A quality inspection method and system for integrated circuit boards

Abstract

The present invention relates to the technical field of circuit board detection, and in particular to a quality detection method and system for integrated circuit boards, the method comprising: collecting a surface image of a circuit board, pre-processing to obtain a grayscale image; traversing all pixel points of the grayscale image, and calculating the noise coefficient of the pixel points; selecting any pixel point as a central pixel point, determining the number of pixel points selected from each sub-neighborhood within a preset neighborhood of the central pixel point, obtaining the pixel point with the smallest Euclidean distance from the central pixel point among the pixel points selected from each sub-neighborhood, establishing a neighborhood for filtering the central pixel point, and using a bilateral filtering algorithm to filter the central pixel point to obtain a processed image, and performing cold welding defect detection on the processed image. The present invention uses an adaptive filtering algorithm to more effectively remove noise in an image while retaining important edge and detail information, thereby improving the quality of the collected image and making subsequent circuit board quality detection more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit board detection, and more specifically, to a quality detection method and system for an integrated circuit board. Background Art

[0002] As the core component of electronic equipment, the quality of integrated circuit boards directly affects the performance and reliability of the entire product. During the welding process, if the welding temperature does not reach the critical temperature required for the solder (solder) to melt, the solder cannot be fully melted. Since the solder is not fully melted, the formed welding point may not be smooth or even firm. This phenomenon is called cold welding defect. Cold welding not only affects the electrical connection (because a weak solder joint may cause the circuit to be disconnected), but also may affect the firmness of the mechanical connection, thereby affecting the reliability and performance of the product.

[0003] Prior art, such as a patent application document with publication number CN117571721A, discloses a circuit board pad surface defect detection method, device and storage medium. The circuit board pad surface defect detection method includes: collecting a front image of the circuit board to be tested as an original image, and constructing a pad defect recognition model based on computer vision; grayscale processing the original image to obtain a grayscale image to be tested, and using the model to identify whether there are pad features in the grayscale image to be tested; marking the pad position and segmenting the image with pads; using the pad defect recognition model to perform feature recognition on each segmented image, counting the actual number of pads, and calculating the coverage area and position of each pad, and judging whether the circuit board is qualified based on the calculation results.

[0004] However, the above patent application documents do not take into account the existence of adverse factors such as power supply interference and transmission line interference, which results in data loss or errors in the grayscale image obtained by collecting the front image of the circuit board to be inspected and performing grayscale processing, further affecting the detection of circuit board defects. Summary of the invention

[0005] In order to solve the above-mentioned technical problem that the quality of collected images is reduced and the defect detection results are affected, the present invention provides solutions in the following aspects.

[0006] In a first aspect, a quality inspection method for an integrated circuit board includes:

[0007] Collect the surface image of the circuit board and pre-process it to obtain a grayscale image; traverse all the pixels of the grayscale image, perform filtering processing, obtain a processed image, and perform cold welding defect detection on the processed image;

[0008] The filtering process includes:

[0009] Calculate the noise coefficient of the pixel point; select any pixel point as the central pixel point, and determine the number of pixels selected from each sub-neighborhood within the preset neighborhood of the central pixel point, and the number of pixels satisfies the relationship:

[0010] ; In the formula, To preset the first pixel in the neighborhood from the center pixel The number of pixels selected by the sub-neighborhood, The center pixel The sum of the noise coefficients of all pixels in the sub-neighborhood, is the total number of sub-neighborhoods, The total number of pixels in the neighborhood is preset for the central pixel. is an exponential function with the natural constant e as base;

[0011] The pixel point with the smallest Euclidean distance to the central pixel point is obtained from the pixels selected from each sub-neighborhood, the neighborhood of the central pixel point for filtering is established, and the central pixel point is filtered using the bilateral filtering algorithm.

[0012] The present invention calculates the corresponding noise coefficient by analyzing the degree to which each pixel point in the circuit board image is affected by noise, selects a specific number of pixel points from a preset neighborhood of a central pixel point according to the noise coefficient, and then adaptively adjusts the range of filtering processing according to the noise situation, so that the filtering effect is more accurate, and further selects the pixel point with the smallest Euclidean distance to the central pixel point from each sub-neighborhood to construct a neighborhood for filtering processing, which helps to retain the edge information of the image and reduce the influence of noise. Through filtering processing, the noise and interference in the image can be effectively removed, making the image clearer and smoother, thereby improving the accuracy of subsequent cold welding defect detection.

[0013] Preferably, the process of acquiring the noise coefficient of the pixel point includes:

[0014] Set the reference area for each pixel and calculate the gray value range within the reference area corresponding to the pixel;

[0015] Calculate the grayscale mean of all pixels in the grayscale image,

[0016] The product of the absolute value of the difference between the grayscale value of the pixel point and the grayscale mean and the grayscale value range is taken as the noise coefficient of the corresponding pixel point.

[0017] By calculating the difference in grayscale value between a pixel and its reference area, as well as the ratio of this difference to the grayscale value extreme, the noise level of the pixel can be quantified. The larger the noise coefficient, the more serious the noise interference to the pixel. In subsequent image processing, such as filtering and enhancement, targeted processing can be performed based on the noise coefficient of the pixel.

[0018] Preferably, after calculating the noise coefficient of the pixel point, the noise coefficient is corrected, and the corrected noise coefficient satisfies the following relationship:

[0019] ; In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the The number of pixels in the connected domain corresponding to the pixel points, For the The number of pixels contained in the minimum circumscribed circle of the connected domain corresponding to the pixels, It is an exponential function with the natural constant e as its base.

[0020] Through the above correction formula, the noise coefficient of each pixel can be adjusted so that the corrected noise coefficient not only depends on the original noise coefficient, but also takes into account the structural characteristics of the connected domain where the pixel is located. This adjustment helps to more accurately reflect the actual impact of noise in fields such as image processing and computer vision, thereby improving image quality and subsequent processing effects.

[0021] Preferably, after determining the number of pixel points selected from each sub-neighborhood within the preset neighborhood of the central pixel point, the method further includes:

[0022] Count the sum of the pixels selected from all sub-neighborhoods. If the sum of the selected pixels is less than the total number of neighborhood pixels required for filtering the central pixel, add one to the number of pixels selected from the sub-neighborhood with the smallest sum of pixel noise coefficients until the sum of the selected pixels is equal to the total number of neighborhood pixels required for filtering the central pixel.

[0023] If the sum of the selected pixels is less than the required total number of neighborhood pixels, the filtering process may result in inaccurate results due to insufficient information. This can be avoided by adjusting the number of pixels selected in the sub-neighborhood, thereby ensuring the accuracy and reliability of the filtering process.

[0024] Preferably, the number of sub-neighborhoods within the preset neighborhood of the central pixel is set to 4.

[0025] Preferably, after calculating the noise coefficient of the pixel point, the noise coefficient is corrected, and the corrected noise coefficient satisfies the following relationship:

[0026] ; In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the first The grayscale mean of all pixels in the window constructed with pixels as the center, is the correction factor, is a constant.

[0027] Correction Factor and constant The introduction of provides more flexibility for the correction of noise coefficient. It can be adjusted according to the actual application scenario to meet different image processing requirements, and the constant This helps to maintain a certain stability during the correction process and prevent the noise coefficient from being too large or too small due to the grayscale mean being too small.

[0028] Preferably, after determining the number of pixel points selected from each sub-neighborhood within the preset neighborhood of the central pixel point, the method further includes:

[0029] Calculate the weight of each sub-neighborhood; count the sum of the pixels selected from all sub-neighborhoods. If the sum of the selected pixels is less than the total number of neighborhood pixels required for filtering the central pixel, increase the number of selected pixels for the sub-neighborhood with the highest weight until the sum of the selected pixels is equal to the total number of neighborhood pixels required for filtering the central pixel.

[0030] In a second aspect, a quality inspection system for an integrated circuit board includes: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the above-mentioned quality inspection method for the integrated circuit board is implemented.

[0031] The beneficial effects of the present invention are:

[0032] The present invention improves the quality of the image by processing the circuit board surface image using an adaptive filtering algorithm, and performs cold welding defect detection on the processed image on the basis of filtering, thereby greatly improving the accuracy and efficiency of cold welding defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0034] Figure 1It is a method flow chart of step S1 to step S2 in a quality inspection method for an integrated circuit board according to an embodiment of the present invention.

[0035] Figure 2 It is a method flow chart of steps S20 to S22 in a quality inspection method for an integrated circuit board according to an embodiment of the present invention.

[0036] Figure 3 It is a distribution diagram of each sub-neighborhood in a quality detection method for an integrated circuit board according to an embodiment of the present invention. DETAILED DESCRIPTION

[0037] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0038] The embodiment of the present invention discloses a quality inspection method for an integrated circuit board, referring to Figure 1 , including steps S1-S2, which are as follows:

[0039] S1: Collect the surface image of the circuit board and pre-process it to obtain a grayscale image.

[0040] Use a high-definition camera to vertically shoot the surface of the integrated circuit board to ensure that the details in the image are clearly visible, and convert the collected color image into a grayscale image.

[0041] Furthermore, the FCN (Fully Convolutional Network) semantic segmentation model is used to extract the integrated circuit board area in the grayscale image. Through semantic segmentation, the integrated circuit board area in the image is distinguished from the background or other non-target areas, providing an accurate target area for detailed analysis in subsequent steps.

[0042] S2: Traverse all the pixels of the grayscale image, perform filtering processing, obtain a processed image, and perform cold welding defect detection on the processed image.

[0043] In the process of collecting images of integrated circuit boards, since the data transmission lines or interfaces may be affected by various external factors (such as electromagnetic interference, physical damage, etc.), the signals may be disturbed during the transmission process. Such interference may cause partial loss of image data or errors, thereby forming speckle noise on the image. The presence of speckle noise will significantly reduce the quality of the image, making it blurred or unclear. In the detection of potential cold welding defects on the surface of the integrated circuit board, the speckle noise may mask or mislead the detection algorithm, resulting in the defect being misjudged or missed. In an embodiment of the present invention, a bilateral filtering algorithm is used to preprocess the surface image of the integrated circuit board. The algorithm not only takes into account the spatial distance between pixels, but also takes into account the intensity difference between pixels (i.e., the difference in color or grayscale value). Therefore, the algorithm can significantly reduce the speckle noise in the image while retaining the detail information of important edges in the image.

[0044] Specifically, all pixels of the grayscale image are traversed and filtered to obtain a processed image. Figure 2 The steps of filtering processing include step S20-step S22.

[0045] S20: Calculate the noise coefficient of the pixel.

[0046] Since speckle noise is more prominent in the grayscale value of the image, for each pixel in the grayscale image, the noise characteristics in its surrounding neighborhood are analyzed.

[0047] Specifically, for each pixel in the grayscale image, determine an area around the pixel as a reference area. The reference area can be a rectangular or circular window of a fixed size, or other shapes, which can be selected according to the needs. In an embodiment of the present invention, for each pixel in the grayscale image, select a 5×5 window centered on itself as the reference area; in the reference area of ​​each pixel, find the maximum and minimum values ​​of the grayscale value, and further calculate the difference between the two values, that is, the grayscale value range. This range reflects the fluctuation range of the grayscale value in the reference area; traverse the entire grayscale image and calculate the average grayscale value of all pixels. This mean value provides a reference for the overall brightness of the image; for each pixel, calculate the absolute value of the difference between its grayscale value and the grayscale mean, multiply this absolute value by the grayscale value range of the corresponding reference area, and the multiplication result is used as the noise coefficient of the pixel.

[0048] Exemplarily, the noise coefficient of a pixel satisfies the relationship:

[0049]

[0050] In the formula, For the The noise coefficient of each pixel is For the The gray value in the reference area corresponding to the pixel point is extremely different. For the The gray value of a pixel, is the mean grayscale value of all pixels in the grayscale image.

[0051] The calculation process of the noise coefficient of a pixel combines local (greyscale value range in the reference area) and global (greyscale mean of the entire image) information. By comparing the grayscale value of a pixel with the overall average grayscale value and taking into account the fluctuation range of the local grayscale value, an indicator reflecting the noise characteristics of the pixel can be obtained. Pixels with higher noise coefficients may correspond to edges, textures, or noisy areas in the image, while pixels with lower noise coefficients may be located in smooth, uniform areas.

[0052] It should be noted that in addition to speckle noise, there are also cold weld areas. Relying solely on the noise coefficient of the above-mentioned calculated pixel points cannot accurately distinguish different types of noise and defects. Therefore, the noise coefficient needs to be corrected in order to improve the accuracy of identifying other types of noise such as cold weld areas and speckle noise in the image.

[0053] Specifically, the above Take pixels as an example to construct its connected domain. The steps to construct a connected domain are as follows:

[0054] Step 1: The first pixel point is taken as the starting point, and starting from the starting point, a search is performed in the eight pixel points around the starting point; the second step is to include the pixel points whose grayscale value difference between the eight pixel points around the starting point and the grayscale value of the starting point is less than a preset threshold value (the threshold value is set to 3 in the embodiment of the present invention) into the connected domain of the starting point; the third step is to repeat the first and second steps for each pixel point included in the connected domain until no new pixel point is included in the connected domain, thereby obtaining the first pixel point. The connected domain corresponding to the pixels.

[0055] In getting the After the connected domain of pixels, considering that noise usually appears as isolated pixels or small areas, the calculation of the The number of pixels contained in the minimum circumscribed circle of the connected domain of a pixel point is calculated to evaluate whether the connected domain around the pixel point is large enough to distinguish whether the pixel point is noise or a valid part of the image.

[0056] Furthermore, the above The noise coefficient of each pixel is corrected, and the corrected noise coefficient satisfies the relationship:

[0057]

[0058] In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the The number of pixels in the connected domain corresponding to the pixel points, For the The number of pixels contained in the minimum circumscribed circle of the connected domain corresponding to the pixels, It is an exponential function with the natural constant e as its base.

[0059] in, The larger the The more likely a pixel and the pixels in its reference area are to be speckle noise from the perspective of grayscale value expression, the larger the corresponding corrected noise coefficient will be. Expressed as The circularity of the connected domain of pixels. The larger the value, the The more connected domains of pixels are, the more likely they are to belong to the speckle noise region (because speckle noise usually appears as a compact circular or elliptical region).

[0060] when When it is smaller, that is, The connected domain of a pixel is small, which means that the pixel is at the edge of the noise or an isolated noise point. The influence of is small (because the negative power of the base e of the exponential function is close to 1 when the input is small), which helps to suppress the excessive influence of large-area noise areas on the overall noise evaluation. It will not increase infinitely due to the large connected domain.

[0061] The above correction process takes into account the influence of the size of the connected domain and the extension range of the connected domain on the noise coefficient. By adjusting these two factors, the corrected noise coefficient can more accurately reflect the actual situation.

[0062] According to the above calculation The noise coefficient and the corrected noise coefficient of each pixel can be calculated in the same way to obtain the noise coefficient and the corrected noise coefficient corresponding to all pixels in the grayscale image. The corrected noise coefficient can be further used as the final noise coefficient of the pixel.

[0063] In another embodiment, another method for calculating the corrected noise coefficient is provided, and the expression is as follows:

[0064]

[0065] In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the first The grayscale mean of all pixels in a window (specifically, a window of size 3×3 can be constructed) with pixels as the center, is the correction factor, is a constant.

[0066] in, Reflects the The grayscale distribution of all pixels in the window built with pixels as the center is introduced As part of the correction, the grayscale changes in the area around the pixel are better considered, thereby more accurately reflecting the noise level of the pixel.

[0067] in, A value of 0.5 is used to avoid over-correction or under-correction.

[0068] S21: Select any pixel point as the central pixel point, and determine the number of pixel points selected from each sub-neighborhood within a preset neighborhood of the central pixel point.

[0069] After the noise coefficients of all pixels are calculated in the above step S20, the neighborhood position used for filtering and the number of pixels included therein are determined according to the noise coefficients of the pixels.

[0070] Specifically, any pixel point in the grayscale image is selected as the central pixel point, a neighborhood range is preset (set to 7×7 in the embodiment of the present invention), and the preset neighborhood is divided into multiple sub-neighborhoods, each sub-neighborhood represents a different direction or position. Specifically, the preset neighborhood can be divided into 4 sub-neighborhoods, and the corresponding positions are as follows: Figure 3 As shown in the figure, the red dot represents the central pixel, and the directions indicated by the arrows correspond to the locations of the four sub-neighborhoods.

[0071] For each sub-neighborhood, calculate the number of pixels selected from the sub-neighborhood, that is, satisfy the relationship:

[0072]

[0073] In the formula, To preset the first pixel in the neighborhood from the center pixel The number of pixels selected by the sub-neighborhood, The center pixel The sum of the noise coefficients of all pixels in the sub-neighborhood, is the total number of sub-neighborhoods (i.e. =4), The total number of pixels in the neighborhood of the central pixel is preset (since the neighborhood range of the central pixel is set to 7×7 in the above process, =49), It is an exponential function with the natural constant e as its base.

[0074] The larger the sum of the noise coefficients is, that is, the more noise there is, the fewer pixels are selected in the corresponding sub-neighborhood.

[0075] It should be noted that since the above-mentioned preset neighborhood from the central pixel The calculation formula for the number of pixels selected by each sub-neighborhood contains a rounding operation, which results in the sum of the number of pixels initially selected by all sub-neighborhoods being slightly less than the total number of pixels required in special cases. To solve this problem, the number of pixels selected by the sub-neighborhood with the smallest sum of pixel noise coefficients is increased by one until the sum of the selected pixels equals the total number of neighborhood pixels required for filtering the center pixel; or the weight of each sub-neighborhood is calculated, and the number of pixels selected by the sub-neighborhood with the highest weight is increased until the sum of the selected pixels equals the total number of neighborhood pixels required for filtering the center pixel.

[0076] The difference between the grayscale value of the pixel in each sub-neighborhood and the grayscale value of the central pixel is calculated. The smaller the difference, the higher the similarity between the sub-neighborhood and the central pixel, so a higher weight is given to the sub-neighborhood.

[0077] S22: Obtain the pixel point with the smallest Euclidean distance from the central pixel point from the pixels selected from each sub-neighborhood, establish a neighborhood for filtering the central pixel point, and use a bilateral filtering algorithm to filter the central pixel point.

[0078] In the previous step, the number of pixels that need to be selected in each sub-neighborhood has been determined. When selecting neighborhood pixels, if the selected pixels are far away from the center pixel, the correlation between them may be reduced, thereby affecting the filtering effect. Using Euclidean distance to select the pixel closest to the center pixel can ensure that the selected pixel has a strong correlation with the center pixel, thereby improving the filtering effect.

[0079] Specifically, the Euclidean distance between the selected pixel points in each sub-neighborhood and the central pixel point is calculated, and the pixel point closest to the central pixel point is selected according to the size of the Euclidean distance. Then, a certain number of pixel points are selected in all sub-neighborhoods, and these pixel points are combined with the central pixel point to form the filtering neighborhood of the central pixel point.

[0080] Similarly, the filtering neighborhood corresponding to all pixels in the grayscale image can be obtained.

[0081] It should be noted that the Euclidean distance is only a selection criterion for selecting a number of pixels closest to the central pixel from each sub-neighborhood. Selecting pixels based on the Euclidean distance does not limit the number of pixels to be selected. The number of pixels is determined by the requirements of the filtering process and the image features, and the number of pixels to be selected in each sub-neighborhood can be adjusted to flexibly adapt to different filtering scenarios. Therefore, when establishing a neighborhood, the pixel closest to the central pixel can be selected based on the Euclidean distance, while ensuring that the number of selected pixels meets the requirements of the filtering process.

[0082] Furthermore, in an embodiment of the present invention, the spatial standard deviation is preset to an empirical value of 2, the grayscale standard deviation is preset to an empirical value of 10, the neighborhood range of each pixel point in the grayscale image is preset to 7×7, and each pixel point is filtered using the existing steps of the bilateral filtering algorithm to further obtain a processed image. Based on the processed image, the existing template matching method is used to detect potential cold welding defects on the surface of the integrated circuit board.

[0083] It should be noted that setting the 7×7 neighborhood range is a global preset, which stipulates how much other pixels around each pixel need to consider when filtering. However, this global preset does not exclude more detailed adjustments in the local range. Calculating the number of pixels selected in the sub-neighborhood is actually to further divide the global 7×7 neighborhood into smaller sub-neighborhoods, and determine how many pixels to select for subsequent filtering based on the noise characteristics of each sub-neighborhood, so that in areas with less noise, more pixels can be selected to maintain image details; while in areas with more noise, fewer pixels can be selected to reduce the impact of noise. This local adaptive adjustment helps to improve the effect of filtering.

[0084] For example, the filtered surface image of the integrated circuit board is compared with the template, and the similarity or difference between the image and the template is calculated to determine whether there is a defect in the image that matches the template. If a region in the image has a high similarity or a small difference with the template, then the region may be considered to have a cold weld defect.

[0085] The embodiment of the present invention further discloses a quality inspection system for an integrated circuit board, including a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the quality inspection method for the integrated circuit board according to the present invention is implemented.

[0086] The system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface. The configuration and functions of these components are known in the art and will not be described in detail here.

[0087] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module or both. Any such computer storage medium may be part of a device or accessible or connectable to a device.

[0088] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0089] 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 quality inspection method for an integrated circuit board, characterized in that: include: Collect the surface image of the circuit board and pre-process it to obtain a grayscale image; Traverse all the pixels of the grayscale image, perform filtering processing, obtain a processed image, and perform cold welding defect detection on the processed image; The filtering process includes: Calculate the noise coefficient of the pixel point; select any pixel point as the central pixel point, and determine the number of pixels selected from each sub-neighborhood within the preset neighborhood of the central pixel point, and the number of pixels satisfies the relationship: ; In the formula, To preset the first pixel in the neighborhood from the center pixel The number of pixels selected by the sub-neighborhood, The center pixel The sum of the noise coefficients of all pixels in the sub-neighborhood, is the total number of sub-neighborhoods, The total number of pixels in the neighborhood is preset for the central pixel. is an exponential function with the natural constant e as base; The pixel point with the smallest Euclidean distance to the central pixel point is obtained from the pixels selected from each sub-neighborhood, the neighborhood of the central pixel point for filtering is established, and the central pixel point is filtered using the bilateral filtering algorithm.

2. The quality inspection method of an integrated circuit board according to claim 1, characterized in that: The process of obtaining the noise coefficient of the pixel point includes: Set the reference area for each pixel and calculate the gray value range within the reference area corresponding to the pixel; Calculate the grayscale mean of all pixels in the grayscale image, The product of the absolute value of the difference between the grayscale value of the pixel point and the grayscale mean and the grayscale value range is taken as the noise coefficient of the corresponding pixel point.

3. The quality inspection method of an integrated circuit board according to claim 2, characterized in that: After calculating the noise coefficient of the pixel point, the noise coefficient is corrected, and the corrected noise coefficient satisfies the following relationship: ; In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the The number of pixels in the connected domain corresponding to the pixel points, For the The number of pixels contained in the minimum circumscribed circle of the connected domain corresponding to the pixels, It is an exponential function with the natural constant e as its base.

4. The quality inspection method of an integrated circuit board according to claim 3, characterized in that: After determining the number of pixel points selected from each sub-neighborhood within the preset neighborhood of the central pixel point, the method further includes: Count the sum of the pixels selected from all sub-neighborhoods. If the sum of the selected pixels is less than the total number of neighborhood pixels required for filtering the central pixel, add one to the number of pixels selected from the sub-neighborhood with the smallest sum of pixel noise coefficients until the sum of the selected pixels is equal to the total number of neighborhood pixels required for filtering the central pixel.

5. The quality inspection method of an integrated circuit board according to claim 4, characterized in that: The number of sub-neighborhoods within the preset neighborhood of the central pixel is set to 4.

6. The quality inspection method of an integrated circuit board according to claim 2, characterized in that: After calculating the noise coefficient of the pixel point, the noise coefficient is corrected, and the corrected noise coefficient satisfies the following relationship: ; In the formula, For the The noise coefficient after pixel correction is For the The noise coefficient of each pixel is For the first The grayscale mean of all pixels in the window constructed with pixels as the center, is the correction factor, is a constant.

7. The quality inspection method of an integrated circuit board according to claim 1, characterized in that: After determining the number of pixel points selected from each sub-neighborhood within the preset neighborhood of the central pixel point, the method further includes: Calculate the weight of each sub-neighborhood; count the sum of the pixels selected from all sub-neighborhoods. If the sum of the selected pixels is less than the total number of neighborhood pixels required for filtering the central pixel, increase the number of selected pixels for the sub-neighborhood with the highest weight until the sum of the selected pixels is equal to the total number of neighborhood pixels required for filtering the central pixel.

8. A quality inspection system for integrated circuit boards, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the quality detection method for the integrated circuit board according to any one of claims 1 to 7 is implemented.

Images