Method for detecting appearance quality of heat exchanger, electronic equipment, air conditioning system and vehicle

Abstract

The invention discloses a method for detecting the appearance quality of a heat exchanger, electronic equipment, an air conditioning system and a vehicle, and the method comprises the steps that a fin appearance detection result of the heat exchanger is obtained based on the output of a fin appearance detection model, and the fin appearance detection model takes the appearance image of the heat exchanger as the input. By adopting the method, the appearance defect detection of the heat exchanger can be realized through the fin appearance detection model, the condition of missing detection during the defect detection of the heat exchanger can be avoided, and the condition of false detection between fin defects with similar forms can also be avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicles, and in particular to a method for detecting the appearance quality of a heat exchanger, an electronic device, an air-conditioning system and a vehicle. Background Art

[0002] In the related art, the appearance quality of the heat exchanger is inspected by machine vision, that is, the edge contour of the heat exchanger is identified by extracting texture features and detecting straight lines on the heat exchanger image taken by a high-resolution camera, and judging whether there are defects based on manually designed rules. However, the manually designed rules are difficult to cover all possible defects of the heat exchanger, resulting in missed detections and false detections when performing defect detection on the heat exchanger. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, one object of the present invention is to provide a method for inspecting the appearance quality of a heat exchanger. This method can detect appearance defects in heat exchangers using a fin appearance inspection model, thereby avoiding missed defects and false detections of similar fin defects.

[0004] A second objective of the present invention is to provide an electronic device.

[0005] A third objective of the present invention is to provide a computer-readable storage medium.

[0006] A fourth object of the present invention is to provide an air conditioning system.

[0007] A fifth object of the present invention is to provide a vehicle.

[0008] In order to solve the above problems, an embodiment of the first aspect of the present invention provides a method for detecting the appearance quality of a heat exchanger, and obtains the fin appearance detection result of the heat exchanger based on the output of a fin appearance detection model, wherein the fin appearance detection model takes the appearance image of the heat exchanger as input.

[0009] According to the method for detecting the appearance quality of a heat exchanger according to an embodiment of the present invention, an appearance image of the heat exchanger is input into a fin appearance detection model, so as to judge whether there are defects in the appearance of the heat exchanger through the output of the fin appearance detection model. Therefore, compared with the method of detecting heat exchanger defects through machine vision processing technology in the prior art, the fin appearance detection model is used in the present application to realize heat exchanger appearance defect detection. Since the fin appearance detection model does not rely on specific or hypothetical rules to determine whether there is a heat exchanger appearance defect, it avoids missed detection when performing defect detection on the heat exchanger, and can also avoid false detection between fin defects with similar morphology.

[0010] In some embodiments, the fin appearance inspection result includes a fin appearance defect location and a fin appearance defect type at the fin appearance defect location.

[0011] In some embodiments, the training data set of the fin appearance detection model includes fin appearance defect simulation data of different fin shapes and / or different fin environmental conditions obtained through data simulation.

[0012] In some embodiments, the fin morphology includes at least one of fin deformation, fin size, and fin position.

[0013] In some embodiments, the environmental conditions of the fins include lighting conditions.

[0014] In some embodiments, the ROI labels of various types of fin appearance defects in the fin appearance detection model are obtained by merging corresponding types of fin defect data into the target fin region of the clean source image of the heat exchanger.

[0015] In some embodiments, the target fin region is obtained by separating the harmonica tube region of the heat exchanger from the binarized image of the clean source image of the heat exchanger; the harmonica tube region is obtained by performing an image morphological opening operation on the binarized image along the vertical direction of the harmonica tube.

[0016] In some embodiments, the fin appearance defect ROI label includes a fin lodging defect ROI label; the fin defect data includes a target fin lodging defect image, which is a fin lodging defect image that is scaled based on the height of the fin area in the clean source image of the heat exchanger.

[0017] In some embodiments, the fin lodging defect ROI label is obtained by randomly merging the target fin lodging defect image into the fin region in the clean source image of the heat exchanger using a Poisson fusion method.

[0018] In some embodiments, the fin appearance defect ROI label includes a fin skew defect ROI label; the fin defect data includes a first target transformation matrix, the first target transformation matrix is ​​obtained based on a first target matching point, and the first target matching point is determined based on a target fin area of ​​the clean source image of the heat exchanger.

[0019] In some embodiments, the first target transformation matrix is ​​obtained by performing an affine transformation calculation based on the first target matching point.

[0020] In some embodiments, the first target matching point is obtained based on the first original point and the second target point of the target fin area; wherein the target fin area is a rectangle, the first original point is the four vertices of the target fin area, and the first target point is the four vertices of a parallelogram with the same base as the target fin area.

[0021] In some embodiments, the fin appearance defect ROI label includes a fin deformation defect ROI label and / or a fin shrinkage defect ROI label; the fin defect data includes a second target transformation matrix, the second target transformation matrix is ​​obtained based on a second target matching point, and the second target matching point is determined based on a target fin area of ​​the clean source image of the heat exchanger.

[0022] In some embodiments, the second target transformation matrix is ​​calculated using a thin plate spline interpolation method based on N groups of the second target matching points, where N is a positive integer greater than or equal to 2.

[0023] In some embodiments, the second target matching point is obtained based on N second original points and N second target points of the target fin area; wherein, the N second original points and the N second target points are randomly selected from the target fin area.

[0024] In some embodiments, for the fin deformation defect ROI label, the second target matching point is determined based on the bending shape of the fin.

[0025] In some embodiments, for the fin shrinkage defect ROI label, the second target matching point is determined based on the fin spacing.

[0026] In some embodiments, the target fin region is a rectangular region of any size randomly demarcated at the fin region on the clean original image of the heat exchanger.

[0027] In some embodiments, the fin appearance detection model is obtained by training a segmentation deep learning network based on a training set.

[0028] In some embodiments, the method further includes: obtaining a core appearance inspection result of the heat exchanger based on the appearance image of the heat exchanger.

[0029] In some embodiments, the core appearance inspection result includes the presence of core tilting deformation; the core tilting deformation is determined based on the angle between the side plate fitting straight line and the header fitting curve.

[0030] In some embodiments, the presence of the core dumping deformation includes the angle exceeding a threshold angle.

[0031] In some embodiments, the core appearance inspection result includes the presence of side panel bending; the side panel bending is determined based on the distance residual between the side panel contour point and the side panel fitting line.

[0032] In some embodiments, the existence of edge plate bending includes the distance residual being greater than a residual threshold.

[0033] In some embodiments, the side plate fitting straight line is obtained based on the outer contour fitting of the side plate; the manifold fitting straight lines are all obtained based on the outer contour fitting of the manifold; the outer contour of the side plate and the outer contour of the manifold are obtained based on edge detection of the initial positioning area of ​​the side plate and the manifold.

[0034] In some embodiments, the initial positioning region is obtained by separating the maximum connected domains in the horizontal and vertical directions along the edges of the pre-processed image of the heat exchanger.

[0035] In some embodiments, the heat exchanger pre-processed image is obtained by performing a binarization operation and morphological processing on the grayscale image of the heat exchanger clean source image.

[0036] In some embodiments, the clean source image of the heat exchanger is a front image of the heat exchanger.

[0037] A second aspect of the present invention provides an electronic device comprising: at least one processor; a memory communicatively connected to the at least one processor; a computer program executable by the at least one processor stored in the memory, wherein the at least one processor implements the method for detecting the appearance quality of a heat exchanger as described in the above embodiment when executing the computer program.

[0038] According to the electronic device of an embodiment of the present invention, by executing the method for detecting the appearance quality of a heat exchanger according to the above embodiment, the appearance defect detection of a heat exchanger can be realized through a fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0039] A third aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method for detecting the appearance quality of a heat exchanger as described in the above embodiment.

[0040] A fourth aspect of the present invention provides an air-conditioning system, comprising a heat exchanger and a controller, wherein the controller is configured to execute the method for detecting the appearance quality of the heat exchanger described in the above embodiment.

[0041] According to the air-conditioning system of an embodiment of the present invention, by executing the method for detecting the appearance quality of the heat exchanger of the above embodiment, the appearance defect detection of the heat exchanger can be realized through the fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0042] In some embodiments, the air conditioning system further includes: an image acquisition device connected to the controller, for acquiring an appearance image of the heat exchanger.

[0043] A fifth aspect of the present invention provides a vehicle, which includes the electronic device described in the above embodiment; or, the vehicle includes the air-conditioning system described in the above embodiment.

[0044] According to the vehicle of the embodiment of the present invention, through the electronic equipment or air-conditioning system of the above embodiment, the heat exchanger appearance defect detection can be realized through the fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0045] In some embodiments, the controller of the air-conditioning system is a domain controller or a vehicle controller of the vehicle.

[0046] Additional aspects and advantages of the present invention will be set forth in part in the description which follows and, in part, will be obvious from the description which follows, or may be learned by practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments with reference to the accompanying drawings, in which: Figure 1 is a flow chart of a method for detecting the appearance quality of a heat exchanger according to one embodiment of the present invention; Figure 2 is a flow chart of a method for detecting the appearance quality of a heat exchanger according to another embodiment of the present invention; Figure 3 is a flow chart of a method for detecting the appearance quality of a heat exchanger according to another embodiment of the present invention; Figure 4 is a flow chart of a method for generating fin defect data according to one embodiment of the present invention; Figure 5 is a structural block diagram of an electronic device according to an embodiment of the present invention; Figure 6 is a structural block diagram of an air conditioning system according to an embodiment of the present invention; Figure 7 is a structural block diagram of a heat exchanger according to one embodiment of the present invention; Figure 8 is a structural block diagram of a vehicle according to one embodiment of the present invention; Figure 9 is a structural block diagram of a vehicle according to another embodiment of the present invention.

[0048] Reference numerals: Vehicle 100; air conditioning system 20; electronic device 10; Processor 1; memory 2; heat exchanger 3; controller 4; image acquisition device 5; Fins 31; harmonica tube 32; side plate 33; collecting pipe 34. Specific implementation method; The embodiments of the present invention will be described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. The embodiments of the present invention will be described in detail below.

[0050] Heat exchangers are key components in automotive air conditioning systems, enabling both cooling and heating. Fins, a crucial component of automotive air conditioning heat exchangers, increase the surface area in contact with the air, improving heat transfer efficiency and contributing to a more uniform temperature distribution across the heat exchanger surface. The shape and arrangement of the fins can disrupt the airflow, creating turbulence and further enhancing heat transfer efficiency. Fin quality and performance directly impact the efficiency and reliability of the entire air conditioning system, making quality inspection of automotive air conditioning heat exchanger fins particularly important.

[0051] Traditional image processing technologies have a significant limitation when applied to heat exchanger fin defect detection: they are highly dependent on good lighting and imaging environment conditions. Specifically, when the ambient light is insufficient or the shooting effect is poor, the contrast between the target detection object and the background will be significantly reduced, and more noise interference will be mixed into the image. Faced with this situation, conventional methods such as image threshold segmentation, edge detection, and morphological processing are often unable to effectively deal with noise problems due to their inherent sensitivity, thereby affecting the accuracy and reliability of image processing. Especially for structures such as heat exchanger fins with thin edges and high-frequency dense arrangement, the unclear edge and detail information makes it difficult for traditional algorithms to effectively separate the target area.

[0052] In order to solve the above problems, the first embodiment of the present invention provides a method for detecting the appearance quality of a heat exchanger. This method can realize the detection of appearance defects of the heat exchanger through a fin appearance detection model, thereby avoiding missed detections when performing defect detection on the heat exchanger, and avoiding false detections between fin defects with similar shapes.

[0053] Reference below Figure 1A method for detecting the appearance quality of a heat exchanger according to an embodiment of the present invention is described. As shown in the figure, the method at least includes: steps S1 and S2.

[0054] Step S1: using the appearance image of the heat exchanger as input to obtain a fin appearance detection model.

[0055] The heat exchanger's appearance image may be an image showing various appearance defects. The heat exchanger's appearance image is input into the fin appearance inspection model, which learns the image features of the heat exchanger's appearance image to identify the characteristics of the heat exchanger's appearance defects. The fin appearance inspection model is used to identify the image features of the appearance image. The fin appearance inspection model may be a deep learning model, such as a deep learning detection network. In step S2, based on the output of the fin appearance inspection model, an appearance inspection result of the heat exchanger's fins is obtained.

[0056] Among them, the appearance defect of the heat exchanger fin refers to the defect that the original upright state of the fin is broken due to external pressure during the production and transportation of the heat exchanger.

[0057] Specifically, an appearance image of the heat exchanger to be inspected is input into a fin appearance inspection model to obtain an output of the fin appearance inspection model. The output of the fin appearance inspection model can be appearance image information, which can be input appearance image feature information. The fin appearance inspection result of the heat exchanger fin is determined based on the output of the fin appearance inspection model. Specifically, the fin appearance inspection result of the heat exchanger fin is determined based on the appearance image information, and the presence of a heat exchanger appearance defect is determined based on the appearance image information. Therefore, compared to the prior art method of detecting heat exchanger defects using machine vision processing technology, the present application utilizes a fin appearance inspection model to implement heat exchanger appearance defect detection. Because the fin appearance inspection model does not rely on specific or hypothetical rules to determine whether a defect is a heat exchanger appearance defect, it avoids missed detections during defect detection of heat exchangers and false detections of fin defects with similar morphologies. Furthermore, it provides greater robustness when detecting defects of new morphologies. Furthermore, the present application utilizes a fin appearance inspection model to implement heat exchanger appearance defect detection, eliminating the need to identify edges and detail information in the appearance image to achieve defect detection, making defect detection independent of favorable lighting and imaging environment conditions.

[0058] According to the method for detecting the appearance quality of a heat exchanger according to an embodiment of the present invention, an appearance image of the heat exchanger is input into a fin appearance detection model, so as to judge whether there are defects in the appearance of the heat exchanger through the output of the fin appearance detection model. Therefore, compared with the method of detecting heat exchanger defects through machine vision processing technology in the prior art, the fin appearance detection model is used in the present application to realize heat exchanger appearance defect detection. Since the fin appearance detection model does not rely on specific or hypothetical rules to determine whether there is a heat exchanger appearance defect, it avoids missed detection when performing defect detection on the heat exchanger, and can also avoid false detection between fin defects with similar morphology.

[0059] In some embodiments, the fin appearance inspection result includes a fin appearance defect location and a fin appearance defect type at the fin appearance defect location.

[0060] Specifically, the appearance image of the inspected heat exchanger is input into the fin appearance inspection model, and the output of the fin appearance inspection model can be appearance image information, so as to determine the appearance inspection result of the heat exchanger fin through the appearance image information, that is, to judge whether there is a defect in the appearance of the heat exchanger through the appearance image information, and to determine the location of the fin appearance defect and the type of fin appearance defect at the fin appearance defect location. Therefore, in this application, the fin appearance inspection model is used to implement heat exchanger appearance defect type and positioning mask detection. Since the fin appearance inspection model does not rely on specific or hypothetical rules to determine whether a defect is a heat exchanger appearance defect, it avoids missed detection when inspecting defects in the heat exchanger, and can also avoid false detection between fin defects with similar morphology. In addition, it has better robustness when detecting defects of new morphologies.

[0061] In some embodiments, the training data set of the fin appearance detection model includes simulated data of fin appearance defects under different fin shapes and / or different environmental conditions of the fins obtained through data simulation. The environmental conditions may be lighting conditions, weather conditions, etc., which are not limited to these.

[0062] Specifically, since the frequency of appearance defects of certain specific fins is low in actual production environments, the training samples of the model are scarce. In order to solve this problem, the present application uses data simulation to generate fin appearance defect simulation data of different fin shapes and / or different fin environmental conditions, so as to add the fin appearance defect simulation data to the training data set. The training data set may include fin appearance defect simulation data of different fin shapes, or include fin appearance defect simulation data of different fin environmental conditions, or include fin appearance defect simulation data of different fin shapes and different fin environmental conditions, so as to solve the problems in actual scenarios. The problem of insufficient sample size of defect data is solved. Therefore, in this application, the amount of training data is expanded through data simulation, and defect simulation data of different fin shapes and different environmental conditions of the fins can be generated to construct a diversified training data set, which can help the fin appearance detection model to better learn various different defect characteristics, so that when the fin appearance detection model identifies fin defects, it can identify defects of different fin shapes and different environmental conditions of the fins, making the model more adaptable to different environments, and having higher stability and reliability when facing complex environments, improving the robustness and generalization ability of the model in detecting various defects, and reducing the risk of overfitting.

[0063] In some embodiments, fin morphology includes at least one of fin deformation, fin size, and fin position. That is, fin morphology includes fin deformation, fin size, or fin position, or fin morphology includes fin deformation and fin position, or fin morphology includes fin deformation, fin size, and fin position. Thus, in this application, data simulation can be used to generate simulation data of fin appearance defects under different fin deformation degrees, shapes, sizes, positions, and lighting conditions. Fin morphology may also include adjacent fin spacing, fin arrangement, fin surface structure, etc., without limitation.

[0064] In some embodiments, the environmental conditions of the fins include lighting conditions. Alternatively, the environmental conditions of the fins also include weather conditions. That is, through data simulation, simulated data of fin appearance defects of different fins under lighting conditions or weather conditions are obtained to construct a diverse training data set, thereby helping the fin appearance detection model to better learn various defect characteristics, making the model more adaptable to different environments, improving the robustness and generalization ability of the model in detecting various defects, and reducing the risk of overfitting.

[0065] In some embodiments, various types of fin appearance defect ROI (Region of Interest) labels of the fin appearance detection model are obtained by marking the defect locations of the heat exchanger. In an embodiment, the corresponding type of fin defect data can be merged into the target fin area of ​​the clean source image of the heat exchanger.

[0066] The clean source image of the heat exchanger is an image of the heat exchanger without any defects.

[0067] In some embodiments, the target fin region is obtained by separating the harmonica tube region of the heat exchanger from the binarized image of the clean source image of the heat exchanger; the harmonica tube region is obtained by performing an image morphological opening operation on the binarized image along the vertical direction of the harmonica tube.

[0068] Image morphology is a shape-based image processing method that uses structuring elements to detect specific shapes in an image, thereby enabling image analysis and recognition. Basic morphological operations include erosion, dilation, opening, closing, morphological gradient, top-hat operations, and black-hat operations.

[0069] Specifically, in this application, an image morphological opening operation is performed on the binarized image of the clean source image of the heat exchanger along the vertical direction of the harmonica tube to locate the harmonica tube area of ​​the binarized image, and the harmonica tube area is separated from other areas in the binarized image to obtain the target fin area. In this way, even when the ambient light for the fin image is insufficient or the shooting effect is poor, the target area can be effectively separated, thereby improving the accuracy and reliability of image processing.

[0070] In some embodiments, the fin appearance defect ROI tag includes a fin lodging defect ROI tag; the fin defect data includes a target fin lodging defect image, which is a fin lodging defect image that is scaled based on the height of the fin area in the clean source image of the heat exchanger.

[0071] Fin lodging defects occur when a fin fails to maintain its original upright position and instead collapses and clings to adjacent components, creating an irregular, shiny surface. In machine vision and image processing, a region of interest (ROI) is an area of ​​interest (ROI) that is defined in the image being processed using a box, circle, ellipse, or irregular polygon.

[0072] Specifically, an industrial 2D camera is used to capture actual heat exchanger images, wherein the actual heat exchanger images may be images of different fin shapes and / or different environmental conditions of the fins, and then the fin lodging defects in the actual heat exchanger images are annotated with ROI masks to obtain fin lodging defect ROI labels, and then several fin lodging defect images are extracted based on the fin lodging defect ROI labels, and then part or all of the several fin lodging defect images are scaled according to the height of the fin area in the clean source image of the heat exchanger, and then the above images are randomly combined to obtain a target fin lodging defect image, and then the target fin lodging defect image is merged into the target fin area of ​​the clean source image of the heat exchanger to obtain different fin lodging defect simulation images, wherein the deformation degree, shape, size, and position of the fin lodging in the fin lodging defect simulation image are different, and the fin lodging defect simulation image is annotated with ROI masks to obtain the fin lodging defect ROI label. Therefore, in this application, the amount of data in the training dataset is expanded by randomly combining and calculating the simulated images of fin tilt defects and generating corresponding fin tilt defect ROI labels, thereby solving the problem of insufficient defect data samples in actual scenarios. In addition, in this application, the heat exchanger appearance defect detection is realized by using a fin appearance detection model trained by the simulated images of fin collapse defects, thereby eliminating the need to detect defects by identifying the edges and details of the appearance image, so that defect detection does not rely on good lighting and imaging environment conditions.

[0073] Furthermore, it should be noted that if the environmental conditions in the actual heat exchanger image are different, the generated fin collapse defect simulation image will have different environmental conditions. Alternatively, the color channel of the fin collapse defect simulation image can be changed to generate fin collapse defect simulation images under different lighting conditions.

[0074] In some embodiments, the fin lodging defect ROI label is obtained by randomly merging a target fin lodging defect image into the fin region in the clean source image of the heat exchanger using a Poisson fusion method. That is, the size of the fin lodging defect image is scaled by the height of the fin region in the clean source image of the heat exchanger to obtain the target fin lodging defect image. The target fin lodging defect image is then randomly merged into the fin region in the clean source image of the heat exchanger using a Poisson fusion method to generate the fin lodging defect ROI label.

[0075] In some embodiments, the fin appearance defect ROI label includes a fin slant defect ROI label; the fin defect data includes a first target transformation matrix, the first target transformation matrix is ​​obtained based on a first target matching point, and the first target matching point is determined based on a target fin area of ​​a clean source image of the heat exchanger.

[0076] The oblique fin defect refers to a change in the original shape of the fin perpendicular to the harmonica tube, resulting in an excessive tilt. The first target matching point is used to calculate the shape of the oblique fin defect, and the first target matching point can reflect the change in the shape of the oblique fin defect.

[0077] Specifically, in order to form a fin oblique defect simulation image and a fin oblique defect ROI label to enrich the data volume of the training data set, the first target matching point is determined based on the target fin area of ​​the clean source image of the heat exchanger, that is, a part of the area where the fin is located in the clean source image of the heat exchanger is arbitrarily selected as the target fin area, and then the first target matching point is determined according to the target fin area, and then the first target transformation matrix is ​​obtained according to the first target matching point, that is, the transformation relationship between a pair of points in the first target matching point is calculated according to the position information of the two points, and the transformation relationship is the first target transformation matrix, and finally the first target transformation matrix and the clean source image of the heat exchanger are multiplied to calculate the fin oblique defect simulation image, that is, the first target transformation matrix is ​​applied to the clean source image of the heat exchanger without appearance defects to obtain a fin oblique defect simulation image, and the first target transformation matrix determines the fin. The deformation degree, shape and size of the fin bevel are determined so that the deformation degree, shape and size of the fin bevel in the fin bevel defect simulation image are different, and the fin bevel defects in the fin bevel defect simulation image are annotated with ROI masks to obtain fin bevel defect ROI labels, wherein the fin bevel defect ROI label can be understood as a defect in the target fin area, and since different areas can be selected as target fin areas in the clean source image of the heat exchanger where the fins are located, multiple different fin bevel defect simulation images and corresponding bevel defect ROI labels are generated, so that the positions of the bevel defects in the fin bevel defect simulation image are different. Therefore, in this application, the fin bevel defect simulation image is obtained by calculating the first target transformation matrix and the clean source image of the heat exchanger and the corresponding fin bevel defect ROI label is generated to expand the data volume of the training data set, thereby solving the problem of insufficient defect data sample size in actual scenarios.

[0078] In addition, the color channel of the fin oblique chip defect simulation image can be changed to generate the fin oblique chip defect simulation images under different lighting conditions.

[0079] In some embodiments, the first target transformation matrix is ​​obtained by performing an affine transformation on the first target matching point. In other words, the first target transformation matrix of the first target matching point is calculated using an affine transformation.

[0080] In some embodiments, the first target matching point is obtained based on the first original point and the second target point of the target fin area; wherein the target fin area is a rectangle, the first original point is the four vertices of the target fin area, and the first target point is the four vertices of a parallelogram with the same base as the target fin area.

[0081] Specifically, a part of the area where the fins are located in the clean source image of the heat exchanger is arbitrarily selected as the target fin area. Since the fin area is a rectangle, the target fin area is a rectangle so that the target fin area can contain all the fins. The four vertices of the target fin area are selected as the first original points. The first original points are used to represent the fins without the oblique fin defect, and the four vertices of the parallelogram with the same base as the target fin area are used as the first target points. The first target point is used to represent the fins with the oblique fin defect, so that the first original point and the first target point are used as the first target matching points, wherein the first target matching point can be 4 groups of matching points, thereby obtaining the first target matching point.

[0082] In some embodiments, the fin appearance defect ROI label includes a fin deformation defect ROI label and / or a fin shrinkage defect ROI label, that is, the fin appearance defect ROI label may include a fin deformation defect ROI label, or include a fin shrinkage defect ROI label, or include a fin deformation defect ROI label and a fin shrinkage defect ROI label; the fin defect data includes a second target transformation matrix, the second target transformation matrix is ​​obtained based on a second target matching point, and the second target matching point is determined based on a target fin area of ​​a clean source image of the heat exchanger.

[0083] Deformation defects refer to fins that are bent from their original upright state. Fin shrinkage defects refer to changes in the fin spacing compared to the original design. Fin shrinkage defects include increases or decreases in the distance between fins.

[0084] Specifically, in order to form a fin deformation defect simulation image and a fin deformation defect ROI label to enrich the data volume of the training data set, a second target matching point is determined based on the target fin area of ​​the clean source image of the heat exchanger, that is, a part of the area where the fin is located in the clean source image of the heat exchanger is arbitrarily selected as the target fin area, and then the second target matching point is determined based on the target fin area, wherein the second target matching point is a matching point selected based on the morphological change that can reflect the fin deformation, and then the second target transformation matrix is ​​obtained based on the second target matching point, that is, the transformation relationship between a pair of points in the second target matching point is calculated based on the position information of the two points, and the transformation relationship is the second target transformation matrix, and finally the second target transformation matrix and the clean source image of the heat exchanger are multiplied to calculate the fin deformation defect simulation image, that is, the second target transformation matrix is ​​applied to the clean source image of the heat exchanger without appearance defects to obtain a fin deformation defect simulation image. , and the second target transformation matrix determines the deformation degree, shape, and size of the fin deformation, so that the deformation degree, shape, and size of the fin oblique blade in the fin deformation defect simulation image are different, and the fin deformation defect of the fin deformation defect simulation image is annotated with an ROI mask to obtain a fin deformation defect ROI label, wherein the fin deformation defect ROI label can be understood as the defect ROI label of the target fin area, and since different areas can be selected as target fin areas in the clean source image of the heat exchanger where the fin is located, multiple different fin deformation defect simulation images and corresponding fin deformation defect ROI labels are generated, so that the position of the deformation defect in the fin deformation defect simulation image is different. Therefore, in this application, the fin deformation defect simulation image is obtained by calculating the second target transformation matrix and the clean source image of the heat exchanger and generating the corresponding fin deformation defect ROI label to expand the data volume of the training data set, thereby solving the problem of insufficient defect data sample size in actual scenarios.

[0085] And / or, in order to form a fin shrinkage defect simulation image and a fin shrinkage defect ROI label to enrich the data volume of the training data set, a second target matching point is determined based on the target fin area of ​​the clean source image of the heat exchanger, that is, a part of the area where the fin is located in the clean source image of the heat exchanger is arbitrarily selected as the target fin area, and then the second target matching point is determined based on the target fin area, wherein the second target matching point is a matching point selected based on the morphological change that can reflect the fin shrinkage, and then a second target transformation matrix is ​​obtained based on the second target matching point, that is, the transformation relationship between a pair of points in the second target matching point is calculated based on the position information of the two points, and the transformation relationship is the second target transformation matrix, and finally the second target transformation matrix and the clean source image of the heat exchanger are multiplied to calculate the fin shrinkage defect simulation image, that is, the second target transformation matrix is ​​applied to the clean source image of the heat exchanger without appearance defects to obtain a fin shrinkage defect simulation image. , and the second target transformation matrix determines the deformation degree, shape, and size of the fin deformation of the fin, so that the deformation degree, shape, and size of the fin in the fin tightening defect simulation image are different, and the fin tightening defect of the fin tightening defect simulation image is annotated with an ROI mask to obtain a fin tightening defect ROI label, wherein the fin tightening defect ROI label can be understood as the defect ROI label of the target fin area, and since different areas can be selected as target fin areas in the clean source image of the heat exchanger where the fin is located, multiple different fin tightening defect simulation images and corresponding fin tightening defect ROI labels are generated, so that the position of the tightening defect in the fin tightening defect simulation image is different. Therefore, in this application, the fin tightening defect simulation image is obtained by calculating the second target transformation matrix and the clean source image of the heat exchanger and generating the corresponding fin tightening defect ROI label to expand the data volume of the training data set, thereby solving the problem of insufficient defect data sample size in actual scenarios.

[0086] In addition, the color channels of the fin deformation defect simulation image and / or the fin shrinkage defect simulation image may be changed to generate fin deformation defect simulation images and / or fin shrinkage defect simulation images under different lighting conditions.

[0087] In some embodiments, the second target transformation matrix is ​​obtained by performing a thin plate spline interpolation method (TPS) based on N groups of second target matching points, where N is a positive integer greater than or equal to 2, for example, N can be 2, 3, or 4, without limitation. The thin plate spline interpolation method is an interpolation method used to find a "minimum bending" smooth surface passing through all given points. The thin plate spline interpolation method determines the interpolation function by minimizing the bending energy. Therefore, the present application uses the thin plate spline interpolation method to calculate the second target transformation matrix of the N groups of second target matching points.

[0088] In some embodiments, the second target matching point is obtained based on N second original points and N second target points in the target fin area; wherein the N second original points and the N second target points are randomly selected in the target fin area.

[0089] Specifically, a part of the area where the fins are located in the clean source image of the heat exchanger is arbitrarily selected as the target fin area, and N second original points and N second target points are randomly selected from all points in the target fin area, wherein the second original points are used to represent the fins when no deformation defects or shrinkage defects occur, and the second target points are used to represent the fins with deformation defects or shrinkage defects, and the N second original points and N second target points are used as second target matching points, wherein the second target matching points are N groups of matching points.

[0090] In some embodiments, for the fin deformation defect ROI label, the second target matching point is determined based on the fin bending shape.

[0091] In some embodiments, for the fin shrinkage defect ROI label, the second target matching point is determined based on the fin spacing.

[0092] In some embodiments, the target fin region is a rectangular region of any size randomly demarcated at the fin region on the clean source image of the heat exchanger.

[0093] In some embodiments, the fin appearance detection model is obtained by training a segmentation deep learning network based on a training set. The segmentation deep learning network includes an image encoder module and a mask decoder module. The image encoder module is used to extract image features, and the mask decoder module uses the extracted features and reconstructs the segmentation mask. Among them, the image encoder module includes but is not limited to structures such as ResNet (Residual Network), ViT (Vision Transformer) and Swin Transformer; the mask decoder module includes but is not limited to structures such as HRNett (High-Resolution Network), BiSeNet (Bilateral Segmentation Network) and Mask2ForMer (Masked-attention Mask Transformer). Therefore, the segmentation deep learning network is used to realize the positioning and type determination of the fin defect area.

[0094] Therefore, this application adopts a segmentation deep learning network to realize the detection of structural defects of the heat exchanger fin appearance, so that defect detection does not rely on good lighting and imaging environment conditions.

[0095] In the embodiment, the fin appearance detection model is trained using a training data set. The training process is as follows: Dice Loss (similarity coefficient loss) is defined as and Cross-Entropy Loss As the loss function, the Adam (Adaptive Moment Estimation) optimizer is used to calculate the gradient information and update the network parameters until the fin appearance detection model converges.

[0096]

[0097] Among them, P is the predicted mask and T is the true mask.

[0098]

[0099] in, is the class prediction probability of the mask pixel, is the true label.

[0100] In some embodiments, a core appearance inspection result of the heat exchanger is obtained based on an appearance image of the heat exchanger.

[0101] Specifically, an image of the heat exchanger's appearance is input into the fin appearance inspection model, which then outputs appearance image information. This information is then used to determine the heat exchanger's fin appearance inspection results. Specifically, the appearance image information determines whether the heat exchanger's appearance has defects, and also determines the location of any fin defects and the type of defect at the location. Thus, the present application utilizes the fin appearance inspection model to implement heat exchanger appearance inspection.

[0102] In some embodiments, the core appearance inspection result includes the presence of core tilting deformation; the core tilting deformation is determined based on the angle between the side plate fitting straight line and the header fitting curve.

[0103] Specifically, due to excessive press-fitting of the heat exchanger during production, the appearance of the heat exchanger core will change from a matrix to a parallelogram. That is to say, when the core is not tilted and deformed, the side plate is perpendicular to the manifold. Therefore, the angle between the side plate fitting line and the manifold fitting curve can be judged to determine whether the core is tilted and deformed. For example, if the angle between the side plate fitting line and the manifold fitting curve is not a right angle, or the angle between the side plate fitting line and the manifold fitting curve is very different from the right angle, it means that the core is tilted and deformed into a parallelogram, and the core appearance inspection result can be determined to be a core tilt deformation. Therefore, in this application, the angle between the side plate fitting line and the manifold fitting curve is used to detect the core tilt deformation problem.

[0104] In some embodiments, the presence of core body tilting deformation includes an angle exceeding an angle threshold. Since the side plate is perpendicular to the manifold when the core body is not tilted and deformed, and considering the accuracy limitations of the measuring tool and the difficulty of achieving absolute verticality between the side plate and the manifold during installation, the angle between the side plate and the manifold is allowed to have a certain angle deviation range, and the angle threshold can be set to [85, 95]. Based on this, if it is detected that the angle between the side plate fitting line and the manifold fitting curve exceeds the angle threshold, it means that the core body tilting deformation is a parallelogram, and the core body appearance inspection result can be determined to be a core body tilting deformation. Therefore, in this application, the detection of the core body tilting deformation problem is achieved through the angle between the side plate fitting line and the manifold fitting curve and the angle threshold.

[0105] In some embodiments, the core appearance inspection result includes the presence of side panel bending; the side panel bending is determined based on the distance residual between the side panel contour points and the side panel fitting line.

[0106] Specifically, heat exchanger side plate deformation refers to bending of the side plates due to collisions and extrusion during production or transportation. Determination of side plate bending can be achieved by estimating the side plate contour points and the side plate fitting line. The side plate contour points reflect the actual shape of the side plate, while the side plate fitting line reflects the distribution trend of the side plate contour points. The specific steps are as follows: Select K edge plate contour points (x k ,y k ), calculate the mean square deviation d of the distance residual between the edge plate contour point and the edge plate fitting line y = ax + b MSE :

[0107] Among them, the mean square deviation d MSEThe method can reflect the degree of difference between the edge panel contour points and the edge panel fitting line, that is, quantify the degree of deviation between the edge panel contour points and the edge panel fitting line through the mean square deviation between the edge panel contour points and the edge panel fitting line, so as to determine the degree of bending of the edge panel at each edge panel contour point. Therefore, whether the edge panel is bent can be determined by the distance residual between the edge panel contour points and the edge panel fitting line, thereby improving the accuracy of edge panel bending detection.

[0108] In some embodiments, the presence of side panel curvature includes a distance residual greater than a residual threshold. The distance residual can reflect the degree of difference between the side panel contour points and the side panel fitted line. A larger distance residual indicates a greater degree of difference between the side panel contour points and the side panel fitted line, i.e., a greater degree of curvature of the side panel. If the distance residual is greater than the residual threshold, it indicates that the side panel fitted line has significant curvature.

[0109] In an embodiment, if the side panel is curved, the deviation change rate can be further analyzed, that is, how the deviation value of each point along the side panel curve changes, that is, the side panel contour is divided into multiple segments, and a straight line fitting is performed on each small segment to obtain multiple side panel fitting straight lines, and then the deviation values ​​between the adjacent side panel fitting straight lines are calculated, and finally the deviation values ​​are accumulated along the curve to obtain the degree of curvature of the overall side panel.

[0110] In some embodiments, the side plate fitting straight line is obtained based on the outer contour fitting of the side plate; the manifold fitting straight line is obtained based on the outer contour fitting of the manifold; the outer contour of the side plate and the outer contour of the manifold are obtained based on edge detection of the initial positioning area of ​​the side plate and the manifold, that is, in some embodiments, the side plate fitting straight line is the side plate edge fitting straight line, and the manifold fitting straight line is the manifold edge fitting straight line.

[0111] In some embodiments, the initial positioning region is obtained by separating the maximum connected domains in the horizontal and vertical directions along the edges of the pre-processed image of the heat exchanger.

[0112] Exemplarily, after obtaining the initial positioning area by separating the maximum connected domains in the horizontal and vertical directions along the edge of the heat exchanger preprocessed image, the preprocessed image is binarized and edge detected again along the image to extract the outer contours of the side plate and the manifold, so as to obtain the side plate fitting straight line by fitting the outer contour of the side plate, and obtain the manifold fitting straight line by fitting the outer contour of the manifold.

[0113] In some embodiments, the heat exchanger pre-processed image is obtained by binarizing and morphologically processing the grayscale image of the heat exchanger clean source image. In other words, the heat exchanger clean source image is converted into a grayscale image, binarized, and morphologically processed to obtain the heat exchanger pre-processed image.

[0114] In some embodiments, the clean source image of the heat exchanger is an image of the front of the heat exchanger. Specifically, an industrial 2D camera and light source are used to capture images of the front of the heat exchanger, obtaining multiple images of the front of the heat exchanger. The image of the front of the heat exchanger without defects is selected as the clean source image of the heat exchanger. A polarizing filter is added to the front image of the heat exchanger to filter out metal reflections in the image, resulting in a high-definition image of the heat exchanger.

[0115] Reference below Figure 2 The method for detecting the appearance quality of a heat exchanger according to an embodiment of the present invention is illustrated as an example, and the specific content is as follows.

[0116] Step S3: constructing a training data set for the fin appearance detection model using the fin appearance defect images and the fin appearance defect ROI labels.

[0117] Step S4: constructing a fin appearance detection model using the training data set.

[0118] Step S5: Use the training data set to train the fin appearance detection model, calculate the loss function, and update the weights of the fin appearance detection model until convergence.

[0119] Step S6: input the appearance image of the heat exchanger as the image to be inspected into the fin appearance inspection model to obtain the fin appearance inspection result of the heat exchanger.

[0120] Reference below Figure 3 The method for detecting the appearance quality of a heat exchanger according to an embodiment of the present invention is illustrated as an example, and the specific content is as follows.

[0121] Step S7: using an image processing method to preliminarily locate the side plate and header area of ​​the heat exchanger appearance image.

[0122] Step S8: fitting the edge straight lines of the side plates and the manifold in the initial positioning area.

[0123] Step S9: Calculate the angle between the side plate and the edge straight line of the manifold to determine whether the core is tilting and deforming.

[0124] Step S10: Calculate the distance residual between the edge panel contour point and the edge panel fitting line to determine whether there is a core edge panel deformation problem.

[0125] Reference below Figure 4 The fin defect data generating method according to an embodiment of the present invention is illustrated as an example, and the specific contents are as follows.

[0126] In step S11 , a front image of a heat exchanger without defects is selected to locate the fin area as a clean source image of the heat exchanger, and steps S12 , S14 and S16 are executed.

[0127] Step S12: Select the fin collapse defect ROI label, scale it, and integrate it into the clean source image of the heat exchanger.

[0128] The fin lodging defect ROI label is a fin lodging defect ROI label corresponding to any image containing the fin lodging defect.

[0129] Step S13: Generate a fin lodging defect simulation image and a corresponding fin lodging defect ROI label.

[0130] Step S14 : determining a first target matching point based on the target fin region of the clean source image of the heat exchanger, and then calculating a first target transformation matrix according to the first target matching point.

[0131] Step S15 , generating a fin oblique-piece defect simulation image and a fin oblique-piece defect ROI label according to the first target transformation matrix.

[0132] Step S16: determining a second target matching point based on the target fin region of the clean source image of the heat exchanger, and then calculating a second target transformation matrix according to the second target matching point.

[0133] Step S17 : generating a fin deformation defect simulation image / or a fin shrinkage defect simulation image, a fin deformation defect ROI label and / or a fin shrinkage defect ROI label according to the second target transformation matrix.

[0134] A second aspect of the present invention provides an electronic device, such as Figure 5 As shown, the electronic device 10 includes: at least one processor 1 and a memory 2 communicatively connected to the at least one processor 1 .

[0135] The memory stores a computer program that can be executed by at least one processor, and when the at least one processor executes the computer program, the method for detecting the appearance quality of the heat exchanger according to the above embodiment is implemented.

[0136] According to the electronic device of an embodiment of the present invention, by executing the method for detecting the appearance quality of a heat exchanger according to the above embodiment, the appearance defect detection of a heat exchanger can be realized through a fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0137] A third aspect of the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method for detecting the appearance quality of a heat exchanger according to the above embodiment.

[0138] A fourth aspect of the present invention provides an air conditioning system, such as Figure 7As shown, the air conditioning system 20 includes a heat exchanger 3 and a controller 4 .

[0139] The controller is used to execute the method for detecting the appearance quality of the heat exchanger according to the above embodiment.

[0140] According to the air-conditioning system of an embodiment of the present invention, by executing the method for detecting the appearance quality of the heat exchanger of the above embodiment, the appearance defect detection of the heat exchanger can be realized through the fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0141] In some embodiments, as Figure 7 As shown, the air conditioning system 20 further includes: an image acquisition device 5 .

[0142] The image acquisition device is connected to the controller and is used to acquire an appearance image of the heat exchanger.

[0143] A fifth aspect of the present invention provides a vehicle, such as Figure 8 As shown, the vehicle 100 includes the electronic device 10 of the above embodiment; or Figure 9 As shown, a vehicle 100 includes the air conditioning system 20 of the above embodiment. The air conditioning system includes a heat exchanger, and the heat exchanger includes fins 31, harmonica tubes 32, side plates 33 and a header 34.

[0144] According to the vehicle of the embodiment of the present invention, through the electronic equipment or air-conditioning system of the above embodiment, the heat exchanger appearance defect detection can be realized through the fin appearance detection model, thereby avoiding missed detection when performing defect detection on the heat exchanger, and also avoiding false detection between fin defects with similar shapes.

[0145] In some embodiments, the controller of the air conditioning system is a domain controller or a vehicle controller of the vehicle.

[0146] In the description of this specification, any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present invention includes additional implementations in which functions may be performed in a different order than shown or discussed, including performing functions in a substantially simultaneous manner or in a reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0147] The logic and / or steps represented in a flowchart or otherwise described herein, for example, can be considered a sequenced list of executable instructions for implementing the logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (e.g., a computer-based system, a system including a processor, or other system that can fetch and execute instructions from an instruction execution system, apparatus, or device). For purposes of this specification, a "computer-readable medium" is any device that can contain, store, communicate, propagate, or transport a program for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (not exhaustive) of computer-readable media include: an electrical connection with one or more wires (electronic devices), a portable computer disk cartridge (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and programmable read-only memory (EPROM or flash memory), fiber optic devices, and a portable compact disc read-only memory (CDROM). Furthermore, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing it in another suitable manner if necessary, and then storing it in a computer memory.

[0148] It should be understood that various components of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the aforementioned embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any of the following technologies known in the art, or a combination thereof, may be used: a discrete logic circuit having logic gates for implementing logic functions on data signals, an application-specific integrated circuit having suitable combinational logic gates, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0149] Those skilled in the art will understand that all or part of the steps in the method of the above embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0150] In addition, the functional units in the various embodiments of the present invention may be integrated into a single processing module, or each unit may exist physically separately, or two or more units may be integrated into a single module. The aforementioned integrated modules may be implemented in the form of hardware or in the form of software functional modules. If the integrated modules are implemented in the form of software functional modules and sold or used as independent products, they may also be stored in a computer-readable storage medium.

[0151] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and are not to be construed as limiting the present invention. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

[0152] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "illustrative embodiments," "example," "specific example," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example.

[0153] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A method for detecting the appearance quality of a heat exchanger, characterized in that: Based on the output of the fin appearance detection model, a fin appearance detection result of the heat exchanger is obtained, wherein the fin appearance detection model takes the appearance image of the heat exchanger as input.

2. The method according to claim 1, characterized in that The fin appearance inspection result includes a fin appearance defect position and a fin appearance defect type at the fin appearance defect position.

3. The method according to claim 1, characterized in that The training data set of the fin appearance detection model includes fin appearance defect simulation data of different fin shapes and / or different environmental conditions of the fins obtained through data simulation.

4. The method according to claim 3, characterized in that The fin shape includes at least one of fin deformation, fin size, and fin position.

5. The method according to claim 3, characterized in that The environmental conditions of the fins include light conditions.

6. The method according to claim 1, wherein The ROI labels of various types of fin appearance defects of the fin appearance detection model are obtained by merging the corresponding types of fin defect data into the target fin region of the clean source image of the heat exchanger.

7. The method according to claim 6, characterized in that The target fin region is obtained by separating the harmonica tube region of the heat exchanger from the binarized image of the clean source image of the heat exchanger; The harmonica tube region is obtained by performing an image morphological opening operation on the binary image along a vertical direction of the harmonica tube.

8. The method according to claim 6, characterized in that The fin appearance defect ROI label includes a fin lodging defect ROI label; The fin defect data includes a target fin lodging defect image, which is a fin lodging defect image that is scaled based on the height of a fin region in the clean source image of the heat exchanger.

9. The method according to claim 8, characterized in that The fin lodging defect ROI label is obtained by randomly merging the target fin lodging defect image into the fin region in the clean source image of the heat exchanger using a Poisson fusion method.

10. The method according to claim 6, characterized in that The fin appearance defect ROI label includes a fin oblique defect ROI label; The fin defect data includes a first target transformation matrix, which is obtained based on a first target matching point, which is determined based on a target fin region of the clean source image of the heat exchanger.

11. The method according to claim 10, characterized in that The first target transformation matrix is ​​obtained by performing affine transformation calculation based on the first target matching point.

12. The method according to claim 10, characterized in that The first target matching point is obtained based on the first original point and the second target point of the target fin area; The target fin region is a rectangle, the first original points are the four vertices of the target fin region, and the first target points are the four vertices of a parallelogram having the same base as the target fin region.

13. The method according to claim 6, characterized in that The fin appearance defect ROI label includes a fin deformation defect ROI label and / or a fin shrinkage defect ROI label; The fin defect data includes a second target transformation matrix, which is obtained based on second target matching points, and the second target matching points are determined based on a target fin region of the clean source image of the heat exchanger.

14. The method according to claim 13, characterized in that The second target transformation matrix is ​​obtained by calculating the thin plate spline interpolation method based on N groups of the second target matching points, where N is a positive integer greater than or equal to 2.

15. The method according to claim 13, characterized in that The second target matching point is obtained based on the N second original points and the N second target points of the target fin area; The N second original points and the N second target points are randomly selected from the target fin area.

16. The method according to claim 13, characterized in that For the fin deformation defect ROI label, the second target matching point is determined based on the bending shape of the fin.

17. The method according to claim 13, wherein For the fin shrinkage defect ROI label, the second target matching point is determined based on the fin spacing.

18. The method according to any one of claims 10 to 17, characterized in that: The target fin area is a rectangular area of ​​any size randomly demarcated at the fin area on the clean original image of the heat exchanger.

19. The method according to any one of claims 1 to 17, characterized in that The fin appearance detection model is obtained by training a segmentation deep learning network based on a training set.

20. The method according to claim 1, wherein The method further comprises: Based on the appearance image of the heat exchanger, a core appearance inspection result of the heat exchanger is obtained.

21. The method according to claim 20, characterized in that The core body appearance inspection result includes the presence of core body tilting and deformation; The core body tilting deformation is determined based on the angle between the side plate fitting straight line and the header fitting curve.

22. The method according to claim 21, characterized in that The existence of the core dumping deformation includes the angle exceeding a threshold value.

23. The method according to claim 20, characterized in that The core appearance inspection result includes the presence of bent side panels; The edge panel curvature is determined based on the distance residual between the edge panel contour points and the edge panel fitting line.

24. The method according to claim 23, wherein The existence of edge plate bending includes the distance residual being greater than a residual threshold.

25. The method according to any one of claims 21 to 24, characterized in that The side plate fitting straight line is obtained based on the outer contour fitting of the side plate; The manifold fitting straight lines are all obtained based on the outer contour fitting of the manifold; The outer contours of the side plates and the manifold are obtained by performing edge detection on the initial positioning areas of the side plates and the manifold.

26. The method according to claim 25, characterized in that The initial positioning area is obtained by separating the maximum connected domains in the horizontal and vertical directions along the edges of the heat exchanger preprocessing image.

27. The method according to claim 26, characterized in that The heat exchanger preprocessed image is obtained by performing a binarization operation and morphological processing on the grayscale image of the heat exchanger clean source image.

28. The method according to any one of claims 6 to 17 and 27, characterized in that The clean source image of the heat exchanger is a front image of the heat exchanger.

29. An electronic device, characterized in that: include: at least one processor; a memory communicatively coupled to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, and when the at least one processor executes the computer program, the method for detecting the appearance quality of a heat exchanger according to any one of claims 1 to 28 is implemented.

30. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for detecting the appearance quality of a heat exchanger according to any one of claims 1 to 28 is implemented.

31. An air conditioning system, characterized in that: The method comprises a heat exchanger and a controller, wherein the controller is used to execute the method for detecting the appearance quality of a heat exchanger according to any one of claims 1 to 28.

32. The air conditioning system according to claim 31, characterized in that The air conditioning system further comprises: An image acquisition device is connected to the controller and is used to acquire an appearance image of the heat exchanger.

33. A vehicle, characterized in that: The vehicle includes the electronic device of claim 29; Alternatively, the vehicle includes the air conditioning system of claim 31.

34. The vehicle according to claim 33, characterized in that The controller of the air-conditioning system is a domain controller or a vehicle controller of the vehicle.

Images