A bulge heat dissipation fin quality detection method based on image processing

Abstract

The present application relates to the technical field of image processing, and particularly relates to a kind of based on image processing's bulge heat dissipation fin quality detection method, comprising: the original image of fin is collected and pretreated to obtain the independent area of each layer fin;Obtain the candidate bulge area in the independent area of target fin;Get each projection search line in the independent area of the adjacent fin below target fin;Along each projection search line is searched to obtain each preset radial direction pixel point sequence;Obtain the complete profile function of each preset radial direction of each candidate bulge area, and then obtain the maximum height of bulge physical uplift, bulge edge slope and bulge top flatness in each preset radial direction of each candidate bulge area, then calculate the comprehensive hazard score of each candidate bulge area;The quality of target fin is detected using the comprehensive hazard score.The present application can effectively detect the quality of heat dissipation fin with bulge.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and more specifically to a method for detecting the quality of bulging heat sink fins based on image processing. Background Technology

[0002] Heat sink fins are core components in heat exchangers, automotive radiators, and air conditioning condensers. They are typically formed by continuous stamping of thin aluminum or copper plates using a high-speed stamping machine and assembled in a multi-layered stacked structure. A narrow airflow channel of approximately 1-3 mm exists between adjacent fins, carrying away heat as the airflow passes through. During the stamping and subsequent brazing assembly processes, localized bulge defects (i.e., "bulges") may appear on the fin surface. These defects reduce the cross-sectional area of ​​the airflow channel, increase flow resistance, and in severe cases, completely block the corresponding channel, directly causing a precipitous drop in heat dissipation performance.

[0003] Currently, machine vision-based defect detection for fin surfaces is being increasingly widely adopted. However, existing methods rely solely on the area or grayscale difference of bulges in two-dimensional images to determine the severity of defects, and the resolution of general-purpose three-dimensional sensors is insufficient to operate effectively within the narrow space of fins. When using two-dimensional vision methods for analysis, they only focus on the grayscale anomaly of the fin itself where the bulge is located, ignoring the fact that in multi-layered fin structures, when a bulge appears on the surface of a certain layer of fins, the bulge's elevation blocks the incident light from a light source at a fixed pitch angle, creating a projected shadow on the surface of the adjacent lower fin. This occlusion effect results in a shadow area with a significantly lower grayscale than the background on the surface of the adjacent fin directly below the bulge. The horizontal projection length of this shadow has a geometric function relationship with the physical elevation height of the bulge, uniquely determined by the pitch angle of the light source. However, if the shadow is simply considered as an "accompanying phenomenon of the bulge" and its length is measured, the fundamental measurement difficulties arise due to the limited interlayer space. The shadowing surface of the bulge is the surface of the adjacent lower fin. This shadowing surface does not extend infinitely in the horizontal direction, but is limited by the fin's own boundaries, the peaks and troughs of the stamping corrugations, and the spatial obstruction of multiple layers of stacked fins. As a result, the shadow is often not a complete free projection, but a "truncated shadow" that is truncated by the interlayer boundaries, cut by the corrugation edges, and superimposed with the structural shadow of the lower fin itself. Specifically, when the bulge is large, its projected length may exceed the effective shadowing range of the adjacent fin, and the end of the shadow is cut off by the fin edge or the adjacent corrugated structure. When there are periodic corrugations on the surface of the lower fin, the gray distribution of the shadow is mixed with the bright and dark stripes of the corrugations themselves, making it difficult to identify the true boundary of the shadow. When the bulge is located near the edge of the fin, the shadow may partially fall outside the fin gap, and only a part of the projected segment can be observed. Summary of the Invention

[0004] To address the aforementioned technical problems, the present invention aims to provide a method for detecting the quality of bulging heat sink fins based on image processing. The specific technical solution adopted is as follows:

[0005] One embodiment of the present invention provides a method for detecting the quality of bulging heat sink fins based on image processing, the method comprising:

[0006] The original image of the fin is acquired and preprocessed to obtain the independent region of each fin layer, and any fin layer is recorded as the target fin; the local grayscale changes of each pixel in the independent region of the target fin are analyzed to obtain the candidate bulge region and the geometric center of the candidate bulge region.

[0007] Project the line segments between the geometric center of a candidate bulge region and the intersection of each preset radial direction with the boundary of the candidate bulge region onto the independent region of the adjacent fin below the target fin to obtain each projection search line; search along each projection search line to obtain the pixel sequence in each preset radial direction.

[0008] Using a partial contour function corresponding to a pixel sequence in a preset radial direction of a candidate bulge region, a cubic spline extrapolation method is used to obtain the complete contour function of the candidate bulge region in that preset radial direction. The complete contour function is then used to obtain the maximum physical height of the bulge, the bulge edge slope, and the bulge top surface flatness in each preset radial direction of each candidate bulge region. The comprehensive hazard score of a candidate bulge region is calculated using the maximum physical height of the bulge, the bulge edge slope, and the bulge top surface flatness in each preset radial direction. The quality of the target fin is then inspected using the comprehensive hazard score.

[0009] Preferably, the original image of the fin is acquired and preprocessed to obtain independent regions for each fin layer, including:

[0010] The original image is filtered to obtain the filtered image. The cumulative gray value of each row is calculated along the vertical direction of the filtered image, and a gray-scale projection curve is drawn. The horizontal axis of the gray-scale projection curve is the row number of each row in the filtered image, and the vertical axis is the cumulative gray value of all pixels in the row. The range between every two adjacent valley points in the gray-scale projection curve is the area where a fin is located. The filtered image is segmented using the horizontal coordinates of each valley point in the gray-scale projection curve as the segmentation boundary to obtain the independent regions of each fin layer.

[0011] Preferably, candidate bulge regions are obtained by analyzing the local grayscale changes of each pixel within an independent region of the target fin, including:

[0012] A window is constructed centered on a pixel within an independent region of the target fin. The judgment threshold for that pixel is obtained by adding the mean of the gray values ​​of all pixels within the window to the sensitivity coefficient and the standard deviation of the gray values ​​of all pixels. If the gray value of the pixel is greater than the judgment threshold, the pixel is a candidate abnormal pixel. Connectivity analysis is performed on all candidate abnormal pixels within the independent region of the target fin to obtain different connected components. Connectivity components with a number of pixels within the area constraint are retained as candidate bulge regions.

[0013] Preferably, the process of searching along each projection search line to obtain the pixel sequence in each preset radial direction includes:

[0014] Starting from the beginning of a projection search line, pixels are searched sequentially along the projection search line. When the gray value of the newly searched pixel is less than the gray value of the previously searched pixels, or when the absolute value of the difference between the first-order difference of the newly searched pixel and the first-order difference of the previous search is less than a preset continuity threshold, the search stops and the previously searched pixels are combined to form a pixel sequence in the preset radial direction corresponding to the projection search line.

[0015] Preferably, the partial contour function is

[0016] ,

[0017] in, For partial contour functions, This refers to the projection position of a pixel in the pixel sequence corresponding to a radial direction of a candidate bulge region of the target fin. This refers to the position of a point on the contour surface of the candidate bulge region corresponding to that pixel in the pixel sequence. The pitch angle of the light source.

[0018] Preferably, the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the bulge top surface are obtained in each preset radial direction of each candidate bulge region using the complete contour function, including:

[0019] The maximum physical height of the bulge in the preset radial direction of a candidate bulge region is obtained by converting the pixel length of the pixel sequence in a preset radial direction of the candidate bulge region into a physical length and multiplying it by the tangent of the light source pitch angle. The arctangent of the maximum absolute value of the first derivative of the complete contour function in the preset radial direction of the candidate bulge region is obtained as the edge slope of the bulge in the preset radial direction of the candidate bulge region. Starting from the complete contour function in the preset radial direction of the candidate bulge region, the length of the interval x in which the contour function value is greater than or equal to the product of a preset percentage and the maximum physical height of the bulge is found. The flatness of the top surface of the bulge in the preset radial direction of the candidate bulge region is obtained by dividing the length of the interval by the bottom radius of the candidate bulge region in the preset radial direction.

[0020] Preferably, the comprehensive hazard score of a candidate bulge area is calculated using the maximum physical height of the bulge in each preset radial direction, the slope of the bulge edge, and the flatness of the bulge top surface, including:

[0021] For a candidate bulge area, the median of the maximum physical height of the bulge, the median of the bulge edge slope, and the median of the bulge top surface flatness in each radial direction are taken as the final maximum physical height of the bulge, the final bulge edge slope, and the final bulge top surface flatness; the comprehensive hazard score is calculated using the final maximum physical height of the bulge, the final bulge edge slope, and the final bulge top surface flatness.

[0022] Preferably, the calculation model for the comprehensive hazard score is as follows:

[0023] ,

[0024] in, The overall hazard score for the j-th candidate bulge region; The final physical height of the bulge in the j-th candidate bulge region; Let be the final bulge edge slope of the j-th candidate bulge region; Let be the flatness of the top surface of the j-th candidate bulge region.

[0025] The embodiments of the present invention have at least the following beneficial effects: The present application acquires the original image of the fins and performs preprocessing to obtain independent regions of each layer of fins. Then, it analyzes the local grayscale changes of each pixel in the independent region of the target fin to obtain candidate bulge regions and their geometric centers. Then, it uses the line segment between the geometric center and the boundary of the candidate bulge region to project onto the independent region of the adjacent fin below the target fin to obtain each projection search line. Then, it searches along each projection search line to obtain the pixel sequence in each preset radial direction. Finally, it uses a cubic spline extrapolation method based on a partial contour function to extrapolate and obtain the complete contour function. From the complete contour function, it obtains the maximum height of the physical bulge, the slope of the bulge edge, and the flatness of the bulge top surface. This overcomes the problem that the bulge shadow is easily truncated by the interlayer boundary in the narrow interlayer channel of the stacked heat dissipation fins in traditional two-dimensional detection, solves the problem of the assessment disconnect between two-dimensional morphology and three-dimensional hazards, and the problem of underestimation of the bulge contour height caused by truncated shadows, thereby improving the accuracy of bulge heat dissipation fin quality detection. Attached Figure Description

[0026] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 The image of the heat sink fins is provided in an embodiment of the present invention for a method of detecting the quality of bulging heat sink fins based on image processing.

[0028] Figure 2 This invention provides a standard gauge block and its projected image for a method of detecting the quality of bulging heat sink fins based on image processing, as provided in an embodiment of the present invention. Detailed Implementation

[0029] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an image processing-based method for detecting the quality of bulging heat sink fins proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0031] The following description, in conjunction with the accompanying drawings, details a specific scheme for a method for detecting the quality of bulging heat sink fins based on image processing, provided by the present invention.

[0032] In this embodiment, the main application scenario of the present invention is to analyze the bulging of heat sink fins, thereby achieving the purpose of detecting the quality of heat sink fins.

[0033] Step S1: Acquire the original image of the fin and preprocess it to obtain the independent region of each fin layer, and record any fin layer as the target fin; analyze the local grayscale changes of each pixel in the independent region of the target fin to obtain the candidate bulge region and the geometric center of the candidate bulge region.

[0034] To establish the geometric mapping relationship between the bulge surface and the shadow of adjacent layers and perform inverse inversion, it is first necessary to acquire an image of the fin stack assembly supporting the bulge and accurately calibrate the core imaging parameters that determine the projection geometry. First, the heat dissipation fin stack assembly to be inspected is placed horizontally on the inspection station, ensuring that the relative angle between the fin plane and the camera imaging plane is fixed. In this embodiment, the camera optical axis is perpendicular to the fin surface (i.e., a vertical overhead shot) to ensure that the acquired two-dimensional image has no perspective distortion on the scale.

[0035] A strip LED light source is fixedly installed on one side of the camera, illuminating the fin surface at a fixed pitch angle θ. The pitch angle θ is defined as the angle between the direction of the light source and the normal to the fin surface. In this embodiment, θ is set to 45°. The value of θ directly affects the contrast and length of the projected shadow. If θ is too large (close to 90°), the projected length tends to be infinite, and the shadow may extend beyond the adjacent fins; if θ is too small (close to 0°), the projected length tends to be zero, and the shadow is almost invisible. This embodiment selects 45° to achieve the best balance between shadow visibility and measurement feasibility.

[0036] The specific process of light source calibration is as follows: Place a standard gauge block of known height (height) on the testing station. Precisely known, such as The image of the gauge block and its projection onto the plane below it is acquired under a fixed light source. The projection length is measured. (in pixels), combined with block height Calculate pixel equivalent Conversion relationship with actual physical dimensions:

[0037] ,

[0038] In the formula, Pixel equivalent, unit: ; The known physical height of the standard gauge block, in units of ; To calibrate the pixel length of the block projection in the image (referring to the continuous pixel length of the block shadow from the beginning to the end of the shadow along the direction of light projection, rather than the area of ​​the entire shadow region, starting from the bottom edge of the block along the direction of light projection (calibration direction), trace pixel by pixel along this direction to the end of the shadow, and record the number of pixels in this continuous distance to obtain the pixel length), that is, the number of pixels occupied by the block shadow, in pixels; Let be the pitch angle of the light source. This formula establishes a quantitative conversion relationship between the pixel distance in the image domain and the actual size of the physical domain by combining the ratio of the known physical height to the observed pixel length with the tangent of the pitch angle. Pixel equivalent. The precise calibration directly affects the absolute accuracy of all subsequent height and topography parameter inversions. If there is a 5% calibration error, the inversion results of all bulge heights will simultaneously produce a 5% systematic deviation. Therefore, the calibration process uses high-precision gauge blocks and takes multiple measurements to obtain the average value in order to control the calibration error within 1%.

[0039] Simultaneously, the projection direction vector of the light source's projection direction onto the image plane is determined from the calibration image. In this embodiment, the calibration result is along the vertical downward direction of the image, that is... .

[0040] Through the above calibration, the following fixed parameters are obtained: light source pitch angle. (nominal value) ), projection direction vector Pixel equivalent and the height of the gap between adjacent fins (Given directly from fin design parameters).

[0041] After acquiring the original image of the fins (the image of the fin stack assembly supporting the bulge, which is a grayscale image), the image needs to be preprocessed. Specifically, the original image is filtered to obtain the filtered image; the cumulative grayscale value of each row is calculated along the vertical direction of the filtered image, and a grayscale projection curve is drawn. The horizontal axis of the grayscale projection curve is the row number of each row in the filtered image (numbered sequentially from top to bottom), and the vertical axis is the cumulative grayscale value of all pixels in a row; the range between every two adjacent valley points in the grayscale projection curve is the area where a fin is located. The filtered image is segmented using the horizontal coordinates of each valley point in the grayscale projection curve as the segmentation boundary to obtain the independent regions of each fin layer.

[0042] In this embodiment, median filtering (with a 3×3 pixel filter window) is used to remove salt-and-pepper noise and sensor noise introduced during image acquisition. Compared with Gaussian filtering, median filtering can better preserve the details of the bulge edges. The median filter window supports adaptive adjustment. For thin and small fin texture scenes, a 5×3 rectangular filter window can be replaced to further suppress subtle periodic ripple noise. The fin layer segmentation is performed by grayscale projection. The air gap between adjacent fins presents obvious troughs on the projection curve. Based on this, the vertical position range of each fin layer in the filtered image is determined. The fin area contains a highly reflective metal surface, resulting in a higher grayscale accumulation value. The air gap area has no obvious reflectors, resulting in a significantly lower grayscale accumulation value, forming a trough. The horizontal coordinate (row number) of the trough point is used as the interlayer boundary, and the image of each fin layer can be cropped from the filtered image to obtain the independent region of each fin layer (i.e., the region in the image containing only a single fin layer). The aforementioned imaging hardware typically employs an industrial 2K / 4K monochrome global shutter camera paired with a strip-shaped infrared LED oblique light source, with a frame rate ≥30fps, further matching the continuous acquisition requirements of high-speed stamping and brazing production lines. Simultaneously, any layer of fins is designated as the target fin, and the analysis will focus on the target fin as an example.

[0043] Since the surface of the target fin is not perfectly flat, fine ripples formed by the stamping process, surface oxidation color differences, and genuine bulge defects will all produce grayscale changes in the image. Directly inspecting the entire area can easily misjudge normal textures as defects. Therefore, a method is needed to effectively distinguish genuine bulges from surface textures. A bulge, as a localized raised structure, appears in the image as a region with a significantly higher grayscale value than its local background, and this region is spatially connected. Anomalies are identified by calculating the grayscale deviation of each pixel relative to its neighboring background. Connectivity analysis is then used to aggregate these anomaly pixels into a complete bulge contour. Finally, area constraints are used to eliminate noise points and non-isolated deformations, focusing the detection results on candidate regions that match the bulge size range. Since shadow curves need to be extracted along multiple radial directions of the bulge contour, it is also necessary to further extract the contour boundary point set of the bulge to provide a geometric basis for determining multiple cross-sectional directions.

[0044] First, the local grayscale changes of each pixel within an independent region of the target fin are analyzed to obtain candidate bulge regions.

[0045] Specifically, a window is constructed centered on a pixel within an independent region of the target fin. The judgment threshold for that pixel is obtained by adding the mean of the gray values ​​of all pixels within the window to the sensitivity coefficient and the standard deviation of the gray values ​​of all pixels. If the gray value of the pixel is greater than the judgment threshold, the pixel is a candidate abnormal pixel. Connectivity analysis is performed on all candidate abnormal pixels within the independent region of the target fin to obtain different connected components. Connectivity components with a number of pixels within the area constraint are retained as candidate bulge regions.

[0046] In this embodiment, the window is a square window with a side length of 31 pixels, approximately 1.55 mm, covering the typical diameter range of the bulge. The judgment threshold is... , and Here, α represents the mean and standard deviation of the grayscale values ​​of the pixels within the window corresponding to the w-th pixel, respectively. α is the sensitivity coefficient, which is set to 2.5 in this embodiment. This criterion means that when the grayscale value of a pixel is significantly higher than its local background (exceeding 2.5 times the standard deviation), it is considered that there may be light convergence or enhanced reflection at that location due to a bulge. The value of α is determined as follows: if α is too small, it will misjudge a large number of normal surface texture undulations as abnormal; if α is too large, it will miss weak bulges with low grayscale contrast. In this embodiment, 2.5 is used to achieve a balance between the detection rate and the false detection rate. The sensitivity coefficient α=2.5 is the initial fixed threshold. The system is equipped with an adaptive calibration process. After collecting a batch of standard qualified and defective fin samples for training, this threshold can be automatically optimized to adapt to fin products of different materials and stamping textures.

[0047] After obtaining different connected components, the number of pixels covered by each connected component is used as its area representation, retaining those that satisfy the area constraint. The connected components are used as candidate bulge regions. The area constraint range is determined by combining the physical size range of the bulge with the pixel equivalent. Based on the conversion, the diameter of the bulge in this embodiment ranges from 0.3mm to 2.0mm, corresponding to a pixel diameter range of 6 to 40 pixels and an area range of 28 to 1256 pixels. Therefore, we take... For 20 pixels, The resolution is 1500 pixels. Connected regions that are too small are noise points or dust particles, while connected regions that are too large are wavy overall deformations rather than isolated bumps, and are therefore excluded.

[0048] For each candidate bulge region selected by area... ( , (Total number of candidate drums), record their set of pixel coordinates. And calculate the geometric center. , and The first The x-coordinates and y-coordinates of the geometric centers of the candidate bulge regions are determined. The calculation of the geometric center coordinates is based on existing technology and will not be elaborated upon here. The geometric center serves as the starting reference point for subsequent searching of the projected shadows of adjacent layers along the light source projection direction. Since the highest point of the bulge is usually located near the geometric center, using the geometric center as the starting point maximizes the probability of covering the starting end of the shadow. Simultaneously, the contour boundaries of the candidate bulge regions are extracted, thus obtaining the contour boundary point set of the candidate bulge regions. The points in the point set are the boundary points of the candidate bulge region, which are used to determine the direction of the subsequent multi-section.

[0049] Step S2: Project the line segment between the geometric center of a candidate bulge region and the intersection of each preset radial direction with the boundary of the candidate bulge region onto the independent region of the adjacent fin below the target fin to obtain each projection search line; search along each projection search line to obtain the pixel sequence of each preset radial direction.

[0050] In a heat sink fin stack structure, a bulge on one layer of fins will cast a shadow on the surface of the adjacent lower fin under fixed-direction illumination. Unlike free-space projection, the shadow's supporting surface is the surface of the adjacent lower fin. This surface does not extend infinitely in the horizontal direction. The fin's own fin width (typically only 20-40 mm), the periodic embossed ripples on the surface (with a peak-to-trough height difference of 0.1-0.3 mm), and the edge shading between multiple layers of stacked fins collectively constitute a "restricted supporting surface." When the bulge is large or located near the fin edge, the bulge's projection onto the adjacent layer surface may exceed the effective supporting range, causing the shadow to be truncated by the fin boundary. When there are periodic ripples on the lower fin surface, the shadow's grayscale mixes with the bright and dark stripes of the ripples themselves, preventing the shadow's end from naturally returning to a uniform background grayscale. If any of these situations occur, directly measuring the total length of the shadow segment and calculating the bulge height using traditional methods will inevitably lead to a systematic underestimation. Therefore, it is necessary to start from the premise that "the shadow may be incomplete" and design a method to recover the complete bulge outline from incomplete shadow observations. First, starting from the beginning of the shadow (corresponding to the highest point of the bulge), analyze the gray-level changes pixel by pixel along the projection direction to identify the effective projection segment that is not affected by truncation and aliasing. The basis for this judgment is that the gray-level changes monotonically with the projection position and the rate of change is continuous, which is the essential characteristic of the continuous curved surface projection of the bulge. Then, based on the physical continuity of the bulge cross-sectional outline (i.e., the bulge surface is continuous and uninterrupted in space), the gray-level-projection position mapping relationship within the effective segment is used to recursively extend backward to recover the outline of the truncated part until the outline function completely covers the bottom width of the bulge. This fundamentally solves the problem of height underestimation caused by truncated shadows.

[0051] After obtaining the geometric center and contour boundary of each candidate bulge region, the projected shadows of these regions on adjacent layers are often not complete free projections. The shadowing surface is the lower fin surface, which is constrained by factors such as fin width, periodic ripples, and edge occlusion. Instead, these shadows are often truncated, cut, or superimposed with structural shadows – “incomplete shadows.” Traditional shadow height measurement methods directly measure the total shadow length and calculate the height accordingly, systematically underestimating the true height of the bulge in this scenario. Therefore, to address this problem, considering that the bulge is a continuous curved surface formed by stamping, its cross-sectional profile function spatially satisfies physical continuity. That is, the cross-sectional profile function and its first derivative are continuous throughout the entire domain, and at the bulge edge, both the cross-sectional profile function and its derivative smoothly approach zero. This continuity is an inherent property of the stamping process and does not change due to whether the shadow is truncated. Therefore, even if the shadow observations are incomplete, the complete cross-sectional profile can still be recursively recovered through the gray-projection mapping relationship within the effective segment.

[0052] Furthermore, different radial directions are set to... Eight radial directions are selected evenly within its contour plane. ( ), For the k-th radial direction, for each candidate bulge region, starting from its geometric center along a radial direction until the boundary of the candidate bulge region, obtain the line segment between the intersection of the geometric center along the preset radial direction and the boundary of the candidate bulge region. This line segment is denoted as the radial cross-section line. Then, for each radial direction of a candidate bulge region, the corresponding radial cross-section line can be obtained. Project the radial cross-section lines corresponding to each radial direction of a candidate bulge region of the target fin onto the independent region of the adjacent fin below the target fin to obtain each projection search line. This indicates that the starting point of the projection search line is the projection of the geometric center onto the independent region of the adjacent fin below the target fin; before performing the pixel-by-pixel grayscale search, a binary mask of the effective surface of the lower fin is pre-generated. Based on the grayscale projection segmentation results between fin layers obtained above, the left and right lateral boundaries of the fins are extracted. Then, the lower layer image is segmented by brightness and darkness periodicity thresholding to distinguish periodic ripple interference areas and automatically output a binary mask. There is no need to import manually drawn fin drawings throughout the process.

[0053] For a projection search line, starting from the starting point of the projection search line, the pixels are searched sequentially along the projection search line. When the gray value of the newly searched pixel is less than the gray value of the previously searched pixels, or when the absolute value of the difference between the first-order difference of the newly searched pixel and the first-order difference of the previous search is less than a preset continuity threshold, the search stops and the previously searched pixels are combined to form a pixel sequence in the preset radial direction corresponding to the projection search line.

[0054] During the search, the shadow's starting point (the beginning of the pixel sequence) corresponds to the projection of the highest point of the bulge, where the grayscale value is lowest (light is completely blocked). As the search progresses, the height of the bulge corresponding to the projection position decreases, the degree of occlusion weakens, and the grayscale value monotonically increases. For a complete shadow, the grayscale value eventually recovers to the background baseline value and remains stable. For truncated shadows, the grayscale value may be cut off by the fin boundary before reaching the baseline (t reaches the fin edge), or it may overlap with the wavy shadow, causing abnormal fluctuations in grayscale.

[0055] Two constraints are set here. The first ensures that the grayscale values ​​of pixels in the pixel sequence are monotonically non-decreasing. This is an essential characteristic of the continuous curved surface projection of the bulge. Along the projection direction, away from the highest point of the bulge, the occlusion gradually weakens, and the transmitted light intensity monotonically increases. That is, for any pixel in the pixel sequence... ,have , and Indicates the order in which the searches are performed. For the previous search, For the subsequent search, The grayscale values ​​of the pixels that have already been searched. This refers to the grayscale value of the pixel being searched later or in the latest search.

[0056] Another constraint ensures the continuity of pixel grayscale. To preset the continuity threshold, this embodiment takes... Grayscale / pixel², where this condition ensures that grayscale changes are smooth, eliminating abrupt changes caused by wavy edge cutting or structural shadow overlay. The value is based on the gray-level change rate corresponding to the maximum curvature of the fin surface ripples. The gray-level change rate generated by the ripple peak-valley transition area in the image usually does not exceed 2 gray levels / pixel². Taking 3 gray levels / pixel² can eliminate ripple interference without over-filtering the real defect signal; the first-order difference is the first-order difference. ,when Stop the search, and form a pixel sequence from the pixels up to the (t+1)th search, where t+1 and t are the (t+1)th and (t)th searches, respectively. and Let be the grayscale values ​​of the pixels in the (t+1)th search and the tth search; =3 grayscale / pixel² is the initial fixed parameter. The system supports an adaptive calibration process. It automatically updates the threshold by collecting fin samples with different ripple cycles to adapt to various stamped fins.

[0057] Each preset radial direction sequence of pixels is considered a reliable observation that is not subject to truncation or aliasing interference.

[0058] Step S3: Using the partial contour function corresponding to the pixel sequence in a preset radial direction of a candidate bulge region, extrapolate using the cubic spline extrapolation method to obtain the complete contour function of the candidate bulge region in that preset radial direction; use the complete contour function to obtain the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the bulge top surface in each preset radial direction of each candidate bulge region; calculate the comprehensive hazard score of the candidate bulge region using the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the bulge top surface in each preset radial direction of a candidate bulge region; use the comprehensive hazard score to detect the quality of the target fin.

[0059] After obtaining the effective projection segment (pixel sequence), the truncated shadow recovery stage begins. This involves recursively recovering the complete contour based on the physical continuity of the bulge cross-sectional profile and utilizing the gray-scale-projection mapping relationship within the effective segment. First, a forward geometric mapping model for inter-layer constrained projection shadows is established. Consider the boundary line between the candidate bulge region's cross-section along the current radial direction and the target fin, denoted as... The axis, with the geometric center of the candidate region of the bulge as the origin, The axis is perpendicular to the fin plane. The cross-sectional profile of a candidate bulge region is as follows: , For partial contour functions, This represents the maximum height of the candidate bulge area. Light source pitch angle. Below, surface points of the candidate bulge region The projection position within the independent area of ​​the adjacent fin below the target fin is:

[0060] ,

[0061] That is, the surface height of the candidate bulge area is The projection offset of the point, in the independent region of the adjacent fin below the target fin, is... The offset direction is along the projection direction of the light source. The larger the value, the greater the projection offset; The smaller the value (the more oblique the light), the greater the offset at the same height. When analyzing independent regions of adjacent fins below the target fin, the coordinate system used is the same as the coordinate system corresponding to each preset radial direction of each candidate bulge region in the target fin. These coordinate systems all have the geometric center as the origin, and the intersection of the cross-section of the candidate bulge region along the current radial direction and the target fin is the x-axis. The axis is perpendicular to the fin plane. The intersection points obtained by progressively translating the bulge surface corresponding to the candidate bulge region upwards along the x-axis are projected onto the pixel sequence. After projection, all pixels are projected onto the line segment obtained by translating the x-axis to an independent region of the adjacent fin below the target fin. Thus, the coordinate system corresponding to a radial direction of a candidate bulge region of the target fin is translated to an independent region of the adjacent fin below the target fin. Let x be the abscissa of a pixel in the pixel sequence in this coordinate system, denoted as the projection position of the pixel. Let x be the abscissa of the pixel in the pixel sequence in the coordinate system corresponding to the radial direction of the candidate bulge region when the pixel is not projected. That is, it is the abscissa of the point on the contour surface of the candidate bulge region corresponding to the pixel in the pixel sequence, denoted as the position of the point on the contour surface of the candidate bulge region corresponding to the pixel.

[0062] After establishing the geometric correspondence between the points on the bulge surface and the projection positions, it is necessary to further clarify from a physical perspective that the grayscale value observed at the projection position is not independently determined by a single isolated height point on the bulge, but rather is the result of the combined effect of all positions on the bulge surface that satisfy the same projection geometric constraint. Light rays projected from different height points on the bulge overlap on adjacent layer surfaces, and the occlusion contribution of each height point accumulates and superimposes according to the projection geometry, ultimately forming a continuous grayscale distribution. This relationship establishes a positive quantitative mapping between the bulge cross-sectional profile function and the shadow grayscale distribution, indicating that the 3D morphological information of the bulge is analyzably encoded in the shadow grayscale, thus providing a theoretical basis for subsequently reconstructing the profile using grayscale observations within the effective projection segment. Based on this mapping, given that the grayscale within the effective projection segment monotonically changes with position, the cross-sectional profile can be reconstructed pixel-by-pixel along the projection direction, starting from the observations.

[0063] Starting from the observations within the effective projection segment (a sequence of pixels corresponding to a radial direction), the contour function is recursively recovered through the inverse process of forward mapping. In the local projection geometric coordinate system, let x=0 correspond to the center of the bulge. The position of this point projected onto the surface of the adjacent layer is denoted as the reference point. Moving pixel by pixel from this reference point along the light source projection direction, the grayscale gradually increases from the lowest value at the starting point until it returns to the background grayscale value. The length of this continuous grayscale variation region is the effective shadow length L, which is also the pixel length corresponding to the pixel sequence. Therefore:

[0064] ,

[0065] In the formula, H is the maximum physical height of the bulge; L is the effective shadow length (converted to physical length after multiplying by the pixel equivalent). Since the shadow start end is always observable, the height inversion is not affected by shadow truncation.

[0066] For any position in the pixel sequence ( ), its corresponding height (Partial contour functions) are recursively recovered by the following formula:

[0067] ,

[0068] In the formula, This refers to the projection position of a pixel in the pixel sequence corresponding to a radial direction of a candidate bulge region of the target fin. The position of the point on the contour surface of the candidate bulge region corresponding to the pixel in the pixel sequence, and... The relationship is a positive mapping Sure; Let be the pitch angle of the light source. The physical meaning of this recursive recovery process is: to reconstruct the observed shadow grayscale distribution along... The direction advances pixel by pixel, with each step based on the currently recovered data. Reverse the corresponding bulge surface position (Given) Given the previous neighborhood, the current x is solved iteratively using fixed-point iteration (which converges quickly), thus gradually reconstructing the complete cross-sectional profile from the shadow observations.

[0069] When the effective projection segment is truncated (That is, the effective segment length is less than the complete projection length). At this point, the physical boundary conditions of the edges of the corresponding bulges in the candidate bulge region are used. and ,in Given the bottom radius of the bulge corresponding to the candidate bulge region, extrapolate the partially recovered contour function within the effective segment. The specific method is as follows:

[0070] Based on projection geometric constraints The observed coordinates of the end point of the effective shadow segment Mapping to the coordinates of the bulge section surface, the corresponding surface position can be obtained by solving. The position satisfies:

[0071] Due to the projection offset effect of the oblique light source, the object surface coordinates The value is always less than the shadow observation coordinates This is a normal geometric mapping result. Using the grayscale observations within the effective projection segment, the partial contour function h(x) of the candidate bulge region has been directly recovered within the interval... The reliable value within. However, because the shadow is truncated by the interlayer boundary, the effective projection segment endpoint... Corresponding bulge surface location Smaller than the actual bottom radius R of the bulge (i.e. ),therefore The contours within the interval cannot be directly recovered from the shadowed observations. Therefore, using... Using the physical continuity of the bulge surface as the starting point for extrapolation, cubic spline extrapolation is employed to complete the contour within this interval along the increasing x-direction. The extrapolation constraints are: maintaining C1 continuity with the existing contour at the starting point, and satisfying the edge boundary conditions at the ending point x=R. and The initial estimate of the bottom radius R of the candidate bulge region is determined by the two-dimensional contour boundary of the candidate bulge region extracted in the above steps. If the initial estimate is unreliable due to shadow truncation of the two-dimensional contour, an iterative correction strategy is adopted: the current R is used to complete the contour extrapolation, and the consistency coefficient corresponding to the multi-radial section reconstruction results is calculated. ;like If the value falls below a preset threshold, R is fine-tuned and the contour extrapolation is re-executed, iterating until the consistency coefficient is reached. It has reached its maximum value.

[0072] Through the above extrapolation, the complete contour function of each candidate bulge region in the radial direction is obtained. , indicating the first The candidate bulge area in the first A complete contour function in the radial direction, whose domain covers ,in For the first The candidate bulge area in the first The bottom radius in the radial direction. This complete profile function fully reflects the cross-sectional shape of the bulge in this direction, even if the shadowed observations are incomplete.

[0073] Furthermore, the complete contour function is used to obtain the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the top surface of the bulge in each preset radial direction of each candidate bulge region.

[0074] Specifically, the maximum physical height of the bulge in the preset radial direction of a candidate bulge region is obtained by converting the pixel length of the pixel sequence in a preset radial direction of the candidate bulge region into a physical length multiplied by the tangent of the light source pitch angle; the arctangent of the maximum absolute value of the first derivative of the complete contour function in the preset radial direction of the candidate bulge region is obtained as the edge slope of the bulge in the preset radial direction of the candidate bulge region; starting from the complete contour function in the preset radial direction of the candidate bulge region, the length of the interval x in which the contour function value is greater than or equal to the product of a preset percentage and the maximum physical height of the bulge is found, and the interval length is divided by the bottom radius of the candidate bulge region in the preset radial direction to obtain the flatness of the top surface of the bulge in the preset radial direction of the candidate bulge region.

[0075] Slope of the bulge edge The edge slope is defined as the maximum slope angle of the outline of the edge region of the bulge:

[0076] ,

[0077] in, For the first The candidate bulge area in the first Edge slope in the radial direction; The first derivative of the complete contour function. This represents the maximum absolute value of the derivative of the contour function. Since the complete contour function fully covers the edge region through extrapolation, the extraction of the edge slope is unaffected by shadow truncation. The edge slope reflects the steepness of the edge surface corresponding to the candidate bulge region. The closer The greater the likelihood of flow separation as airflow passes through, the higher the local drag coefficient.

[0078] Flatness of the top surface of the bulge Top surface flatness is defined as the relative width of the flat area at the top of the bulge. This is derived from the recovered complete contour function. Starting from, find a contour function value greater than or equal to of Interval length , For the first The candidate bulge area in the first Maximum height of the bulge in each radial direction, and flatness of the bulge top surface:

[0079] ,

[0080] in, For the first The first bulge was on the... Top surface flatness in each radial direction; The length of the interval where the height of the top of the bulge is not less than 95% of the maximum height; For the first The first bulge was on the... The radius of the bottom of the bulge in the radial direction, passing through the geometric center along the radial direction. Here, the flatness definition quantifies the morphology of the top surface of the bulge as a continuous value between 0 and 1. The closer to 1, the flatter the top and the wider the extent of laminar boundary layer disruption. The closer to 0, the sharper the top, the more concentrated the airflow turning point, but the higher the local drag coefficient. The 95% threshold is selected based on the fact that the area above this height can be physically identified as a "flat-topped area," which can effectively identify the typical flat features of a flat-topped bulge, and will not cause drastic fluctuations in the length of the interval due to noise.

[0081] Therefore, for a candidate bulge region, the maximum physical height of the bulge in each radial direction, the slope of the bulge edge, and the flatness of the bulge top surface constitute the following sets: , , The median of the sets corresponding to the three parameters is taken as the final physical maximum height of the bulge in the candidate bulge region. The final slope of the bulge edge and the final flatness of the top surface of the bulge The reason for choosing the median instead of the mean is that the median is naturally robust to outliers. When the recovery results deviate significantly in some of the eight directions due to local disturbances, the median can effectively suppress the impact of the outlier on the final estimate, making the morphological parameter estimation more robust.

[0082] This can then be determined based on the maximum physical height of the final bulge. The final slope of the bulge edge and the final flatness of the top surface of the bulge Calculate the overall hazard score. Before calculating the overall hazard score, it is necessary to verify the reliability and effectiveness of the above-mentioned quantitative assessment three-dimensional topography inversion results.

[0083] Specifically, the following is a set of recovery results for each bulge in 8 directions, obtained from the above. , , Then, the coefficients of variation (the ratio of standard deviation to median) for the three parameters were calculated respectively:

[0084] ,

[0085] in, , , The first The coefficients of variation for the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the bulge top surface in each candidate bulge region; "std" is the standard deviation operator. (If Hj < 0.15mm, it is directly marked as unmeasurable and not included in the CV calculation to avoid numerical anomalies where the denominator approaches 0.) , These are the median estimates of the sets corresponding to the three parameters. The coefficient of variation eliminates the differences in the dimensions and numerical ranges of the different parameters, enabling the comparison and fusion of the dispersion of the three physical quantities with huge spans—height (on the order of 1 mm), slope (on the order of 60°), and flatness (on the order of 0.5)—on a unified scale.

[0086] Define the consistency coefficient of the candidate bulge region. :

[0087] ,

[0088] in, For the first The consistency coefficient of each bulge, with a range of values. ; , , The preset weighting coefficients are set to 3.0, 2.0, and 2.0 respectively. The weights can be customized between 1 and 5 based on the fin model used in the production line. The weighted sum of the three coefficients of variation is mapped to an exponential function. Interval – When the recovery results in all eight directions are highly consistent, the coefficients of variation approach 0. It approaches 1; as the dispersion increases, Monotonically decreasing. Weighting coefficients. The basis for taking the maximum value (3.0) is that height is the most critical indicator in the assessment of bulge hazards, and its measurement dispersion has the greatest impact on the final classification decision. Therefore, it is given the highest weight in the consistency assessment.

[0089] when (This embodiment) The value can be set according to the actual situation; the larger the value, the higher the strictness of the judgment, and vice versa. When the value is set, the detection result of the bulge is considered reliable, and the estimated value of the three-dimensional morphology parameters is retained for subsequent grading. If the bulge is found to have serious asymmetry or detection interference, it is marked as "insufficient consistency of three-dimensional morphology inversion, requiring manual verification".

[0090] Furthermore, the three-dimensional morphological parameters of each bulge are obtained from the above, ultimately determining the maximum physical height of the bulge. The final slope of the bulge edge and the final flatness of the top surface of the bulge and consistency coefficient Subsequently, these physical parameters need to be further transformed into engineering-executable decision-making. Unlike existing methods that rely solely on a single indicator of two-dimensional area, this invention obtains three independent morphological parameters, each describing the impact of the bulge on heat dissipation performance from different dimensions. The bulge height directly affects the blockage ratio of the airflow channel and is the most critical factor; the edge slope affects the degree of flow separation when airflow passes through the bulge, with steep edges leading to greater local resistance; the flatness of the top surface affects the extent of laminar boundary layer disruption, with a flatter top surface causing a wider range of airflow disturbance. These three factors are weighted and combined into a hazard score, and an A / B / C three-level classification is performed based on preset grading thresholds. At the same time, bulges with insufficient consistency or unmeasurable data are clearly marked as "uncertain," ensuring that only reliable data enters the automatic decision-making process.

[0091] Among them, bulges marked as "unmeasurable three-dimensional morphology" or "insufficient consistency of three-dimensional morphology inversion" are classified as "uncertain" and marked as requiring manual review, and are not included in the automatic grading and judgment process.

[0092] For bulges with reliable detection results, a grading assessment is performed based on comprehensive three-dimensional morphological parameters. According to heat sink fin engineering design specifications and wind tunnel experimental data, this embodiment will use a comprehensive hazard score. Defined as:

[0093] ,

[0094] in, The overall hazard score for the i-th candidate bulge region; The final physical maximum height of the bulge, in units of ; The final slope of the bulge edge, in degrees (°). The flatness of the final bulge top surface is dimensionless. The physical meaning of this formula is: As a basic item, it directly reflects the compression of the channel cross-sectional area; This is a penalty term for edge steepness; the steeper the edge ( The closer The greater the likelihood of flow separation when airflow passes over the bulge, the higher the local drag coefficient. The penalty weight of 0.3 is derived from the experimental correlation of the flow resistance around a local obstacle in a straight channel (the steeper the edge of an obstacle (here, the bulge)...). The closer to 90°, the easier it is for "flow separation" to occur behind the obstacle when airflow passes over it, meaning the airflow detaches from the wall and forms a vortex region. This vortex region generates additional drag, which is approximately equivalent to a portion of the base drag (drag caused by the height itself), so a weight of 0.3 is used (this scheme uses 0.3, but can be adjusted based on actual experiments). The flatness of the top surface is a penalty term. The flatter the top surface, the wider the laminar boundary layer is disrupted and the larger the turbulent region. The penalty weight of 0.2 is also derived from the experimental correlation.

[0095] After obtaining the overall hazard score for each bulge, it can be classified according to actual needs, such as: Acceptable (Grade A): ,or and The impact of bulges within this range on airflow is negligible; re-inspection is required (Grade B). Manual verification is required; direct scrapping (Level C): The implementers may adjust the classification based on the actual situation or determine it manually.

[0096] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0097] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for quality inspection of bulging heat sink fins based on image processing, characterized in that, The method includes: The original image of the fin is acquired and preprocessed to obtain the independent region of each fin layer, and any fin layer is recorded as the target fin; the local grayscale changes of each pixel in the independent region of the target fin are analyzed to obtain the candidate bulge region and the geometric center of the candidate bulge region. Project the line segments between the geometric center of a candidate bulge region and the intersection of each preset radial direction with the boundary of the candidate bulge region onto the independent region of the adjacent fin below the target fin to obtain each projection search line; search along each projection search line to obtain the pixel sequence in each preset radial direction. Using a partial contour function corresponding to a pixel sequence in a preset radial direction of a candidate bulge region, a cubic spline extrapolation method is used to obtain the complete contour function of the candidate bulge region in that preset radial direction. The complete contour function is then used to obtain the maximum physical height of the bulge, the bulge edge slope, and the bulge top surface flatness in each preset radial direction of each candidate bulge region. The comprehensive hazard score of a candidate bulge region is calculated using the maximum physical height of the bulge, the bulge edge slope, and the bulge top surface flatness in each preset radial direction. The quality of the target fin is then inspected using the comprehensive hazard score.

2. The method for detecting the quality of bulging heat sink fins based on image processing according to claim 1, characterized in that, The process of acquiring the original image of the fins and preprocessing it to obtain independent regions for each fin layer includes: The original image is filtered to obtain the filtered image. The cumulative gray value of each row is calculated along the vertical direction of the filtered image, and a gray-scale projection curve is drawn. The horizontal axis of the gray-scale projection curve is the row number of each row in the filtered image, and the vertical axis is the cumulative gray value of all pixels in the row. The range between every two adjacent valley points in the gray-scale projection curve is the area where a fin is located. The filtered image is segmented using the horizontal coordinates of each valley point in the gray-scale projection curve as the segmentation boundary to obtain the independent regions of each fin layer.

3. The method for quality inspection of bulging heat sink fins based on image processing according to claim 1, characterized in that, The analysis of local grayscale changes of each pixel within an independent region of the target fin to obtain candidate bulge regions includes: A window is constructed centered on a pixel within an independent region of the target fin. The judgment threshold for that pixel is obtained by adding the mean of the gray values ​​of all pixels within the window to the sensitivity coefficient and the standard deviation of the gray values ​​of all pixels. If the gray value of the pixel is greater than the judgment threshold, the pixel is a candidate abnormal pixel. Connectivity analysis is performed on all candidate abnormal pixels within the independent region of the target fin to obtain different connected components. Connectivity components with a number of pixels within the area constraint are retained as candidate bulge regions.

4. The method for detecting the quality of bulging heat sink fins based on image processing according to claim 1, characterized in that, The step of searching along each projection search line to obtain the pixel sequence in each preset radial direction includes: Starting from the beginning of a projection search line, pixels are searched sequentially along the projection search line. When the gray value of the newly searched pixel is less than the gray value of the previously searched pixels, or when the absolute value of the difference between the first-order difference of the newly searched pixel and the first-order difference of the previous search is less than a preset continuity threshold, the search stops and the previously searched pixels are combined to form a pixel sequence in the preset radial direction corresponding to the projection search line.

5. The method for quality detection of bulging heat sink fins based on image processing according to claim 1, characterized in that, The partial contour function is: ， in, For partial contour functions, This refers to the projection position of a pixel in the pixel sequence corresponding to a radial direction of a candidate bulge region of the target fin. This refers to the position of a point on the contour surface of the candidate bulge region corresponding to that pixel in the pixel sequence. The pitch angle of the light source.

6. The method for detecting the quality of bulging heat sink fins based on image processing according to claim 1, characterized in that, The process of obtaining the maximum physical height of the bulge, the slope of the bulge edge, and the flatness of the bulge top surface in each preset radial direction of each candidate bulge region using the complete contour function includes: The maximum physical height of the bulge in the preset radial direction of a candidate bulge region is obtained by converting the pixel length of the pixel sequence in a preset radial direction of the candidate bulge region into a physical length and multiplying it by the tangent of the light source pitch angle. The arctangent of the maximum absolute value of the first derivative of the complete contour function in the preset radial direction of the candidate bulge region is obtained as the edge slope of the bulge in the preset radial direction of the candidate bulge region. Starting from the complete contour function in the preset radial direction of the candidate bulge region, the length of the interval x in which the contour function value is greater than or equal to the product of a preset percentage and the maximum physical height of the bulge is found. The flatness of the top surface of the bulge in the preset radial direction of the candidate bulge region is obtained by dividing the length of the interval by the bottom radius of the candidate bulge region in the preset radial direction.

7. The method for quality inspection of bulging heat sink fins based on image processing according to claim 1, characterized in that, The comprehensive hazard score of a candidate bulge region is calculated using the maximum physical height of the bulge in each preset radial direction, the slope of the bulge edge, and the flatness of the bulge top surface. This includes: For a candidate bulge area, the median of the maximum physical height of the bulge, the median of the bulge edge slope, and the median of the bulge top surface flatness in each radial direction are taken as the final maximum physical height of the bulge, the final bulge edge slope, and the final bulge top surface flatness; the comprehensive hazard score is calculated using the final maximum physical height of the bulge, the final bulge edge slope, and the final bulge top surface flatness.

8. The method for detecting the quality of bulging heat sink fins based on image processing according to claim 7, characterized in that, The calculation model for the comprehensive hazard score is as follows: ， in, The overall hazard score for the j-th candidate bulge region; The final physical height of the bulge in the j-th candidate bulge region; Let be the final bulge edge slope of the j-th candidate bulge region; Let be the flatness of the top surface of the j-th candidate bulge region.

Images