A method and system for measuring the color of environmentally friendly paint
Through the combination of image data and external environmental data, the VOCS release rate model and multivariate linear regression model are used to solve the multi-angle measurement and detection speed problems of environmentally friendly paint color measurement, and the rapid and accurate measurement of environmentally friendly paint color and Consistency detection improves detection efficiency and environmental performance.
Patent Information
- Application Number
- CN202410920311.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-10
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-07-10
AI Technical Summary
The existing color measurement methods for environmentally friendly paint paints have difficulties in multi-angle measurement and color consistency detection, and require professional instruments. The detection speed is slow and the flexibility is poor, and it is impossible to achieve fast and accurate color measurement, especially for the color of metal paints. The measurement effect is not good.
By acquiring image data and external environment data, the VOCS release rate model, multivariate linear regression model and feedforward neural network model are used, combined with image processing at different light intensities, the interference scattering effect is identified to achieve accurate measurement of the color of environmentally friendly paints.
It realizes rapid and accurate measurement of the color of environmentally friendly paints, improves detection efficiency and consistency, reduces dependence on professional instruments, and enhances the balance of environmentally friendly performance.
Smart Images

Figure CN118655094B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image data processing, and in particular to a method and system for measuring the color of environmentally friendly coatings. Background Art
[0002] From the perspective of color measurement methods for environmentally friendly paints, today's color measurement of environmentally friendly paints is usually based on image threshold segmentation technology. When faced with large color consistency or small differences, it is difficult to find color inconsistencies. Especially for the color measurement of environmentally friendly paints and metallic paints, facing the difficulties of multi-angle measurement and the need for professional instruments, the existing methods cannot achieve rapid color measurement of environmentally friendly paints and metallic paints, and the environmental performance is poorly taken into account during the color measurement process. Today's color measurement methods for environmentally friendly paints not only require the support of accurate mathematical models, but also have problems such as slow detection speed, poor flexibility, and poor versatility. In the context of the rapid development of intelligence, the color measurement of environmentally friendly paints is gradually developing in the direction of complexity and intelligence. Summary of the invention
[0003] In order to overcome the disadvantages of difficulty in multi-angle measurement and high instrument dependence, the present invention provides a method and system for measuring the color of environmentally friendly coatings.
[0004] The technical implementation scheme of the present invention is: a method for measuring the color of environmentally friendly coatings, comprising the following steps:
[0005] S1: Acquire image data and external environment data;
[0006] S2: Obtaining a first inference result according to the image data;
[0007] S3: Obtaining a first judgment result according to the first inference result and the external environment data;
[0008] S4: selecting a first captured image and a second captured image according to the first judgment result;
[0009] S5: Acquire a first confrontation image according to the first captured image and the second captured image;
[0010] S6: Acquire environmentally friendly paint color measurement information according to the first adversarial image;
[0011] S7: Determine the environmentally friendly paint color measurement result according to the environmentally friendly paint color measurement information.
[0012] Preferably, the acquisition of image data and external environment data includes: acquiring the image at the current moment t1, and acquiring the paint images at intervals t2, t3, t4, t5..., and acquiring an image inconsistent with the image at moment t1 based on the paint images at intervals t2, t3, t4, t5...; acquiring the external environment data at the current moment t1, and acquiring the external environment data at intervals t2, t3, t4, t5..., and acquiring the time when the temperature difference changes with the current moment t1; based on the image inconsistent with the image at moment t1 and the time when the temperature difference changes at the current moment t1, acquiring image data and temperature difference change data at the moment when the image changes with the moment t1 and the temperature difference changes.
[0013] Preferably, obtaining the first inference result according to the image data includes: assuming that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color, obtaining the VOCS release amount according to the VOCS release rate model, and the mathematical expression of the VOCS release amount is: ,in, : represents the release rate of VOCS; : is a pre-exponential factor related to materials and environmental conditions, including parameters related to material properties and boundary layer mass transfer; ( ): The saturated VOCS concentration inside the material, that is, the maximum VOCS concentration reached by the material; : The concentration of VOCS on the surface of the material or in the environment changes over time, reflecting the decrease in concentration caused by the release of VOCS; : Release kinetic index, reflecting the nonlinear characteristics of the release process; ( ): Activation energy, usually in joules per mole (J / mol), represents the energy barrier that VOCS molecules must overcome to transfer from the interior of the material to the external environment; : Ideal gas constant, with a value of 8.314 J / (mol·K), is a physical constant; : Absolute temperature, measured in Kelvin (K), affects the kinetic energy and release rate of VOCS molecules.
[0014] Preferably, obtaining the first judgment result according to the first inference result and the external environment data includes: analyzing the cause of color change in the environmentally friendly paint image according to the VOCS release amount combined with the change of temperature increase to obtain the first judgment result; specifically: calculating the VOCS release amount according to the VOCS release amount calculation formula and the temperature change data in the external environment data; then collecting data of a large number of paint samples; and using a multivariate linear regression model to simplify the prediction of VOCS release amount ( ) and color parameters ( , , ), the model is expressed as: ,in,( ) is the VOC emission, ( ), ), ) are color parameters, ( , , , ) are model parameters, ( ) is the error term; finally, through the feedforward neural network model, the input layer receives the paint formula and environmental factors, the hidden layer performs complex nonlinear transformation, and the output layer gives the predicted color change value ( , , ), the formula is: ,in,( ) represents the function of the neural network, ( ) are the variables of the input layer.
[0015] Preferably, the selecting the first captured image and the second captured image according to the first judgment result includes: determining whether to shoot according to the first judgment result, the first captured image and the second captured image are captured images of adjacent parts of the same area under different light intensities, the captured images of adjacent parts of the area under the first light intensity are recorded as the first captured image and the second captured image, and the captured images of adjacent parts of the area under the second light intensity are also recorded as the first captured image and the second captured image, the first light intensity and the second light intensity are: among the light intensities that cause interference and scattering of the environmentally friendly paint, the light intensity with a large light intensity is selected as the first light intensity, and the light intensity with a light intensity less than the first light intensity is selected as the second light intensity, and the relationship between the first captured image and the second captured image is determined by taking the camera in the area as the center, and the part close to the left side of the area is the second light intensity. A shooting area, the right side of the first shooting area is the second shooting area; specifically: if the reason for the image color change is the change in VOCS temperature, which leads to an increase in the release amount and thus causes the color of the environmentally friendly paint to change, then the first shooting image and the second shooting image are no longer selected; otherwise, the image is preprocessed, specifically using Gaussian filtering to reduce the noise in the image; converting the RGB image to the HSV image; selecting a seed point, and expanding the area according to the color similarity criterion until a significantly different color is encountered or the preset condition is met; applying the region growing algorithm, setting the color similarity threshold, specifically: 1, initialization: select one or more pixels as seed points, these points are located inside the region of interest; 2, similarity judgment: define a neighborhood, for each neighborhood pixel of the seed point, calculate the similarity between the neighborhood pixel color and the seed point color; similarity measurement formula: ,in,( )and( ) are the seed point and the ( ) component values, ( ) is the color space dimension; 3. Expanding the area: if the similarity of the neighborhood pixels is less than the preset threshold, the neighborhood pixels are marked as part of the segmented area, and the neighborhood pixels are added to the list of pixels to be checked, and step 2 is repeated; 4. Stop condition: when no more pixels are added to the area, the growth process of the seed point is ended; multiple seed points are selected for growth separately until the entire image is covered or a specific goal is achieved; finally, each area is marked or numbered to distinguish the first shooting area from the second shooting area.
[0016] Preferably, the method of obtaining the first adversarial image according to the first captured image and the second captured image includes: determining the area where the interference scattering effect occurs according to the first captured image and the second captured image, determining the area where the interference scattering effect affects each other according to the area where the interference scattering effect occurs, and determining the first adversarial image according to the effect of the area that affects each other under different illumination intensities; specifically, firstly denoising and graying the images of the first captured area and the second captured area captured under two different illumination conditions; converting the image from the spatial domain to the frequency domain using Fourier transform; identifying features related to the interference scattering effect in the frequency domain, such as the intensity difference of specific frequency components; using the image segmentation algorithm Otsu threshold method based on the frequency domain features or in combination with the feature color histogram of the spatial domain to identify and mark the area with inconsistent colors; optimizing the segmentation results, removing noise and misclassified areas, and ensuring smooth and accurate boundaries; the two-dimensional discrete Fourier transform formula is: ,in,( ) is the spatial domain pixel value of the original image, ( ) is the complex coefficient in the frequency domain, ( )and( ) are the width and height of the image respectively;
[0017] •The formula used by Otsu threshold method is: ,in,( ) is the gray level ( ) of pixels, ( ) is the total number of gray levels, ( , ) are the average grayscale values of the two types of pixels, ( ) is the global average gray value, ( ) is the between-class variance, by maximizing ( ) to find the optimal threshold ( ).
[0018] Preferably, the first adversarial image includes: observing the metal paint surface with different light intensities, recording the light intensity, angle and physical properties of the metal paint; using a camera to capture the distribution of reflected light intensity on the metal paint surface; extracting features from the collected data, including but not limited to light intensity, observation angle, and reflected light intensity; standardizing all numerical features to facilitate model training; constructing new optical path difference features based on optical theory; using a support vector machine model to distinguish whether there is significant interference or scattering phenomenon under different light intensities; and then using a polynomial regression model to predict the intensity of reflected light and the intensity of interference scattering effects under specific conditions; dividing the data set into a training set and a test set, using the training set data to train the model, evaluating the generalization ability of the model through the K-fold cross-validation method, and selecting network search for hyperparameter tuning.
[0019] Preferably, the method of obtaining the color measurement information of the environmentally friendly paint according to the first adversarial image includes: according to the first adversarial image obtained under the first light intensity and the second light intensity, changing the light intensity one by one, recording the interference fringes or the area with the most obvious color change under different light intensities, and gradually narrowing the range of the multiple first adversarial images obtained thereby, and determining the inconsistent color area of the environmentally friendly paint; specifically: based on the random forest, isolating the data points by constructing a series of random decision trees; the data points that are easily isolated are considered abnormal; the abnormal score is calculated by the average path length of the tree: the calculation formula is: ,in,( ) represents the sample ( )’s abnormal probability score, ( ) is the sample ( ) is the path length in the decision tree, ( ) is the given sample number ( ) is the average path length.
[0020] Preferably, the method of determining the color measurement result of the environmentally friendly paint type based on the environmentally friendly paint color measurement information includes: presetting a paint color consistency estimation index threshold based on experience, if the paint color consistency estimation index is less than the paint color consistency estimation index threshold, it means that the color of the paint is consistent; conversely, when the paint color consistency estimation index is greater than the paint color consistency estimation index threshold, the color of the paint is inconsistent.
[0021] An environmentally friendly paint color measurement system, comprising:
[0022] Image data acquisition and preprocessing module: used to collect and preprocess environmentally friendly paint image data;
[0023] External environment data collection and preprocessing module: used to collect external environment data of environmentally friendly paint and perform preprocessing;
[0024] The first inference result building module: used to assume that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color;
[0025] The first judgment result building module is used to verify the first inference result;
[0026] A first captured image acquisition module: used for capturing images under a first light intensity;
[0027] A second captured image acquisition module: used for capturing images under a second light intensity;
[0028] The first adversarial image acquisition module is used to find abnormal images in the first shooting area and the second shooting area images taken under different lighting conditions;
[0029] VOCS release rate model building module: used to calculate the VOCS release amount;
[0030] Multiple linear regression model building module: used to predict VOCS release ( ) and color parameters ( , , )
[0031] Feedforward neural network model building module: the input layer receives the paint formula and environmental factors, and the hidden layer performs complex nonlinear transformations.
[0032] Beneficial effects: The present invention compares the image within the interval with the temperature difference change to obtain the image data and the temperature difference data when the image and the temperature difference change simultaneously within the interval; the cause of the image change is considered to be the change in the chemical properties of the coating under the temperature difference change, thereby changing the color of the image. First, it is assumed that the chemical properties of the coating are changed under high temperature, thereby changing the color of the coating, and the VOCS release of the coating at high temperature is calculated, and combined with the temperature difference change, it is analyzed whether the cause of the color change of the coating is the influence of high temperature; if it is the influence of high temperature, color measurement is no longer performed, otherwise color measurement is continued; the coating color consistency measurement result is obtained by analyzing the image threshold segmentation results under different light intensities and the interference and scattering conditions under different light intensities; and according to the continuously narrowing light irradiation range and light intensity, the area causing color inconsistency is accurately located; the accuracy and efficiency of color consistency measurement are achieved through the joint use of multiple algorithms. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 The present invention is a flow chart of a method and system for measuring the color of environmentally friendly coatings. DETAILED DESCRIPTION
[0034] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0035] Example 1: A method for measuring the color of environmentally friendly paints, such as Figure 1 As shown, the following steps are included:
[0036] S1: Acquire image data and external environment data;
[0037] S2: Obtaining a first inference result according to the image data;
[0038] S3: Obtaining a first judgment result according to the first inference result and the external environment data;
[0039] S4: selecting a first captured image and a second captured image according to the first judgment result;
[0040] S5: Acquire a first confrontation image according to the first captured image and the second captured image;
[0041] S6: Acquire environmentally friendly paint color measurement information according to the first adversarial image;
[0042] S7: Determine the environmentally friendly paint color measurement result according to the environmentally friendly paint color measurement information.
[0043] Acquiring image data and external environment data, including: acquiring the image at the current moment t1, and acquiring the paint images at intervals t2, t3, t4, t5..., and acquiring an image inconsistent with the image at moment t1 based on the paint images at intervals t2, t3, t4, t5...; acquiring the external environment data at the current moment t1, and acquiring the external environment data at intervals t2, t3, t4, t5..., and acquiring the time when the temperature difference changes with the current moment t1; acquiring image data and temperature difference change data at the time when the image changes with the moment t1 and the temperature difference changes based on the image inconsistent with the image at moment t1 and the time when the temperature difference changes at the current moment t1.
[0044] Among them, the times t1, t2, t3, t4, t5... are counted from the current time t1, and the intervals t2, t3, t4, t5 are intervals of 1 second, 2 seconds, 1 minute, 2 minutes or other intervals.
[0045] Obtaining a first inference result according to the image data includes: assuming that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color, obtaining the VOCS release amount according to the VOCS release rate model, and the mathematical expression of the VOCS release amount is: ,in, : represents the release rate of VOCS; : is a pre-exponential factor related to materials and environmental conditions, including parameters related to material properties and boundary layer mass transfer; ( ): The saturated VOCS concentration inside the material, that is, the maximum VOCS concentration reached by the material; : The concentration of VOCS on the surface of the material or in the environment changes over time, reflecting the decrease in concentration caused by the release of VOCS; : Release kinetic index, reflecting the nonlinear characteristics of the release process; ( ): Activation energy, usually in joules per mole (J / mol), represents the energy barrier that VOCS molecules must overcome to transfer from the interior of the material to the external environment; : Ideal gas constant, with a value of 8.314 J / (mol·K), is a physical constant; : Absolute temperature, measured in Kelvin (K), affects the kinetic energy and release rate of VOCS molecules.
[0046] Among them, under high temperature conditions outdoors, high temperature will change the chemical balance of the paint film and cause VOCS release, which in turn will cause color instability; the existing measurement method for high temperature changing the chemical balance of the paint film and causing VOCS release, which in turn will cause color instability has inaccurate volume and flow rate monitoring data of VOCS release, and thus inaccurately judges whether the change in paint color is caused by high temperature causing VOCS release, which in turn affects the color change of the paint.
[0047] The first judgment result is obtained according to the first inference result and the external environment data, including: analyzing the cause of color change in the environmentally friendly paint image according to the VOCS release amount combined with the change of temperature increase to obtain the first judgment result; specifically: calculating the VOCS release amount according to the VOCS release amount calculation formula and the temperature change data in the external environment data; then collecting a large number of paint sample data; and using a multivariate linear regression model to simplify the prediction of VOCS release amount ( ) and color parameters ( , , ), the model is expressed as: ,in,( ) is the VOC emission, ( ), ), ) are color parameters, ( , , , ) are model parameters, ( ) is the error term; finally, through the feedforward neural network model, the input layer receives the paint formula and environmental factors, the hidden layer performs complex nonlinear transformation, and the output layer gives the predicted color change value ( , , ), the formula is: ,in,( ) represents the function of the neural network, ( ) are the variables of the input layer.
[0048] Selecting a first captured image and a second captured image according to a first judgment result includes: determining whether to shoot according to the first judgment result, the first captured image and the second captured image are captured images of adjacent parts of the same area under different light intensities, the captured images of adjacent parts of the area under the first light intensity are recorded as the first captured image and the second captured image, and the captured images of adjacent parts of the area under the second light intensity are also recorded as the first captured image and the second captured image, the first light intensity and the second light intensity are: among the light intensities that cause interference and scattering of the environmentally friendly paint, a light intensity with a large light intensity is selected as the first light intensity, and a light intensity with a light intensity less than the first light intensity is selected as the second light intensity, and the relationship between the first captured image and the second captured image is determined by taking the camera in the area as the center, and the part close to the left side of the area is the first light intensity. Shooting area, the right side of the first shooting area is the second shooting area; specifically: if the reason for the image color change is the change in VOCS temperature, which leads to an increase in the release amount and thus causes the color of the environmentally friendly paint to change, then the first shooting image and the second shooting image are no longer selected; otherwise, the image is preprocessed, specifically using Gaussian filtering to reduce the noise in the image; converting the RGB image to the HSV image; selecting the seed point, and expanding the area according to the color similarity criterion until a significantly different color is encountered or the preset condition is met; applying the region growing algorithm, setting the color similarity threshold, specifically: 1, initialization: select one or more pixels as seed points, these points are located inside the region of interest; 2, similarity judgment: define a neighborhood, for each neighborhood pixel of the seed point, calculate the similarity between its color and the seed point color; similarity measurement formula: ,in,( )and( ) are the seed point and the ( ) component values, ( ) is the color space dimension (for example, in HSV space (n=3)); 3. Expanding the area: if the similarity of the neighborhood pixels is less than a preset threshold, the neighborhood pixels are marked as part of the segmented area, and the neighborhood pixels are added to the list of pixels to be checked, and step 2 is repeated; 4. Stop condition: when no more pixels are added to the area, the growth process of the seed point is ended; multiple seed points are selected for growth separately until the entire image is covered or a specific goal is achieved; finally, each area is marked or numbered to distinguish the first shooting area from the second shooting area.
[0049] The first adversarial image is obtained according to the first captured image and the second captured image, including: determining the area where the interference scattering effect occurs according to the first captured image and the second captured image, determining the area where the interference scattering effect affects each other according to the area where the interference scattering effect occurs, and determining the first adversarial image according to the effect of the area that affects each other under different illumination intensities; specifically: firstly, denoising and graying the first captured area and the second captured area images captured under two different illumination conditions; using Fourier transform to convert the image from the spatial domain to the frequency domain; identifying features related to the interference scattering effect in the frequency domain, such as the intensity difference of specific frequency components; based on the frequency domain features or in combination with the feature color histogram of the spatial domain, using the image segmentation algorithm Otsu threshold method to identify and mark the area with inconsistent colors; optimizing the segmentation result, removing noise and misclassified areas, and ensuring smooth and accurate boundaries; the two-dimensional discrete Fourier transform formula is: ,in,( ) is the spatial domain pixel value of the original image, ( ) is the complex coefficient in the frequency domain, ( )and( ) are the width and height of the image respectively;
[0050] •The formula used by Otsu threshold method is: ,in,( ) is the gray level ( ) of pixels, ( ) is the total number of gray levels, ( , ) are the average grayscale values of the two types of pixels, ( ) is the global average gray value, ( ) is the between-class variance, by maximizing ( ) to find the optimal threshold ( ).
[0051] The first adversarial image includes: observing the metal paint surface with different light intensities, recording the light intensity, angle and physical properties of the metal paint; using a camera to capture the distribution of reflected light intensity on the metal paint surface; extracting features from the collected data, including but not limited to light intensity, observation angle, and reflected light intensity; standardizing all numerical features to facilitate model training; constructing new optical path difference features based on optical theory; using a support vector machine model to distinguish whether there is significant interference or scattering under different light intensities; and then using a polynomial regression model to predict the intensity of reflected light and the intensity of interference and scattering effects under specific conditions; dividing the data set into a training set and a test set, using the training set data to train the model, evaluating the generalization ability of the model through the K-fold cross-validation method, and selecting network search for hyperparameter tuning.
[0052] Among them, the formula and theoretical basis are, Interference: When two or more light waves meet on the surface of metallic paint, their peaks and troughs overlap, resulting in an increase or decrease in light intensity. The interference effect is understood using the basic principle of Young's double-slit experiment, where the visibility of the interference fringes depends on the optical path difference (ΔL=nλ, n is an integer, λ is the wavelength). Scattering: The phenomenon of light changing direction when encountering uneven surfaces or particles in metallic paint is described by sharp scattering or Mie scattering theory, where the scattering intensity is inversely proportional to the fourth power of the wavelength of the incident light ( ).
[0053] Obtaining the color measurement information of the environmentally friendly paint according to the first adversarial image, including: according to the first adversarial image obtained under the first light intensity and the second light intensity, changing the light intensity one by one, recording the interference fringes or the area with the most obvious color change under different light intensities, and gradually narrowing the range of the multiple first adversarial images obtained thereby, and determining the inconsistent color area of the environmentally friendly paint; specifically: based on the random forest, isolating the data points by constructing a series of random decision trees; the data points that are easily isolated are considered to be abnormal; the abnormal score is calculated by the average path length of the tree: the calculation formula is, ,in,( ) represents the sample ( )’s abnormal probability score, ( ) is the sample ( ) is the path length in the decision tree, ( ) is the given sample number ( ) is the average path length.
[0054] Determining the color measurement result of the environmentally friendly paint according to the color measurement information of the environmentally friendly paint, including: presetting the paint color consistency estimation index threshold according to experience, if the paint color consistency estimation index is less than the paint color consistency estimation index threshold, it means that the color of the paint is consistent; conversely, when the paint color consistency estimation index is greater than the paint color consistency estimation index threshold, the color of the paint is inconsistent.
[0055] Embodiment 2: Based on embodiment 1, a system for measuring the color of environmentally friendly coatings includes:
[0056] Image data acquisition and preprocessing module, used to acquire and preprocess environmentally friendly paint image data;
[0057] External environment data collection and preprocessing module, used to collect and preprocess the external environment data of environmentally friendly paint;
[0058] The first inference result building module is used to assume that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color;
[0059] A first judgment result building module, used to verify the first inference result;
[0060] A first captured image acquisition module, used for capturing images under a first light intensity;
[0061] A second captured image acquisition module, used for capturing images under a second light intensity;
[0062] A first adversarial image acquisition module is used to find abnormal images in the first shooting area and the second shooting area images taken under different lighting conditions;
[0063] VOCS release rate model building module, used to calculate the VOCS release amount;
[0064] Multiple linear regression model building module for predicting VOCS release ( ) and color parameters ( , , )
[0065] Feedforward neural network model building module, where the input layer receives the paint formula and environmental factors, and the hidden layer performs complex nonlinear transformations.
[0066] The above is a detailed introduction to the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A method for measuring the color of environmentally friendly paint, characterized in that: The following steps are involved: S1: Acquire image data and external environment data; S2: Obtaining a first inference result according to the image data; S3: Obtaining a first judgment result according to the first inference result and the external environment data; S4: selecting a first captured image and a second captured image according to the first judgment result; S5: Acquire a first confrontation image according to the first captured image and the second captured image; S6: Acquire environmentally friendly paint color measurement information according to the first adversarial image; S7: Determine the environmentally friendly paint color measurement result according to the environmentally friendly paint color measurement information; The acquisition of image data and external environment data includes: acquiring the image at the current moment t1, and acquiring the paint images at intervals t2, t3, t4, t5..., and acquiring an image inconsistent with the image at moment t1 according to the paint images at intervals t2, t3, t4, t5...; acquiring the external environment data at the current moment t1, and acquiring the external environment data at intervals t2, t3, t4, t5..., and acquiring the time when the temperature difference changes with the current moment t1; acquiring the image data and temperature difference change data at the time when the image changes with the moment t1 and the temperature difference changes according to the image inconsistent with the image at moment t1 and the time when the temperature difference changes at the current moment t1; The first inference result is obtained according to the image data, including: assuming that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color, and obtaining the VOCS release amount according to the VOCS release rate model. The mathematical expression of the VOCS release amount is: ,in, : represents the release rate of VOCS; : It is a pre-exponential factor related to materials and environmental conditions, and includes parameters related to material properties and boundary layer mass transfer; : The saturated VOCS concentration inside the material, that is, the maximum VOCS concentration reached by the material; : The concentration of VOCS on the surface of the material or in the environment changes over time, reflecting the decrease in concentration caused by the release of VOCS; : Release kinetic index, which reflects the nonlinear characteristics of the release process and depends on the release mechanism; : Activation energy, in joules per mole J / mol, represents the energy barrier that VOCS molecules must overcome to transfer from the interior of the material to the external environment; : Ideal gas constant, with a value of 8.314 J / (mol·K), is a physical constant; : Absolute temperature, in Kelvin K, affects the kinetic energy and release rate of VOCS molecules; The first judgment result is obtained according to the first inference result and the external environment data, including: analyzing the cause of color change in the environmentally friendly paint image according to the VOCS release amount combined with the change of temperature increase to obtain the first judgment result; specifically: calculating the VOCS release amount according to the VOCS release amount calculation formula and the temperature change data in the external environment data; then collecting a large number of paint sample data; and using a multivariate linear regression model to simplify the prediction of VOCS release amount. With color parameters , , The relationship between them is expressed as: ,in, is the VOC emission, , , They are color parameters, , , , are model parameters, is the error term; finally, through the feedforward neural network model, the input layer receives the paint formula and environmental factors, the hidden layer performs complex nonlinear transformation, and the output layer gives the predicted color change value , , , the formula is: ,in, A function representing a neural network, , is the variable of the input layer; The selecting of the first captured image and the second captured image according to the first judgment result includes: determining whether to shoot according to the first judgment result, the first captured image and the second captured image are captured images of adjacent parts of the same area under different light intensities, the captured images of the adjacent parts of the area under the first light intensity are recorded as the first captured image and the second captured image, and the captured images of the adjacent parts of the area under the second light intensity are also recorded as the first captured image and the second captured image, the first light intensity and the second light intensity are: among the light intensities that cause interference and scattering of the environmentally friendly paint, the light intensity with a large light intensity is selected as the first light intensity, and the light intensity with a light intensity less than the first light intensity is selected as the second light intensity, and the relationship between the first captured image and the second captured image is determined by taking the camera in the area as the center, and the part close to the left side of the area is the first captured image. The first shooting area is the second shooting area to the right of the first shooting area; specifically: if the reason for the image color change is the change in VOCS temperature, which leads to an increase in the release amount and thus causes the color of the environmentally friendly paint to change, then the first shooting image and the second shooting image are no longer selected; otherwise, the image is preprocessed, specifically using Gaussian filtering to reduce the noise in the image; converting the RGB image to the HSV image; selecting the seed point, and expanding the area according to the color similarity criterion until a significantly different color is encountered or the preset condition is met; applying the region growing algorithm, setting the color similarity threshold, specifically: 1, initialization: select one or more pixels as seed points, these points are located inside the region of interest; 2, similarity judgment: define a neighborhood, for each seed point's neighborhood pixels, calculate the similarity between the neighborhood pixel color and the seed point color; similarity measurement formula: ,in, and are the seed point and the pixel currently considered in the color space. Component value, is the color space dimension; 3. Expand the area: if the similarity of the neighboring pixels is less than the preset threshold, mark the neighboring pixels as part of the segmented area, add the neighboring pixels to the list of pixels to be checked, and repeat step 2; 4. Stop condition: when no more pixels are added to the area, end the growth process of the seed point; select multiple seed points to grow separately until the entire image is covered or a specific goal is achieved; finally, mark or number each area to distinguish the first shooting area from the second shooting area; The method of obtaining a first adversarial image based on a first captured image and a second captured image includes: determining an area where an interference scattering effect occurs based on the first captured image and the second captured image, determining an area where the interference scattering effect mutually affects each other based on the area where the interference scattering effect occurs, and determining the first adversarial image based on the effects of the mutually affected area under different illumination intensities; specifically, firstly denoising and graying the first captured area and the second captured area images captured under two different illumination conditions; converting the image from a spatial domain to a frequency domain using a Fourier transform; identifying features related to the interference scattering effect in the frequency domain, and using an image segmentation algorithm, an Otsu threshold method, to identify and mark areas with inconsistent colors based on frequency domain features and a feature color histogram combined with a spatial domain; and optimizing the segmentation result to remove noise and misclassified areas to ensure smooth and accurate boundaries.
2. A method for measuring the color of an environmentally friendly coating paint as claimed in claim 1, characterized in that: The method of obtaining the color measurement information of the environmentally friendly paint according to the first adversarial image includes: according to the first adversarial image obtained under the first light intensity and the second light intensity, changing the light intensity one by one, recording the interference fringes or the area with the most obvious color change under different light intensities, and gradually narrowing the range of the multiple first adversarial images obtained thereby, and determining the area with inconsistent color of the environmentally friendly paint; specifically: based on the random forest, a series of random decision trees are constructed to isolate data points; data points that are easily isolated are considered abnormal; and the abnormal score is calculated by the average path length of the tree: the calculation formula is, ,in, Representation sample The abnormal probability score of For sample The path length in the decision tree, is the given number of samples The average path length.
3. The method for measuring the color of an environmentally friendly coating paint as claimed in claim 1, characterized in that: The method of determining the environmentally friendly paint color measurement result based on the environmentally friendly paint color measurement information includes: presetting a paint color consistency estimation index threshold based on experience, if the paint color consistency estimation index is less than the paint color consistency estimation index threshold, it means that the color of the paint is consistent; conversely, when the paint color consistency estimation index is greater than the paint color consistency estimation index threshold, the color of the paint is inconsistent.
4. A measuring system for the method for measuring the color of environmentally friendly coating paints according to any one of claims 1 to 3, characterized in that: Included are: Image data acquisition and preprocessing module: used to collect and preprocess environmentally friendly paint image data; External environment data collection and preprocessing module: used to collect external environment data of environmentally friendly paint and perform preprocessing; The first inference result building module: used to assume that the change in temperature difference causes the change in VOCS release rate and thus causes the change in paint color; The first judgment result building module is used to verify the first inference result; A first captured image acquisition module: used for capturing images under a first light intensity; A second captured image acquisition module: used for capturing images under a second light intensity; The first adversarial image acquisition module is used to find abnormal images in the first shooting area and the second shooting area images taken under different lighting conditions; VOCS release rate model building module: used to calculate the VOCS release amount; Multiple linear regression model building module: used to predict VOCS release With color parameters , , the relationship between; Feedforward neural network model building module: the input layer receives the paint formula and environmental factors, and the hidden layer performs complex nonlinear transformations.
