Coating aging color feature evaluation method and system based on machine learning

Abstract

The invention relates to the technical field of coating aging evaluation, and discloses a coating aging color feature evaluation method and system based on machine learning, and the method comprises the steps: collecting macroscopic morphology images of a coating sample in the aging process before and after corrosion; performing chromatic aberration measurement to obtain chromaticity coordinate values of the coating sample before and after corrosion, and calculating a chromatic aberration value delta E; marking and segmenting chromatic aberration measurement positions of the macroscopic morphology image; extracting R, G and B component values of each image, and calculating delta R, delta G and delta B of the images before and after corrosion; on the basis of a machine learning algorithm, taking delta R, delta G and delta B as independent variables, taking the color difference value delta E as a dependent variable, and building and training a color difference prediction model based on colors; and collecting standard target images before and after corrosion of the to-be-evaluated coating, and performing chromatic aberration prediction and aging grade evaluation by using the trained model. According to the invention, quantitative identification and aging grade determination of coating aging color features are carried out by using a machine learning method, and the method has the characteristics of high efficiency and high precision.

Description

Technical Field

[0001] The present invention relates to the technical field of coating aging evaluation, and in particular to a coating aging color feature evaluation method and system based on machine learning. Background Art

[0002] The marine atmosphere is a typical atmospheric corrosion environment with the highest corrosion level, characterized by high radiation, high humidity and heat, and high salt spray. Marine engineering steel structures are the most important basic structures in modern engineering equipment and are widely used in corrosion environments of C5-M to CX levels. Coating systems are often used for corrosion protection. However, the coating will degrade under the influence of various environmental factors, and aging phenomena such as powdering, loss of gloss, and fading will occur.

[0003] In the industry, the evaluation of coating color aging characteristics mainly relies on the visual colorimetry of professionals or the use of instrumental determination methods to measure chromaticity coordinates and calculate color difference values. After that, the aging grade is determined according to national and ISO standards. Although this method has the advantages of low threshold and easy operation, it has the disadvantages of low efficiency and long detection cycle. In addition, the visual colorimetry method is subject to subjective judgment influenced by human factors and is not suitable for batch detection of large-scale projects, large areas, and high-altitude areas that are difficult for personnel to reach.

[0004] In recent years, with the vigorous development of emerging industries such as information science and technology and artificial intelligence, machine vision technology has been widely used in fields such as image processing and non-destructive testing. Color classification and prediction based on machine learning has made breakthrough progress in color difference prediction in color space due to its advantages of high efficiency, high precision and scalability. Domestic scholars have used models for color quantification and classification in other industries, which has the characteristics of easy data acquisition, on-site measurement and relatively reliable calculation results. However, in coating color evaluation, the research on the quantification and grade determination of coating aging characteristics using machine learning methods is still in its infancy, and further in-depth research is required by technical personnel in this field. Summary of the invention

[0005] The purpose of the present invention is to provide a coating aging color feature evaluation method and system based on machine learning, which uses machine learning methods to quantitatively identify coating aging color features and determine aging levels, which can reduce manual intervention and greatly improve the efficiency of coating color difference evaluation.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A coating aging color feature evaluation method based on machine learning, the method comprising the following steps:

[0008] S1, macroscopic morphological images of coating samples during the aging process before and after corrosion are collected from different angles to form a standardized aging image dataset;

[0009] S2, measure the color difference of the coating sample, obtain the chromaticity coordinate value data of the coating sample before and after corrosion, calculate the color difference value ΔE, and determine the color difference aging grade based on the national standard and ISO standard;

[0010] S3, marking and segmenting the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set to form a corrosion image database corresponding to the color difference measurement positions;

[0011] S4, extracting the R, G, and B component value data of each image in the corrosion image database, and completing the calculation of ΔR, ΔG, and ΔB of the images before and after corrosion;

[0012] S5, based on the machine learning algorithm, with ΔR, ΔG, ΔB as independent variables and color difference value ΔE as dependent variable, build and train a color-based color difference prediction model, and evaluate the model performance;

[0013] S6, collecting standard target images of the coating to be evaluated before and after corrosion, extracting ΔR, ΔG, and ΔB data of the standard target image as input, using the trained color-based color difference prediction model to perform color difference prediction, obtaining the color difference value prediction result of the coating to be evaluated, and performing aging grade assessment based on the aging characteristic color grade assessment standard.

[0014] Furthermore, in S1, a standardized image acquisition device is used to acquire macroscopic morphological images of the coating sample during the aging process before and after corrosion from different angles;

[0015] The standardized image acquisition device acquires images in a dark room protected from light using a uniform fixed light source.

[0016] Furthermore, in said S1, the aging process includes natural aging and accelerated aging process;

[0017] The S1 also includes storing the standardized aging image data set into a database and recording associated data information, wherein the stored standardized aging image data set includes a chemical formula with a known coating composition and associated data information related to the coating, and the associated data information includes a standardized macroscopic image, coating thickness, coating time, location, coating corrosion test type and time.

[0018] Furthermore, in S2, a traditional color difference measurement tool is used to measure the color difference of the coating sample, and the traditional color difference measurement tool includes at least one of a colorimeter and a spectrophotometer that can measure the L, a, and b data of the coating sample before and after aging;

[0019] The chromaticity coordinate value data includes L, a, and b data, wherein L represents lightness, a represents red and green, and b represents yellow and blue;

[0020] Calculate the color difference ΔE: According to the formula Calculate the color difference value.

[0021] Furthermore, in S3, the image annotation tool and computer vision algorithm are used to mark and segment the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set;

[0022] The segmentation method includes one or a combination of cropping, image segmentation algorithm, image annotation tool, and masking method.

[0023] Furthermore, in S5, the machine learning algorithm includes one or more combinations of linear regression, polynomial regression, ridge regression, LASSO regression and neural network prediction algorithms.

[0024] Furthermore, in S5, ΔR, ΔG, and ΔB calculated by S4 are used as independent variables, and the color difference value ΔE calculated by S2 is used as a dependent variable, and matching is performed, and the training set and the test set are divided into a training set and a test set according to a set ratio. The training set is used for iteratively training a color-based color difference prediction model, and the test set is used for evaluating model performance.

[0025] Further, in S5, the model performance evaluation is performed, including: using the test set to predict the trained color difference prediction model based on color to obtain a prediction result, and evaluating the model performance based on multiple evaluation indicators, wherein the multiple evaluation indicators include mean square error, root mean square error, determination coefficient, and mean absolute error;

[0026] The model performance evaluation also includes: for the test set, taking the color difference aging level determined by S2 as the accurate result, comparing the color difference aging level determined based on the prediction result, and determining the accuracy of the color difference prediction model based on color to evaluate the color characteristics of the coating.

[0027] The present invention also discloses a coating aging color feature evaluation system based on machine learning, which is applied to execute the above-mentioned coating aging color feature evaluation method based on machine learning, comprising:

[0028] An image acquisition unit is used to acquire macroscopic morphological images of the coating sample during the aging process before and after corrosion from different angles to form a standardized aging image data set; it is also used to acquire standard target images of the coating to be evaluated before and after corrosion;

[0029] The color difference measurement unit is used to measure the color difference of the coating sample, obtain the chromaticity coordinate value data of the coating sample before and after corrosion, calculate the color difference value ΔE, and determine the color difference aging grade based on the national standard and ISO standard;

[0030] An image matching unit is used to mark and segment the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set, so as to form a corrosion image database corresponding to the color difference measurement positions one by one;

[0031] The image processing unit extracts the R, G, and B component value data of each image in the corrosion image database and completes the calculation of ΔR, αG, and αB of the images before and after corrosion; it is also used to extract the R, G, and B data of the standard target image and calculate the ΔR, ΔG, and ΔB corresponding to the standard target image;

[0032] The model training unit is used to build and train a color-based color difference prediction model based on a machine learning algorithm, using ΔR, ΔG, and ΔB as independent variables and the color difference value ΔE as a dependent variable, and to evaluate the model performance;

[0033] The model prediction unit is used to use the trained color-based color difference prediction model to predict the color difference, obtain the color difference value prediction result of the coating to be evaluated, and perform aging grade assessment based on the aging characteristic color grade assessment standard.

[0034] Furthermore, the image acquisition unit includes a standardized image acquisition device, which acquires images in a dark room with a uniform fixed light source.

[0035] The color difference measurement unit includes at least one of the tools selected from the group consisting of a colorimeter and a spectrophotometer, which can measure the L, a, and b data of the coating sample before and after aging.

[0036] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects: the coating aging color feature evaluation method and system based on machine learning provided by the present invention, pre-establishes a color-based color difference prediction model based on the machine learning method, and uses the color-based color difference prediction model to realize the quantitative evaluation of the coating aging color feature, which can avoid the problems of low measurement accuracy and large amount of calculation when directly using traditional measurement tools in a field environment; the present invention simplifies the color difference calculation process through a machine learning model, avoids complex calculations, and the image data can be measured on site and easily obtained, is non-destructive to the coating itself, and is suitable for batch detection of large-scale projects, large areas, and high-altitude areas that are difficult for personnel to reach.

[0037] The present invention further improves the accuracy of quantitative identification and grade determination of aging color characteristics by establishing a color-based color difference prediction model, which can provide an important basis for the design and selection of protective coatings for metal materials. To a certain extent, it helps to select materials that are durable, reliable and adaptable to specific environments in different application requirements. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0039] Figure 1 This is a flow chart of a coating aging color feature evaluation method based on machine learning according to an embodiment of the present invention;

[0040] Figure 2 A portion of the image after the standardized image acquisition according to the embodiment of the present invention;

[0041] Figure 3 This is a diagram showing the result of image segmentation of the colorimeter measurement area according to an embodiment of the present invention;

[0042] Figure 4 This is a diagram showing the training result of a color difference prediction model based on color according to an embodiment of the present invention;

[0043] Figure 5 This is a test result diagram of a color difference prediction model based on color according to an embodiment of the present invention;

[0044] Figure 6 1 is a segmentation diagram of the Q235 / epoxy zinc-rich / acrylic polyurethane coating before and after corrosion in an embodiment of the present invention, wherein (a1) is the segmentation image before corrosion, and (b1) is the corresponding segmentation image after corrosion. DETAILED DESCRIPTION

[0045] The embodiments of the present invention are described in detail below. The embodiments are intended to explain the present invention and should not be construed as limiting the present invention. If no specific techniques or conditions are specified in the embodiments, the techniques or conditions described in the literature in the art or the product specifications are used. If no manufacturer is specified for the materials or instruments used, they are all conventional products that can be obtained commercially.

[0046] When measuring color difference, it is difficult to ensure the measurement accuracy of traditional measuring tools in the field environment; there is currently a lack of effective measuring tools for batch testing of large-scale projects, large areas and high-altitude parts that are difficult for personnel to reach; in response to the problems existing in the prior art, the present invention discloses a coating aging color feature evaluation method and system based on machine learning, which realizes efficient and high-precision batch testing by establishing a color-based color difference prediction model, and is suitable for large-scale coating quality control.

[0047] like Figure 1 As shown, the coating aging color feature evaluation method based on machine learning provided by an embodiment of the present invention includes the following steps:

[0048] Step 1, using a standardized image acquisition device to collect macroscopic morphological images of coating samples in natural aging, accelerated aging and other processes before and after corrosion from different angles, together forming a standardized aging image data set, and storing the standardized aging image data set in a database, recording associated data information; wherein the stored standardized aging image data set includes a chemical formula with a known coating composition, and associated data information matching related to the coating, and the associated data information includes standardized macroscopic images, coating thickness, coating time, location, coating corrosion test type and time; illustratively, in the step 1, the standardized image acquisition device includes performing standardized image acquisition under a light-proof darkroom condition using a unified fixed light source irradiation condition, and the standardized image acquisition device includes industrial cameras, zoom lenses, copy tables, fixed light sources, shutter cables and other equipment.

[0049] Take images of each sample before and after corrosion, and take multiple images from different angles to further ensure the comprehensiveness of the data. Image preprocessing can also be performed: the collected images are preprocessed (such as cropping, adjusting brightness and contrast, etc.) to meet standardization requirements.

[0050] In this embodiment, a carbon steel Q235 / epoxy zinc-rich primer / acrylic polyurethane topcoat composite coating system was selected to conduct a natural environment corrosion test in the Qingdao marine atmosphere environment. At the same time, three artificial accelerated corrosion conditions were conducted: salt spray test, ultraviolet aging test, and multi-factor coupling acceleration test. A self-built standardized image acquisition device was used to collect standardized aging image data before and after corrosion under conditions such as a dark room and a fixed light source. Some of the collected standardized image data are shown in FIG. Figure 2 shown.

[0051] In the present embodiment, Q235 steel is selected, and after the substrate is sandblasted and the cleaning quality reaches S2.5 grade, epoxy zinc-rich primer is sprayed on the substrate surface by spraying, and acrylic polyurethane topcoat is sprayed after drying at room temperature for 24 hours, and it is standby after drying in dark at room temperature for 7 days. TT260 coating thickness gauge is used to measure the primer dry film thickness of about 40 μm, the topcoat dry film thickness of about 80 μm, and the total thickness of the paint film is controlled at 120 ± 10 μm. In order to quickly obtain aging characteristics, the spraying of acrylic polyurethane topcoat is carried out separately on the substrate surface by spraying, and after drying in dark at room temperature for 7 days, the total thickness of the paint film is measured and controlled at 30 ± 5 μm.

[0052] Step 2, using traditional color difference measurement tools to measure the color difference of the coating sample, and obtain the chromaticity coordinate value data L, a, and b of the coating sample before and after corrosion, where L represents lightness, a represents red and green, and b represents yellow and blue. The value range of L is from 0 to 100. When L=50, it is equivalent to 50% black. The value ranges of a and b are from +127 to -128, where +127a is red, and gradually transitions to -128a when it turns green. Similarly, +127b is yellow and -128b is blue. All colors are composed of the interactive changes of these three values. The color difference value ΔE is calculated based on these three values, and the color difference aging grade is determined based on the national standard and ISO standard;

[0053] For example, in step 2, the conventional color difference measurement tool includes at least one of a colorimeter and a spectrophotometer that can measure the L, a, and b data of the coating sample before and after aging. The L, a, and b values ​​at each color difference measurement position are measured using a colorimeter (spectrophotometer);

[0054] Calculate the color difference ΔE: According to the formula Calculate the color difference value, ΔL, Δa, Δb are the corresponding changes of L, a, and b.

[0055] The specific national standards and ISO standards are: GB / T 1766-2008 "Rating method for aging of paint and varnish coatings" and ISO4628 "Rating method for aging of paint and varnish coatings".

[0056] In this embodiment, the chromaticity coordinate values ​​L, a, and b of the coating sample before and after corrosion are obtained by using an X·rite colorimeter to measure the coating sample at the same fixed position to obtain the L, a, and b data, and according to the formula Calculate color difference value;

[0057] In this embodiment, the color difference aging grade is determined by referring to the aging characteristic determination method of GB / T 1766-2008 "Rating method for aging of paint and varnish coatings" and ISO4628 "Rating method for aging of paint and varnish coatings".

[0058] The present invention uses a colorimeter to measure chromaticity coordinates and calculate ΔΔE. Part of the results are shown in Table 1.

[0059] Table 1 Chromaticity coordinates and ΔE results of coating before and after corrosion

[0060]

[0061]

[0062] Step 3, using image annotation tools and computer vision algorithms, marking and segmenting the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set, and forming a corrosion image database corresponding to the color difference measurement positions;

[0063] Specifically, the locations for color difference measurement are marked on each image to ensure that these locations are consistent in the images before and after corrosion.

[0064] The segmentation method includes one or more combinations of cropping, image segmentation algorithm, image annotation tool, and masking method. The measurement area is marked using an image annotation tool (such as a Labelme annotation tool), and the pixel characteristics of the marked area can be referred to, and the color difference measurement area can be extracted and segmented in the original image using code.

[0065] In this embodiment, the Labelme annotation tool is used to mark the color difference measurement position and obtain the corresponding label image. Among them, the area size of a single color difference measurement position point of the X·rite colorimeter is a circular area with a diameter of 453 pixels. In this embodiment, the labeled label image and the corresponding original image are used as input, and the segmentation and extraction of the marked area in the original image are performed using the self-written Matlab code to obtain the macro image of the color difference measurement position in the standardized image. Figure 3 The macroscopic morphological images of the segmented and extracted color difference measurement positions of 10 groups of samples after different corrosion tests are shown.

[0066] In this embodiment, according to the label image obtained by the Labelme annotation tool, the pixel value of the color difference measurement area is displayed as 38, and the pixel value of the non-measurement area is 0. The code sets a mask in the original image according to the area with a pixel value of 0 in the label image, and only extracts the area with a pixel value of 38, completing the segmentation extraction corresponding to the specified position.

[0067] Step 4, extract the R, G, B component value data of each image in the corrosion image database, and complete the calculation of ΔR, ΔG, ΔB of the images before and after corrosion; in the step 4, define the R, G, B component value data in the optical image and the data changes such as ΔR, ΔG, ΔB as feature description information.

[0068] Specifically, the code for extracting the R, G, and B component value data of the image in step 4 can be completed by a variety of programming tools, including but not limited to Python and Matlab. In the process of using the code to complete the extraction of the R, G, and B component values ​​of the image, each image read from the database is separated into three independent channels: R (red channel), G (green channel), and B (blue channel). For each channel, each pixel is traversed to complete the extraction of the component value.

[0069] In this embodiment, Matlab programming tool is selected to compile code to complete the extraction of R, G, and B component value data of the macro image. The output result includes the R, G, and B values ​​of each image, and the corresponding data results before and after corrosion are sorted, and the obtained data are shown in Table 2.

[0070] Table 2 Results of R, G, and B component values ​​before and after coating corrosion

[0071]

[0072] Step 5: Based on the machine learning algorithm, with ΔR, ΔG, and ΔB as independent variables and the color difference value ΔE as the dependent variable, a color difference prediction model based on color is built and trained, and the model performance is evaluated;

[0073] For example, in step 5, the machine learning algorithm includes linear regression, polynomial regression, ridge regression, LASSO regression and neural network prediction algorithm, etc., and the model performance is improved by optimizing training methods such as data augmentation to realize color model training and quantification of coating color characteristics and determination of aging level.

[0074] In step 5, the ΔR, ΔG, and ΔB calculated in step 4 are used as independent variables, and the color difference value ΔE calculated in step 2 is used as the dependent variable, and matching is performed, and the training set and the test set are divided according to the set ratio. The training set is used to iteratively train the color difference prediction model based on color, and the test set is used to evaluate the model performance. In order to make the model training results more representative, the database data is randomly allocated and divided according to the ratio of training set: test set = 3:1.

[0075] In the step 5, the model performance is evaluated, including: using the test set to predict the trained color difference prediction model based on color to obtain a prediction result, and evaluating the model performance based on multiple evaluation indicators, wherein the multiple evaluation indicators include mean square error, root mean square error, determination coefficient, and mean absolute error;

[0076] The mean square error expression is: Among them, y i is the observed value, is the model prediction value, and n represents the number of sample sets. MSE measures the mean of the sum of squares of the differences between the predicted value and the actual value. The smaller the MSE, the better the model.

[0077] The root mean square error expression is: RMSE is the square root of MSE and represents the difference between the predicted value and the actual value.

[0078] The determination coefficient expression is: in, is the mean of the observed values, R 2 The proportion of data variability that the model can explain, R 2 The closer it is to 1, the stronger the explanatory power of the model.

[0079] The mean absolute error expression is: MAE measures the average of the absolute values ​​of the differences between the predicted values ​​and the actual values, which reflects the average level of model prediction error.

[0080] In this embodiment, a total of 42 groups of initial image data were collected without any corrosion test, and a total of 576 groups of image data were obtained after salt spray test, UV aging test, multi-factor coupling accelerated test and natural environment corrosion data of different lengths. The data were arranged and combined to obtain a total of 24,192 groups of data corresponding to the areas before and after corrosion. The ΔR, ΔG, and ΔB obtained by the above calculations were used as independent variables, and the color difference value ΔE was used as the dependent variable. The data were randomly assigned and divided into training set and test set in a ratio of 3:1, and a color difference prediction model was established through iterative training. Among them, the training set data is used for color difference prediction model training, and the test set data is used for model evaluation.

[0081] In this embodiment, the mean square error MSE and the determination coefficient R are selected from the error evaluation indicators. 2 Two indicators are used to evaluate the model. The 18144 sets of training data are input into the network for model training. After the model training is completed, the trained model is saved. Among them, the R corresponding to the training set data 2 =0.972, MSE = 0.262, and the model training results are as follows Figure 4As shown. After that, the test set data is used as input and the saved prediction model is called to evaluate the model. Among them, the R corresponding to the test set data is 2 =0.969, MSE = 0.344, and the model prediction results are as follows Figure 5 shown.

[0082] In addition, the model performance evaluation also includes: for the test set, the color difference aging level determined by S2 is used as the accurate result, and the color difference aging level determined based on the prediction result is compared to determine the accuracy of the color-based color difference prediction model in evaluating the color characteristics of the coating.

[0083] Specifically, based on the aging characteristic color grade assessment standard, the color difference prediction value Pre(ΔE) of the color-based color difference prediction model is used to perform aging grade assessment to determine the accuracy of the quantitative results of the coating color characteristics;

[0084] In this embodiment, the aging characteristic determination method of GB / T 1766-2008 "Rating method for aging of paint and varnish coatings" and ISO4628 "Rating method for aging of paint and varnish coatings" is referred to, and the quantitative results predicted by the model are used to evaluate the aging grade.

[0085] In this embodiment, 200 data of the test set are randomly selected to determine the accuracy of the model. The aging grade result measured by the colorimeter is taken as the most accurate result, and the aging grade of the quantitative result predicted by the model is compared. Among them, a total of 194 data are completely consistent, with an accuracy rate of about 97%.

[0086] Step 6: for the coating to be evaluated, at least one set of standard target images before and after corrosion is collected, and the image ΔR, ΔG, and ΔB data are extracted and used as input, and the color difference prediction is performed using the trained color-based color difference prediction model to obtain the coating color difference value ΔE prediction result;

[0087] In this embodiment, the present invention uses a set of standardized image data of the Q235 / epoxy zinc-rich / acrylic polyurethane coating system before the test and after 8 weeks of salt spray test as the target image to complete image segmentation and obtain ΔR, ΔG, and ΔB data. The results are as follows Figure 6 shown. Figure 6 (a1) is the initial aging feature image. Figure 6 (b1) is the aging characteristic image corresponding to the corrosion test. Afterwards, ΔR, ΔG, and ΔB are used as inputs to the color difference prediction model based on color to obtain the color difference prediction value.

[0088] In this embodiment, the ΔR, ΔG, and ΔB data corresponding to the group of images are 0.68, -3.19, and -8.45, respectively. When they are input into the model, the color difference prediction value ΔE=3.68 is obtained.

[0089] Step 7, match the model prediction result ΔE with the aging judgment level, store the output result in the database, establish a data association relationship, output the corresponding aging feature level, and realize aging level judgment.

[0090] In this embodiment, the present invention matches the color difference prediction value according to the grade judgment criteria in the relevant standards and regulations, and outputs the composite coating after 8 weeks of salt spray test, with reference to the aging characteristic judgment mark of GB / T 1766-2008 "Rating Method for Aging of Paint and Varnish Coatings" and ISO 4628 "Rating Method for Aging of Paint and Varnish Coatings", and the color aging grade is rated as 2. Comparative judgment shows that the aging grade determined by the instrumental determination method matches the aging characteristic grade of this method.

[0091] The aging status of the epoxy zinc-rich primer and polyurethane topcoat system with a total thickness of about 120μm in Qingdao’s marine atmospheric environment, as well as the color aging characteristics corresponding to the salt spray test, UV aging test, and multi-factor coupled accelerated corrosion test are saved in the database for subsequent query and analysis.

[0092] The present invention also discloses a coating aging color feature evaluation system based on machine learning, which is applied to execute the above-mentioned coating aging color feature evaluation method based on machine learning, comprising:

[0093] An image acquisition unit is used to acquire macroscopic morphological images of the coating sample during the aging process before and after corrosion from different angles to form a standardized aging image data set; it is also used to acquire standard target images of the coating to be evaluated before and after corrosion;

[0094] The color difference measurement unit is used to measure the color difference of the coating sample, obtain the chromaticity coordinate value data of the coating sample before and after corrosion, calculate the color difference value ΔE, and determine the color difference aging grade based on the national standard and ISO standard;

[0095] An image matching unit is used to mark and segment the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set, so as to form a corrosion image database corresponding to the color difference measurement positions one by one;

[0096] The image processing unit extracts the R, G, and B component value data of each image in the corrosion image database and completes the calculation of ΔR, ΔG, and ΔB of the images before and after corrosion; it is also used to extract the R, G, and B data of the standard target image and calculate the ΔR, ΔG, and ΔB corresponding to the standard target image;

[0097] The model training unit is used to build and train a color-based color difference prediction model based on a machine learning algorithm, taking ΔΔR, ΔG, and ΔB as independent variables and the color difference value ΔE as a dependent variable, and to evaluate the model performance;

[0098] The model prediction unit is used to use the trained color-based color difference prediction model to predict the color difference, obtain the color difference value prediction result of the coating to be evaluated, and perform aging grade assessment based on the aging characteristic color grade assessment standard.

[0099] Wherein, the image acquisition unit comprises a standardized image acquisition device, and the standardized image acquisition device acquires images under the condition of a light-proof darkroom and a uniform fixed light source;

[0100] The color difference measurement unit includes at least one of the tools selected from the group consisting of a colorimeter and a spectrophotometer, which can measure the L, a, and b data of the coating sample before and after aging.

[0101] The coating aging color feature evaluation system based on machine learning also includes a database module for storing chemical formulas of known coating compositions and related information such as standardized macroscopic aging images, coating thickness, coating time, and location that match the coating, as well as for storing and collecting aging features and color difference aging data to realize the storage of basic data and the sorting and mining of data association relationships.

[0102] Regarding the system in the above embodiment, the specific manner in which each unit and module performs operations has been described in detail in the embodiment of the method, and will not be elaborated again here.

[0103] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the software portion of the logic in the coating aging color characteristic evaluation method based on machine learning as described above is implemented.

[0104] Matters not covered by the present invention are known technologies.

[0105] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A coating aging color feature evaluation method based on machine learning, characterized in that: The following steps are involved: S1, macroscopic morphological images of coating samples during the aging process before and after corrosion are collected from different angles to form a standardized aging image dataset; S2, measure the color difference of the coating sample, obtain the chromaticity coordinate value data of the coating sample before and after corrosion, calculate the color difference value ΔE, and determine the color difference aging grade based on the national standard and ISO standard; S3, marking and segmenting the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set to form a corrosion image database corresponding to the color difference measurement positions; S4, extracting the R, G, and B component value data of each image in the corrosion image database, and completing the calculation of ΔR, ΔG, and ΔB of the images before and after corrosion; S5, based on the machine learning algorithm, with ΔR, ΔG, ΔB as independent variables and color difference value ΔE as dependent variable, build and train a color-based color difference prediction model, and evaluate the model performance; S6, collecting standard target images of the coating to be evaluated before and after corrosion, extracting ΔR, ΔG, and ΔB of the standard target image as input, using the trained color-based color difference prediction model to perform color difference prediction, obtaining the color difference value prediction result of the coating to be evaluated, and performing aging grade assessment based on the aging characteristic color grade assessment standard.

2. The coating aging color feature evaluation method based on machine learning according to claim 1 is characterized in that: In S1, a standardized image acquisition device is used to acquire macroscopic morphological images of the coating sample during the aging process before and after corrosion from different angles; The standardized image acquisition device acquires images in a dark room protected from light using a uniform fixed light source.

3. The coating aging color feature evaluation method based on machine learning according to claim 1 is characterized in that: In S1, the aging process includes natural aging and accelerated aging; The S1 also includes storing the standardized aging image data set into a database and recording associated data information, wherein the stored standardized aging image data set includes a chemical formula with a known coating composition and associated data information related to the coating, and the associated data information includes a standardized macroscopic image, coating thickness, coating time, location, coating corrosion test type and time.

4. The coating aging color feature evaluation method based on machine learning according to claim 1 is characterized in that: In S2, a traditional color difference measurement tool is used to measure the color difference of the coating sample, and the traditional color difference measurement tool includes at least one of a colorimeter and a spectrophotometer that can measure the L, a, and b data of the coating sample before and after aging; The chromaticity coordinate value data includes L, a, and b data, wherein L represents lightness, a represents red and green, and b represents yellow and blue; Calculate the color difference ΔE: According to the formula Calculate the color difference value.

5. The coating aging color feature evaluation method based on machine learning according to claim 1, characterized in that: In S3, using image annotation tools and computer vision algorithms, the macroscopic morphology images in the standardized aging image data set are marked and segmented for color difference measurement positions; The segmentation method includes one or a combination of cropping, image segmentation algorithm, image annotation tool, and masking method.

6. The coating aging color feature evaluation method based on machine learning according to claim 1, characterized in that: In S5, the machine learning algorithm includes one or more combinations of linear regression, polynomial regression, ridge regression, LASSO regression and neural network prediction algorithms.

7. The coating aging color feature evaluation method based on machine learning according to claim 1, characterized in that: In S5, ΔR, ΔG, and ΔB calculated by S4 are used as independent variables, and the color difference value ΔE calculated by S2 is used as the dependent variable, and matching is performed. The training set and the test set are divided into a training set and a test set according to a set ratio. The training set is used to iteratively train the color-based color difference prediction model, and the test set is used to evaluate the model performance.

8. The coating aging color feature evaluation method based on machine learning according to claim 7 is characterized in that: In S5, the model performance evaluation is performed, including: using the test set to predict the trained color difference prediction model based on color to obtain a prediction result, and evaluating the model performance based on multiple evaluation indicators, wherein the multiple evaluation indicators include mean square error, root mean square error, determination coefficient, and mean absolute error; The model performance evaluation also includes: for the test set, taking the color difference aging level determined by S2 as the accurate result, comparing the color difference aging level determined based on the prediction result, and determining the accuracy of the color difference prediction model based on color to evaluate the color characteristics of the coating.

9. A coating aging color feature evaluation system based on machine learning, applied to execute the coating aging color feature evaluation method based on machine learning according to any one of claims 1 to 8, characterized in that: include: An image acquisition unit is used to acquire macroscopic morphological images of the coating sample during the aging process before and after corrosion from different angles to form a standardized aging image data set; it is also used to acquire standard target images of the coating to be evaluated before and after corrosion; The color difference measurement unit is used to measure the color difference of the coating sample, obtain the chromaticity coordinate value data of the coating sample before and after corrosion, calculate the color difference value AP, and determine the color difference aging grade based on the national standard and ISO standard; An image matching unit is used to mark and segment the color difference measurement positions of the macroscopic morphology images in the standardized aging image data set, so as to form a corrosion image database corresponding to the color difference measurement positions one by one; The image processing unit extracts the R, G, and B component value data of each image in the corrosion image database and completes the calculation of ΔR, ΔG, and ΔB of the images before and after corrosion; it is also used to extract the R, G, and B data of the standard target image and calculate the ΔR, ΔG, and ΔB corresponding to the standard target image; The model training unit is used to build and train a color-based color difference prediction model based on a machine learning algorithm, using ΔR, ΔG, and ΔB as independent variables and the color difference value ΔE as a dependent variable, and to evaluate the model performance; The model prediction unit is used to use the trained color-based color difference prediction model to predict the color difference, obtain the color difference value prediction result of the coating to be evaluated, and perform aging grade assessment based on the aging characteristic color grade assessment standard.

10. The coating aging color feature evaluation system based on machine learning according to claim 9, characterized in that: The image acquisition unit includes a standardized image acquisition device, which acquires images in a dark room with a uniform fixed light source. The color difference measurement unit includes at least one of the tools selected from the group consisting of a colorimeter and a spectrophotometer, which can measure the L, a, and b data of the coating sample before and after aging.

Images