A method and system for detecting the production quality of finned tubes of an air cooler

Abstract

The present invention belongs to the technical field of image processing, and particularly relates to a method and system for detecting the production quality of air cooler fins. The method includes: converting the binary image corresponding to each gray value in the air cooler fin image to the Hough space, clustering the high-brightness points corresponding to all binary images in the Hough space, grouping the pixel points in the air cooler fin image according to the clustering result, determining the first defect probability of each group according to the distribution shape and distribution direction of all pixel points in each group, determining the second defect probability of the pixel points in each group according to the gray distribution law of all pixel points in each group, screening the first defect area according to the magnitudes of the first defect probability and the second defect probability, and determining the second defect area according to the distribution position relationship between each connected domain composed of all pixel points that do not belong to any group and each group. The present invention eliminates the influence of uneven illumination and improves the accuracy of detecting the production quality of air cooler fins.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing. More specifically, the present invention relates to a method and system for detecting the production quality of air cooler fins. Background Art

[0002] Air cooler fins are key components of air coolers. During the production process of air cooler fins, defects such as deformation, cracks, scratches, and uneven coatings may appear on the fin surface. These defects directly affect the heat dissipation efficiency and service life of the equipment.

[0003] With the improvement of industrial automation and intelligence levels, the detection of defects such as deformation, cracks, scratches, and uneven coatings on air cooler fins has gradually shifted to automated detection methods based on image recognition. For example, the Chinese patent application document with the application publication number CN119147543A and the name of "Defect Detection Method and Device for Air Conditioner Fins", and the Chinese patent application document with the authorization announcement number CN114943736B and the name of "A Method and System for Detecting the Production Quality of Automobile Radiator Fins" both use image recognition technology to detect defects in air cooler fin images.

[0004] However, the clarity and distinguishability of defects in air cooler fin images are often affected by lighting conditions. Different angles and intensities of light will cause shadows and highlight areas to appear on the fin surface, making it difficult to clearly identify the edges of defects on the fin surface. Uneven lighting will cause some small defects to be obscured or blurred, and may even lead to misdetection or missed detection.

[0005] Therefore, how to overcome the impact of lighting changes on the quality of air cooler fin images and improve the accuracy of air cooler fin defect detection has become an urgent problem to be solved in current image processing technology. Summary of the Invention

[0006] To solve the above technical problem of lighting changes affecting the defect detection of air cooler fin images, the present invention provides solutions in the following aspects.

[0007] In a first aspect, the present invention provides a method for detecting the production quality of air cooler fins, including:

[0008] Collect the fin images of the air-cooled heat exchanger; obtain the binary images corresponding to each gray value in the fin images of the air-cooled heat exchanger, and convert the binary images to the Hough space; cluster the high-brightness points corresponding to all the binary images in the Hough space, and group the pixel points in the fin images of the air-cooled heat exchanger according to the clustering results; determine the first defect probability of each group according to the distribution shape and distribution direction of all the pixel points in each group, determine the second defect probability of each pixel point in each group according to the gray distribution law of all the pixel points in each group, and screen the first defect regions according to the magnitudes of the first defect probability and the second defect probability; perform connected component analysis on all the pixel points that do not belong to any group to obtain multiple connected components; determine the second defect regions according to the distribution position relationship between the connected components and each group.

[0009] In the present invention, by clustering the high-brightness points corresponding to the binary images of each gray value in the Hough space, the interference of light is excluded, so that each clustering result represents one or more nearly parallel straight lines in the image. In the case of no defects, the pixel points of each group obtained according to the clustering results are regularly arranged along the direction of the fin tube, effectively reflecting the structural characteristics of the fin tube, providing a basis for subsequent defect detection; the pixel points in the normal region will be arranged along the direction of the fin tube, while the pixel points in the defect region may be distributed deviating from the fin tube. By analyzing the distribution shape and distribution direction of all the pixel points in each group, the present invention obtains the first defect probability of each group and can identify the defect regions inconsistent with the direction of the fin tube; under the influence of the fins, the gray values of the pixel points on the fin tube are regularly distributed in a cycle. According to the gray distribution law of all the pixel points in each group, the present invention determines the second defect probability of each pixel point in each group and can screen out the defect regions in the group with the distribution direction consistent with the direction of the fin tube; the defects that do not belong to any group may change the straight line of the originally regularly arranged fin tube into multiple interrupted line segments. Therefore, by comparing the distribution position relationship between the connected components composed of the pixel points that do not belong to any group and each group, the present invention can locate the defects outside the group. In summary, the present invention excludes the interference of light, can stably extract the defect characteristics of the fins under different lighting conditions, and ensures the high efficiency and accuracy of the detection system in various environments.

[0010] Preferably, the obtaining of the binary images corresponding to each gray value in the fin images of the air-cooled heat exchanger includes: taking any gray value as the target gray value, setting the target gray value in the fin images of the air-cooled heat exchanger to 1, and setting the remaining gray values to 0 to obtain the binary image corresponding to the target gray value.

[0011] Preferably, the grouping of the pixel points in the fin images of the air-cooled heat exchanger according to the clustering results includes: dividing the pixel points corresponding to all the high-brightness points in each category in the clustering results in the fin images of the air-cooled heat exchanger into one group.

[0012] The arrangement trend of pixel points in each group divided according to the clustering results can reflect the arrangement characteristics of the finned tubes. When there are no defects, the pixel points within each group are neatly arranged along the direction of the finned tubes, which can effectively reduce the interference caused by noise or uneven illumination, ensure the accurate identification of the arrangement direction of the finned tubes, and help identify defects in combination with the arrangement direction of the finned tubes in the subsequent process.

[0013] Preferably, the first defect probability satisfies the expression: ; where is the first defect probability of the th group; , are respectively the length of the short side and the length of the long side of the minimum bounding rectangle of the th group; is the acute angle between the straight line where the long side of the minimum bounding rectangle of the th group is located and the straight line where the long side of the minimum bounding rectangle of the th group is located; is the number of groups.

[0014] When calculating the first defect probability, the present invention considers the ratio of the lengths of the short side and the long side of the minimum bounding rectangle of each group and the angular difference from other groups, providing a geometric basis for defect identification. If the distribution direction of a certain group is significantly different from the trend of the finned tubes and the angular difference is large, then this group is more likely to be a defect area.

[0015] Preferably, determining the second defect probability of each pixel point in each group includes: performing a Z-shaped scan on the minimum bounding rectangles of each group, and forming a gray level sequence of each group with the gray level values of the pixel points belonging to each group in the minimum bounding rectangles; performing STL decomposition on the gray level sequences of each group; determining the second defect probability of the th pixel point in the th group according to the periodic term and the residual term obtained from the STL decomposition : , is the residual term corresponding to the th pixel point in the th group; , are respectively the average difference between the maximum value point closest to the periodic term corresponding to the th pixel point and the remaining maximum value points, and the average difference between the minimum value point closest to the periodic term corresponding to the th pixel point and the remaining minimum value points in the periodic term curve of the , are hyperparameters, is the exponential function with the natural constant as the base, is the hyperbolic tangent function.

[0016] Uneven illumination will cause brightness differences at different positions of the finned tube, disturbing the periodic distribution law of the gray values on the finned tube. Through STL decomposition, the present invention removes the trend term, i.e., the influence of external factors such as illumination, and can more accurately capture the true periodic law of the pixel points of the finned tube. By combining the periodic term and the residual term, it is possible to effectively distinguish normal pixel points from defective pixel points that do not conform to the periodic law.

[0017] Preferably, the screening of the first defect area according to the first defect probability magnitude and the second defect probability magnitude includes: in response to the first defect probability being greater than a preset first defect threshold, the area formed by all pixel points in the corresponding group is the first defect area; or in response to the second defect probability being greater than a preset second defect threshold, the corresponding pixel points are taken as defective pixel points, and the area formed by all defective pixel points is the first defect area.

[0018] Preferably, the determination of the second defect area according to the distribution position relationship between the connected component and each group includes: in response to the number of pixel points included in the connected component being greater than 2, the boundary pixel points of the connected component are divided into multiple pixel pairs, the pixel pair includes two boundary pixel points, and the two boundary pixel points in the pixel pair and the centroid of the connected component are on the same straight line; in response to any adjacent pixel point of one boundary pixel point in the pixel pair and any adjacent pixel point of the other boundary pixel point belonging to the same group, the connected component is taken as the second defect area.

[0019] By analyzing the distribution position relationship between the connected component with the number of pixel points greater than 2 and each group, the present invention not only excludes the interference of noise points but also can accurately identify defects outside the group.

[0020] Preferably, the mean shift clustering algorithm is used for the clustering.

[0021] Preferably, the Two-Pass algorithm is used for the connected component analysis.

[0022] In a second aspect, the present invention provides an air-cooled fin production quality detection system, including a processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned air-cooled fin production quality detection method is implemented.

[0023] By adopting the above technical solution, the above-mentioned air-cooled fin production quality detection method is generated into a computer program and stored in the memory to be loaded and executed by the processor, so as to manufacture a terminal device according to the memory and the processor, which is convenient to use.

[0024] The beneficial effects of the present invention are as follows: The present invention can eliminate the influence of light on the image quality of the fins of the air cooler, exclude the interference of light on the defect detection of the fins of the air cooler, and can stably extract the defect features of the fins under different lighting conditions, ensuring the high efficiency and accuracy of the detection system in various environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 FIG. is a flowchart schematically showing a method for detecting the production quality of fins of an air cooler in the present invention;

[0026] Figure 2 FIG. is a schematic diagram of an image of the fins of an air cooler;

[0027] Figure 3 FIG. is a schematic diagram of a binary image obtained by segmenting the image of the fins of the air cooler using the Otsu threshold segmentation algorithm;

[0028] Figure 4 FIG. is a schematic diagram of a binary image with a gray value of 37;

[0029] Figure 5 FIG. is a schematic diagram of the Hough space corresponding to the binary image with a gray value of 37;

[0030] Figure 6 FIG. is a schematic diagram of a finned tube;

[0031] Figure 7 FIG. is a schematic diagram of a straight line on the finned tube;

[0032] Figure 8 FIG. is a schematic diagram of the high-brightness points corresponding to the binary image with a gray value of 37 in the Hough space returned to the image of the fins of the air cooler;

[0033] Figure 9 FIG. is a flowchart of step S4 of a method for detecting the production quality of fins of an air cooler in the present invention;

[0034] Figure 10 FIG. is a schematic diagram of a pixel pair. DETAILED DESCRIPTION OF THE INVENTION

[0035] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0036] Next, the specific implementation manners of the present invention will be described in detail with reference to the accompanying drawings.

[0037] An embodiment of the present invention discloses a method for detecting the production quality of air cooler fins, referring to Figure 1 , including steps S1 - S5:

[0038] S1. Collect images of air cooler fins.

[0039] Lay the air cooler fins flat and use a camera to take a top - down image of the air cooler fins. In the present invention, for the convenience of subsequent processing, the captured image of the air cooler fins is a grayscale image and only contains the fin area, without including other areas. Figure 2 is a schematic diagram of the air cooler fin image.

[0040] S2. Obtain the binary image corresponding to each grayscale value in the air cooler fin image and convert the binary image to the Hough space.

[0041] It should be noted that the air cooler is composed of a base tube and fins. The fins are tightly attached to the base tube through processes such as welding and expansion jointing to form finned tubes. Multiple finned tubes in the air cooler are distributed in parallel, and the fins on the finned tubes are regularly spaced. The straight lines in the air cooler fin image can reflect the arrangement trend of the finned tubes. The present invention uses the Hough line detection algorithm to identify the straight lines in the air cooler fin image, so as to identify the defects of the air cooler fins according to the arrangement trend of the finned tubes. The Hough line detection algorithm is for binary images. Due to the complex shooting environment and uneven light distribution, the brightness of different positions of the finned tubes in the obtained air cooler fin image is different. Directly performing binarization on the air cooler fin image has a poor effect and it is difficult to completely identify the arrangement trend of the finned tubes in the air cooler fin image. For example Figure 3 is a schematic diagram of the binary image obtained by segmenting the air cooler fin image using the Otsu threshold segmentation algorithm. The finned tube in the upper right corner of the image cannot be recognized. Therefore, the present invention obtains the binary image for each grayscale value respectively, so as to subsequently completely identify the arrangement trend of the finned tubes in the air cooler fin image according to the distribution of the high - light points corresponding to all binary images in the Hough space.

[0042] Specifically, take any grayscale value as the target grayscale, set the target grayscale in the air cooler fin image to 1, and the remaining grayscales to 0 to obtain the binary image corresponding to the target grayscale. For example Figure 4 is a schematic diagram of the binary image with a grayscale value of 37.

[0043] Similarly, obtain the binary image for each grayscale in the air cooler fin image.

[0044] Use the Hough line detection algorithm to convert the binary image of each grayscale into the Hough space. For example Figure 5 is a schematic diagram of the Hough space corresponding to the binary image with a grayscale value of 37.

[0045] S3. Cluster the high-brightness points corresponding to all binary images in the Hough space, and group the pixel points in the fin image of the air-cooled heat exchanger according to the clustering results.

[0046] It should be noted that each high-brightness point in the Hough space represents a straight line in the image space. The brightness of the high-brightness point represents the number of pixel points on the corresponding straight line in the image space. The abscissa of the high-brightness point represents the angle of the straight line in the image space, and the ordinate represents the distance from the straight line in the image space to the origin. When the distance between high-brightness points is closer, the corresponding straight lines in the image space are more similar.

[0047] In this embodiment, cluster the high-brightness points corresponding to all binary images with different gray values in the Hough space, and divide the high-brightness points into multiple categories. The clustering method adopted in this embodiment is the mean shift clustering algorithm. In other embodiments, the implementer can select a clustering algorithm according to needs, such as the Density-Based Spatial Clustering of Applications with Noise (DBSCAN).

[0048] It should be noted that a high-brightness point corresponds to multiple pixel points on the same straight line in the binary image, and the binary image is converted from the fin image of the air-cooled heat exchanger. Each pixel point in the binary image corresponds one-to-one to each pixel point in the fin image of the air-cooled heat exchanger. Therefore, a high-brightness point corresponds to multiple pixel points on the same straight line in the fin image of the air-cooled heat exchanger.

[0049] In this embodiment, divide the pixel points corresponding to all high-brightness points in each category in the fin image of the air-cooled heat exchanger into one group.

[0050] It should be noted that due to the curvature of the surface of the finned tube of the air-cooled heat exchanger, the reflection angles of the finned tube at different positions for light are different. Therefore, the finned tube shows the characteristics of being bright in the middle position and dark on both sides. Figure 6 As shown in the schematic diagram of the finned tube, it can be seen that the middle position of the finned tube is bright and the two sides are dark. In the case of absolutely uniform light and no noise, the gray values of the pixel points belonging to the base tube in the middle position of the finned tube are the same, the gray values of the pixel points belonging to the fins are the same, the gray values of the pixel points belonging to the base tube on both sides are the same, and the gray values of the pixel points belonging to the fins are the same. This makes each high-brightness point in the Hough space correspond to multiple straight lines parallel to the finned tube on the finned tube. For example Figure 7 As shown in the schematic diagram of the straight lines on the finned tube. Due to uneven light and the existence of noise, in the case of no defects, each gray value is concentrated in a local position, and the distribution direction is parallel to the direction of the finned tube. Furthermore, after the high-brightness points corresponding to each binary image with a certain gray value in the Hough space are returned to the fin image of the air-cooled heat exchanger, they are relatively short line segments parallel to the direction of the finned tube on the finned tube. For exampleFigure 8 It is a schematic diagram of the high-brightness points corresponding to the binary image with a gray value of 37 in the Hough space returned to the air-cooled fin image. The short line segments with different gray values together form multiple straight lines parallel to the fin tube direction. Therefore, the high-brightness points in each category obtained by clustering correspond to a straight line or multiple approximately parallel straight lines with close positions in the air-cooled fin image. When there are no defects, the pixel points in the group divided according to each category are arranged along the direction of the fin tube, which can completely reflect the arrangement trend of the fin tubes.

[0051] S4. Identify defects for each group.

[0052] The flowchart of step S4 is referred to Figure 9 , including step S401 to step S403, specifically:

[0053] S401. Determine the first defect probability of each group according to the distribution shape and distribution direction of all pixel points in each group.

[0054] It should be noted that when there are defects, if there are continuously distributed pixel points with the same gray value in the defects, they may also be recognized as straight lines. After clustering in step S3, the defects inconsistent with the fin tube direction correspond to a separate group. Therefore, the present invention judges each group to identify defects. Since the defects are usually located in a local area of the air-cooled cooler and the overall distribution direction is different from the fin tube direction, while the pixel points in the group reflecting the arrangement trend of the fin tubes usually cover the entire fin tube and the overall distribution direction is consistent with the fin tube direction, the present invention determines the first defect probability of each group according to the distribution shape and distribution direction of each group.

[0055] Specifically, for any group, obtain the minimum circumscribed rectangle of all pixel points in the group, and determine the first defect probability of the group according to the shape and distribution direction of the minimum circumscribed rectangle:

[0056] ;

[0057] Among them, is the first defect probability of the th group; is the length of the short side of the minimum circumscribed rectangle of the th group; is the length of the long side of the minimum circumscribed rectangle of the th group; is the acute angle between the straight line where the long side of the minimum circumscribed rectangle of the th group is located and the straight line where the long side of the minimum circumscribed rectangle of the th group is located; is the number of groups;

[0058] Able to reflect the shape of the minimum bounding rectangle of the th group. When is closer to 1, the minimum bounding rectangle of the th group is closer to a square. When is smaller, the minimum bounding rectangle of the i-th group is more slender. After clustering in step S3, the groups obtained are probably one or more approximately parallel straight lines along the fin tube direction, and their minimum bounding rectangles are slender. Defects are usually concentrated, so when is larger, the th group is more likely to be a defect. When is smaller, the

[0059] th group is more likely to be the pixel points corresponding to the normal fin tube. The direction of the straight line where the long side of the minimum bounding rectangle of each group is located reflects the distribution direction of the pixel points of that group. After clustering in step S3, the groups obtained are probably one or more approximately parallel straight lines along the fin tube direction. The distribution directions of the minimum bounding rectangles of these groups are basically parallel to the fin tube direction, that is, the distribution directions of the minimum bounding rectangles between these groups are basically the same; while the distribution directions of defects are usually uncertain and may not be consistent with the fin tube direction. Let be the included angle between the distribution direction of the minimum bounding rectangle of the th group and the distribution directions of the minimum bounding rectangles of the other groups. When the included angle is larger, the th group is more likely to be a defect. When the included angle is smaller, the

[0060] th group is more likely to be one or more approximately parallel straight lines along the fin tube direction.

[0061] S402. Determine the second defect probability of each pixel point in each group according to the gray distribution law of all pixel points in each group.

[0062] It should be noted that when the angle of the defect is exactly the same as the direction of the fin tube, it may be grouped with the normal pixel points on one or more approximately parallel straight lines along the direction of the fin tube, and it is difficult to identify using the first defect probability. Since the fins protrude compared to the fin tubes, the fins reflect more light, so the pixel points at the fin positions on the fin tubes are brighter than the pixel points at the base tube positions. Moreover, the fins on the fin tubes are regularly spaced. Therefore, in the absence of the influence of light and defects, the gray values of the pixel points in the group corresponding to one or more approximately parallel straight lines along the direction of the fin tube are periodically distributed. When there are defects in the group, the gray distribution of the defects is complex and does not conform to the periodic distribution law of the gray values of the pixel points on the fin tubes. Therefore, the defects in the group can be identified according to the periodic distribution law of the gray values of the pixel points. However, due to the complex shooting environment and uneven light distribution, the brightness of different positions of the fin tubes in the finned-tube image of the air cooler obtained is different, which interferes with the periodic distribution law of the gray values of the pixel points on the fin tubes, resulting in the inability to directly analyze the periodic distribution law of the gray values of the pixel points on the fin tubes. Therefore, the present invention first performs STL decomposition on the gray distribution of the pixel points in each group to remove the trend term of the gray distribution of the pixel points in each group to eliminate the influence of light, and then determines the second defect probability of each pixel point in each group according to the periodic law of the gray distribution of the pixel points in each group and the residual term of the pixel points.

[0063] Specifically, perform a Z-shaped scan on the minimum bounding rectangle of each group, and form a gray sequence of each group with the gray values of the pixel points belonging to each group in the minimum bounding rectangle. Perform STL (Seasonal-Trend Decomposition using LOESS) decomposition on the gray sequences of each group to obtain a trend term, a periodic term, and a residual term. Each pixel point in each group corresponds to a periodic term and a residual term, and the periodic terms corresponding to all the pixel points in each group form a periodic term curve of each group.

[0064] Determine the second defect probability of each pixel point in each group according to the periodic term and the residual term obtained by STL decomposition:

[0065] ;

[0066] where is the second defect probability of the th pixel point in the th group; is the residual term corresponding to the th pixel point in the th group; is the average difference between the maximum value point closest to the periodic term corresponding to the th pixel point in the periodic term curve of the th group and the other maximum value points; is the The average difference between the minimum value point closest to the periodic term corresponding to the th pixel point in the periodic term curve of a group and the remaining minimum value points; , is a hyperparameter, used to prevent from being too large, resulting in always approaching 0, with an empirical value of 5, used to prevent from being too large, resulting in always approaching 1, with an empirical value of 3. Implementers can set and according to the actual implementation situation; is the absolute value symbol; is the exponential function with the natural constant as the base, used to perform a negative correlation mapping on and limit within the range of [0, 1]; is the hyperbolic tangent function, used to perform a positive correlation mapping on and limit within the range of [0, 1].

[0067] When the absolute value of the residual term corresponding to the th pixel point in the th group is larger, the th pixel point conforms less to the periodic law of the gray level distribution of the pixel points in the th group. At this time, the second defect probability of the th pixel point is larger, and the th pixel point is more likely to be a defective pixel point. When the average difference between the maximum value point closest to the periodic term corresponding to the th pixel point in the periodic term curve and the remaining maximum value points is larger, or the average difference between the minimum value point closest to the periodic term corresponding to the th pixel point in the periodic term curve and the remaining minimum value points is larger, it indicates that the period in which the th pixel point is located is more different from the other periods, and the period in which the th pixel point is located conforms less to the periodic law of the gray level distribution of the pixel points in the th group. At this time, the th pixel point is more likely to be located inside the defect. Therefore, the present invention uses as the exponent of to increase . When is larger, is smaller, and for The greater the degree of increase, the greater the second defect probability of the th pixel in the th group.

[0068] S403. Screen the first defect region according to the magnitudes of the first defect probability and the second defect probability.

[0069] In response to the first defect probability being greater than a preset first defect threshold, the region formed by all pixel points in the corresponding group is the first defect region; or in response to the second defect probability being greater than a preset second defect threshold, the corresponding pixel points are regarded as defect pixel points, and the region formed by all defect pixel points is the first defect region.

[0070] Conversely, in response to the first defect probability not being greater than the preset first defect threshold and the second defect probability of all pixel points in the corresponding group not being greater than the preset second defect threshold, the corresponding group is not defective.

[0071] Among them, the specific values of the first defect threshold and the second defect threshold are set by the implementer according to the actual application scenario and requirements. For example, the first defect threshold is 0.2 and the second defect threshold is 0.9. However, it should be noted that since the value ranges of the first defect probability and the second defect probability are [0, 1], the setting ranges of the first defect threshold and the second defect threshold are also [0, 1].

[0072] S5. Perform connected component analysis on all pixel points that do not belong to any group to obtain multiple connected components, and determine the second defect region according to the distribution position relationship between the connected components and each group.

[0073] It should be noted that the gray levels inside some defects are complex, and the gray levels inside them cannot be recognized as straight lines. Therefore, the corresponding pixel points are not included in the groups obtained by clustering in step S3. In the present invention, defect analysis is performed on pixel points that do not belong to any group. In the case of no defect interference, each group obtained in step S3 is one or more approximately parallel straight lines along the fin tube direction. When there are defects, the defects may truncate the straight lines, making the straight lines become multiple line segments that are disconnected in the middle. Therefore, the present invention determines the second defect region according to the distribution position relationship between all pixel points that do not belong to any group and each group.

[0074] Specifically, the algorithm adopted by the present invention for performing connected component analysis on all pixel points that do not belong to any group is the two-pass method, and the implementer can select a connected component analysis algorithm according to the actual implementation situation, such as the seed filling method.

[0075] For each connected component, if the number of pixel points contained in the connected component is no more than 2, the connected component is considered noise. If the number of pixel points contained in the connected component is greater than 2, defect analysis is performed on the connected component:

[0076] First, the boundary pixel points of the connected component are divided into multiple pixel pairs, where each pixel pair contains two boundary pixel points, and the two boundary pixel points in the pixel pair and the centroid of the connected component are on the same straight line. For example Figure 10 is a schematic diagram of a pixel pair, Figure 10 the gray area in is the connected component, O is the centroid of the connected component, H1 and H2 are a pair of pixel pairs, and H3 and H4 are a pair of pixel pairs.

[0077] In response to any adjacent pixel point of one boundary pixel point in the pixel pair and any adjacent pixel point of the other boundary pixel point belonging to the same group, the connected component is taken as the second defect area.

[0078] Thus far, the quality inspection of the fins of the air cooler has been realized.

[0079] An embodiment of the present invention also discloses a quality inspection system for the production of fins of an air cooler, including a processor and a memory. The memory stores computer program instructions, and when the computer program instructions are executed by the processor, a quality inspection method for the production of fins of an air cooler according to the present invention is realized.

[0080] The above system also includes other components well-known to those skilled in the art such as a communication bus and a communication interface. Their settings and functions are known in the art, so they will not be elaborated here.

Claims

1. A method for detecting the production quality of fins of an air cooler, characterized in that, Including: Collecting fin images of an air-cooled heat exchanger; Obtaining a binary image corresponding to each gray value in the fin image of the air-cooled heat exchanger, and converting the binary image into the Hough space; Clustering the highlight points corresponding to all binary images in the Hough space, and grouping the pixel points in the fin image of the air-cooled heat exchanger according to the clustering result; Determining the first defect probability of each group according to the distribution shape and distribution direction of all pixel points in each group, determining the second defect probability of each pixel point in each group according to the gray distribution law of all pixel points in each group, and screening the first defect area according to the magnitudes of the first defect probability and the second defect probability, including: In response to the first defect probability being greater than a preset first defect threshold, the area composed of all pixel points in the corresponding group is the first defect area; or in response to the second defect probability being greater than a preset second defect threshold, taking the corresponding pixel points as defect pixel points, and the area composed of all defect pixel points is the first defect area; Performing connected component analysis on all pixel points that do not belong to any group to obtain a plurality of connected components; determining the second defect area according to the distribution position relationship between the connected components and each group, including: In response to the number of pixel points included in the connected component being greater than 2, dividing the boundary pixel points of the connected component into a plurality of pixel pairs, where the pixel pair includes two boundary pixel points, and the two boundary pixel points in the pixel pair and the centroid of the connected component are on the same straight line; In response to any adjacent pixel point of one boundary pixel point in the pixel pair and any adjacent pixel point of the other boundary pixel point belonging to the same group, taking the connected component as the second defect area.

2. The quality inspection method for the production of fins of an air cooler according to claim 1, characterized in that, The obtaining a binary image corresponding to each gray value in the fin image of the air-cooled heat exchanger includes: Taking any gray value as the target gray value, setting the target gray value in the fin image of the air-cooled heat exchanger to 1 and the remaining gray values to 0 to obtain the binary image corresponding to the target gray value.

3. A quality inspection method for the production of air cooler fins according to claim 1, characterized in that, The grouping the pixel points in the fin image of the air-cooled heat exchanger according to the clustering result includes: Dividing the pixel points corresponding to all highlight points in each category in the clustering result into one group in the fin image of the air-cooled heat exchanger.

4. A method for detecting the production quality of fin of an air cooler according to claim 1, characterized in that, The first defect probability satisfies the expression: ； Among them, is the first defect probability of the th group; , are respectively the lengths of the short side and the long side of the minimum bounding rectangle of the th group; is the acute angle between the straight line where the long side of the minimum bounding rectangle of the th group is located and the straight line where the long side of the minimum bounding rectangle of the th group is located; is the number of groups.

5. A method for detecting the production quality of fins of an air cooler according to claim 1, characterized in that The determining the second defect probability of each pixel point in each group includes: Perform a Z - type scan on the minimum bounding rectangles of each group, and form the gray - level sequences of each group from the gray - level values of the pixel points belonging to each group in the minimum bounding rectangles; perform STL decomposition on the gray - level sequences of each group; determine the second defect probability of the th pixel point in the th group according to the periodic term and the residual term obtained from the STL decomposition : , is the residual term corresponding to the th pixel point in the th group; , are respectively the average differences between the maximum value point closest to the periodic term corresponding to the th pixel point and the other maximum value points, and between the minimum value point closest to the periodic term corresponding to the th pixel point and the other minimum value points in the periodic term curve of the th group; are hyperparameters, is the exponential function with the natural constant as the base, is the hyperbolic tangent function.

6. The quality inspection method for the production of fin of an air cooler according to claim 1, wherein The clustering uses the mean shift clustering algorithm.

7. A quality inspection method for the production of finned tubes of an air cooler according to claim 1, characterized in that The connected component analysis uses the Two-Pass algorithm.

8. An air cooler fin production quality inspection system, characterized in that, Including: A processor and a memory, where the memory stores computer program instructions, and when the computer program instructions are executed by the processor, it implements a method for detecting the production quality of fins of an air-cooled heat exchanger according to any one of claims 1-7.

Images