Copper and aluminum dissimilar metal welding quality detection method based on image processing

Abstract

The present invention relates to the field of image processing, and more specifically, to a copper-aluminum dissimilar metal welding quality detection method based on image processing, the method comprising: obtaining a grayscale image of a copper-aluminum dissimilar metal welding image, constructing a grayscale co-occurrence matrix according to a preset direction, clustering the grayscale co-occurrence matrix in each direction to obtain a plurality of cluster clusters; obtaining the position of a pixel point in the grayscale image corresponding to each cluster cluster to construct a scatter plot, obtaining a local area with a preset radius of the pixel point to calculate the outlier index and abnormal possibility of each pixel point in the scatter plot, marking it as an abnormal point in response to the abnormal possibility being greater than an abnormal threshold, obtaining a connected domain composed of all abnormal points, obtaining an abnormal area in the grayscale image, and evaluating the copper-aluminum welding quality of the abnormal area in the welding image. The present invention can accurately identify abnormal textures in the welding area and improve the detection accuracy of welding quality by constructing a grayscale co-occurrence matrix and cluster analysis.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and more specifically, to a copper-aluminum dissimilar metal welding quality detection method based on image processing. Background Art

[0002] Ultrasonic welding of copper and aluminum dissimilar metals is an effective connection technology. Ultrasonic welding is a method of welding that uses the mechanical vibration energy of ultrasonic frequency (over 15KHz) to produce plastic deformation on the surface of the workpiece and destroy the surface layer under pressure. It is determined by three factors: vibration shear force, static pressure, and temperature rise in the welding area. Ultrasonic welding technology is used in many industries.

[0003] The existing Chinese patent application document with publication number CN115330780A discloses a method for rapid detection of metal welding slag inclusion defects, which involves the field of artificial intelligence. It includes: obtaining a weld grayscale image, obtaining the grayscale value of each pixel in the grayscale image; calculating the probability that each grayscale level in the grayscale image belongs to the background grayscale level, and taking the grayscale level corresponding to the maximum probability as the background grayscale level of the grayscale image; calculating the attention level of each pixel point; calculating the degree of confusion of each pixel point; calculating the abnormal probability value of each pixel point; calculating the probability of slag inclusion defects in the weld grayscale image; when the slag inclusion probability is greater than a preset threshold, the weld grayscale image has a slag inclusion defect; determining the slag inclusion defect area of ​​the weld grayscale image according to the abnormal probability value of each pixel point and the average value of the abnormal probability.

[0004] The above application documents only analyze the degree of attention of each pixel through the probability that each grayscale level in the grayscale image belongs to the background grayscale level. At present, in the process of welding quality inspection, there will be a large grayscale change in the junction area of ​​the two metals at the weld. When the quality assessment of the weld area is performed based on the local texture characteristics of the grayscale image, it will be greatly affected. Since the grayscale change caused by the junction of the two metals at the weld is different from the texture of the weld area, it is easy to be misdetected as an abnormal texture, resulting in inaccurate assessment results of the welding quality of copper and aluminum dissimilar metals. Summary of the invention

[0005] In order to solve the problem that the grayscale change caused by the junction of two metals at the welding point is different from the texture of the welding area and is easily misdetected as abnormal texture, the present invention provides solutions in the following aspects.

[0006] The copper-aluminum dissimilar metal welding quality detection method based on image processing comprises: obtaining a grayscale image of a copper-aluminum dissimilar metal welding image; constructing a grayscale co-occurrence matrix according to a preset direction according to the correlation of the grayscale values ​​of the pixels in the grayscale image, clustering the grayscale co-occurrence matrix in each direction to obtain a plurality of cluster clusters; obtaining the position of the pixel points in the grayscale image corresponding to each cluster cluster to construct a scatter plot, obtaining a local area with a preset radius of the pixel points, so as to calculate the outlier index of each pixel point in the scatter plot; obtaining the abnormal possibility of the pixel points in the grayscale image according to the outlier index corresponding to the pixel points in the preset direction, marking the pixel points corresponding to the abnormal possibility as abnormal points in response to the abnormal possibility being greater than the abnormal threshold, obtaining a connected domain formed by all abnormal points, obtaining an abnormal area in the grayscale image, and evaluating the copper-aluminum welding quality of the abnormal area in the welding image; wherein the abnormal possibility satisfies the following relationship: , where Represents the grayscale image OK The probability of abnormality of the column pixel, Represents the grayscale image OK The pixel point in the column The outlier index corresponding to the preset direction, Represents the grayscale image OK The pixel point in the column The dispersion degree of the gray level co-occurrence matrix in a preset direction.

[0007] The effect is as follows: by constructing a grayscale co-occurrence matrix and performing cluster analysis on it, the abnormal texture features in the welding area can be accurately identified; by calculating the outlier index and abnormal possibility of each pixel in the scatter plot, the pixels that are inconsistent with the surrounding texture can be efficiently identified; by obtaining the connected domain composed of all abnormal points, the actual size and shape of the abnormal area can be further confirmed, and then the copper-aluminum welding quality in the welding image can be effectively evaluated, and repaired or improved.

[0008] Preferably, constructing a gray level co-occurrence matrix according to a preset direction includes:

[0009] Determine the preset direction and distance of pixel pairs, create a matrix with the same size as the grayscale level, count the frequency of grayscale values ​​of pixel pairs in the preset direction in the grayscale image, update the co-occurrence matrix, divide all values ​​in the co-occurrence matrix by the sum of all values ​​in the co-occurrence matrix to obtain a grayscale co-occurrence matrix, wherein the grayscale co-occurrence matrix represents the probability distribution of the co-occurrence of pixel pairs.

[0010] The effect is that by analyzing the gray-level co-occurrence matrix, the texture properties of the image can be quantified to identify abnormal textures in the welding area, thereby detecting potential defects or uneven areas. By constructing a gray-level co-occurrence matrix in multiple preset directions, the directional characteristics of the image texture can be more fully understood.

[0011] Preferably, the gray level co-occurrence matrix in each direction is clustered to obtain a plurality of clusters, including:

[0012] Taking any preset direction as the target direction, a preset window is set as the central element value according to any element value in the gray-level co-occurrence matrix corresponding to the target direction, and in response to the central element value in the window being a maximum value, a plurality of clustering centers in the gray-level co-occurrence matrix of the target direction are selected as the clustering center of the gray-level co-occurrence matrix, and the Euclidean distances between the spatial positions of the clustering centers and other element values ​​in the gray-level co-occurrence matrix of the target direction are calculated respectively, wherein the element value is the frequency value of the occurrence of the corresponding pixel value of the pixel pair;

[0013] The other element values ​​are divided into the clusters corresponding to the cluster centers closest to them. In response to the other element values ​​being equidistant from the two cluster centers, the other element values ​​are divided into the clusters with smaller indexes, thereby obtaining multiple clusters.

[0014] The effect is that through cluster analysis, the characteristics of different textures in grayscale images can be identified, pixel pairs with similar texture properties can be classified into the same class, and abnormal areas in grayscale images can be identified, such as welding defects or other irregular textures, which may appear as outliers in the cluster.

[0015] Preferably, obtaining the position of the pixel points in the grayscale image corresponding to each cluster to construct a scatter plot includes:

[0016] A scatter plot is created according to each cluster in the gray level co-occurrence matrix to mark the position of the pixel in the gray image, wherein the marked pixel is the first pixel in each pixel pair in the gray level co-occurrence matrix.

[0017] Preferably, obtaining the outlier index of each pixel in the scatter plot includes:

[0018] Taking any pixel point in the scatter plot as the central pixel point with a preset radius, obtaining the ratio of the number of pixels in the local area of ​​the central pixel point to the average number of pixels in the local area of ​​each pixel point in the scatter plot, and taking the complement of the normalized ratio as the local density difference;

[0019] Obtaining the complement of the ratio between the number of pixels in the connected domain where the central pixel is located and the number of pixels in the largest connected domain in the scatter plot, to obtain a connected domain difference;

[0020] The product of the local density difference and the connected domain difference is used as the outlier index of the central pixel.

[0021] The effect is that the distribution and aggregation of texture features are made intuitively visible through the scatter plot, which helps to understand the spatial distribution of image texture. By calculating the local density difference and the connected domain difference, those pixels that are inconsistent with the surrounding texture can be identified.

[0022] Preferably, the dispersion degree of the gray level co-occurrence matrix in the preset direction is obtained as The average value of the discrete index of each pixel in the scatter plot of the clusters.

[0023] Preferably, the step of obtaining a connected domain formed by all abnormal points further includes:

[0024] The number of pixels in each connected domain is obtained, and the connected domains with less pixel numbers than the preset minimum number of pixels in the connected domain are recorded as isolated connected domains, which are removed from the abnormal data points to obtain the abnormal pixel points after the isolated connected domains are removed.

[0025] The effect is that by eliminating isolated connected domains, it can ensure that the identified abnormal areas are actually present, rather than being composed of noise or small abnormal points, thereby improving the accuracy of detection and thus enhancing the robustness of the detection system.

[0026] Preferably, evaluating the copper-aluminum welding quality of the abnormal area in the welding image includes:

[0027] The ratio of the number of abnormal pixels to the total number of pixels in the copper-aluminum welding quality image is obtained. In response to the ratio being greater than the abnormal area proportion threshold, the quality of the copper-aluminum welded part corresponding to the welding image is poor.

[0028] The present invention has the following effects:

[0029] 1. By constructing a gray-level co-occurrence matrix and cluster analysis, the present invention can accurately identify abnormal textures in the welding area, calculate the outlier index and abnormal possibility of each pixel point, and effectively mark potential quality problems, thereby improving the detection accuracy of welding quality.

[0030] 2. The present invention analyzes the grayscale co-occurrence matrix in multiple preset directions. This method can comprehensively capture the texture features in the image, including directionality and distribution patterns, automatically evaluate the welding quality, reduce the need for manual inspection, improve the efficiency of the inspection process, and speed up the production process. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0032] Figure 1 It is a method flow chart of steps S1 to S4 in the copper-aluminum dissimilar metal welding quality detection method based on image processing in an embodiment of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0034] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0035] Reference Figure 1 The copper-aluminum dissimilar metal welding quality detection method based on image processing includes steps S1 to S4, which are as follows:

[0036] S1: Obtain a grayscale image of a welding image of copper and aluminum dissimilar metals.

[0037] It should be noted that the grayscale image of the welding image is obtained. The grayscale image well preserves the detailed structure and grayscale features in the image. While reducing the image storage space, the detailed features in the image are retained, and the welding quality can still be well detected.

[0038] Furthermore, in the welding images of copper and aluminum dissimilar metals, similar textures are often the same connected domain area in the image. If there are pixels with a large outlier degree or that do not conform to the distribution law of a type of texture pixels in the features corresponding to the pixels of this type, these pixels are considered to be abnormal pixels. Based on this, a grayscale co-occurrence matrix is ​​constructed for the pixels in the grayscale image, and the degree of abnormality of the pixels is measured by the grayscale co-occurrence matrix.

[0039] S2: According to the correlation of the grayscale values ​​of the pixels in the grayscale image, a grayscale co-occurrence matrix is ​​constructed according to a preset direction, and the grayscale co-occurrence matrix in each direction is clustered to obtain multiple clusters.

[0040] Determine the preset direction and distance of pixel pairs, create a matrix with the same size as the grayscale level, count the frequency of grayscale values ​​of pixel pairs in the preset direction in the grayscale image, update the co-occurrence matrix, divide all values ​​in the co-occurrence matrix by the sum of all values ​​in the co-occurrence matrix to obtain the grayscale co-occurrence matrix, where the grayscale co-occurrence matrix represents the probability distribution of the co-occurrence of pixel pairs.

[0041] That is to say, according to the 8 neighboring pixels of each pixel, the gray level co-occurrence matrix of four directions can be obtained. Specifically, in this embodiment, the preset directions are: 0 degree, 45 degree, 90 degree, and 135 degree, the pixel pair distance is 1, and the gray level co-occurrence matrix of four scanning directions is constructed. Since the four scanning directions of 180 degree, 225 degree, 270 degree, and 315 degree have the same texture features as the four selected scanning directions, there is no need to consider the textures of these four scanning directions;

[0042] To further illustrate, create a matrix with the same size as the number of gray levels. The size of the matrix is .

[0043] For example, count any pixel point in the directions of 0 degrees, 45 degrees, 90 degrees, and 135 degrees respectively The gray value is , according to the preset distance point The gray value is Probability, statistics The frequency of occurrence, among which, , Represents the minimum gray value in the grayscale image. Indicates the maximum gray value in the edge image.

[0044] Taking any preset direction as the target direction, a preset window is set according to any element value in the gray level co-occurrence matrix corresponding to the target direction as the central element value, and in response to the central element value in the window being the maximum value, it is used as the cluster center of the gray level co-occurrence matrix, and multiple cluster centers in the gray level co-occurrence matrix of the target direction are selected, and the Euclidean distances between the spatial positions of other element values ​​in the gray level co-occurrence matrix of the target direction and the cluster centers are calculated respectively, wherein the element value is the frequency value of the corresponding pixel value of the pixel pair;

[0045] Further explanation, the default window is , this embodiment does not limit the preset window range, and the implementer can set the preset window range according to the actual implementation situation.

[0046] The other element values ​​are divided into the clusters corresponding to the cluster centers closest to them. In response to the other element values ​​being equidistant from the two cluster centers, the other element values ​​are divided into the clusters with smaller indexes, thereby obtaining multiple clusters.

[0047] It should be noted that for any pixel in the scatter plot, since the distribution patterns of different textures in the image are different, directly measuring the outlier degree based on the absolute density cannot meet the distribution characteristics of different images. Therefore, it is necessary to compare the local density of each pixel with the density of other pixels in the local area to obtain the relative outlier index of each pixel.

[0048] S3: Obtain the position of the pixel points in the grayscale image corresponding to each cluster to construct a scatter plot, and obtain the local area with a preset radius of the pixel points to calculate the outlier index of each pixel point in the scatter plot.

[0049] A scatter plot is created for each cluster in the gray-level co-occurrence matrix to mark the position of the pixel in the gray-level image, where the marked pixel is the first pixel in each pixel pair in the gray-level co-occurrence matrix.

[0050] Taking any pixel point in the scatter plot as the central pixel point with a preset radius, obtain the ratio of the number of pixels in the local area of ​​the central pixel point and the average number of pixels in the local area of ​​each pixel point in the scatter plot, and normalize the complement of the ratio as the local density difference;

[0051] That is, the preset radius is This embodiment does not limit the preset radius, and the implementer can set the preset radius according to the actual implementation situation.

[0052] Obtain the complement of the ratio between the number of pixels in the connected domain where the central pixel is located and the number of pixels in the largest connected domain in the scatter plot to obtain the connected domain difference;

[0053] The product of the local density difference and the connected domain difference is taken as the outlier index of the central pixel.

[0054] Specifically, the outlier index satisfies the following relationship:

[0055] ;

[0056] In the formula, Indicates The gray level co-occurrence matrix in the direction The scatter plot of the clusters The outlier index of the pixel, Indicates The gray level co-occurrence matrix in the direction The scatter plot of the clusters The number of pixels corresponding to other pixels in the local area. Indicates The gray level co-occurrence matrix in the direction The average number of other pixels in the local area of ​​each pixel in the scatter plot of the clusters, Indicates The gray level co-occurrence matrix in the direction In the scatter plot of clusters The number of pixels in the connected domain where the pixel is located, Indicates The gray level co-occurrence matrix in the direction The number of scattered points contained in the largest connected domain in the scatter plot of clusters, Indicates The gray level co-occurrence matrix in the direction The total number of pixels in the scatter plot of clusters, represents the linear normalization function.

[0057] That is to say, if the local density difference is greater than 1, the normalized difference is smaller, indicating that the local density of the pixel is higher; the smaller the connected domain difference, the pixel may be isolated or in a smaller connected domain; the higher the outlier index, the greater the difference between the pixel and the pixels in the same cluster, and it may be an outlier.

[0058] S4: According to the outlier index corresponding to the pixel point in the preset direction, the abnormal possibility of the pixel point in the grayscale image is obtained. In response to the abnormal possibility being greater than the abnormal threshold, the pixel point corresponding to the abnormal possibility is marked as an abnormal point, and the connected domain formed by all abnormal points is obtained to obtain the abnormal area in the grayscale image, and the copper-aluminum welding quality of the abnormal area in the welding image is evaluated.

[0059] Further explanation: the abnormal threshold is , This embodiment does not limit the abnormal threshold, and the implementer can set it according to the actual implementation situation.

[0060] Specifically, the abnormal probability satisfies the following relationship:

[0061] ;

[0062] In the formula, Represents the grayscale image OK The probability of abnormality of the column pixel, Represents the grayscale image OK The pixel point in the column The outlier index corresponding to the preset direction, Represents the grayscale image OK The pixel point in the column The dispersion degree of the gray level co-occurrence matrix in a preset direction.

[0063] It should be noted that for a scatter plot in any direction, the discrete degree of the scatter points in the grayscale image itself is relatively high, so the discrete degree of each pixel point will be too high, resulting in inaccurate calculation. Therefore, it is necessary to obtain the dispersion degree of the scatter points of various types of textures corresponding to the grayscale co-occurrence matrix in each direction.

[0064] Get the dispersion degree of the gray level co-occurrence matrix of the preset direction as The average value of the discrete index of each pixel in the scatter plot of the clusters.

[0065] The number of pixels in each connected domain is obtained, and the connected domains with less pixel numbers than the preset minimum number of pixels in the connected domain are recorded as isolated connected domains, which are removed from the abnormal data points to obtain the abnormal pixel points after the isolated connected domains are removed.

[0066] It should be noted that the minimum number of pixels in the preset connected domain is 30.

[0067] The ratio of the number of abnormal pixels to the total number of pixels in the copper-aluminum welding quality image is obtained. In response to the ratio being greater than the abnormal area proportion threshold, the quality of the copper-aluminum welded part corresponding to the welding image is poor.

[0068] Further explanation: the abnormal area ratio threshold is This embodiment does not limit the threshold of abnormal area proportion, and the implementers can set it according to the actual implementation situation.

[0069] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly defined.

[0070] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A copper-aluminum dissimilar metal welding quality detection method based on image processing, characterized in that: include: Obtaining a grayscale image of a welding image of copper and aluminum dissimilar metals; According to the correlation of the grayscale values ​​of the pixels in the grayscale image, a grayscale co-occurrence matrix is ​​constructed according to a preset direction, and the grayscale co-occurrence matrix in each direction is clustered to obtain a plurality of clusters; Obtain the position of the pixel points in the grayscale image corresponding to each cluster to construct a scatter plot, and obtain the local area with a preset radius of the pixel points to calculate the outlier index of each pixel point in the scatter plot; According to the preset direction, the outlier index corresponding to the pixel point is obtained to obtain the abnormal possibility of the pixel point in the grayscale image. In response to the abnormal possibility being greater than the abnormal threshold, the pixel point corresponding to the abnormal possibility is marked as an abnormal point, and the connected domain formed by all abnormal points is obtained to obtain the abnormal area in the grayscale image, and the copper-aluminum welding quality of the abnormal area in the welding image is evaluated; The abnormal probability satisfies the following relationship: , where Represents the grayscale image OK The probability of abnormality of the column pixel, Represents the grayscale image OK The pixel point in the column The outlier index corresponding to the preset direction, Represents the grayscale image OK The pixel point in the column The dispersion degree of gray-level co-occurrence matrix in preset directions; Obtaining the outlier index of each pixel in the scatter plot includes: Taking any pixel point in the scatter plot as the preset radius of the central pixel point, obtain the ratio of the number of pixels in the local area of ​​the central pixel point and the average value of the number of pixels in the local area of ​​each pixel point in the scatter plot, and use the complement of the ratio after normalization as the local density difference; obtain the complement of the ratio between the number of pixels in the connected domain where the central pixel point is located and the number of pixels in the largest connected domain in the scatter plot to obtain the connected domain difference; and use the product of the local density difference and the connected domain difference as the outlier index of the central pixel point.

2. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: Construct a gray-level co-occurrence matrix according to the preset direction, including: Determine the preset direction and distance of pixel pairs, create a matrix with the same size as the grayscale level, count the frequency of grayscale values ​​of pixel pairs in the preset direction in the grayscale image, update the co-occurrence matrix, divide all values ​​in the co-occurrence matrix by the sum of all values ​​in the co-occurrence matrix to obtain a grayscale co-occurrence matrix, wherein the grayscale co-occurrence matrix represents the probability distribution of the co-occurrence of pixel pairs.

3. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: The grayscale co-occurrence matrix in each direction is clustered to obtain multiple clusters, including: taking any preset direction as the target direction, presetting a window as the central element value according to any element value in the grayscale co-occurrence matrix corresponding to the target direction, in response to the central element value in the window being a maximum value, taking it as the cluster center of the grayscale co-occurrence matrix, selecting multiple cluster centers in the grayscale co-occurrence matrix of the target direction, and calculating the Euclidean distance between the spatial position of other element values ​​in the grayscale co-occurrence matrix of the target direction and the cluster center, respectively, wherein the element value is the frequency value of the occurrence of the corresponding pixel value of the pixel pair; dividing the other element values ​​into the cluster cluster corresponding to the cluster center closest to it, and in response to the other element values ​​being equal to the distance between the two cluster centers, dividing the other element values ​​into the cluster cluster with the smaller index, and obtaining multiple cluster clusters.

4. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: Get the position of the pixel points in the grayscale image corresponding to each cluster to build a scatter plot, including: A scatter plot is created according to each cluster in the gray level co-occurrence matrix to mark the position of the pixel in the gray image, wherein the marked pixel is the first pixel in each pixel pair in the gray level co-occurrence matrix.

5. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: The dispersion degree of the gray level co-occurrence matrix in the preset direction is obtained as The average value of the discrete index of each pixel in the scatter plot of the clusters.

6. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: The step of obtaining a connected domain formed by all abnormal points further includes: The number of pixels in each connected domain is obtained, and the connected domains with less pixel numbers than the preset minimum number of pixels in the connected domain are recorded as isolated connected domains, which are removed from the abnormal data points to obtain the abnormal pixel points after the isolated connected domains are removed.

7. The copper-aluminum dissimilar metal welding quality detection method based on image processing according to claim 1 is characterized in that: Evaluate the copper-aluminum welding quality of abnormal areas in welding images, including: The ratio of the number of abnormal pixels to the total number of pixels in the copper-aluminum welding quality image is obtained. In response to the ratio being greater than the abnormal area proportion threshold, the quality of the copper-aluminum welded part corresponding to the welding image is poor.

Images