Color palette surface appearance measuring method and related device

Abstract

The invention discloses a color palette surface appearance measurement method and a related device, and belongs to the technical field of appearance measurement, and the color palette surface appearance measurement method comprises the steps: calculating texture similarity, and marking a screened sample color palette picture as a first sample color palette picture based on the texture similarity; calculating color similarity, and marking the screened first sample color plate picture as a second sample color plate picture based on the color similarity; calculating particle similarity, and marking the screened second sample color plate picture as a third sample color plate picture based on the particle similarity; calculating the overall similarity between the to-be-tested color plate picture and the third sample color plate picture; and outputting a third sample palette picture corresponding to the to-be-tested palette picture and related matching information based on the overall similarity. The color palette surface appearance measurement method disclosed by the invention has the advantages that the accuracy and efficiency of color palette surface appearance analysis are improved, the overall appearance characteristics of the color palette can be comprehensively reflected, and more accurate appearance measurement is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of appearance measurement, and particularly to a method for measuring the surface appearance of color swatches and related devices. Background Art

[0002] In current color swatch management practices, finding items similar to a specific color often relies on manual experience and visual judgment. This method not only makes the search process complex and difficult, but also, due to the subjectivity of visual comparison, it is difficult to guarantee accuracy. In addition, the lack of quantitative means for measuring color differences also makes the evaluation of color similarity unreliable. Therefore, this process is not only inefficient but also quite challenging to execute. With the improvement of people's living standards, the demand for personalized product appearance is increasing day by day, and traditional appearance comparison methods are difficult to meet the needs of modern industrial production.

[0003] Through retrieval, we found that the patent with the publication number CN118427387A discloses a color swatch management method. This patent measures the color quantization value of an external color swatch by using a color quantization measurement instrument (such as a color difference meter) and conducts a quantitative difference comparison analysis with the existing color quantization values in the database.

[0004] However, although the above-mentioned method for measuring the surface appearance of color swatches can obtain accurate data when measuring a uniform surface, in the coating industry, the demand for customization is extremely extensive, and the product appearance shows extremely rich diversity, covering various types such as solid colors (homogeneous surfaces) and metallic colors (non-homogeneous surfaces), flat surfaces and textures, high gloss and low gloss, bright colors and black, white, and gray. When this method is applied to the measurement of non-homogeneous surfaces, the results obtained often deviate significantly from the intuitive perception of the human eye, or even are completely opposite. With the continuous improvement of living standards, people's pursuit of personalized product appearance is becoming more intense, and traditional single appearance comparison means can no longer meet the urgent needs of industrial production for diverse and precise appearance detection. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for measuring the surface appearance of color swatches and related devices, which has the advantages of improving the accuracy and efficiency of measuring the surface appearance of color swatches, being able to comprehensively reflect the overall appearance characteristics of color swatches, and achieving more precise appearance measurement.

[0006] In a first aspect, the present invention provides a method for measuring the surface appearance of color swatches, including the following steps: Calculate the texture similarity between the picture of the color swatch to be measured and the pictures of all sample color swatches; Based on the texture similarity, screen all the sample color swatch pictures, and mark the screened sample color swatch pictures as the first sample color swatch pictures; Calculate the color similarity between the to-be-tested color plate image and the first sample color plate image; Based on the color similarity, screen the first sample color plate images, and mark the screened first sample color plate images as second sample color plate images; Calculate the particle similarity between the to-be-tested color plate image and the second sample color plate images; Based on the particle similarity, screen the second sample color plate images, and mark the screened second sample color plate images as third sample color plate images; Calculate the overall similarity between the to-be-tested color plate image and the third sample color plate images; Based on the overall similarity, determine the third sample color plate image corresponding to the to-be-tested color plate image, and output the matching information between the corresponding third sample color plate image and the to-be-tested color plate image.

[0007] The color plate surface appearance measurement method provided by the present invention calculates the texture similarity, color similarity, and particle similarity between the to-be-tested color plate image and the sample color plate image, and performs multiple rounds of screening based on these similarities. Finally, the overall similarity is calculated and the matching result is output, realizing the comprehensive analysis and accurate measurement of the color plate surface appearance, solving the inaccurate problem caused by single-feature analysis in the prior art, improving the accuracy and efficiency of the color plate surface appearance measurement, being able to comprehensively reflect the overall appearance characteristics of the color plate, and realizing more accurate appearance measurement.

[0008] Further, the calculating the texture similarity between the to-be-tested color plate image and all sample color plate images includes the following steps: Input the to-be-tested color plate image into the visual self-attention pre-trained model to obtain the first feature vector of the to-be-tested color plate image output by the visual self-attention pre-trained model; Input the sample color plate image into the visual self-attention pre-trained model to obtain the second feature vector of the sample color plate image output by the visual self-attention pre-trained model; Calculate the cosine similarity between the first feature vector and the second feature vector; Generate the texture similarity based on the cosine similarity.

[0009] Adopting the above technical solution, by using the visual self-attention pre-trained model and the method of calculating the cosine similarity, the texture similarity between the to-be-tested color plate image and the sample color plate image can be calculated more accurately. Compared with the traditional method, it can better solve the subjectivity problem brought by artificial experience and visual judgment, and improve the accuracy and reliability of the color plate surface appearance measurement.

[0010] Further, the method further includes the following steps: Generate a first similarity distribution curve based on the cosine similarity; Obtain the concentrated distribution interval of the first similarity distribution curve; Perform spatial expansion on the cosine similarity within the concentrated distribution interval; Generate a second similarity distribution curve based on the expanded cosine similarity.

[0011] Adopting the above technical solution, by performing spatial expansion on the cosine similarity within the concentrated distribution interval and generating a second similarity distribution curve based on the expanded cosine similarity, the sensitivity of texture similarity can be effectively improved, making it more in line with the intuitive perception of the human eye, thereby reducing the occurrence of misjudgment during use.

[0012] Further, the calculation of the color similarity between the to-be-tested color plate image and the first sample color plate image includes the following steps: Convert the to-be-tested color plate image and multiple first sample color plate images into the HSV color space; Obtain the hue histogram of the HUE channel in the HSV color space; Calculate the color similarity based on the hue histogram.

[0013] Adopting the above technical solution, using the HSV color space is more in line with the human eye's perception of color and can more accurately reflect the color differences between images; by statistically analyzing the color distribution through the hue histogram, the color similarity can be quantified, improving the accuracy of the calculation.

[0014] Further, the calculation of the particle similarity between the to-be-tested color plate image and the second sample color plate image includes the following steps: Obtain the main colors of the to-be-tested color plate image and the second sample color plate image; Generate a main color range based on the main colors; Perform binarization processing on the to-be-tested color plate image and the second sample color plate image according to the main color range to obtain corresponding binary images; Obtain the number and size of particles of the to-be-tested color plate image and the second sample color plate image based on the binary images; Generate one-dimensional arrays of the to-be-tested color plate image and the second sample color plate image based on the number and size of particles. The length of the one-dimensional array is the corresponding number of particles, and the element value of the one-dimensional array is the size of the corresponding particle; align the one-dimensional array of the to-be-tested color plate image and the one-dimensional array of the second sample color plate image, and fill 0 for the one-dimensional array with insufficient length; After rounding both of the one-dimensional arrays, obtain the maximum element value M2 and the minimum element value M1 of the one-dimensional array. Using all integers within the range [M1, M2] as the grouping values, obtain the number of elements in the rounded one-dimensional array that are equal to each grouping value, and use these to form the first particle size distribution matrix corresponding to the one-dimensional array. Calculate the Euclidean distance between the two first particle size distribution matrices and normalize it to obtain the particle similarity.

[0015] Adopting the above technical solution, by obtaining the main colors of the color plate image to be measured and the second sample color plate image, and generating a main color range based on the main colors, it is possible to more accurately obtain the number and size of particles outside the main color range. By generating one-dimensional arrays from the particle number and size, and performing alignment and padding with 0 on them. Convert the two one-dimensional arrays into the first particle size distribution matrices, calculate the Euclidean distance between the two first particle size distribution matrices and normalize it to obtain the particle similarity. Through the above steps, it is possible to more accurately calculate the particle similarity between the color plate image to be measured and the second sample color plate image, thereby improving the accuracy and reliability of the surface appearance measurement of the color plate.

[0016] Further, calculating the particle similarity between the color plate image to be measured and the second sample color plate image includes the following steps: Obtain the main colors of the color plate image to be measured and the second sample color plate image; Generate a main color range based on the main colors; Perform binary processing on the color plate image to be measured and the second sample color plate image according to the main color range to obtain the corresponding binary images; Based on the binary images, obtain the number and size of particles in the color plate image to be measured and the second sample color plate image; Generate one-dimensional arrays for the color plate image to be measured and the second sample color plate image based on the particle number and size. The length of the one-dimensional array is the corresponding particle number, and the element value of the one-dimensional array is the size of the corresponding particle; align the one-dimensional array of the color plate image to be measured and the one-dimensional array of the second sample color plate image, and pad 0 to the one-dimensional array with insufficient length; After rounding both of the one-dimensional arrays, obtain the maximum element value M2 and the minimum element value M1 of the one-dimensional array. Using all integers within the range [M1, M2] as the grouping values, obtain the number of elements in the rounded one-dimensional array that are equal to each grouping value, and use these to form the second particle size distribution matrix corresponding to the one-dimensional array. Construct a weight matrix based on the particle size. Among them, the size of the weight matrix is the same as the size of the one-dimensional array, the initial value of the weight is 1, and for each 1 increase in the particle size, the weight is incremented by 1; Multiply the two second particle size distribution matrices by the weight matrix to respectively obtain weighted third particle size distribution matrices for each; Calculate the Euclidean distance between the two third particle size distribution matrices and normalize it to obtain the particle similarity.

[0017] With the above technical solution, by constructing a weight matrix and assigning higher weights to large particles, the impact brought by large particles can be more focused on. This weight matrix can significantly improve the discrimination of the particle size distribution, enabling particles of different sizes to play different roles in the similarity calculation process. With the help of the weight matrix, the particle similarity calculation is more sensitive to changes in large particles, which is beneficial for more accurately detecting and comparing the differences in particle size distribution between two samples. The introduction of the weight matrix effectively takes into account the influence of different particle sizes on the particle similarity calculation, thereby greatly improving the accuracy of the calculation results.

[0018] Further, determining the third sample color plate image corresponding to the color plate image to be measured based on the overall similarity and outputting the matching information between the corresponding third sample color plate image and the color plate image to be measured includes the following steps: Obtain a preset number of the third sample color plate images based on the overall similarity; Output the texture similarity, color similarity, and particle similarity between the preset number of the third sample color plate images and the color plate image to be measured respectively.

[0019] With the above technical solution, a comprehensive similarity analysis can be carried out on the color plate image to be measured and the third sample color plate image in multiple dimensions. This not only improves the accuracy of color plate matching but also better meets the diverse needs of the coating industry for the appearance detection of color plates.

[0020] In a second aspect, a color plate surface appearance measurement device provided by the present invention includes: A first calculation module for calculating the texture similarity between the color plate image to be measured and all sample color plate images; A first screening module for screening all the sample color plate images based on the texture similarity and marking the screened sample color plate images as the first sample color plate images; A second calculation module for calculating the color similarity between the color plate image to be measured and the first sample color plate images; A second screening module for screening the first sample color plate images based on the color similarity and marking the screened first sample color plate images as the second sample color plate images; A third calculation module for calculating the particle similarity between the color plate image to be measured and the second sample color plate images; A third screening module, configured to screen the second sample color plate image based on the particle similarity, and mark the screened second sample color plate image as a third sample color plate image; A fourth calculation module, configured to calculate the overall similarity between the color plate image to be measured and the third sample color plate image; A color plate matching module, configured to determine the third sample color plate image corresponding to the color plate image to be measured based on the overall similarity, and output the matching information between the corresponding third sample color plate image and the color plate image to be measured.

[0021] In a third aspect, the present application provides an electronic device, including a processor and a memory, where the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect above are run.

[0022] In a fourth aspect, the present application provides a storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps in the method provided in the first aspect above are run.

[0023] As can be seen from the above, the color plate surface appearance measurement method provided by the present invention calculates the texture similarity, color similarity, and particle similarity between the color plate image to be measured and the sample color plate image, and performs multiple rounds of screening based on these similarities. Finally, the overall similarity is calculated and the matching result is output, realizing a comprehensive analysis and accurate measurement of the color plate surface appearance, solving the inaccurate problem caused by single-feature analysis in the prior art, improving the accuracy and efficiency of color plate surface appearance measurement, being able to comprehensively reflect the overall appearance characteristics of the color plate, and realizing more accurate appearance measurement.

[0024] Other features and advantages of the present application will be described in the subsequent specification, and, in part, will be obvious from the specification, or can be understood by implementing the embodiments of the present application. The objectives and other advantages of the present application can be achieved and obtained by the structures specifically pointed out in the written specification and the drawings. Description of the Drawings

[0025] Figure 1 It is a schematic structural diagram of a color plate surface appearance measurement method proposed by the present invention.

[0026] Figure 2 It is a function schematic diagram of a first similarity distribution curve and a second similarity distribution curve.

[0027] Figure 3 It is a block diagram of a color plate surface appearance measurement device proposed by the present invention.

[0028] Figure 4Schematic diagram of the structure of the electronic device provided by the embodiment of the present application.

[0029] In the drawings: 10, the first calculation module; 20, the first screening module; 30, the second calculation module; 40, the second screening module; 50, the third calculation module; 60, the third screening module; 70, the fourth calculation module; 80, the color palette matching module; 3, the electronic device; 301, the processor; 302, the memory; 303, the communication bus. Detailed implementation manners

[0030] The following describes in detail the embodiments of the present invention. Examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

[0031] The following disclosure provides many different embodiments or examples for implementing different structures of the present invention. To simplify the disclosure of the present invention, the components and settings of specific examples are described below. Of course, they are only examples and are not intended to limit the present invention. In addition, the present invention may repeat reference numerals and / or reference letters in different examples. This repetition is for the purpose of simplification and clarity and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0032] In current color palette management practices, finding items similar to a specific color often relies on manual experience and visual judgment. This method not only makes the search process complex and difficult, but also, due to the subjectivity of vision, it is difficult to guarantee accuracy. In addition, the lack of a quantitative measure of color difference also makes the evaluation of color similarity unreliable. Therefore, this process is not only inefficient but also challenging to execute. With the improvement of people's living standards, the demand for product appearance personalization is increasing day by day, and traditional appearance comparison methods are difficult to meet the needs of modern industrial production.

[0033] Referring to the attached Figure 1 , to solve the above technical problems, the present application proposes a method for measuring the surface appearance of a color palette, the method comprising the following steps: S100. Calculate the texture similarity between the picture of the color palette to be measured and the pictures of all sample color palettes; S200. Screen all the sample color palette pictures based on the texture similarity, and mark the screened sample color palette pictures as the first sample color palette pictures; S300. Calculate the color similarity between the picture of the color palette to be measured and the first sample color palette pictures; S400. Screen the first sample color plate image based on color similarity, and mark the screened first sample color plate image as the second sample color plate image; S500. Calculate the particle similarity between the color plate image to be measured and the second sample color plate image; S600. Screen the second sample color plate image based on the particle similarity, and mark the screened second sample color plate image as the third sample color plate image; S700. Calculate the overall similarity between the color plate image to be measured and the third sample color plate image; S800. Determine the third sample color plate image corresponding to the color plate image to be measured based on the overall similarity, and output the matching information between the corresponding third sample color plate image and the color plate image to be measured.

[0034] Specifically, during the color matching and production process of the paint, the prepared paint is sprayed on a card, and the card sprayed with the paint is called a color plate.

[0035] Both the sample color plate image and the color plate image to be measured are collected by a vision instrument, which includes a light source system, a lens, a camera, a backend image signal processor, a sample holder, a memory, a display screen, and accessories.

[0036] The collection process of the sample color plate image and the color plate image to be measured is as follows: The light source system simulates sunlight and projects it onto the color plate. A lens with a resolution of tens of millions of pixels and a magnification of 2x projects the color information on the surface of the color plate onto the full-color optical sensor of a 12-megapixel camera. The full-color optical sensor of the camera converts the light and shadow signal into an electrical signal, and the backend image signal processor then performs white balance calibration and color calibration to obtain a color plate image that is extremely close to the visual effect of the human eye.

[0037] It should be noted that both the sample color plate image and the color plate image to be measured are in png format. Since the png format uses lossless compression and the color space is "rgba", it can retain the image information to the greatest extent.

[0038] The texture similarity is a value obtained by performing texture analysis on the texture information obtained from the color plate image using a texture algorithm. Its value range can be between 0 and 100. The higher the value, the more similar the textures are. This parameter can be used to formulate recipes in big data color matching to assist the color matching master in quickly handling color abnormalities.

[0039] The color similarity is a value obtained by performing color similarity analysis on the color space information obtained from the color plate image using a distance algorithm with other color plate images. Its value range can be between 0 and 100. The higher the value, the more similar the colors are. This parameter can be used to formulate recipes in big data color matching to assist the color matching master in quickly handling color abnormalities.

[0040] The number of particles is the number of particles obtained by the particle (irregular) analysis algorithm for the color swatch image of the metallic powder coating; the particle size distribution is a two-dimensional array distribution data obtained by grouping the number of particles according to the size of the particles (pixel area) per unit area; the particle similarity is a value obtained by performing particle similarity analysis on the color swatch image with other color swatch images using a distance algorithm based on the particle number and particle size distribution information obtained from the color swatch image. Its value ranges from 0 to 100. The higher the value, the more similar the particle number and its distribution are. This parameter can be used to formulate a recipe through big data color matching, assisting the color matching master to quickly handle color anomalies.

[0041] In the actual application of color swatch management, the traditional method mainly relies on manual experience and visual judgment. The complexity and subjectivity of this method make the search process difficult and inaccurate. By introducing computer vision technology and quantitative analysis means, the method of this application can effectively solve these problems. First, by calculating the texture similarity between the color swatch image to be measured and the sample color swatch image, the sample color swatch images with similar texture features to the color swatch image to be measured can be preliminarily screened out. Then, further calculate the color similarity and particle similarity, gradually narrow the screening range, and finally obtain the sample color swatch image that is most similar to the color swatch image to be measured.

[0042] As can be seen from the above, the color swatch surface appearance measurement method provided by the present invention calculates the texture similarity, color similarity and particle similarity between the color swatch image to be measured and the sample color swatch image, and performs multiple rounds of screening based on these similarities. Finally, the overall similarity is calculated and the matching result is output, realizing a comprehensive analysis and accurate measurement of the color swatch surface appearance, solving the inaccurate problem caused by single feature analysis in the prior art, improving the accuracy and efficiency of color swatch surface appearance measurement, being able to comprehensively reflect the overall appearance characteristics of the color swatch, and realizing more accurate appearance measurement.

[0043] In one of the embodiments, calculating the texture similarity between the color swatch image to be measured and all sample color swatch images includes the following steps: Input the color swatch image to be measured into the vision self-attention pre-trained model to obtain the first feature vector of the color swatch image to be measured output by the vision self-attention pre-trained model; Input the sample color swatch image into the vision self-attention pre-trained model to obtain the second feature vector of the sample color swatch image output by the vision self-attention pre-trained model; Calculate the cosine similarity between the first feature vector and the second feature vector; Generate the texture similarity based on the cosine similarity.

[0044] Specifically, the visual self-attention pre-training model can be an open-source visual self-attention pre-training model such as Vision Transformer (ViT). Specifically, using the ViT model, an image is divided into 16x16 blocks, and each block is embedded into a 768-dimensional space. Through the self-attention mechanism, a 197x768 feature matrix (196 blocks + 1 class embedding) is generated. Finally, the feature matrix is encoded into a fixed-length feature vector through a fully connected layer.

[0045] The steps for extracting the feature vectors of the color swatch pictures are as follows: I. Keep the aspect ratio of the original picture and reduce the maximum width of the picture to 1000 pixels, and then convert it into a standard picture in the "JPEG" format.

[0046] II. By loading Vision Transformer (ViT), use the color swatch picture as the model input to obtain the corresponding feature vector of the picture.

[0047] III. Save the picture number and the feature vector, and store other color swatch picture metadata in a relational database. At this time, the feature vectors of all color swatch pictures have been obtained. If we want to obtain similar color swatches, we can continue the following operations: IV. If it is a new color swatch picture, obtain the feature vector of the standard picture according to steps "I" and "II". If it is an existing picture, the saved feature vector can be directly obtained.

[0048] The calculation of cosine similarity can be achieved through vector dot product and norm. This is a commonly used method for measuring vector similarity, which has the characteristics of simple calculation and intuitive results.

[0049] Cosine similarity , where is the dot product of the first feature vector and the second feature vector, , are the norms (lengths) of the first feature vector and the second feature vector respectively.

[0050] The value range of cosine similarity is [-1, 1], where 1 means exactly the same, -1 means exactly opposite, and 0 means orthogonal and no similarity.

[0051] Since the value range of the texture similarity s1 is [0, 100], therefore, the method of linear transformation can be used to convert the range of cosine similarity [-1, 1] to the texture similarity s1 [0, 100].

[0052] With the above technical solution, by using the method of visual self-attention pre-training model and cosine similarity calculation, the texture similarity between the to-be-tested color plate image and the sample color plate image can be calculated more accurately. Compared with the traditional method, it can better solve the subjectivity problem brought by artificial experience and visual judgment, and improve the accuracy and reliability of the surface appearance measurement of the color plate.

[0053] Refer to the appendix Figure 2 , the cosine similarity distribution curve y' = f'(x) generally shows a sigmoid function distribution (the distribution diagram in the above figure where both x and y are greater than 0, where the y-axis represents the cosine similarity and the x-axis represents different sample color plate images). The distribution is extremely uneven and mostly concentrated in the interval above 95%. In order to overcome this situation, it is necessary to perform a spatial mapping on the original Vits similarity value to expand the space in the interval above 95% to make it sensitive, so as to obtain a similarity value that is convenient for observation and judgment; in the paint appearance detection and analysis, in the same paint color matching, its contour and texture are often highly similar, so the given cosine similarity values are often concentrated in the interval above 95%. As a result, in high-precision color matching, it often takes 3 to 4 decimal places to make a difference, which is extremely inconvenient in the use process and is prone to misjudgment, and does not conform to the intuitive feeling of the human eye.

[0054] Therefore, in one of the embodiments, the method further includes the following steps: Generate a first similarity distribution curve based on the cosine similarity; Obtain the concentrated distribution interval of the first similarity distribution curve; Perform spatial expansion on the cosine similarity within the concentrated distribution interval; Generate a second similarity distribution curve based on the expanded cosine similarity.

[0055] Specifically, through a large number of simulation tests, obtain the accurate distribution curve of the first similarity distribution curve on the paint appearance; then plot the curve, observe its distribution space, and find out the concentrated distribution interval; then, according to the actual use situation and the distribution space, expand its concentrated distribution interval, manually plot the expanded curve, and record its (x, y) coordinate values. Finally, write a program to fit the expanded curve to obtain a polynomial fitting function y = f(y'), such as the adjusted second similarity curve in the above figure. In this way, the similarity that was originally concentrated in the interval above 95% is expanded to the interval above 20%, the similarity sensitivity is greatly improved, it is more convenient in the use process, not prone to misjudgment, and conforms to the intuitive feeling of the human eye.

[0056] By adopting the above technical solution, through spatially expanding the cosine similarity within the concentrated distribution interval and generating the second similarity distribution curve based on the expanded cosine similarity, the sensitivity of the texture similarity can be effectively improved, making it more in line with the intuitive perception of the human eye, thereby reducing the occurrence of misjudgment during use.

[0057] In one embodiment, calculating the color similarity between the color plate image to be measured and the first sample color plate image includes the following steps: Convert the color plate image to be measured and multiple first sample color plate images into the HSV color space; Obtain the hue histogram of the HUE channel in the HSV color space; Calculate the color similarity based on the hue histogram.

[0058] Specifically, first convert the PNG image into the HSV color space to reduce the interference of light on the color, and then obtain the hue histogram of the HUE channel. The hue histogram of the HUE channel can represent the number of pixels of different colors in the image. Then, use the histogram intersection method to calculate the color similarity between the two images.

[0059] Exemplarily, the HUE histogram of the color plate image to be measured is as follows: H1 = [10, 20, 30, 40, 50], indicating that the hue range of [0°, 360°] is evenly divided into 5 intervals, each interval has a range of 72°, and there are 10 pixels in the interval [0°, 72°), indicating that there are 20 pixels in the interval [72°, 144°), 30 pixels in the interval [144°, 216°), indicating that there are 40 pixels in the interval [216°, 288°), indicating that there are 50 pixels in the interval [288°, 360°); The HUE histogram of the first sample color plate image is as follows: H2 = [15, 25, 35, 45, 55]; indicating that the hue range of [0°, 360°] is evenly divided into 5 intervals, each interval has a range of 72°, and there are 15 pixels in the interval [0°, 72°), indicating that there are 25 pixels in the interval [72°, 144°), 35 pixels in the interval [144°, 216°), indicating that there are 45 pixels in the interval [216°, 288°), indicating that there are 55 pixels in the interval [288°, 360°); Calculate the minimum value of different intervals of the two HUE histograms to obtain the intersection value of different intervals: min(10, 15) = 10 min(20, 25) = 20 min(30, 35) = 30 min(40, 45) = 40 min(50, 55) = 50; Then sum up the intersection values within all intervals to obtain the total intersection value: t = 10 + 20 + 30 + 40 + 50 = 150; Finally, perform normalization: Sum up the pixel values within H1: ∑H1 = 10 + 20 + 30 + 40 + 50 = 150; Sum up the pixel values within H2: ∑H2 = 15 + 25 + 35 + 45 + 55 = 175; The color similarity s2 = t / min(∑H1, ∑H2) * 100 = (150 / 150) = 100, indicating that the test color plate image and the sample color plate image are very similar in hue distribution.

[0060] Adopting the above technical solution, using the HSV color space is more in line with the human eye's perception of colors and can more accurately reflect the color differences between images; by statistically analyzing the color distribution through the hue histogram, the color similarity can be quantified, improving the accuracy of calculation.

[0061] In one of the embodiments, calculating the particle similarity between the test color plate image and the second sample color plate image includes the following steps: Obtain the main colors of the test color plate image and the second sample color plate image; Generate a main color range based on the main colors; Perform binarization on the test color plate image and the second sample color plate image according to the main color range to obtain the corresponding binary images; Obtain the number and size of particles of the test color plate image and the second sample color plate image based on the binary images; Generate one-dimensional arrays for the test color plate image and the second sample color plate image based on the number and size of particles. The length of the one-dimensional array is the corresponding number of particles, and the element values of the one-dimensional array are the sizes of the corresponding particles; align the one-dimensional arrays of the test color plate image and the second sample color plate image, and fill in 0 for the one-dimensional array with insufficient length; After rounding all the elements of the two one-dimensional arrays, obtain the maximum element value M2 and the minimum element value M1 of the one-dimensional array. Using all integers within the range [M1, M2] as grouping values, obtain the number of elements in the rounded one-dimensional array that are equal to each grouping value, so as to form the first particle size distribution matrix of the corresponding one-dimensional array; Calculate the Euclidean distance between the two particle size distribution matrices and normalize it to obtain the first particle similarity.

[0062] Exemplarily, first, the particles in the picture need to be obtained. The definition of a particle is a color block that is significantly different from the main color of the coating. Therefore, the first step is to determine the main color of the coating and the color range of the main color. First, the median-cut algorithm is used to extract the main color (here, the true color is converted into 256 colors, and the main color is one of the 256 colors, in the [r, g, b] format). Then, through testing, the main color range is adjusted (within the main color range, the human eye cannot perceive that the color range exceeds the main color). Here, the main color range is adjusted to [r, g, b] ± [20, 20, 20]. Then, a binary image of the main color + 20 is obtained. At this time, the picture has only black and white colors. Among them, pixels whose colors are not within the main color range can be set to white, and pixels whose colors are within the main color range can be set to black (or vice versa). Through relevant tool functions, the number and size of all white blocks (particles) can be counted. So far, the number and corresponding size of all particles in the picture have been obtained and saved as a one-dimensional array (in the one-dimensional array, the length of the array represents the number of particles, and the element value represents the size of the particle). Then, in the second step, the particle similarity needs to be calculated according to the number and size of the particles. First, align the lengths of the two one-dimensional arrays to be compared (padding with 0 if insufficient), then take the integer part of all values, and obtain their maximum and minimum values. Then, group them with a particle diameter (particle size) of 1 as the increment value; then, convert the two one-dimensional arrays into matrices respectively to obtain their respective first particle size distribution matrices, and then calculate the Euclidean distance between the two matrices and normalize it to obtain their particle similarity.

[0063] Exemplarily, the one-dimensional array of the picture of the color plate to be measured can be: [1.1, 2.3, 2.2, 3.3, 3.2, 3.1, 4.2, 4.3, 4.4, 4.5]; the one-dimensional array of the picture of the sample color plate can be: [1, 1, 2, 2, 3, 3, 4, 4, 4, 5]; specifically, the size of the particle is obtained by calculating the contour area. Since the calculation of the contour area involves information such as the coordinates of pixel points, the finally obtained area is a value accurate to decimals. For example, the contour area of a particle may be square pixels with a decimal point. Therefore, in order to simplify data processing and storage, the particle size can be converted into an integer.

[0064] The one-dimensional array of the picture of the color plate to be measured after taking the integer part is [1, 2, 2, 3, 3, 3, 4, 4, 4, 4]; the one-dimensional array of the picture of the sample color plate is [1, 1, 2, 2, 3, 3, 4, 4, 4, 5]; After taking the integer part, the maximum element value M2 of the particle size in the two one-dimensional arrays is 5, and the minimum element value M1 is 1; Thus, the grouping values include {1, 2, 3, 4, 5}; The particle size distribution matrix A of the color plate image to be measured is [1, 2, 3, 4, 0]. Matrix A indicates that there is 1 particle with a particle size of 1, 2 particles with a particle size of 2, 3 particles with a particle size of 3, 4 particles with a particle size of 4, and 0 particles with a particle size of 5; The particle size distribution matrix B of the second sample color plate image is [2, 2, 2, 3, 1]; Matrix B indicates that there are 2 particles with a particle size of 1, 2 particles with a particle size of 2, 3 particles with a particle size of 2, 3 particles with a particle size of 4, and 1 particle with a particle size of 5; The Euclidean distance d(A, B)= ; where is the value of the i-th in matrix A, the i-th value in matrix B, and n is the length of the first particle size distribution matrix (which is also equal to the number of grouping values). In the above embodiment, n = 5; Normalization means normalizing the Euclidean distance to the interval [0, 1]. The calculation formula for the normalized distance is dn ; The particle similarity s3=(1 - dn)*100.

[0065] Adopting the above technical solution, by obtaining the main colors of the color plate image to be measured and the second sample color plate image, and generating the main color range based on the main colors, it is possible to more accurately obtain the number and size of particles outside the main color range. By generating a one-dimensional array from the number and size of particles and performing alignment and zero-padding processing on it. Converting the two one-dimensional arrays into the first particle size distribution matrix, calculating the Euclidean distance between the two first particle size distribution matrices and normalizing it to obtain the particle similarity. Through the above steps, it is possible to more accurately calculate the particle similarity between the color plate image to be measured and the second sample color plate image, thereby improving the accuracy and reliability of the surface appearance measurement of the color plate.

[0066] In one of the embodiments, calculating the particle similarity between the color plate image to be measured and the second sample color plate image includes the following steps: Obtain the main colors of the color plate image to be measured and the second sample color plate image; Generate the main color range based on the main colors; Perform binarization processing on the color plate image to be measured and the second sample color plate image according to the main color range to obtain the corresponding binary images; Obtain the number and size of particles of the color plate image to be measured and the second sample color plate image based on the binary images; Generate a one-dimensional array of the test color plate image and the second sample color plate image based on the number and size of particles. The length of the one-dimensional array is the corresponding number of particles, and the element value of the one-dimensional array is the size of the corresponding particle; align the one-dimensional arrays of the test color plate image and the second sample color plate image, and fill in 0 for the one-dimensional array with insufficient length; After taking the integer values of both one-dimensional arrays, obtain the maximum element value M2 and the minimum element value M1 of the one-dimensional array. Using all integers in the range [M1, M2] as grouping values, obtain the number of elements in the rounded one-dimensional array that are equal to each grouping value, so as to form the second particle size distribution matrix of the corresponding one-dimensional array; Construct a weight matrix based on the particle size. Among them, the size of the weight matrix is the same as that of the one-dimensional array, the initial weight value is 1, and for each increase of 1 in the particle size, the weight is increased by 1; Multiply the two second particle size distribution matrices by the weight matrix to obtain their respective weighted third particle size distribution matrices; Calculate the Euclidean distance between the two third particle size distribution matrices and normalize it to obtain the particle similarity.

[0067] Exemplarily, the one-dimensional array of the test color plate image is [1, 2, 2, 3, 3, 3, 4, 4, 4, 4], and the one-dimensional array of the sample color plate image is [1, 1, 2, 2, 3, 3, 4, 4, 4, 5]. Grouping with a particle size of 1 as the increment value gives {1, 2, 3, 4, 5}; the second particle size distribution matrix A of the test color plate image is [1, 2, 3, 4, 0]; the second particle size distribution matrix B of the second sample color plate image is [2, 2, 2, 3, 1]; then the weight matrix is [1, 2, 3, 4, 5]; the one-dimensional array of the weighted test color plate image is [1, 4, 9, 16, 0]; the one-dimensional array of the weighted sample color plate image is [2, 4, 6, 12, 5].

[0068] Adopting the above technical solution, by constructing a weight matrix and assigning higher weights to larger particles, the influence brought by larger particles can be focused more. This weight matrix can significantly improve the discrimination of the particle size distribution, so that particles of different sizes play different roles in the similarity calculation process. With the help of the weight matrix, the particle similarity calculation is more sensitive to the changes of larger particles, which is conducive to more accurately detecting and comparing the differences in particle size distribution between two samples. The measure of introducing the weight matrix effectively takes into account the influence of different particle sizes on the particle similarity calculation, and thus greatly improves the accuracy of the calculation results.

[0069] In one of the embodiments, determining the third sample color plate image corresponding to the test color plate image based on the overall similarity and outputting the matching information between the corresponding third sample color plate image and the test color plate image includes the following steps: Obtain a preset number of third sample color swatch images based on the overall similarity; Output the texture similarity, color similarity, and particle similarity between the preset number of third sample color swatch images and the color swatch image to be measured respectively.

[0070] Specifically, the preset number can be 5. That is, when comparing with the sample color swatch image, obtain the 5 sample color swatch images with the highest overall similarity to the color swatch image to be measured, and display the information such as "texture similarity", "color similarity", and "particle similarity" of the 5 sample color swatch images in sequence.

[0071] Adopting the above technical solution, the color swatch image to be measured and the third sample color swatch image can be comprehensively analyzed for similarity in multiple dimensions. This not only improves the accuracy of color swatch matching, but also better meets the diverse needs of the coating industry for the appearance detection of color swatches.

[0072] In one embodiment, the overall similarity S = s1*w1 + s2*w2 + s3*w3, where w1 + w2 + w3 = 1, s1 is the texture similarity, s2 is the color similarity, s3 is the particle similarity, w1 is the texture similarity weight, w2 is the color similarity weight, and w3 is the particle similarity weight.

[0073] Specifically, after multiple tests, w1 can be 0.4, w2 can be 0.3, and w3 can be 0.3.

[0074] Adopting the above technical solution, considering the texture similarity, color similarity, and particle similarity comprehensively, the results can be given and judgments can be made more conveniently and intuitively.

[0075] In one embodiment, the method further includes the following steps: Calculate the gloss similarity between the color swatch image to be measured and the sample color swatch image; The overall similarity S = s1*w1 + s2*w2 + s3*w3 + s4*w4, where w1 + w2 + w3 + w4 = 1, s1 is the texture similarity, s2 is the color similarity, s3 is the particle similarity, s4 is the gloss similarity, w1 is the texture similarity weight, w2 is the color similarity weight, w3 is the particle similarity weight, and w4 is the gloss similarity weight.

[0076] Specifically, the gloss similarity is the similarity degree of the reflectance of the object surface at 60° measured by a glossmeter; this parameter can be used for formulating recipes in big data color matching to assist the color matching master to quickly handle color abnormalities.

[0077] By adopting the above technical solution, considering texture similarity, color similarity, particle similarity, and gloss similarity comprehensively, it is more convenient and intuitive to give results and make judgments.

[0078] Refer to the appendix Figure 3 , the appendix Figure 3 is a color plate surface appearance measurement device in some embodiments of the present application, including a first calculation module 10, a first screening module 20, a second calculation module 30, a second screening module 40, a third calculation module 50, a third screening module 60, a fourth calculation module 70, and a color plate matching module 80.

[0079] Among them, the first calculation module 10 is used to calculate the texture similarity between the picture of the color plate to be measured and all sample color plate pictures; the first screening module 20 is used to screen all sample color plate pictures based on the texture similarity, and mark the screened sample color plate pictures as the first sample color plate pictures; the second calculation module 30 is used to calculate the color similarity between the picture of the color plate to be measured and the first sample color plate pictures; the second screening module 40 is used to screen the first sample color plate pictures based on the color similarity, and mark the screened first sample color plate pictures as the second sample color plate pictures; the third calculation module 50 is used to calculate the particle similarity between the picture of the color plate to be measured and the second sample color plate pictures; the third screening module 60 is used to screen the second sample color plate pictures based on the particle similarity, and mark the screened second sample color plate pictures as the third sample color plate pictures; the fourth calculation module 70 is used to calculate the overall similarity between the picture of the color plate to be measured and the third sample color plate pictures; the color plate matching module 80 is used to determine the third sample color plate picture corresponding to the picture of the color plate to be measured based on the overall similarity, and output the matching information between the corresponding third sample color plate picture and the picture of the color plate to be measured.

[0080] Please refer to Figure 4 , Figure 4A structural schematic diagram of an electronic device provided by an embodiment of the present application. The present application provides an electronic device 3, including: a processor 301 and a memory 302. The processor 301 and the memory 302 are interconnected and communicate with each other through a communication bus 303 and / or other forms of connection mechanisms (not marked). The memory 302 stores computer-readable instructions executable by the processor 301. When the computing device runs, the processor 301 executes the computer-readable instructions to execute the method in any optional implementation manner of the above embodiment to achieve the following functions: by calculating the texture similarity, color similarity, and particle similarity between the to-be-tested color plate picture and the sample color plate picture, and performing multiple rounds of screening based on these similarities, finally calculating the overall similarity and outputting a matching result, which realizes the comprehensive analysis and accurate measurement of the surface appearance of the color plate, solves the inaccurate problem caused by single-feature analysis in the prior art, improves the accuracy and efficiency of the surface appearance measurement of the color plate, can comprehensively reflect the overall appearance characteristics of the color plate, and realizes more accurate appearance measurement.

[0081] An embodiment of the present application provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, it executes the method in any optional implementation manner of the above embodiment to achieve the following functions: by calculating the texture similarity, color similarity, and particle similarity between the to-be-tested color plate picture and the sample color plate picture, and performing multiple rounds of screening based on these similarities, finally calculating the overall similarity and outputting a matching result, which realizes the comprehensive analysis and accurate measurement of the surface appearance of the color plate, solves the inaccurate problem caused by single-feature analysis in the prior art, improves the accuracy and efficiency of the surface appearance measurement of the color plate, can comprehensively reflect the overall appearance characteristics of the color plate, and realizes more accurate appearance measurement. Among them, the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0082] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For another example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some communication interfaces. The indirect coupling or communication connection of the devices or units can be in electrical, mechanical or other forms.

[0083] In addition, the units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0084] Furthermore, in each embodiment of this application, the functional modules can be integrated together to form an independent part, or each module can exist alone, or two or more modules can be integrated to form an independent part.

[0085] In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0086] The above are only the embodiments of this application and are not used to limit the protection scope of this application. For those skilled in the art, this application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included in the protection scope of this application.

Claims

1. A method for measuring the surface appearance of a color plate, characterized in that: The method comprises the following steps: Calculate the texture similarity between the color plate image to be tested and all sample color plate images; Screening all the sample color palette images based on the texture similarity, and marking the screened sample color palette images as first sample color palette images; Calculating the color similarity between the color palette image to be tested and the first sample color palette image; Filtering the first sample color palette pictures based on the color similarity, and marking the filtered first sample color palette pictures as second sample color palette pictures; Calculating the particle similarity between the color plate image to be tested and the second sample color plate image; Screening the second sample color palette picture based on the particle similarity, and marking the screened second sample color palette picture as a third sample color palette picture; Calculating the overall similarity between the color palette image to be tested and the third sample color palette image; A third sample color palette picture corresponding to the color palette picture to be tested is determined based on the overall similarity, and matching information between the corresponding third sample color palette picture and the color palette picture to be tested is output.

2. A color plate surface appearance measurement method according to claim 1, characterized in that: The calculation of the texture similarity between the color plate image to be tested and all sample color plate images comprises the following steps: Inputting the color plate picture to be tested into a visual self-attention pre-training model, and obtaining a first feature vector of the color plate picture to be tested output by the visual self-attention pre-training model; Inputting the sample color plate image into a visual self-attention pre-training model, and obtaining a second eigenvector of the sample color plate image output by the visual self-attention pre-training model; Calculate the cosine similarity between the first eigenvector and the second eigenvector; A texture similarity is generated based on the cosine similarity.

3. A color plate surface appearance measurement method according to claim 2, characterized in that: The method further comprises the following steps: generating a first similarity distribution curve based on the cosine similarity; Obtaining a concentrated distribution interval of the first similarity distribution curve; Performing spatial expansion on the cosine similarity within the concentrated distribution interval; A second similarity distribution curve is generated based on the expanded cosine similarity.

4. A color plate surface appearance measurement method according to claim 1, characterized in that: The calculating of the color similarity between the color palette image to be tested and the first sample color palette image comprises the following steps: Convert the color palette image to be tested and the plurality of the first sample color palette images into an HSV color space; Get the hue histogram of the HUE channel in the HSV color space; The color similarity is calculated based on the hue histogram.

5. A color plate surface appearance measurement method according to claim 1, characterized in that: The calculating of the particle similarity between the color plate image to be tested and the second sample color plate image comprises the following steps: Obtaining the main colors of the color palette image to be tested and the second sample color palette image; generating a primary color range based on the primary color; Binarization is performed on the color palette image to be tested and the second sample color palette image according to the main color range to obtain corresponding binary images; Acquire the number and size of particles of the color plate image to be tested and the second sample color plate image based on the binary image; Generate a one-dimensional array of the color palette image to be tested and the second sample color palette image based on the number and size of the particles, wherein the length of the one-dimensional array is the corresponding number of particles, and the element value of the one-dimensional array is the size of the corresponding particle; align the one-dimensional array of the color palette image to be tested and the one-dimensional array of the second sample color palette image, and fill the one-dimensional array with zeros if the length is insufficient; After rounding off the two one-dimensional arrays, the maximum element value M2 and the minimum element value M1 of the one-dimensional array are obtained, and all integers in the range of [M1, M2] are used as grouping values, and the number of elements in the rounded one-dimensional array that are equal to each grouping value is obtained to form a first particle size distribution matrix corresponding to the one-dimensional array; The Euclidean distance between the two first particle size distribution matrices is calculated and normalized to obtain the particle similarity.

6. A color plate surface appearance measurement method according to claim 1, characterized in that: The calculating of the particle similarity between the color plate image to be tested and the second sample color plate image comprises the following steps: Obtaining the main colors of the color palette image to be tested and the second sample color palette image; generating a primary color range based on the primary color; Binarization is performed on the color palette image to be tested and the second sample color palette image according to the main color range to obtain corresponding binary images; Acquire the number and size of particles of the color plate image to be tested and the second sample color plate image based on the binary image; Generate a one-dimensional array of the color palette image to be tested and the second sample color palette image based on the number and size of the particles, wherein the length of the one-dimensional array is the corresponding number of particles, and the element value of the one-dimensional array is the size of the corresponding particle; align the one-dimensional array of the color palette image to be tested and the one-dimensional array of the second sample color palette image, and fill the one-dimensional array with zeros if the length is insufficient; After rounding off the two one-dimensional arrays, the maximum element value M2 and the minimum element value M1 of the one-dimensional array are obtained, and all integers in the range of [M1, M2] are used as grouping values, and the number of elements in the rounded one-dimensional array that are equal to each grouping value is obtained to form a second particle size distribution matrix corresponding to the one-dimensional array; A weight matrix is ​​constructed based on the particle size, wherein the size of the weight matrix is ​​consistent with the size of the one-dimensional array, the initial value of the weight is 1, and the weight increases by 1 for each increase in particle size; Multiplying the two second particle size distribution matrices by the weight matrix to obtain respective weighted third particle size distribution matrices; The Euclidean distance between the two third particle size distribution matrices is calculated and normalized to obtain the particle similarity.

7. A color plate surface appearance measurement method according to claim 1, characterized in that: The step of determining a third sample color palette picture corresponding to the color palette picture to be tested based on the overall similarity, and outputting matching information between the corresponding third sample color palette picture and the color palette picture to be tested comprises the following steps: Acquire a preset number of the third sample color palette pictures based on the overall similarity; The texture similarity, color similarity and particle similarity between a preset number of the third sample color plate images and the color plate image to be tested are output respectively.

8. A color plate surface appearance measuring device, characterized in that: The device comprises: The first calculation module is used to calculate the texture similarity between the color plate image to be tested and all sample color plate images; A first screening module, used for screening all the sample color palette pictures based on the texture similarity, and marking the screened sample color palette pictures as first sample color palette pictures; A second calculation module, used to calculate the color similarity between the color palette image to be tested and the first sample color palette image; A second screening module, configured to screen the first sample color palette pictures based on the color similarity, and mark the screened first sample color palette pictures as second sample color palette pictures; A third calculation module, used for calculating the particle similarity between the color plate image to be tested and the second sample color plate image; A third screening module, configured to screen the second sample color palette pictures based on the particle similarity, and mark the screened second sample color palette pictures as third sample color palette pictures; A fourth calculation module, used for calculating the overall similarity between the color palette picture to be tested and the third sample color palette picture; A color palette matching module is used to determine a third sample color palette picture corresponding to the color palette picture to be tested based on the overall similarity, and output matching information between the corresponding third sample color palette picture and the color palette picture to be tested.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the steps in the method according to any one of claims 1 to 7 are executed.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are executed.

Images