Building ceramic automatic detection method and system based on visual image analysis

Abstract

The invention discloses an architectural ceramic automatic detection method and system based on visual image analysis, and relates to the technical field of deformation detection. Comprising the steps of preprocessing an original picture and a to-be-detected picture through a preprocessing method; respectively adjusting the complexity coefficient and adjusting the original matrix and the to-be-detected matrix; fourier transform is carried out after adjustment, and a frequency domain matrix and an amplitude spectrum matrix are obtained; and calculating a frequency spectrum energy difference and an overall difference coefficient, and inputting the frequency spectrum energy difference and the overall difference coefficient into a deformation degree calculation model to obtain a deformation degree detection result. According to the method, the complexity coefficients of the original matrix and the to-be-detected matrix are calculated through a complexity calculation method, the size change of the irregular surface and the contour of the architectural ceramic is represented through the difference between the complexity coefficients, the complexity coefficients are combined with the picture data, the deformation degree of the architectural ceramic is detected, and the accuracy of the deformation degree of the architectural ceramic is improved. The irregular surface of the architectural ceramic and the size change of the contour are fully considered, and when the texture on the architectural ceramic is complex, the method has high detection accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of deformation detection, and in particular to an automatic detection method and system for architectural ceramics based on visual image analysis. Background Art

[0002] In order to meet the safety and quality requirements of buildings during long-term use, it is necessary to detect the deformation of building ceramics. For example, patent publication number CN119354091A describes a vision-based automatic deformation detection method for building ceramics, which includes the following steps: S1. Acquire a production digital image of the building ceramic and preprocess the production digital image to generate a standard digital image; S2. Determine several grid models based on several corner points of the standard digital image and calculate the principal component parameters; S3. Generate deformation detection results for the building ceramic based on the principal component parameters of the standard digital image and the edge intensity values ​​of each pixel. The entire deformation detection process of the present invention can be automated, reducing manual intervention and errors, and can capture subtle changes in the image, so the deformation detection results are highly sensitive.

[0003] Deformation testing of building ceramics can ensure the appearance quality and dimensional accuracy of building ceramics, improve their use effect, and thus enhance the overall aesthetics and decorative effect of the building. However, the above-mentioned and similar methods require analysis and calculation through corner points when testing building ceramics. When the texture on the building ceramics is complex, it is difficult to obtain accurate corner points. Therefore, the above-mentioned methods have low detection accuracy when testing building ceramics with complex textures. Summary of the Invention

[0004] The object of the present invention is to provide a method and system for automatic detection of architectural ceramics based on visual image analysis, so as to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: an automatic detection method for architectural ceramics based on visual image analysis, the method comprising:

[0006] Obtain original images and images to be tested of building ceramics;

[0007] Preprocess the original image and the image to be detected by a preprocessing method to obtain a single-channel matrix to be detected and an original matrix;

[0008] The complexity coefficients of the original matrix and the matrix to be tested are calculated respectively by the complexity calculation method. The complexity coefficients are used to quantify the fineness and complexity of the surface texture of architectural ceramics. The differences between the complexity coefficients are used to show the changes in the surface texture of architectural ceramics.

[0009] The original matrix and the matrix to be detected are adjusted using an adjustment method according to the complexity coefficient to obtain the original intermediate matrix and the intermediate matrix to be detected. The complexity coefficient is combined with the image data to consider the irregular surface and dimensional changes of the contour of the building ceramics to improve the detection accuracy.

[0010] Performing Fourier transform on the original intermediate matrix and the intermediate matrix to be detected to obtain frequency domain matrices respectively, and calculating the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected through the frequency domain matrices;

[0011] The spectrum energy difference between the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected is calculated by the spectrum difference calculation method. The overall difference coefficient of the two amplitude spectrum matrices is calculated by the overall difference calculation method. The spectrum energy difference and the overall difference coefficient are input into the deformation calculation model to obtain the deformation detection result.

[0012] Preferably, the complexity calculation method includes:

[0013] Divide the original matrix and the matrix to be detected into several blocks of the same size;

[0014] Count the number of times the eigenvalue of each element in the block appears and the total number of elements in the block, and calculate the probability of the eigenvalue of each element appearing;

[0015] The monomer complexity coefficient of the target block is calculated according to the formula, specifically:

[0016]

[0017] in Indicates the monomer complexity coefficient of the target block, Indicates the probability of occurrence of the characteristic value x of the element in the target block;

[0018] Calculate the individual complexity of all blocks and integrate them as the complexity coefficient of the original image and the image to be detected.

[0019] Preferably, the complexity calculation method includes:

[0020] Obtain the single-channel co-occurrence matrix of the original matrix and the matrix to be detected respectively;

[0021] The complexity coefficient is calculated according to the formula, specifically:

[0022]

[0023] in represents the complexity coefficient, represents the order of magnitude of the eigenvalue of the element, represents the difference in eigenvalues, represents the single-channel co-occurrence matrix, represents the Kronecker function, when = hour, =1, otherwise =0.

[0024] Preferably, the spectrum energy calculation method includes:

[0025] The frequency domain matrices of the original intermediate matrix and the intermediate matrix to be detected are filtered according to the ideal low-pass filter transfer function, and low-frequency matrices are obtained respectively;

[0026] The low-frequency spectrum energy is calculated according to the formula:

[0027]

[0028] in represents the low-frequency spectrum energy of the target low-frequency matrix, Indicates the row magnitude of the target low-frequency matrix, Indicates the column magnitude of the target low-frequency matrix, Indicates that the row label in the target low-frequency matrix is , the column is labeled The characteristic values ​​of the elements of

[0029] Calculate the spectrum energy difference between the low-frequency spectrum energy corresponding to the original intermediate matrix and the intermediate matrix to be detected, specifically:

[0030]

[0031] in represents the difference in spectral energy, Represents the low-frequency spectrum energy corresponding to the original intermediate matrix, Indicates the low-frequency spectrum energy corresponding to the intermediate matrix to be detected.

[0032] Preferably, the overall difference calculation method includes:

[0033] The overall difference coefficient is calculated based on the frequency domain matrix of the original intermediate matrix and the intermediate matrix to be detected, specifically:

[0034]

[0035] in represents the overall coefficient of variation, Indicates the row magnitude of the target frequency domain matrix, Indicates the column magnitude of the target frequency domain matrix, The row label in the frequency domain matrix corresponding to the original intermediate matrix is , the column is labeled The eigenvalues ​​of the elements of The row label in the frequency domain matrix corresponding to the intermediate matrix to be detected is , the column is labeled The eigenvalues ​​of the elements.

[0036] Preferably, the adjustment method includes:

[0037] Randomly generate a random matrix of the same size as the original matrix or the matrix to be tested, and the eigenvalues ​​of each element in the random matrix are uniformly distributed between -1 and 1;

[0038] Adjust the original matrix and the matrix to be detected according to the formula to obtain the original intermediate matrix and the intermediate matrix to be detected respectively. The specific formula is:

[0039]

[0040] in are the original intermediate matrix and the intermediate matrix to be detected, represents the original matrix or the matrix to be tested, represents the complexity coefficient, represents a random matrix.

[0041] Preferably, the pretreatment method comprises:

[0042] Convert the original image and the image to be detected into single-channel images;

[0043] Use edge detection algorithms to perform edge recognition on the single-channel images corresponding to the original image and the image to be detected to determine the detection location;

[0044] Perform perspective transformation on the single-channel image corresponding to the image to be detected so that the edge of the detection part in the single-channel image corresponding to the image to be detected and the edge of the detection part in the single-channel image corresponding to the original image overlap to a set ratio;

[0045] Overall change the eigenvalues ​​of the single-channel image corresponding to the image to be detected so that the average value of the eigenvalues ​​of the single-channel image corresponding to the image to be detected is the same as the average value of the eigenvalues ​​of the single-channel image corresponding to the original image;

[0046] The single-channel image corresponding to the original image is cropped and the complete detection area is retained to establish the intermediate matrix of the original image. At the same time, the single-channel image after perspective transformation is cropped and the complete detection area is retained to establish the intermediate matrix to be detected, thereby reducing the influence of factors outside the detection area on the detection results.

[0047] Preferably, the deformation calculation model is specifically:

[0048]

[0049] in Indicates the degree of deformation, and The coefficients representing the spectral energy difference and the overall difference coefficient, respectively, and They respectively represent the contribution of the spectrum energy difference and the overall difference coefficient to the deformation result in the deformation calculation, and and The sum of is 1, the more delicate and complex the surface of the building ceramics is. The larger the value, the The lower the value, the specific value can be determined by the least squares calculation based on a large amount of known data. represents the difference in spectral energy, represents the overall coefficient of variation, represents the maximum value of the spectrum energy difference, Indicates the maximum value of the overall coefficient of variation.

[0050] Compared with the prior art, the present invention has the following beneficial effects:

[0051] The complexity coefficients of the original matrix and the matrix to be detected are calculated respectively through the complexity calculation method. The irregular surface and contour of building ceramics are quantified by the complexity coefficients. The dimensional changes of the irregular surface and contour of building ceramics are expressed by the difference between the complexity coefficients. The complexity coefficients are then combined with the image data to detect the deformation of building ceramics, fully considering the dimensional changes of the irregular surface and contour of building ceramics. When the texture on the building ceramics is complex, the detection accuracy can be improved.

[0052] At the same time, the original image and the image to be detected are preprocessed by a preprocessing method, and the single-channel image corresponding to the image to be detected is perspective transformed so that the edge of the detection part in the single-channel image corresponding to the image to be detected and the edge of the detection part in the single-channel image corresponding to the original image reach a set ratio. This makes it more accurate when detecting the difference between the original image and the image to be detected, and reduces the influence of external factors such as shooting angle and shooting position on the detection;

[0053] In addition, the spectrum energy difference between the amplitude spectrum matrix of the original intermediate matrix and the intermediate matrix to be detected is calculated through the spectrum difference calculation method. When filtering, only the low-frequency spectrum is retained, so that the detection results can more accurately reflect the small deformations in a large range and are more in line with the deformation law of building ceramics, thereby ensuring the reliability of the detection results. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 Schematic diagram of the process of the automatic deformation detection method of the present invention;

[0055] Figure 2 Schematic diagram of the process of complexity calculation method in embodiment 1 of the present invention;

[0056] Figure 3 Schematic diagram of the process of complexity calculation method in embodiment 2 of the present invention;

[0057] Figure 4 Schematic diagram of the process of the adjustment method of the present invention;

[0058] Figure 5 Schematic diagram of the structure of the pretreatment method of the present invention. DETAILED DESCRIPTION

[0059] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.

[0060] In this application, for ease of understanding, the method steps used do not need to be executed in the order of the steps in this embodiment during actual operation. In other embodiments, these steps may be performed simultaneously or in a different order.

[0061] Example 1:

[0062] Deformation testing of building ceramics can ensure the appearance quality and dimensional accuracy of building ceramics, improve their use effect, and thus enhance the overall aesthetics and decorative effect of the building. When the texture on the building ceramics is complex, deformation testing needs to be combined with its irregular surface and contour to effectively improve detection accuracy.

[0063] like Figure 1 、 Figure 2 、 Figure 4 and Figure 5 As shown, the present invention provides a technical solution: an automatic detection method for architectural ceramics based on visual image analysis, comprising:

[0064] Obtain original images and images to be tested of building ceramics;

[0065] The original image and the image to be detected are preprocessed by a preprocessing method to obtain a single-channel matrix to be detected and an original matrix. It should be noted that the original image obtained can be an output image of building ceramics or an image of building ceramics that meets the standards;

[0066] The complexity coefficients of the original matrix and the matrix to be tested are calculated respectively by the complexity calculation method. The complexity coefficients are used to quantify the fineness and complexity of the surface texture of architectural ceramics. The difference between the complexity coefficients is used to show the changes in the surface texture of architectural ceramics.

[0067] The original matrix and the matrix to be detected are adjusted using an adjustment method according to the complexity coefficient to obtain the original intermediate matrix and the intermediate matrix to be detected. The complexity coefficient is combined with the image data to consider the irregular surface and dimensional changes of the contour of the building ceramics to improve the detection accuracy.

[0068] Performing Fourier transform on the original intermediate matrix and the intermediate matrix to be detected to obtain frequency domain matrices respectively, and calculating the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected through the frequency domain matrices;

[0069] The spectrum energy difference between the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected is calculated by the spectrum difference calculation method. The overall difference coefficient of the two amplitude spectrum matrices is calculated by the overall difference calculation method. The spectrum energy difference and the overall difference coefficient are input into the deformation calculation model to obtain the deformation detection result.

[0070] like Figure 5 As shown, the preprocessing methods include:

[0071] Convert the original image and the image to be detected into single-channel images;

[0072] It should be noted that when converting into a single-channel image, existing technologies can be used, such as using the average value of multiple channel values ​​of a pixel as the grayscale value to obtain a grayscale value image, or selecting the maximum channel value as the grayscale value, etc. The selection can be made based on the actual situation such as the color type and light environment of the building ceramics, and will not be elaborated here.

[0073] Use edge detection algorithms to perform edge recognition on the single-channel images corresponding to the original image and the image to be detected to determine the detection location;

[0074] It should be noted that the edge detection algorithm is an existing technology (such as the Canny edge detection algorithm, Robert operator, etc.), which will not be described in detail here. After edge detection, the building ceramic area to be detected is distinguished from the surrounding area to prevent the surrounding area from affecting the detection of the deformation of the building ceramics, thereby improving the accuracy of the detection.

[0075] Perform perspective transformation on the single-channel image corresponding to the image to be detected so that the edge of the detection part in the single-channel image corresponding to the image to be detected and the edge of the detection part in the single-channel image corresponding to the original image overlap to a set ratio;

[0076] It should be noted that after perspective transformation (existing technology), the shooting angle of the image to be detected can be corrected, reducing the impact of external factors such as shooting angle, position and distance on the detection results.

[0077] Overall change the eigenvalues ​​of the single-channel image corresponding to the image to be detected so that the average value of the eigenvalues ​​of the single-channel image corresponding to the image to be detected is the same as the average value of the eigenvalues ​​of the single-channel image corresponding to the original image;

[0078] It should be noted that, taking grayscale values ​​as an example, increasing or decreasing the grayscale values ​​corresponding to the image to be detected as a whole can further reduce the impact of light intensity in the shooting environment on the detection results. When the shooting environments of the original image and the image to be detected are different (mainly referring to the different brightness of the environment), the detection accuracy can be further improved.

[0079] The single-channel image corresponding to the original image is cropped and the complete detection area is retained to establish the intermediate matrix of the original image. At the same time, the single-channel image after perspective transformation is cropped and the complete detection area is retained to establish the intermediate matrix to be detected, thereby reducing the influence of factors outside the detection area on the detection results.

[0080] like Figure 2 As shown, the complexity calculation method includes:

[0081] Divide the original matrix and the matrix to be detected into several blocks of the same size;

[0082] Count the number of times the eigenvalue of each element in the block appears and the total number of elements in the block, and calculate the probability of the eigenvalue of each element appearing;

[0083] The monomer complexity coefficient of the target block is calculated according to the formula. The formula uses the monotonic function composite information characteristic and considers each probability case. The formula is deduced as follows:

[0084]

[0085] in Indicates the monomer complexity coefficient of the target block, Indicates the probability of occurrence of the characteristic value x of the element in the target block.

[0086] Calculate the individual complexity of all blocks and integrate them as the complexity coefficient of the original image and the image to be detected.

[0087] It should be noted that, in order to facilitate calculation, the simulation data is set as follows:

[0088] The original matrix is ; Divide the original matrix into four 2×2 blocks, with the upper left corner block For example, calculate the monomer complexity coefficient, specifically:

[0089] Among them, 20 and 24 appear once each, and 22 appears twice. Calculate the probability of each gray value: =0.25, =0.5, =0.25, and then the monomer complexity coefficient of the area can be calculated according to the formula =1.5;

[0090] In this embodiment, since the data of the remaining three blocks is consistent with the block in the upper left corner, the individual complexity coefficients of the remaining three blocks are also 1.5. After taking the average value, the complexity coefficient is 1.5 (this embodiment uses the average value of the individual complexity coefficients as the complexity coefficient, and the median or mode can also be used for integration, without limitation).

[0091] The same calculation method is also used for the matrix to be tested, and the complexity coefficients of the original matrix and the matrix to be tested can be obtained respectively. The irregular surface and contour of the building ceramics are quantified by the complexity coefficients, and the dimensional changes of the irregular surface and contour of the building ceramics are expressed by the difference between the complexity coefficients.

[0092] like Figure 4 As shown, the adjustment methods include:

[0093] Randomly generate a random matrix of the same size as the original matrix or the matrix to be tested, and the eigenvalues ​​of each element in the random matrix are uniformly distributed between -1 and 1;

[0094] According to the formula, the original matrix and the matrix to be tested are adjusted to obtain the original intermediate matrix and the intermediate matrix to be tested respectively. The random matrix is ​​weighted by the complexity coefficient, which can bring the complexity coefficient to the original matrix or the matrix to be tested, change the data structure, and then deduce the formula:

[0095]

[0096] in are the original intermediate matrix and the intermediate matrix to be detected, represents the original matrix or the matrix to be tested, represents the complexity coefficient, represents a random matrix.

[0097] It should be noted that, in order to facilitate calculation, the simulation data is set as follows:

[0098] Set the original matrix to ;

[0099] Generate a 4×4 random matrix , the element values ​​are uniformly distributed between [−1,1]. Assume that the generated random matrix is , according to the formula Calculation can be obtained = , in the calculation, the complexity coefficient adopts the value 1.5 obtained by the above calculation. Similarly, the intermediate matrix to be detected can be calculated.

[0100] After the two intermediate matrices are calculated, the frequency domain matrices are obtained through Fourier transform. Taking the above original intermediate matrix as an example, the frequency domain matrix obtained after Fourier transform (existing technology) is: ,

[0101] Its amplitude spectrum matrix (existing technology) is: .

[0102] Spectral energy calculation methods include:

[0103] The frequency domain matrices of the original intermediate matrix and the intermediate matrix to be detected are filtered according to the ideal low-pass filter transfer function, and low-frequency matrices are obtained respectively;

[0104] The low-frequency spectrum energy is calculated according to the formula:

[0105]

[0106] in represents the low-frequency spectrum energy of the target low-frequency matrix, Indicates the row magnitude of the target low-frequency matrix, Indicates the column magnitude of the target low-frequency matrix, Indicates that the row label in the target low-frequency matrix is , the column is labeled The characteristic values ​​of the elements of

[0107] Calculate the spectrum energy difference between the low-frequency spectrum energy corresponding to the original intermediate matrix and the intermediate matrix to be detected, specifically:

[0108]

[0109] in represents the difference in spectral energy, Represents the low-frequency spectrum energy corresponding to the original intermediate matrix, Indicates the low-frequency spectrum energy corresponding to the intermediate matrix to be detected.

[0110] Taking the above data as an example, low-frequency filtering is performed, assuming that the cutoff frequency D0 = 1. The ideal low-pass filter transfer function is , it can be calculated that the corresponding low-frequency matrix is , and then through the formula Can be obtained ≈124190.

[0111] Similarly, the low-frequency spectrum energy of the intermediate matrix to be detected can be calculated. Assuming that the low-frequency spectrum energy of the intermediate matrix to be detected is =141344, then the spectrum energy difference can be calculated =17154.

[0112] The overall difference calculation method includes:

[0113] The overall difference coefficient is calculated based on the frequency domain matrix of the original intermediate matrix and the intermediate matrix to be detected, specifically:

[0114]

[0115] in represents the overall coefficient of variation, Indicates the row magnitude of the target frequency domain matrix, Indicates the column magnitude of the target frequency domain matrix, The row label in the frequency domain matrix corresponding to the original intermediate matrix is , the column is labeled The eigenvalues ​​of the elements of The row label in the frequency domain matrix corresponding to the intermediate matrix to be detected is , the column is labeled The eigenvalues ​​of the elements.

[0116] It should be noted that, for the convenience of calculation, taking the frequency domain matrix of the original intermediate matrix in the above calculation process as an example, it is assumed that the frequency domain matrix of the intermediate matrix to be detected calculated by Fourier transform is , and then according to the formula we can calculate ≈31.92.

[0117] The deformation calculation model is as follows:

[0118]

[0119] in Indicates the degree of deformation, and The coefficients representing the spectral energy difference and the overall difference coefficient, respectively, represents the difference in spectral energy, represents the overall coefficient of variation, represents the maximum value of the spectrum energy difference, Indicates the maximum value of the overall coefficient of variation.

[0120] and They respectively represent the contribution of the spectrum energy difference and the overall difference coefficient to the deformation result in the deformation calculation, and and The sum of is 1, the more delicate and complex the surface of the building ceramics is. The larger the value, the The lower the value, the specific value can be determined by least squares calculation based on a large amount of known data. Assume that for a certain type of building ceramics, there are multiple sample data as shown in Table 1 below:

[0121] Table 1: Sample data

[0122] According to the least squares method, SSE = (1-(0.3 +0.7 ))2+(1-(0.5 +0.5 ))2+(0-(0.2 +0.8 )) The minimum value of 2 is sufficient to determine and The specific data of ≈0.7, ≈0.3, from this we can judge that for this type of building ceramics and The values ​​of are 0.7 and 0.3 respectively, and this way we can determine and The specific value of .

[0123] It should be noted that, in order to facilitate calculation, the overall difference coefficient is set Weight =0.4, spectral energy difference Weight =0.6, =100 (based on statistics of a large number of samples), =20000 (according to a large number of sample statistics), according to the formula, the deformation degree can be obtained. =64.23% (the deformation degree is calculated based on statistics of a large number of samples, indicating the degree of deformation, not the amplitude of deformation). After calculating the deformation degree, it can be determined whether the building ceramics are within the qualified standards based on actual production and construction needs.

[0124] Example 2:

[0125] The irregular surface of building ceramics may have various forms of expression, and the fineness and complexity of the texture will produce different visual perceptions. Example 1 provides a complexity calculation method to perform partition calculation on the original matrix or the matrix to be detected. However, this complexity calculation method is not accurate enough in quantifying the irregular surface and contour of building ceramics when the fineness and complexity of the texture on the surface of building ceramics is unevenly distributed. This embodiment provides another complexity calculation method based on Example 1 to improve the accuracy of quantifying the irregular surface and contour of building ceramics.

[0126] like Figure 3 As shown, the complexity calculation method includes:

[0127] Obtain the single-channel co-occurrence matrix of the original matrix and the matrix to be detected respectively;

[0128] The complexity coefficient is calculated according to the formula, specifically:

[0129]

[0130] in represents the complexity coefficient, represents the order of magnitude of the eigenvalue of the element, Represents the difference in eigenvalues, used to traverse the range of grayscale differences. represents the single-channel co-occurrence matrix, represents the Kronecker function, when = hour, =1, otherwise =0.

[0131] It should be noted that It represents the probability that gray values ​​i and j appear at the same time when the distance is d and the direction is θ.

[0132] In order to facilitate calculation, the simulation data is set as follows:

[0133] Assume that the gray level co-occurrence matrix of the original matrix (the calculation method of the gray level co-occurrence matrix is ​​an existing technology and will not be described here) is:

[0134]

[0135] And in the original matrix, the order of grayscale =4, the complexity coefficient can be calculated according to the formula =1.2.

[0136] According to the gray-level co-occurrence matrix, the irregular surface and contour of the building ceramics are refined and the complexity coefficient is calculated. It can reflect the fineness and complexity of the texture. Compared with Example 1, it is more sensitive to the fineness and complexity of the texture on the surface of building ceramics, and can further improve the quantitative accuracy of the irregular surface and contour of building ceramics.

[0137] Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is limited by the accompanying embodiments and their equivalents.

Claims

1. A method for automatic detection of architectural ceramics based on visual image analysis, comprising: Obtain the original image and the image to be tested of the building ceramics, which is characterized by: Preprocess the original image and the image to be detected by a preprocessing method to obtain a single-channel matrix to be detected and an original matrix; The complexity coefficients of the original matrix and the matrix to be tested are calculated respectively by the complexity calculation method. The complexity coefficients are used to quantify the fineness and complexity of the surface texture of architectural ceramics. The differences between the complexity coefficients are used to show the changes in the surface texture of architectural ceramics. The original matrix and the matrix to be detected are adjusted using an adjustment method according to the complexity coefficient to obtain the original intermediate matrix and the intermediate matrix to be detected. The complexity coefficient is combined with the image data to consider the irregular surface and dimensional changes of the contour of the building ceramics to improve the detection accuracy. Performing Fourier transform on the original intermediate matrix and the intermediate matrix to be detected to obtain frequency domain matrices respectively, and calculating the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected through the frequency domain matrices; The spectrum energy difference between the amplitude spectrum matrices of the original intermediate matrix and the intermediate matrix to be detected is calculated by the spectrum difference calculation method. The overall difference coefficient of the two amplitude spectrum matrices is calculated by the overall difference calculation method. The spectrum energy difference and the overall difference coefficient are input into the deformation calculation model to obtain the deformation detection result.

2. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The complexity calculation method includes: Divide the original matrix and the matrix to be detected into several blocks of the same size; Count the number of times the eigenvalue of each element in the block appears and the total number of elements in the block, and calculate the probability of the eigenvalue of each element appearing; The monomer complexity coefficient of the target block is calculated according to the formula, specifically: in Indicates the monomer complexity coefficient of the target block, Indicates the probability of occurrence of the characteristic value x of the element in the target block; Calculate the individual complexity of all blocks and integrate them as the complexity coefficient of the original image and the image to be detected.

3. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The complexity calculation method includes: Obtain the single-channel co-occurrence matrix of the original matrix and the matrix to be detected respectively; The complexity coefficient is calculated according to the formula, specifically: in represents the complexity coefficient, represents the order of magnitude of the eigenvalue of the element, represents the difference in eigenvalues, represents the single-channel co-occurrence matrix, represents the Kronecker function, when = hour, =1, otherwise =0.

4. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The spectrum energy calculation method includes: The frequency domain matrices of the original intermediate matrix and the intermediate matrix to be detected are filtered according to the ideal low-pass filter transfer function, and low-frequency matrices are obtained respectively; The low-frequency spectrum energy is calculated according to the formula: in represents the low-frequency spectrum energy of the target low-frequency matrix, Indicates the row magnitude of the target low-frequency matrix, Indicates the column magnitude of the target low-frequency matrix, Indicates that the row label in the target low-frequency matrix is , the column is labeled The characteristic values ​​of the elements of ; Calculate the spectrum energy difference between the low-frequency spectrum energy corresponding to the original intermediate matrix and the intermediate matrix to be detected, specifically: in represents the difference in spectral energy, Represents the low-frequency spectrum energy corresponding to the original intermediate matrix, Indicates the low-frequency spectrum energy corresponding to the intermediate matrix to be detected.

5. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The overall difference calculation method includes: The overall difference coefficient is calculated based on the frequency domain matrix of the original intermediate matrix and the intermediate matrix to be detected, specifically: in represents the overall coefficient of variation, Indicates the row magnitude of the target frequency domain matrix, Indicates the column magnitude of the target frequency domain matrix, The row label in the frequency domain matrix corresponding to the original intermediate matrix is , the column is labeled The eigenvalues ​​of the elements of The row label in the frequency domain matrix corresponding to the intermediate matrix to be detected is , the column is labeled The eigenvalues ​​of the elements.

6. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The adjustment method includes: Randomly generate a random matrix of the same size as the original matrix or the matrix to be tested, and the eigenvalues ​​of each element in the random matrix are uniformly distributed between -1 and 1; Adjust the original matrix and the matrix to be detected according to the formula to obtain the original intermediate matrix and the intermediate matrix to be detected respectively. The specific formula is: in are the original intermediate matrix and the intermediate matrix to be detected, represents the original matrix or the matrix to be tested, represents the complexity coefficient, represents a random matrix.

7. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 1, characterized in that: The pretreatment method comprises: Convert the original image and the image to be detected into single-channel images; Use edge detection algorithms to perform edge recognition on the single-channel images corresponding to the original image and the image to be detected to determine the detection location; Perform perspective transformation on the single-channel image corresponding to the image to be detected so that the edge of the detection part in the single-channel image corresponding to the image to be detected and the edge of the detection part in the single-channel image corresponding to the original image overlap to a set ratio; Overall change the eigenvalues ​​of the single-channel image corresponding to the image to be detected so that the average value of the eigenvalues ​​of the single-channel image corresponding to the image to be detected is the same as the average value of the eigenvalues ​​of the single-channel image corresponding to the original image; The single-channel image corresponding to the original image is cropped and the complete detection part is retained to establish the original matrix. At the same time, the single-channel image after perspective transformation is cropped and the complete detection part is retained to establish the matrix to be detected, thereby reducing the influence of factors outside the detection part on the detection results.

8. The automatic inspection method for architectural ceramics based on visual image analysis according to claim 7, characterized in that: The deformation calculation model is specifically: in Indicates the degree of deformation, and The coefficients representing the spectral energy difference and the overall difference coefficient, respectively, and They respectively represent the contribution of the spectrum energy difference and the overall difference coefficient to the deformation result in the deformation calculation, and and The sum of is 1, the more delicate and complex the surface of the building ceramics is. The larger the value, the The lower the value, the specific value can be determined by the least squares calculation based on a large amount of known data. represents the difference in spectral energy, represents the overall coefficient of variation, represents the maximum value of the spectrum energy difference, Indicates the maximum value of the overall coefficient of variation.

9. An automatic inspection system for architectural ceramics based on visual image analysis, characterized by: include: Data collection module: used to obtain original images and images to be tested of building ceramics; Preprocessing module: preprocess the original image and the image to be detected by the preprocessing method to obtain the single-channel matrix to be detected and the original matrix; Data processing module: Calculate the complexity coefficients of the original matrix and the matrix to be detected respectively through the complexity calculation method, quantify the irregular surface and contour of the building ceramics through the complexity coefficient, and express the dimensional changes of the irregular surface and contour of the building ceramics through the difference between the complexity coefficients; adjust the original matrix and the matrix to be detected using the adjustment method according to the complexity coefficient to obtain the original intermediate matrix and the intermediate matrix to be detected, combine the complexity coefficient with the image data, consider the dimensional changes of the irregular surface and contour of the building ceramics, and improve the detection accuracy; perform Fourier transform on the original intermediate matrix and the intermediate matrix to be detected to obtain frequency domain matrices respectively, and calculate the amplitude spectrum matrix of the original intermediate matrix and the intermediate matrix to be detected through the frequency domain matrix; calculate the spectral energy difference of the amplitude spectrum matrix of the original intermediate matrix and the intermediate matrix to be detected through the spectrum difference calculation method, and calculate the overall difference coefficient of the two amplitude spectrum matrices through the difference calculation method; Data output module: input the spectrum energy difference and the overall difference coefficient into the deformation calculation model to obtain the deformation detection result.

Images