A heat exchanger fin deformation degree detection method and system based on machine vision

Abstract

The present application relates to the field of image data processing, and more particularly to a heat exchanger fin deformation degree detection method and system based on machine vision, which comprises the following steps: preprocessing and edge detection of heat exchanger fin image, and obtaining gradient information; integrating discrete edge pixels into parameterized straight line clusters; calculating internal disorder degree to evaluate local structural disorder in the cluster; calculating overall offset degree to evaluate systematic installation deviation, and incorporating reliability weight based on gradient direction consistency to suppress noise misjudgment; finally, fusing scores in two dimensions to obtain a comprehensive score, and determining the deformation degree accordingly. Through this structured multi-dimensional diagnostic method, the present application can accurately identify unqualified areas caused by the accumulation of micro-deformation, and can preliminarily judge the defect cause, thereby improving the accuracy and robustness of detection.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image data processing, and in particular to a heat exchanger fin deformation degree detection method and system based on machine vision. BACKGROUND

[0002] As the core component of heat exchange equipment, the performance of the heat exchanger depends largely on the structural integrity of its internal fin array. In an ideal state, the fins of the heat exchanger should be parallel and equidistantly arranged to ensure smooth airflow. In the production and manufacturing process of the heat exchanger, the slight bending or positional deviation of individual fins is usually within the process tolerance range. However, when a large number of fins in a local area undergo consistent and slight cumulative deformation, this effect will continue to accumulate, resulting in a decrease in the effective cross-sectional area of the airflow passage and forming a local blockage. This unqualified area, which is formed by the accumulation of multiple qualified slight deformations, will disrupt the fluid dynamics of the heat exchanger and is a key concern in product quality control.

[0003] Existing machine vision detection methods usually focus on detecting serious defects of individual fins, such as breakage or large-angle bending, but lack the ability to assess the cumulative deformation effect in this regional area and are difficult to identify unqualified areas formed by the accumulation of a large number of qualified slight deformations.

[0004] In related technologies, the technical means of Hough transform combined with parameter space clustering analysis is used, and its basic principle is: first, the fin edges in the image are detected as points in the parameter space using Hough transform, then the parameter points representing the main fin group are divided into clusters through clustering algorithm, and finally the arrangement state of the fins is evaluated by analyzing the distribution of the parameter points in the cluster. However, this method has defects in actual application: on the one hand, Hough transform is very sensitive to noise and false edges in the image, which can easily lead to inaccurate detection of straight line parameters, affecting the reliability of the clustering results; on the other hand, simply analyzing the compactness of the parameter cluster cannot effectively distinguish between two completely different deformation modes, i.e. overall installation deviation and local structural disorder, and lacks the ability to make deep judgments on the root cause of the deformation, which can easily lead to missed or false judgments. SUMMARY

[0005] To solve the technical problems of the above-mentioned Hough transform combined with parameter space clustering analysis technology, which is sensitive to noise and false edges and lacks the ability to distinguish between different deformation modes and make deep judgments, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application provides a heat exchanger fin deformation degree detection method based on machine vision, which comprises the following steps:

[0007] Collecting an image of a fin of a heat exchanger, and performing edge detection on the image to obtain a set of edge pixels and gradient information of each edge pixel; performing straight line transformation on the set of edge pixels to obtain a set of straight lines; clustering the set of straight lines to obtain a plurality of straight line clusters; calculating an internal disorder degree of each straight line cluster based on a direction distribution and a distance distribution of the straight lines in the straight line cluster; determining an overall deviation degree of each straight line cluster based on a difference between an average position and an average direction of the straight line cluster and a preset ideal reference value, and a consistency between gradient directions of the edge pixels of the straight line cluster and a preset ideal gradient direction of edge pixels; fusing the internal disorder degree and the overall deviation degree to obtain a straight line cluster score of each straight line cluster; and taking a maximum value in all straight line cluster scores as a cumulative deformation score, and determining a deformation degree of the fin of the heat exchanger according to a comparison result of the cumulative deformation score and a preset threshold.

[0008] The application obtains gradient information by pre-processing and edge detection on an image; integrates discrete edge pixels into parameterized straight line clusters, thereby laying a foundation for regional analysis; evaluates local structural disorder caused by factors such as twisting and extrusion by calculating an internal disorder degree; evaluates systematic installation deviation caused by assembly errors by calculating an overall deviation degree, and effectively suppresses misjudgment caused by image noise and artifacts by incorporating the consistency of gradient directions of edge signals as a reliability weight in the calculation; fuses two dimensions of scores independent of each other to obtain a comprehensive score that can comprehensively reflect the health status of a local region. The application can accurately identify unqualified regions caused by a large number of small deformations by using this structured and multi-dimensional judgment method, and can preliminarily judge defect causes, thereby improving the accuracy and robustness of detection.

[0009] Preferably, the calculation of the internal disorder degree of the straight line cluster comprises: calculating a direction deviation degree representing direction consistency according to normal direction angles of the straight lines in the straight line cluster; calculating a distance disorder degree representing distance uniformity according to vertical distances of adjacent straight lines in the straight line cluster after sorting; and determining the internal disorder degree based on the direction deviation degree and the distance disorder degree.

[0010] The application decomposes the complex internal disorder degree index into two independent sub-dimensions of direction deviation degree and distance disorder degree, which enables the detection model to clearly distinguish two main physical modes of deformation: one is direction inconsistency caused by twisting or tilting accumulation of a single fin, and the other is spatial arrangement disorder caused by uneven distance between fins; the application evaluates the two dimensions respectively, thereby improving the recognition ability of internal disorder causes and making the final disorder degree score have a clearer physical meaning.

[0011] Preferably, the calculation of the direction deviation degree comprises: sorting the straight lines according to the normalized values of the perpendicular distances of the straight lines in the straight line cluster to obtain sorted straight lines; calculating the absolute difference of the normalized values of the normal direction angles of each pair of adjacent straight lines in the sorted straight lines; and the direction deviation degree is the average of all the absolute differences.

[0012] The present application firstly sorts the straight lines according to the physical positions of the straight lines, ensures that the analysis order is consistent with the actual spatial arrangement of the fins, and accurately captures the small and continuous changes in the angles between the fins by calculating the absolute difference between the normal direction angles of each pair of adjacent straight lines and averaging all the differences, which is particularly sensitive to the gradual cumulative tilt or disordered local twist and intuitively reflects the local consistency degree of the fin arrangement in the rotation dimension.

[0013] Preferably, the obtaining of the spacing disorder degree comprises: sorting the straight lines according to the normalized values of the perpendicular distances of the straight lines in the straight line cluster to obtain sorted straight lines; determining two adjacent spacings formed by three consecutive straight lines based on the sorted straight lines; calculating the absolute deviation between the ratio of each two adjacent spacings and 1; and the spacing disorder degree is the average of all the absolute deviations.

[0014] The present application firstly sorts the straight lines to ensure that the analysis order is consistent with the actual physical arrangement of the fins; then, by calculating the ratio of the adjacent spacings instead of the spacings themselves, the influence of the overall size specifications of the fins is eliminated, so that the index has better universality; finally, the deviation of the ratio from the ideal value 1 is accumulated and averaged to obtain the final spacing disorder degree score, which can accurately capture any spacing unevenness problems caused by local compression, stretching or even missing, and accurately evaluate the spatial uniformity of the fin arrangement.

[0015] Preferably, the determination of the overall deviation degree of the straight line cluster based on the difference between the average position and the average direction of each straight line cluster and the preset ideal reference value and the consistency of the gradient direction of the edge pixels of the straight line cluster and the preset ideal gradient direction of the edge pixels comprises: determining a systematic deviation term based on the difference between the average position and the average direction and the preset ideal reference value; determining a signal reliability weight based on the consistency of the gradient direction of the edge pixels and the preset ideal gradient direction of the edge pixels; and the overall deviation degree is the product of the systematic deviation term and the signal reliability weight.

[0016] The application establishes a double-checking mechanism considering physical authenticity and signal reliability by decomposing the overall deviation into a systematic deviation term and a signal reliability weight, and finally multiplying the two; the application can clearly distinguish two cases: one is that the fin cluster indeed has overall physical installation deviation, and the other is that false deviation is caused by pseudo-edges due to image noise or reflection; the application multiplies the systematic deviation term reflecting the real physical deviation and the reliability weight evaluating the edge signal quality, so that only the physical deviation proved by high-quality and high-reliability edge signals can produce obvious overall deviation score, thereby enhancing the ability of the system to resist noise interference and effectively avoiding false alarms.

[0017] Preferably, the signal reliability weight is obtained by calculating the average value of the square value of the angle deviation between the gradient direction angle of all edge pixels constituting the straight line cluster and the gradient direction angle of ideal edge pixels, and taking the negative natural exponent of the average value as the signal reliability weight.

[0018] Preferably, the edge detection comprises: performing gray processing on the fin image to obtain a gray image; applying a non-local mean filtering algorithm to the gray image to obtain a denoised image; and applying a Canny edge detection algorithm to the denoised image to obtain an edge pixel set.

[0019] Preferably, the clustering is a density-based spatial clustering algorithm.

[0020] Preferably, the fusion of the internal disorder degree and the overall deviation degree comprises:

[0021] The internal disorder degree and the overall deviation degree are added.

[0022] In a second aspect, the application provides a heat exchanger fin deformation degree detection system based on machine vision, which comprises a memory and a processor, and the memory stores computer program instructions.

[0023] By using the above technical solution, the computer program of the heat exchanger fin deformation degree detection method based on machine vision is generated, and is stored in the memory to be loaded and executed by the processor, so that a terminal device is manufactured according to the memory and the processor, and is convenient to use.

[0024] The beneficial effects of this invention are as follows: First, it integrates discrete edge pixels into a parameterized cluster of lines, deconstructing the fin deformation problem into two independent dimensions: internal disorder and overall offset. Internal disorder is used to assess local structural disorder caused by factors such as torsion and compression; overall offset is used to assess systematic installation deviation caused by assembly errors. By incorporating a reliability weight based on the consistency of edge signal gradient direction, misjudgments caused by image noise and artifacts are effectively suppressed. Finally, by fusing the scores of these two dimensions, it can accurately identify difficult-to-detect defective areas accumulated from numerous minute deformations and make a preliminary judgment on the cause of defects, achieving high-precision and high-robustness judgment of cumulative deformation defects. Attached Figure Description

[0025] Figure 1 A flowchart of a machine vision-based method for detecting the degree of deformation of heat exchanger fins, provided for an embodiment of the present invention;

[0026] Figure 2 This is a structural block diagram of a heat exchanger fin deformation detection system based on machine vision, provided in an embodiment of the present invention. Detailed Implementation

[0027] The first aspect of this invention provides a machine vision-based method for detecting the deformation degree of heat exchanger fins, such as... Figure 1 As shown, the method includes steps S100-S400:

[0028] Step S100: Acquire images of the heat exchanger fins and perform edge detection to obtain a set of edge pixels and gradient information of each edge pixel.

[0029] It should be noted that, in order to detect the deformation degree of the fin of the heat exchanger, the picture thereof needs to be collected first. However, direct analysis on the original image will be interfered by various factors such as uneven environmental light, sensor noise and material reflection, which will affect the accuracy of subsequent edge detection and feature extraction. Therefore, the preprocessing operation needs to be performed first to purify the image information and provide a stable and reliable data basis for the subsequent accurate analysis. Specifically, the grayscale processing needs to be performed first, which can reduce the calculation complexity of the subsequent algorithm and exclude the interference of color information on the fin structure judgment; then, the image noise needs to be eliminated, and the non-local mean filtering algorithm is preferably adopted in the present application. Compared with the traditional local smoothing method such as Gaussian filtering, the non-local mean filtering algorithm can effectively filter out the noise while retaining the fin edge and other high-frequency details to a greater extent, thereby avoiding the edge blurring problem caused by excessive smoothing. Finally, the Canny edge detection algorithm is selected to extract the contour to obtain a plurality of edge points, and the advantage thereof lies in that it can generate a fine edge without direction and with a single-pixel width, which provides a guarantee for the accuracy of the subsequent Hough straight line transformation. Meanwhile, the gradient information of the edge points needs to be obtained and stored to provide key original data for evaluating the edge reliability and calculating the gradient direction deviation degree weighting factor in the subsequent step.

[0030] Specifically, first, an image of the fin area of the heat exchanger to be detected is collected by using an image collection device such as an industrial camera; then, the collected color image is converted into a grayscale image; then, the non-local mean filtering algorithm is applied to the grayscale image for noise reduction processing to obtain a preprocessed image; finally, the following operations are performed on the preprocessed image:

[0031] Edge detection: the Canny edge detection algorithm is applied to the preprocessed image to obtain a single-pixel edge pixel set containing only the fin contour Then, the gradient component of each pixel in the edge pixel set is calculated.

[0032] At this point, the edge pixel set representing the fin contour and the gradient component of each edge pixel thereof are obtained.

[0033] Step S200, performing straight line transformation on the edge pixel set to obtain a straight line set; clustering the straight line set to obtain a plurality of straight line clusters; and calculating the internal disorder degree of the straight line cluster based on the direction distribution and the interval distribution of the straight lines in the straight line cluster.

[0034] It should be noted that, ideally, the fin array of a qualified heat exchanger should exhibit strictly parallel and equally spaced geometric features. Deviations from this ideal arrangement usually indicate a decline in product quality. This step aims to establish an index that can accurately assess the degree of deviation from the ideal arrangement. The Hough transform is a feature detection technique that identifies specific geometric shapes in an image by mapping edge points in image space to a parameter space for voting. Based on this, this invention uses the Hough linear transform to effectively integrate a large number of discrete edge points constituting the fin profile into parameterized straight lines, laying the foundation for subsequent analysis of the entire fin group. After obtaining the set of straight lines representing the fins, clustering is required to distinguish different fin groups and noise. DBSCAN is a density-based clustering algorithm whose core advantage is that it does not require pre-specifying the number of clusters and can automatically identify and mark outliers that cannot be assigned to any cluster as noise. The present invention uses DBSCAN precisely because it takes into account that there may be false edge lines in the heat exchanger image formed by burrs, reflections, etc. DBSCAN can automatically eliminate these interfering lines as noise points, and retain only the cluster of lines with sufficient density composed of real fins, thereby improving the accuracy and robustness of fin group identification.

[0035] Specifically, firstly, it is necessary to obtain a set of lines and cluster them to obtain several line clusters. The detailed process includes: firstly, applying the standard Hough line transform to the set of edge pixels to obtain a set of lines. ,in, It is the total number of lines in the set of lines. It is the first The perpendicular distance from the straight line to the origin of the image coordinate system. It is the first The direction angle of the normal to a straight line; secondly, for a set of lines. of and Standardization is performed on each line to obtain the standardized set of lines. ,in, It is the first The standardized value of the perpendicular distance from the line to the origin of the image coordinate system. It is the first The normal direction angles of the straight lines are standardized to eliminate dimensional differences and ensure that clustering is not affected by numerical range; Z-score standardization is preferred. Then, the DBSCAN clustering algorithm is applied to the standardized set of lines to identify point clusters, resulting in several line clusters where the fins in each cluster have similar orientations and positions. Hough transform and DBSCAN clustering are existing technologies and will not be elaborated upon here.

[0036] As a preferred implementation, the neighborhood radius and the minimum number of points (MinPts) required to form a cluster in the DBSCAN algorithm are set as follows: the neighborhood radius can be determined according to the mapping relationship of the fin spacing in Hough space, for example, a parameter distance corresponding to a physical spacing of 0.3-0.5 units; MinPts can be set according to the expected minimum number of consecutive fins, for example, it can be set to 3-10.

[0037] Secondly, for each cluster of lines, to assess the degree of disorder within its arrangement, two dimensions are considered: directional deviation and spacing disorder. Because within a cluster of lines, these lines... Since it is unordered, to calculate the spacing and directional deviation between adjacent lines, first, for all lines in the line family, according to their... The values ​​are sorted in ascending or descending order to ensure that the adjacency relationships calculated subsequently are consistent with the actual spatial arrangement of the fins.

[0038] Regarding directional deviation, it should be noted that if all lines within a cluster are perfectly parallel, their normal directions will be completely consistent, and the angle difference between adjacent lines after sorting should be 0. If the lines exhibit irregular twisting or tilting, the angle difference between adjacent lines will fluctuate. Therefore, the directional deviation of the fin arrangement can be assessed by analyzing the angle differences of all adjacent lines after sorting within the cluster and calculating their average value.

[0039] Based on the above logic, the directional deviation satisfies the following relationship:

[0040] ;

[0041] in, It is the first Directional deviation of a cluster of straight lines; It is the first The total number of lines in a family of lines; , They are the first Within the nth straight line cluster, the th Article, No. The standardized value of the normal direction angle of a straight line; It is the absolute value symbol.

[0042] In this relation, The calculation is the first The absolute value of the shortest angle difference between the normal directions of two adjacent lines after sorting within a cluster of lines. When all lines within the cluster are perfectly parallel, all adjacent angle differences are 0, thus making... The value is 0. As the fins twist or tilt randomly, the straight-line direction is no longer consistent, and the absolute value of the difference between adjacent angles increases, leading to... The value increases. Therefore, It can accurately assess the degree of directional disorder in fin arrangement.

[0043] The degree of fin spacing disorder can be assessed by comparing the actual spacing differences between adjacent straight lines. Considering that all fins are equidistant under ideal conditions, the ratio of adjacent spacing should be 1. If there is compression, stretching, or local missingness, the ratio of adjacent spacing will deviate from 1. Therefore, the degree of spacing disorder can be assessed by calculating the deviation of the ratio of adjacent spacing from 1.

[0044] Based on the above logic, the spacing disorder satisfies the following relationship:

[0045] ;

[0046] in, It is the first The degree of disorder in the spacing of a cluster of straight lines; It is the first The total number of lines in a family of lines; , , They are the first After sorting the nth straight line cluster, the th , , The standardized value of the perpendicular distance from the line to the origin of the image coordinate system; It is the absolute value symbol; It is a preset microvalue used to prevent the denominator from being 0, and can be set to 0.001.

[0047] In this relation, It is the deviation between the ratio of two adjacent spacings and the ideal ratio of 1. When all fins are equidistant, this deviation is 0, making... The value is 0; when the fins are compressed, stretched, or partially missing, the spacing is no longer uniform, and the absolute value of this deviation will increase, leading to... The value of also increases accordingly, therefore, It can accurately assess the degree of disorder in the spacing of the fin arrangement.

[0048] Finally, the two indicators mentioned above are combined into a comprehensive internal disorder degree, which satisfies the following relationship:

[0049] ;

[0050] in, It is the first The internal disorder of a linear cluster; It is the first Directional deviation of a cluster of straight lines; It is the first The degree of disorder in the spacing of a cluster of straight lines.

[0051] In the relationship, by decomposing the complex internal disorder into two orthogonal dimensions of direction and spacing with clear physical meaning, and by using the form of Euclidean distance to fuse them into a single, assessable internal disorder index Any local regional arrangement disorder caused by the accumulation of a single fin micro-torsion or displacement can be captured with high sensitivity and independence.

[0052] At this point, the internal disorder of each straight line cluster is obtained.

[0053] Step S300, based on the difference between the average position and the average direction of each straight line cluster and the preset ideal reference value, and the consistency of the gradient direction of the edge pixels of the straight line cluster with the preset ideal edge pixel gradient direction, determine the overall offset degree of the straight line cluster.

[0054] It should be noted that the internal disorder alone is not enough to fully evaluate the fin quality, because the fin cluster with internal arrangement regularity may also deviate from its designed standard position as a whole due to assembly errors, which is also a serious defect that needs to be detected. In addition, noise or false edges in the image may be incorrectly identified as fins, leading to misjudgment of the offset. Therefore, the fin overall offset degree is introduced in this step to evaluate such systematic deviation, the idea is to divide the offset degree calculation into two parts: one is the systematic offset term that directly reflects the physical offset, and the other is the signal reliability weight that serves as a signal reliability measure; finally, by multiplying the two, it is ensured that the final offset degree score can not only reflect the true physical deviation, but also effectively suppress false alarms caused by noise or false signal.

[0055] Specifically, for the first straight line cluster, the calculation process of the fin overall offset degree is as follows: first, calculate the systematic offset term for evaluating the physical installation deviation of the straight line cluster.

[0056] It should be noted that the construction logic is to evaluate the overall translation and rotation error by comparing the difference between the average position and the average direction of the straight line cluster and the ideal reference value given by the design pattern. The present application uses the form of normalized sum of squares to fuse the position deviation and direction deviation into a unified measure.

[0057] According to the above logic, the systematic offset term satisfies the relationship:

[0058] ;

[0059] Wherein, is the systematic offset term of the first straight line cluster; is the systematic offset term of the first The standardized value of the average vertical distance from all lines in a cluster of lines to the origin of the image coordinate system; It is the first The standardized value of the average normal direction of all lines within a line cluster; It is the vertical distance from the preset ideal reference position. It is an ideal reference direction; It is a standard normalization function used to eliminate dimensional differences.

[0060] In this relation, It reflects the deviation between the average vertical distance of the straight fin cluster and the ideal reference position. The greater the deviation, the more obvious the overall translation of the fin cluster in the direction perpendicular to the normal. This reflects the deviation between the average normal direction of the linear cluster and the ideal reference direction. The larger the deviation, the more significant the overall rotation angle of the fin cluster. The sum of the two terms constitutes a comprehensive offset term for position and direction. The larger this term is, the more significant the overall displacement and directional deviation of the linear cluster relative to the ideal reference position, i.e., there is a significant translation or rotation, and the greater the overall offset of the fins. Conversely, the smaller this term is, the closer the linear cluster is to the ideal installation state, and the smaller the overall offset of the fins.

[0061] It should be added that, The method for determining the ideal reference direction is as follows: Multiple heat exchanger fin sample images with known healthy conditions are acquired, ensuring the samples are free from deformation, installed correctly, and under normal imaging conditions. The same testing procedure is performed on each sample, extracting the average direction of its main fin linear clusters. Finally, the mean or median of all samples is used as the ideal reference direction. . The determination method is as follows: On the same healthy sample set, extract the average position of the main fin cluster of each sample and record its vertical distance relative to the image coordinate origin; under the premise of ensuring consistency between the imaging field of view and the installation positioning, take the mean or median of the vertical distances of all average positions relative to the image coordinate origin as the vertical distance of the ideal reference position. .

[0062] Regarding signal reliability weights, it's important to note that, theoretically, the gradient direction of each pixel on a true, clear fin edge should align with the normal direction of the line fitted to that edge. If the angle between these two directions deviates significantly, it indicates that these edge points are likely generated by random noise or reflection artifacts. Therefore, reliability can be assessed by calculating the average angle difference between the gradient directions of all edge points and the theoretical normal direction.

[0063] Based on the above logic, the signal reliability weights satisfy the following relationship:

[0064] ;

[0065] wherein, is the signal reliability weight of the th straight line cluster; is the gradient direction angle of the th edge pixel on the th straight line of the th straight line cluster, which is calculated by the horizontal gradient component and the vertical gradient component of the edge pixel through the inverse tangent function . is the gradient direction angle of the ideal edge pixel; is the total number of straight lines of the th straight line cluster; is the total number of edge pixels on the th straight line of the th straight line cluster; is the natural exponential function.

[0066] In the relationship, represents the square term of the directional deviation between the gradient direction of each edge pixel and the gradient direction angle of the ideal edge pixel. The larger the term, the more the point deviates from the ideal direction, and the more likely it is a noise point; the numerator accumulates the directional deviations of all edge pixels on all straight lines in the th straight line cluster to obtain the total square deviation sum, reflecting the gradient direction consistency of the entire cluster; the denominator is a normalization factor that converts the total square deviation sum into the square of the average directional deviation, so that is not affected by the cluster size.

[0067] It should be noted that the gradient direction angle of the ideal edge pixel is obtained as follows: collect multiple heat exchanger fin sample images of known healthy state, ensure that the samples are not deformed, installed normally and imaged under normal conditions; perform edge detection and gradient calculation on each sample, extract the gradient direction of the edge pixel at each corresponding spatial position in the main fin area, and for the edge points at the same spatial position, calculate the mean or median of their gradient directions in multiple healthy samples, and take it as the reference gradient direction angle of that position.

[0068] Finally, multiply the systematic deviation term by the signal reliability weight to obtain the final overall deviation degree . The multiplication fusion introduces the signal reliability weight , so that only the physical deviation confirmed by clear and true edge signals can produce a higher overall deviation degree ; and the false deviation caused by noise or artifacts will be suppressed, thereby effectively avoiding false alarms.

[0069] At this point, the overall deviation degree of each straight line cluster is obtained.

[0070] Step S400: The internal disorder degree and the overall offset degree are fused to obtain the straight line cluster score of each straight line cluster; the maximum value among all straight line cluster scores is taken as the cumulative deformation score, and the degree of deformation of the heat exchanger fins is determined according to the comparison result of the cumulative deformation score and the preset threshold.

[0071] It should be noted that the cumulative deformation of the fins is the result of two factors: firstly, local structural disorder assessed by the internal disorder index, and secondly, systematic installation deviation assessed by the overall offset index. To obtain a comprehensive evaluation of the deformation degree of each local area, it is necessary to integrate the internal disorder and overall offset. Furthermore, the performance of the entire heat exchanger fin area is often determined by its weakest link; any severe local deformation may lead to substandard overall performance. Therefore, the final judgment should be based on the most severe local area. After obtaining the final score representing the worst area, in an industrial online quality inspection scenario, a clear pass / fail conclusion is still required. Therefore, this step introduces a preset judgment threshold to map continuous score values ​​to discrete judgment results, and can locate specific defective areas when a failure is judged.

[0072] Specifically, for the first A linear cluster is used to calculate a linear cluster score by combining the internal disorder degree and the overall fin offset degree. The linear cluster score satisfies the following relationship:

[0073] ;

[0074] in, It is the first The score of a family of lines; It is the first The degree of disorder within a linear cluster; It is the first The overall offset of a cluster of straight lines; It is the hyperbolic tangent function.

[0075] In this relationship, the greater the internal disorder, the more chaotic the fin arrangement within the linear cluster, indicating obvious twisting, compression, or uneven spacing. The linear cluster score... The larger the value, the more regular the internal arrangement; conversely, the smaller the value, the more regular the linear cluster score. The smaller the value, the greater the overall offset of the fins, indicating that the straight cluster... The more significant the overall displacement and directional deviation relative to the ideal position, and the worse the consistency of edge orientation, the higher the score of the straight line cluster. The larger the value, the closer its position and orientation are to the standard installation state, the more reliable the edges, and the better the straight cluster score. The smaller, the two added to get the straight line cluster score The greater the value, the more serious the accumulated deformation of the region.

[0076] Then, one heat exchanger fin area image can have multiple obvious straight line clusters, the accumulated deformation score of the region should be determined by the worst one, so take the maximum value of all the straight line cluster scores as the accumulated deformation score, set a judgment threshold , and compare the accumulated deformation score with it: if the accumulated deformation score , it is determined that the accumulated deformation of the heat exchanger fin area is within an acceptable range, and it is determined to be qualified. If the accumulated deformation score , it is determined that there is obvious accumulated deformation in the heat exchanger fin area, and it is determined to be unqualified. When the judgment result is unqualified, the system will further locate the straight line cluster whose score is equal to the accumulated deformation score, and determine the image area corresponding to the straight line cluster as a defect area.

[0077] The second aspect of the embodiment provides a heat exchanger fin deformation degree detection system based on machine vision, as shown in Figure 2 The heat exchanger fin deformation degree detection system based on machine vision comprises a memory and a processor, and the memory stores computer program instructions, which realize the heat exchanger fin deformation degree detection method based on machine vision of the first aspect of the application when executed by the processor.

[0078] The heat exchanger fin deformation degree detection system based on machine vision further comprises a communication bus and a communication interface and other components familiar to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0079] In the present application, the aforementioned memory can be any tangible medium containing or storing a program, which can be used or combined with an instruction execution system, device or apparatus. For example, the computer readable storage medium can be any appropriate magnetic storage medium or magneto-optical storage medium, such as resistive random access memory, dynamic random access memory, static random access memory, enhanced dynamic random access memory, high bandwidth memory, hybrid memory cube, etc., or any other medium that can be used to store the required information and can be accessed by an application program, module or both. Any such computer storage medium can be part of the device or accessible or connectable to the device.

[0080] The above are the preferred embodiments of the present application, which do not limit the protection scope of the present application, therefore: any equivalent changes made on the structure, shape, principle of the present application should be covered within the protection scope of the present application.

Claims

1. A method for detecting the degree of deformation of heat exchanger fins based on machine vision, characterized in that, Including the following steps: The heat exchanger fin images are acquired, and edge detection is performed to obtain the edge pixel set and the gradient information of each edge pixel; A linear transformation is performed on the edge pixel set to obtain a set of lines; the set of lines is clustered to obtain several line clusters; based on the direction distribution and spacing distribution of each line within a line cluster, the internal disorder of the line cluster is calculated, including: calculating the direction deviation, which represents the consistency of direction, based on the normal direction angle of each line within the line cluster; calculating the spacing disorder, which represents the uniformity of spacing, based on the perpendicular distance between adjacent lines after sorting within the line cluster; and determining the internal disorder based on the direction deviation and spacing disorder. Based on the differences between the average position and average direction of each line cluster and the preset ideal reference value, and the consistency between the gradient direction of the edge pixels of the line cluster and the preset ideal edge pixel gradient direction, the overall offset of the line cluster is determined, including: Based on the differences between the average position and average direction and the preset ideal reference values, a systematic offset term is determined; Based on the consistency between the gradient direction of edge pixels and the preset ideal gradient direction of edge pixels, the signal reliability weight is determined, including: Calculate the average of the squared angular deviations between the gradient direction angles of all edge pixels constituting the straight line cluster and the gradient direction angle of the ideal edge pixel, and take the negative natural exponent of the average value as the signal reliability weight. The overall offset is the product of the systematic offset term and the signal reliability weight; The internal disorder degree and the overall offset degree are fused to obtain the straight line cluster score for each straight line cluster; the maximum value among all straight line cluster scores is taken as the cumulative deformation score, and the degree of deformation of the heat exchanger fins is determined based on the comparison result of the cumulative deformation score and the preset threshold.

2. The method for detecting the degree of deformation of heat exchanger fins based on machine vision according to claim 1, characterized in that, The calculation of the directional deviation includes: The lines are sorted according to the standardized value of the vertical distance between each line in the line cluster, resulting in sorted lines. Calculate the absolute difference of the normal direction angle of each pair of adjacent lines after sorting. The directional deviation is the average of all absolute differences.

3. The method for detecting the degree of deformation of heat exchanger fins based on machine vision according to claim 1, characterized in that, The acquisition of the spacing disorder includes: The lines are sorted according to the standardized value of the vertical distance between each line in the line cluster, resulting in sorted lines. Based on the sorted straight lines, determine the spacing between two adjacent segments formed by three consecutive straight lines one by one; Calculate the absolute deviation between the ratio of each pair of adjacent spacings and 1, where the spacing disorder is the average of all absolute deviations.

4. The method for detecting the degree of deformation of heat exchanger fins based on machine vision according to claim 1, characterized in that, The edge detection performed includes: The fin image is converted to grayscale to obtain a grayscale image; The grayscale image is denoised by applying a nonlocal mean filtering algorithm to obtain the denoised image. The Canny edge detection algorithm is applied to the denoised image to obtain a set of edge pixels.

5. The method for detecting the degree of deformation of heat exchanger fins based on machine vision according to claim 1, characterized in that, The clustering is a density-based spatial clustering algorithm.

6. The method for detecting the degree of deformation of heat exchanger fins based on machine vision according to claim 1, characterized in that, The process of fusing internal disorder with overall offset includes: Add the internal disorder to the overall offset.

7. A machine vision-based system for detecting the deformation degree of heat exchanger fins, characterized in that, The machine vision-based heat exchanger fin deformation detection system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement a machine vision-based heat exchanger fin deformation detection method according to any one of claims 1-6.

Images