Surface defect detection method and system for liquid cooling plate

Abstract

The invention relates to the technical field of image processing, in particular to a surface defect detection method and system for a liquid cooling plate, and the method comprises the steps: collecting a surface image of the liquid cooling plate, and dividing the surface image into a reflective region and a non-reflective region; for the non-reflective area, taking the mean value of the gray defect degree and the texture defect degree of each pixel point as the comprehensive defect degree of each pixel point, and taking the mean value of the comprehensive defect degrees of all the pixel points as the defect degree of the non-reflective area; aiming at the reflective area, extracting all scraped areas, and taking the ratio of the total area of all scraped areas to the total area of the reflective area as the defect degree of the reflective area; and determining the total defect degree of the surface of the liquid cooling plate by integrating the defect degree of the non-reflective area and the defect degree of the reflective area, and performing defect detection according to the total defect degree. According to the detection method, the interference caused by illumination influence and texture of the surface of the liquid cooling plate is overcome, and the accuracy of defect detection is improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to a surface defect detection method and system for a liquid cooling plate. Background Art

[0002] As a key component of liquid cooling radiator, liquid cooling plate plays a vital role in the battery thermal management system of new energy vehicles. The manufacturing of liquid cooling plate requires multiple processes such as cutting, welding, grinding, and polishing. If the process control is not proper, scratches are easy to appear on the surface. Surface scratches will damage the protective layer and increase the risk of coolant leakage. Coolant leakage not only reduces the heat dissipation efficiency, but also may damage electronic equipment. Therefore, liquid cooling plate must be strictly tested before leaving the factory.

[0003] At present, when detecting surface defects of liquid cooling plates by traditional image processing methods, there are the following deficiencies: On the one hand, the flow channel on the surface of the liquid cooling plate is raised, which will produce reflection under lighting conditions. This reflection will interfere with the image acquisition process, making the brightness of some areas in the acquired image abnormally high. In these overly bright areas, if there are scratch defects, the details of the scratches will be covered up due to the excessive brightness, affecting the accuracy of defect detection. On the other hand, the surface of the liquid cooling plate itself has certain textures, which will interfere with the detection process. The surface texture and scratch defects of the liquid cooling plate are often presented in the form of lines or stripes, which are easy to confuse with each other, making it difficult to distinguish between normal textures and scratch defects, resulting in inaccurate surface defect detection results.

[0004] In summary, when detecting surface defects of liquid cooling plates by traditional image processing methods, the detection results of scratch defects are inaccurate due to the influence of light and the texture interference of the surface of the liquid cooling plate itself, which cannot meet the quality inspection requirements of liquid cooling plate manufacturing. Summary of the invention

[0005] In order to solve the problem that when detecting surface defects of liquid cooling plates by traditional image processing methods, the detection results of scratch defects are inaccurate due to the influence of light and the texture interference of the surface of the liquid cooling plate itself, the present invention proposes a surface defect detection method and system for liquid cooling plates.

[0006] In one aspect, the present invention provides a method for detecting surface defects of a liquid cooling plate, comprising: Collect the surface image of the liquid cooling plate for preprocessing and divide it into reflective area and non-reflective area; For each non-reflective area: The grayscale defect degree of each pixel is determined based on the local grayscale fluctuation degree of each pixel and the contrast of each non-reflective area; the texture defect degree of each pixel is determined based on the gradient consistency of each pixel and the edge regularity of each non-reflective area; the edge regularity of each non-reflective area is the normalized mean of the curvature change values ​​of all edge pixels of each non-reflective area; The average of the grayscale defect degree and the texture defect degree of each pixel is taken as the comprehensive defect degree of each pixel, and the average of the comprehensive defect degrees of all pixels is taken as the defect degree of each non-reflective area; For each reflective area: extract all scratched areas in each reflective area, and take the ratio of the total area of ​​all scratched areas to the total area of ​​each reflective area as the defect degree of each reflective area; The total defect degree of the liquid cooling plate is determined by comprehensively considering the defect degree of each non-reflective area and the defect degree of each reflective area, and the defect detection result of the surface of the liquid cooling plate is determined according to the total defect degree of the liquid cooling plate.

[0007] The above technical solution can effectively remove the noise introduced by factors such as the accuracy limitation of the image acquisition equipment and the instability of the ambient light by preprocessing the surface image of the liquid cooling plate, while highlighting the detailed features in the image, providing a more accurate data basis for subsequent analysis. The reflective area and the non-reflective area are divided because these two types of areas are affected by light and their own characteristics, showing completely different characteristics. Separate treatment can avoid mutual interference and significantly improve the pertinence of detection. And further, for the non-reflective area, from the perspective of grayscale, the scratch defect will break the normal grayscale distribution law, causing local grayscale fluctuations, and then change the overall contrast of the non-reflective area; from the perspective of texture, the gradient consistency of the scratch defect and the edge irregularity of the non-reflective area are significantly different from the normal texture. Considering these two perspectives comprehensively and meticulously, the characteristics of each pixel can be characterized, which makes up for the shortcomings of analyzing only from a single grayscale or texture feature, making the evaluation of the degree of defects in the non-reflective area more accurate and comprehensive. Furthermore, for the reflective area, given that the grayscale value of the scratched area changes significantly, the degree of defect is determined by calculating the area ratio, which can intuitively reflect the severity of the defect in the reflective area. Furthermore, the degree of defect in the non-reflective area and the reflective area is combined to determine the total degree of defect of the liquid cooling plate, thereby improving the accuracy of defect detection on the surface of the liquid cooling plate.

[0008] Furthermore, the grayscale defect degree of each pixel satisfies the following formula: ; In the formula, For the The non-reflective area The grayscale defect degree of each pixel, is the normalization function, and They are and The weight of For the The contrast of the non-reflective area For the The non-reflective area The local grayscale fluctuation degree of each pixel.

[0009] The above-mentioned technical scheme can more comprehensively reflect the difference between the pixel points and the normal state from the grayscale perspective by analyzing the severity of the overall grayscale changes in the non-reflective area and the fluctuation of the grayscale values ​​in the local area around the pixel points in the non-reflective area, and judge whether there are grayscale defects, thereby improving the accuracy of judging the grayscale defects of the pixel points.

[0010] Furthermore, the texture defect degree of each pixel satisfies the following formula: ; In the formula, For the The non-reflective area The degree of texture defect of each pixel, is the normalization function, and are weights, It is The non-reflective area The gradient value of a pixel, To obtain the maximum value function, It is The absolute value of the maximum gradient value corresponding to all pixels in the non-reflective area, It is The edge regularity of the non-reflective area, is the absolute value symbol.

[0011] The above technical solution comprehensively considers two key texture features, namely gradient consistency and edge regularity. Compared with judging based on a single texture feature alone, it can more accurately identify scratch defect pixels in non-reflective areas. By quantifying the impact of these two features on the degree of texture defects, it reduces misjudgment caused by texture similarity and improves the ability to recognize tiny scratch defects or those with unclear texture features.

[0012] Furthermore, the overall defect degree of the liquid cooling plate is determined based on the following relationship: ; In the formula, is the total defect level of the liquid cooling plate, for The weight of for The weight of is the normalized mean of the defect levels of all non-reflective areas, It is the normalized mean of the defect degree of all reflective areas.

[0013] The above technical solution incorporates the normalized mean value of the defect degree in the non-reflective area and the normalized mean value of the defect degree in the reflective area, which means that when evaluating the overall defect degree of the liquid cooling plate, the defect information of different characteristic areas on the surface of the liquid cooling plate is taken into account, thereby being able to more comprehensively and accurately reflect the overall defect status of the surface of the liquid cooling plate.

[0014] Furthermore, the method of dividing the reflective area and the non-reflective area is: Performing region growing on the surface image to obtain multiple initial regions, if the average grayscale value of all pixels in a certain initial region is greater than or equal to a preset grayscale threshold, the initial region is used as a reflective region; if the average grayscale value of all pixels in a certain initial region is less than the preset grayscale threshold, the initial region is used as a non-reflective region; If the edge regularity of a certain reflective area is less than or equal to a preset edge regularity threshold, the reflective area is also divided into a non-reflective area.

[0015] The above technical solution uses a region growing algorithm to preliminarily divide the reflective area and the non-reflective area, and then optimizes the preliminary division result by calculating the curvature change value of the edge pixel points of the reflective area, avoiding the errors that are easy to exist when dividing the area only by grayscale values, making the division results of the reflective area and the non-reflective area more accurate.

[0016] Furthermore, the degree of local grayscale fluctuation of each pixel is determined based on the variance of grayscale values ​​of eight neighboring pixels of each pixel.

[0017] Furthermore, the contrast of each non-reflective area is obtained by: Obtain the gray level co-occurrence matrix of each non-reflective area in the directions of 0 degrees, 45 degrees, 90 degrees, 135 degrees and 180 degrees, and obtain the average gray level co-occurrence matrix; The contrast of the non-reflective area is calculated based on the mean gray-level co-occurrence matrix:

[0018] In the formula, For the The contrast of the non-reflective area is the total number of gray levels in the mean gray-level co-occurrence matrix, and All are grayscale. is the mean gray-level co-occurrence matrix The element at the position, the value of the element is the gray value and The gray value combination of The probability of appearing in a non-reflective area.

[0019] Furthermore, the method for extracting all scratched areas in each reflective area is: Each reflective area is binarized to obtain a binary image of each reflective area, pixels with a gray value of 1 in the binary image are marked as connected domains, edge detection is performed on each marked connected domain, and the closed area enclosed by the edge of each connected domain is used as the scratch area.

[0020] Furthermore, the method for determining the defect detection result of the surface of the liquid cooling plate according to the total defect degree is as follows: A first threshold, a second threshold, and a third threshold of the total defect degree are preset, and the first threshold is smaller than the second threshold, and the second threshold is smaller than the third threshold; When the total defect degree is less than or equal to the first threshold, the surface of the liquid cooling plate has no defects; when the total defect degree is greater than the first threshold and less than or equal to the second threshold, there are slight defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than the second threshold and less than or equal to the third threshold, there are moderate defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than the third threshold, there are severe defects on the surface of the liquid cooling plate.

[0021] On the other hand, the present invention also provides a surface defect detection system for a liquid cooling plate, the surface defect detection system comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any step of the surface defect detection method.

[0022] The present invention has the following effects: The present invention collects and preprocesses images, accurately divides reflective areas and non-reflective areas, and then analyzes the degree of defects in different areas from multiple feature angles such as grayscale and texture. Finally, the overall defect degree of the liquid cooling plate is comprehensively obtained. According to the size of the total defect degree, it is judged whether there are scratch defects on the surface of the liquid cooling plate. This overcomes the problems of light interference and surface texture interference, and improves the accuracy of surface defect detection of the liquid cooling plate. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0025] Reference Figure 1 The present invention provides a method for detecting surface defects of a liquid cooling plate, comprising steps S1 to S5: S1: Collect the surface image of the liquid cooling plate and perform preprocessing.

[0026] A high-resolution camera is used to shoot the surface of the liquid cooling plate vertically to avoid distortion of the surface image of the liquid cooling plate due to shooting angle problems. In addition, considering that during the image acquisition process, factors such as equipment performance differences and environmental changes may cause noise to be mixed into the image, and the noise will interfere with the subsequent feature analysis of the surface of the liquid cooling plate, the collected surface image of the liquid cooling plate is preprocessed, including: first graying the surface image, then using the median filtering algorithm to remove the noise in the surface image, and then using the histogram equalization algorithm to enhance the contrast so that the grayscale distribution of the surface image is more uniform, and finally sharpening is performed to enhance the edges and details of the surface image.

[0027] An accurate and reliable surface image can be obtained through the operation of this step.

[0028] S2: Divide the surface image into reflective and non-reflective areas.

[0029] First, analyze the inherent texture characteristics of the non-reflective area on the surface of the liquid cooling plate: the non-reflective area on the surface of the liquid cooling plate presents a specific texture pattern, showing a texture of black and white dots interlaced with each other. This texture has a high complexity and is manifested as frequent grayscale value fluctuations at the grayscale value level. Although the grayscale value fluctuates frequently, its fluctuation amplitude is relatively small. From the perspective of gradient value, it also presents the characteristics of frequent fluctuations. This inherent texture feature of the non-reflective area provides a basic reference for the subsequent identification of abnormal conditions (such as scratch defects).

[0030] When there are scratches in the non-reflective area on the surface of the liquid cooling plate, the texture characteristics of the scratches are: From the grayscale performance point of view, the color of the scratches is similar to the reflective area on the surface of the liquid cooling plate, which means that in the surface image, the grayscale value of the scratched area is higher and the grayscale value of the reflective area is also higher. Therefore, from the grayscale performance point of view, the scratched area in the non-reflective area presents a similar grayscale performance to the reflective area.

[0031] From the perspective of morphology, the scratched area usually presents a relatively long and thin shape, and the edge of the scratched area is not smooth and continuous, but has burrs, and the overall shape is irregular. This is an intuitive description of the visual morphology of scratches, which is consistent with the general understanding of scratches on the surface of objects. For example, common metal surface scratches in life are often long and thin with irregular edges.

[0032] From the perspective of gradient, in the non-reflective area, the gradient value of the pixel point in the scratch area fluctuates less than the gradient value of the normal texture pixel point in the non-reflective area. Because in the image, the gradient reflects the rate of change of the grayscale value of the pixel point. Due to its own complex structure (such as the interlacing of black and white dots), the grayscale value of the normal texture pixel point in the non-reflective area changes frequently, so the fluctuation of the gradient value is also large. The scratch area is often a highlight area, and the change of the grayscale value is relatively small, resulting in a smaller fluctuation of its gradient value. Therefore, by analyzing the fluctuation of the gradient value, it is helpful to distinguish the normal texture pixel points in the non-reflective area from the pixel points in the scratch area.

[0033] In one embodiment, the method of dividing the reflective area and the non-reflective area is: First, a region growing algorithm is performed on the surface image to generate multiple initial regions. A grayscale threshold is preset to 125 (empirical value). For each initial region, the average grayscale value of all its pixels is calculated. If the average grayscale value of an initial region is greater than or equal to 125, the initial region will be judged as a reflective region; if the average grayscale value is less than 125, it will be regarded as a non-reflective region. This method preliminarily completes the division of reflective and non-reflective regions based on the region growing algorithm and the grayscale values ​​of the pixels.

[0034] However, there are various reasons why the reflective areas obtained by the preliminary division have large grayscale values. On the one hand, it may be due to normal reflection phenomenon that causes the grayscale value to be high; on the other hand, the scratched area also shows high grayscale performance characteristics similar to the reflective area. Therefore, it is necessary to conduct further analysis on all the reflective areas obtained by the preliminary division.

[0035] From the perspective of physical properties, reflection usually occurs at the edges of the raised texture of the liquid cooling plate, so the edge shape of the area with a larger gray value due to reflection is relatively regular. However, due to the randomness of the scratching process, the scratched area will have burrs on the edge and the overall shape is relatively irregular.

[0036] Therefore, in order to effectively distinguish these two situations, the erosion and dilation algorithm is first used to process the surface image. The erosion operation can remove isolated pixels and small burrs at the edge of the object in the image, while the dilation operation can restore important edge information that may be lost due to erosion, while further highlighting the main part of the object. After the processing is completed, the Sobel operator is used to obtain all edge pixels in each reflective area in the surface image.

[0037] Next, take the pixel point in the lower left corner of the surface image as the coordinate origin, set the horizontal right direction as the horizontal axis, and set the vertical upward direction as the vertical axis to construct a two-dimensional coordinate system. For each reflective area, according to the curvature calculation formula, use the coordinates of each edge pixel point in the coordinate system to calculate the curvature value of each edge pixel point. Then, subtract the curvature value of each edge pixel point from the mean of the curvature values ​​of all edge pixels in the reflective area, and the difference value (absolute value of the difference) obtained is the curvature change value of the edge pixel point. Normalize the curvature change values ​​of all edge pixels in the reflective area so that the curvature change value of each edge pixel point is between 0 and 1, and then calculate the average value, which is used as the edge regularity of the reflective area.

[0038] The preset edge regularity threshold is 0.5 (empirical value). If the edge regularity of a reflective area is less than or equal to 0.5, it means that the edge irregularity of the area is relatively high, which is more consistent with the characteristics of a scratched area. Therefore, the reflective area is reclassified as a non-reflective area.

[0039] Through this series of rigorous operating procedures, the reflective area and non-reflective area in the surface image of the liquid cooling plate can be divided more accurately.

[0040] S3: Determine the defect degree of each non-reflective area according to the grayscale defect degree and the texture defect degree of each pixel point in each non-reflective area.

[0041] S31: Analyze the grayscale defect level of each pixel.

[0042] For a pixel in a non-reflective area, the greater the fluctuation of the grayscale value in the local area of ​​the pixel, and the more drastic the overall grayscale value change of all pixels in the non-reflective area where the pixel is located, it means that the pixel is more likely to have a grayscale defect.

[0043] Therefore, the grayscale co-occurrence matrix of each non-reflective area is obtained, and the contrast of each non-reflective area is calculated according to the grayscale co-occurrence matrix. The contrast of each non-reflective area reflects the severity of the overall grayscale value change of all pixels in the non-reflective area. For each pixel in each non-reflective area, the variance of the grayscale values ​​of the pixels in its eight neighborhoods is obtained to reflect the fluctuation of the grayscale value in the local area of ​​the pixel.

[0044] In one embodiment, the contrast of the non-reflective area is obtained by: The gray level co-occurrence matrix of each non-reflective area in the directions of 0 degrees, 45 degrees, 90 degrees, 135 degrees and 180 degrees is obtained to obtain the average gray level co-occurrence matrix.

[0045] The contrast of the non-reflective area is calculated based on the mean gray-level co-occurrence matrix:

[0046] In the formula, For the The contrast of the non-reflective area is the total number of gray levels in the mean gray-level co-occurrence matrix, and All are grayscale. is the mean gray-level co-occurrence matrix The element at the position, the value of the element is the gray value and The gray value combination of The probability of appearing in a non-reflective area.

[0047] In one embodiment, the grayscale defect degree of each pixel in each non-reflective area satisfies the following formula:

[0048] In the formula, For the The non-reflective area The grayscale defect degree of each pixel. This value is used to quantify the possibility of grayscale defects in the pixel. The larger the value, the higher the possibility of grayscale defects in the pixel. is the normalization function, and They are and The weight of ,For each pixel, the local gray scale fluctuation degree of the pixel and the contrast of the non-reflective area where the pixel is located are equally important for measuring the gray scale defect degree. For the The contrast of a non-reflective area is obtained by calculating the grayscale co-occurrence matrix of the non-reflective area. It reflects the severity of the overall grayscale value change of all pixels in the non-reflective area. The higher the contrast, the greater the difference in grayscale values ​​of the pixels in the non-reflective area, and the more drastic the overall grayscale change. For the The non-reflective area The local grayscale fluctuation degree of a pixel is calculated by obtaining the variance of the grayscale values ​​of the pixels in the eight neighborhoods of the pixel. The larger the variance, the greater the fluctuation of the grayscale value in the local area of ​​the pixel.

[0049] This formula comprehensively considers the overall grayscale change of the non-reflective area where the pixel is located and the local grayscale fluctuation of the pixel itself, linearly combines these two factors according to certain weights, and after normalization, obtains the grayscale defect degree of the pixel. This comprehensive consideration is in line with the judgment logic mentioned above, that is, the greater the grayscale value fluctuation in the local area of ​​a pixel, and the more drastic the overall grayscale value change of all pixels in its non-reflective area, the more likely the pixel will have a grayscale defect.

[0050] S32: Analyze the texture defect degree of each pixel.

[0051] For a pixel in a non-reflective area, when the gradient of the pixel is more consistent with the gradient of the non-reflective area, it means that the gradient direction of the pixel is more consistent with the gradient direction of other pixels in the non-reflective area, and is more consistent with the gradient characteristics of the pixels in the scratched area, that is, the pixel is more likely to be in the scratched area. If the edge regularity of the non-reflective area where the pixel is located is smaller, the edge is more irregular, which also increases the possibility that the pixel is located in the scratched area.

[0052] In one embodiment, the texture defect degree of each pixel point in each non-reflective area satisfies the following formula:

[0053] In the formula, For the The non-reflective area The texture defect degree of each pixel point, the larger the value, the The non-reflective area The more pixels there are, the more likely they are to be in the scratched area. is the normalization function, and are weights, , It is The non-reflective area The gradient value of a pixel, To obtain the maximum value function, It is The absolute value of the maximum gradient value corresponding to all pixels in a non-reflective area is used to measure the maximum range of the gradient values ​​of the pixels in the non-reflective area. By comparing it with the gradient value of a single pixel, it can reflect the relative gradient size of the pixel in the area. It is The edge regularity of a non-reflective area is used to measure whether the edge of the non-reflective area is regular. For areas with defects, their edges are usually irregular. The smaller the value, the more irregular the edge of the non-reflective area is, and the more likely it is that there will be texture defects. is the absolute value symbol.

[0054] In this formula, Reflects the The non-reflective area pixels and Therefore, in a non-reflective area, the maximum gradient value of all pixels is usually the case where the grayscale changes most dramatically in a certain direction. For example, in the non-reflective area on the surface of the liquid cooling plate, if there are some long and thin textures or scratches, these features will make the grayscale changes of pixels along their length more obvious, then a larger gradient value may appear in this direction. At this time, the direction corresponding to the maximum gradient value is likely to represent the main direction of the texture or structure in the area, that is, the main gradient direction. The non-reflective area pixels and The greater the gradient consistency of the non-reflective area, the The closer the gradient value of a pixel is to the maximum gradient value of the non-reflective area.

[0055] In summary, this step comprehensively considers the key factors of the pixel's own gradient value, the maximum gradient value of the non-reflective area (used to determine the gradient direction consistency), and the edge regularity of the non-reflective area when analyzing the texture defect degree of the pixel in the non-reflective area. Through the combination of these factors, the possibility of the pixel having defects in the texture angle can be determined. When the gradient direction consistency factor is large, at the same time, if the edge regularity corresponding to the non-reflective area where the pixel is located is small, that is, the edge is irregular, the possibility of the pixel having defects in the texture angle is greater, that is, the greater the degree of texture defects.

[0056] S33: Determine the degree of defects in each non-reflective area.

[0057] In each non-reflective area, the comprehensive defect degree of each pixel is first determined based on the grayscale defect degree and texture defect degree of each pixel, and then the defect degree of the non-reflective area is determined based on the comprehensive defect degree of all pixels.

[0058] In one embodiment, the comprehensive defect degree of each pixel point in each non-reflective area is determined based on the following method: the average of the grayscale defect degree and the texture defect degree of each pixel point in the non-reflective area is used as the comprehensive defect degree of each pixel point. The calculation formula of the comprehensive defect degree of each pixel point is:

[0059] In this formula, For the The non-reflective area The comprehensive defect degree of each pixel, For the The non-reflective area The grayscale defect degree of each pixel, For the The non-reflective area The degree of texture defect of each pixel.

[0060] In one embodiment, the defect degree of each non-reflective region is determined based on the following method: the average of the comprehensive defect degrees of all pixels in each non-reflective region is used as the defect degree of the non-reflective region.

[0061] S4: Determine the degree of defect of each reflective area according to the area ratio of the scratched area in each reflective area.

[0062] Due to its physical structure and light reflection characteristics, the reflective area on the surface of the liquid cooling plate is curved, and the scratched area in the reflective area presents short and irregular white scratches due to physical damage, and the scratched area has been classified as a reflective area in the previous step S2.

[0063] In one embodiment, the scratched area in each reflective area is extracted according to the following method: From an optical point of view, scratched areas will destroy the flatness of the object's surface. When light hits the scratched area, diffuse reflection will occur, making the scratched area appear brighter in the image, and the corresponding grayscale value will be larger. Although the grayscale value of the normal (no scratches) reflective area surface is also high, because its surface is relatively flat and the light reflection is more regular, relatively less light enters the camera, so the grayscale value is lower than that of the scratched area.

[0064] Therefore, the reflective area image is binarized, and the part with a smaller gray value (representing the normal area) is marked as 0, and the part with a larger gray value (most likely the scratched area, because the scratched area has a larger gray value in the gray image) is marked as 1. This step simplifies the image information, making it easier to focus on the scratched area later.

[0065] In the binary image, the parts marked as 1 (i.e. suspected scratch areas) are marked as connected domains, and then the edge detection algorithm is used on these marked connected domains to determine the edges of these connected domains. Finally, the closed area enclosed by the edge of each connected domain is used as each scratch area.

[0066] In one embodiment, the method for obtaining the defect degree of each reflective area is: In each reflective area, all scratched areas are obtained. The number of pixels in each scratched area is counted. The area of ​​each scratched area is equal to the total number of pixels it contains. The areas of all scratched areas are accumulated to obtain the total area of ​​the scratched areas. The ratio of the total area of ​​the scratched areas to the total area of ​​the entire light-emitting area is taken as the defect degree of the reflective area.

[0067] Since the size of the scratched area is directly related to the performance and service life of the liquid cooling plate, this calculation method can intuitively reflect the actual coverage of the reflective area by the scratches, and thus indirectly reflect the impact of the scratches on the overall physical properties of the liquid cooling plate.

[0068] S5: Comprehensively determine the total defect degree of the liquid cooling plate, and determine the defect detection result according to the magnitude of the total defect degree.

[0069] In one embodiment, the overall defect level of the liquid cooling plate is determined based on the following formula:

[0070] In the formula, is the total defect level of the liquid cooling plate, for The weight of for The weight of , , is the normalized mean of the defect levels of all non-reflective areas (the defect levels of all non-reflective areas are normalized first, and then the mean is calculated). It is the normalized mean of the defect levels of all reflective areas (the defect levels of all reflective areas are normalized first, and then the mean is calculated).

[0071] The weight is set in this way to reflect that the degree of defects in the non-reflective area has a greater impact on the total degree of defects. Because the non-reflective area usually represents the normal surface structure of the liquid cooling plate, its defects (such as scratches, pits, corrosion, etc.) directly affect the mechanical strength and thermal conductivity of the liquid cooling plate. Therefore, the degree of defects in the non-reflective area has a more direct impact on the overall performance of the liquid cooling plate.

[0072] The total defect level calculated by this formula can more truly reflect the overall performance of the liquid cooling plate. In actual production and quality inspection, it can provide a more accurate quantitative basis for determining whether the liquid cooling plate is qualified and whether it meets the use requirements. For example, in the cooling system of electronic equipment, the total defect level can be used to determine whether the liquid cooling plate can be used, avoiding problems such as poor equipment cooling due to neglect of local defects.

[0073] In one embodiment, the method for determining the defect detection result according to the magnitude of the total defect degree is: The first threshold of the preset total defect degree is 0.3, the second threshold is 0.5, and the third threshold is 0.7; when the total defect degree is less than or equal to 0.3, the surface of the liquid cooling plate has no defects; when the total defect degree is greater than 0.3 and less than or equal to 0.5, there are slight defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than 0.5 and less than or equal to 0.7, there are moderate defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than 0.7, there are severe defects on the surface of the liquid cooling plate.

[0074] Through this defect detection method, in the production inspection link, the liquid cooling plates can be quickly classified and screened according to the total defect degree. For a large number of products to be inspected, different quality grades can be quickly distinguished, which improves the inspection efficiency, reduces labor costs and time costs, and ensures that only products that meet specific quality requirements enter the next link or market.

[0075] The present invention also provides a surface defect detection system for a liquid cooling plate, the surface defect detection system comprising a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement any step of the surface defect detection method to complete the surface defect detection of the liquid cooling plate.

[0076] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A method for detecting surface defects of a liquid cooling plate, characterized in that: include: Collect the surface image of the liquid cooling plate for preprocessing and divide it into reflective area and non-reflective area; For each non-reflective area: Determine the grayscale defect level of each pixel based on the local grayscale fluctuation level of each pixel and the contrast of each non-reflective area; The texture defect degree of each pixel is determined according to the gradient consistency of each pixel and the edge regularity of each non-reflective area; the edge regularity of each non-reflective area is the average value of the normalized curvature change values ​​of all edge pixels of each non-reflective area; The average of the grayscale defect degree and the texture defect degree of each pixel is taken as the comprehensive defect degree of each pixel, and the average of the comprehensive defect degrees of all pixels is taken as the defect degree of each non-reflective area; For each reflective area: extract all scratched areas in each reflective area, and take the ratio of the total area of ​​all scratched areas to the total area of ​​each reflective area as the defect degree of each reflective area; The total defect degree of the liquid cooling plate is determined by comprehensively considering the defect degree of each non-reflective area and the defect degree of each reflective area, and the defect detection result of the surface of the liquid cooling plate is determined according to the total defect degree.

2. The surface defect detection method of the liquid cooling plate according to claim 1, characterized in that: The grayscale defect degree of each pixel satisfies the following formula: ； In the formula, For the The non-reflective area The grayscale defect degree of each pixel, is the normalization function, and They are and The weight of For the The contrast of the non-reflective area For the The non-reflective area The local grayscale fluctuation degree of each pixel.

3. The surface defect detection method of the liquid cooling plate according to claim 1, characterized in that: The texture defect degree of each pixel satisfies the following formula: ； In the formula, For the The non-reflective area The degree of texture defect of each pixel, is the normalization function, and are weights, It is The non-reflective area The gradient value of a pixel, To obtain the maximum value function, It is The absolute value of the maximum gradient value corresponding to all pixels in the non-reflective area, It is The edge regularity of the non-reflective area, is the absolute value symbol.

4. The surface defect detection method of a liquid cooling plate according to claim 1, characterized in that: The total defect level of the liquid cooling plate is determined based on the following relationship: ； In the formula, is the total defect level of the liquid cooling plate, for The weight of for The weight of is the normalized mean of the defect levels of all non-reflective areas, It is the normalized mean of the defect degree of all reflective areas.

5. The surface defect detection method of the liquid cooling plate according to claim 1, characterized in that: The method of dividing the reflective area and the non-reflective area is: Performing region growing on the surface image to obtain multiple initial regions, if the average grayscale value of all pixels in a certain initial region is greater than or equal to a preset grayscale threshold, the initial region is used as a reflective region; if the average grayscale value of all pixels in a certain initial region is less than the preset grayscale threshold, the initial region is used as a non-reflective region; If the edge regularity of a certain reflective area is less than or equal to a preset edge regularity threshold, the reflective area is also divided into a non-reflective area.

6. The surface defect detection method of the liquid cooling plate according to claim 1, characterized in that: The degree of local grayscale fluctuation of each pixel is determined based on the variance of the grayscale values ​​of the eight neighboring pixels of each pixel.

7. The surface defect detection method of the liquid cooling plate according to claim 1, characterized in that: The contrast of each non-reflective area is obtained as follows: Obtain the gray level co-occurrence matrix of each non-reflective area in the directions of 0 degrees, 45 degrees, 90 degrees, 135 degrees and 180 degrees, and obtain the average gray level co-occurrence matrix; The contrast of the non-reflective area is calculated based on the mean gray-level co-occurrence matrix: ; In the formula, For the The contrast of the non-reflective area is the total number of gray levels in the mean gray-level co-occurrence matrix, and All are grayscale. is the mean gray-level co-occurrence matrix The element at the position, the value of the element is the gray value and The gray value combination of The probability of appearing in a non-reflective area.

8. The surface defect detection method of a liquid cooling plate according to claim 1, characterized in that: The method to extract all scratched areas in each reflective area is: Each reflective area is binarized to obtain a binary image of each reflective area, pixels with a gray value of 1 in the binary image are marked as connected domains, edge detection is performed on each marked connected domain, and the closed area enclosed by the edge of each connected domain is used as the scratch area.

9. The surface defect detection method of a liquid cooling plate according to claim 1, characterized in that: The method for determining the defect detection results of the liquid cooling plate surface according to the total defect degree is: A first threshold, a second threshold, and a third threshold of the total defect degree are preset, and the first threshold is smaller than the second threshold, and the second threshold is smaller than the third threshold; When the total defect degree is less than or equal to the first threshold, the surface of the liquid cooling plate is free of defects; When the total defect degree is greater than the first threshold and less than or equal to the second threshold, there are slight defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than the second threshold and less than or equal to the third threshold, there are moderate defects on the surface of the liquid cooling plate; when the total defect degree of the liquid cooling plate is greater than the third threshold, there are severe defects on the surface of the liquid cooling plate.

10. A surface defect detection system for a liquid cooling plate, characterized in that: The defect detection system comprises a memory and a processor, wherein a computer program is stored in the memory, and the processor executes the computer program to implement the steps of the defect detection method according to any one of claims 1 to 9.

Images