleaning quality assessment model module 500 includes a trained cleaning quality assessment model, and is used to predict the cleaning quality based on the normalized features.

[0065] It should be noted that this invention significantly improves the accuracy of cleaning quality monitoring by comprehensively considering the texture, roughness, and gloss characteristics of the FRP surface and employing a suitable machine learning algorithm to construct an evaluation model. This overall solution enables rapid image processing and monitoring results, meeting the real-time requirements of production lines.

[0066] In a preferred embodiment, the present application further provides an electronic device, comprising: A memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method for monitoring the surface cleaning quality of fiberglass is implemented. The computer device can be broadly defined as a server, a terminal, or any other electronic device with the necessary computing and / or processing capabilities. In one embodiment, the computer device may include a processor, a memory, a network interface, a communication interface, etc. connected via a system bus. The processor of the computer device can be used to provide the necessary computing, processing, and / or control capabilities. The memory of the computer device may include a non-volatile storage medium and an internal memory. An operating system, a computer program, etc. may be stored in or on the non-volatile storage medium. The internal memory can provide an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface and the communication interface of the computer device can be used to connect and communicate with external devices via a network. When the computer program is executed by the processor, the steps of the method of the present invention are performed.

[0067] The present invention can be implemented as a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, causes the steps of the method of an embodiment of the present invention to be performed. In one embodiment, the computer program is distributed on a plurality of computer devices or processors coupled to a network so that the computer program is stored, accessed, and executed in a distributed manner by one or more computer devices or processors. A single method step / operation, or two or more method steps / operations, can be performed by a single computer device or processor or by two or more computer devices or processors. One or more method steps / operations can be performed by one or more computer devices or processors, and one or more other method steps / operations can be performed by one or more other computer devices or processors. One or more computer devices or processors can perform a single method step / operation, or perform two or more method steps / operations.

[0068] Those skilled in the art will appreciate that the method steps of the present invention can be performed by instructing related hardware, such as a computer device or processor, through a computer program. The computer program can be stored in a non-transitory computer-readable storage medium, and when the computer program is executed, the steps of the present invention are performed. Depending on the circumstances, any reference herein to memory, storage, database, or other media may include non-volatile and / or volatile memory. Examples of non-volatile memory include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), flash memory, magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid-state disk, etc. Examples of volatile memory include random access memory (RAM), external cache memory, etc.

[0069] As can be appreciated, the present invention significantly improves the accuracy of cleaning quality monitoring by comprehensively considering the texture, roughness, and gloss characteristics of the FRP surface and employing a suitable machine learning algorithm to construct an evaluation model. The overall solution of the present invention enables rapid image processing and monitoring results, meeting the real-time requirements of production lines.

[0070] The various technical features described above can be combined arbitrarily. Although not all possible combinations of these technical features are described, any combination of these technical features should be considered to be covered by this specification as long as such combination does not conflict.

[0071] The specific embodiments of the present invention described above do not limit the scope of protection of the present invention. Any other corresponding changes and modifications made based on the technical concept of the present invention should be included in the scope of protection of the claims of the present invention.

Claims

1. A method for monitoring the surface cleanliness quality of glass fiber reinforced plastics, characterized in that: The method comprises: S100, image acquisition: acquiring the original image of the fiberglass surface; S200, image preprocessing: grayscale the collected original image, and then use the median filter algorithm to denoise the grayscale image; S300, feature extraction: extracting texture features, surface roughness features and gloss features from the preprocessed image; S400, feature normalization: normalizing the extracted texture features, roughness features, and glossiness features respectively so that their value ranges are within a preset range; S500, constructing a cleaning quality assessment model: using a machine learning algorithm to construct a cleaning quality assessment model, and using the texture features, roughness features, and gloss features of existing FRP surface samples with known cleaning quality levels as training data to train the model; S600, cleaning quality prediction: input the normalized texture features, roughness features, and glossiness features into the trained cleaning quality assessment model to obtain the predicted cleaning quality level of the FRP surface.

2. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 1, wherein: In the image preprocessing of step S200, the grayscale processing adopts the weighted average method, and the calculation formula is Igray=0.299R+0.587G+0.114B, where R, G, and B are the pixel values of the image in the red, green, and blue channels respectively.

3. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 1, wherein: Before performing median filtering denoising, texture feature analysis is performed on the image of the fiberglass surface to determine the region type to which each pixel belongs. The region types include regions with fine texture, regions with coarse texture, or regions with more noise. The gray-level co-occurrence matrix method is used to calculate the texture feature parameters of the image. The texture feature parameters include contrast, energy, and entropy. According to the calculated texture feature parameters, threshold conditions are set to determine the region type of each pixel. For pixels determined to be in regions with fine texture as a result of the texture feature analysis, a 3×3 filter window is used for median filtering. For pixels determined to be in regions with coarse texture or regions with more noise as a result of the texture feature analysis, a 5×5 filter window is used for median filtering.

4. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 3, wherein: The setting of a threshold condition to determine the area type of each pixel point includes: When the contrast is less than the first preset threshold, the energy is greater than the second preset threshold, and the entropy is less than the third preset threshold, the pixel is determined to belong to an area with finer texture; otherwise, the pixel is determined to belong to an area with coarser texture or more noise.

5. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 3, wherein: The method further comprises: The middle value of the grayscale values of the sorted pixels is taken as the value of the pixel after filtering; if the number of pixels is even, the average of the two middle values is taken as the filtered value.

6. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 1, wherein: The method comprises: Laser triangulation is used to extract surface roughness features. A specific area is selected on the FRP surface. The laser transmitter and receiver are installed at a preset height from the FRP surface at a preset angle. The laser transmitter emits a laser beam to irradiate the FRP surface, and the reflected light is received by the receiver. Based on the principle of laser triangulation, the height information h (x, y) of each point on the FRP surface is calculated according to the position change of the laser spot received by the receiver. (x, y) is the coordinate of the point in the surface area. The height data is then converted into a roughness value, including calculating the standard deviation of the height using the following formula, which is then recorded as the roughness data: Among them, M and N are the number of rows and columns of points in the selected area, is the average height of the area.

7. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 1, wherein: The method further comprises: Place the measuring head of the gloss meter vertically on the FRP surface, keep the contact pressure between the measuring head and the surface stable, measure the gloss at the selected measuring point, and obtain the gloss value G.

8. The method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to claim 1, wherein: The construction of the cleaning quality assessment model includes: A three-layer neural network is constructed with 4 input layer nodes, corresponding to the three features of texture, roughness, and glossiness plus a bias node. The number of hidden layer nodes is 15, and the number of output layer nodes is 11. The 11 output layer nodes correspond to the cleaning quality levels 0-10. Back propagation algorithm is used for training, and the learning rate η is set to 0.

01. The loss function is the cross entropy loss function: Among them, y i For a true clean quality grade label, is the probability of the cleaning quality grade predicted by the model. During the training process, when the training error on the validation set is less than 0.01, the training is stopped and the trained neural network cleaning quality assessment model is obtained.

9. A glass fiber reinforced plastic surface cleaning quality monitoring system, characterized in that: include: Image acquisition module, used to acquire images of the fiberglass surface; Image preprocessing module, used to perform grayscale processing and median filtering denoising on the collected images; A feature extraction module is used to extract texture features, surface roughness features and glossiness features from the preprocessed image; The feature normalization module is used to normalize the extracted features so that their value range is within a preset range; The cleaning quality assessment model module contains a trained cleaning quality assessment model, which is used to predict the cleaning quality of the normalized features.

10. An electronic device, characterized in that: include: Memory; and a processor, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method for monitoring the surface cleaning quality of glass fiber reinforced plastics according to any one of claims 1 to 8 is implemented.

Images