A QFP chip image pyramid positioning system

Abstract

The present invention relates to the field of image analysis technology, and specifically to a QFP chip image pyramid positioning system. The QFP chip image pyramid positioning system includes a pyramid construction module, a pin initial positioning module, a pin precise positioning module, and an interference area exclusion module. In the present invention, the original chip image is processed by step-by-step downsampling and Gaussian filtering, effectively improving the image processing speed when constructing a multi-level image pyramid. Binarization processing and a pin neighbor connection strategy are applied to clearly identify the pin connected domain, enhancing the system's pin area recognition capability. An upsampled image of the original image data size is used for area recalibration, enabling accurate pin positioning while maintaining high resolution. By accurately extracting pin contour information and effectively excluding non-target connected domains, the probability of incorrect positioning is significantly reduced, improving chip assembly quality and speed, and bringing significant benefits to high-speed electronic assembly lines.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a QFP chip image pyramid positioning system. Background Art

[0002] Image analysis is a technical field concerned with extracting useful information from images. It encompasses multiple subfields, including image segmentation, edge detection, motion analysis, depth and shape recovery, and color feature measurement. It can identify and classify objects in images, determine their position and size, and analyze motion and interaction information. Image analysis is commonly used in fields such as surveillance, medical imaging, remote sensing, and industrial visual inspection, emphasizing the extraction of as much quantitative and qualitative information as possible from image data.

[0003] The QFP chip image pyramid positioning system primarily utilizes image pyramid technology to precisely locate the position of QFP (quad flat package) chips. By processing images at different resolution levels, it effectively improves image processing speed and positioning accuracy. It is a common application in electronics assembly and quality inspection. Its main uses include automatically detecting chip placement accuracy to ensure the assembly quality of electronic devices, and quickly identifying and correcting chip position errors on the production line, playing a vital role in improving production efficiency and product quality.

[0004] Existing technologies exhibit significant deficiencies in image processing speed and accuracy, particularly when it comes to precisely locating the tiny pins of QFP chips. Relying on single-resolution image processing strategies makes it difficult to quickly and accurately locate chip pins in complex or low-contrast environments. Furthermore, traditional technologies require high computing resources to process large amounts of data, easily creating bottlenecks on the production line. The lack of efficient pin connectivity domain identification and error elimination mechanisms also increases the risk of mislocalization, leading to reduced production efficiency and yield, which in turn impacts the functionality and reliability of electronic devices and increases production and repair costs. Summary of the Invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a QFP chip image pyramid positioning system.

[0006] To achieve the above-mentioned object, the present invention adopts the following technical solution: A QFP chip image pyramid positioning system includes:

[0007] The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling step by step, and applies Gaussian filtering to each level of the image to smooth the image edges and noise, thus constructing a pyramid structure and forming a reduced version of the image set.

[0008] The pin initial positioning module performs binarization processing based on the reduced version image set, converts the image into black and white format, connects touching and adjacent pins, identifies all potential pin connected domains, sets filtering conditions based on the circumscribed rectangle of the largest connected domain based on the potential pin connected domain identification results, selects and merges the connected domains that match the conditions, and creates a pin area initial positioning result;

[0009] The pin precise positioning module uses the initial pin area positioning results to mark the corresponding area in the upsampled image of the original chip image data size. Based on the corresponding area marking results, the module further screens the connected pin shapes according to size and shape standards, identifies the actual outline of each pin, and outputs the pin accurate outline information;

[0010] The interference area exclusion module calculates the center position of each connected domain based on the accurate pin contour information and compares it with the preset center position. According to the comparison result, the connected domains with abnormal deviations are excluded, and the area including only pins is accurately screened to obtain accurate pin positioning results.

[0011] As a further solution of the present invention, the steps of obtaining the reduced version image set are:

[0012] Collect the original image data of the chip and perform step-by-step downsampling, reducing the image resolution to half of the original size at each level. The formula is:

[0013]

[0014] Calculate the size S of the i-th level image i , generate difference resolution image data, where S i-1 Represents the size of the previous level image; Gaussian filtering is applied to the difference resolution image data using the formula:

[0015]

[0016] Calculate the weight value G(x,y) of the Gaussian filter to obtain the Gaussian filtered image data, where x and y represent the pixel positions of the image and σ represents the standard deviation of the Gaussian filter;

[0017] The image data is processed based on the Gaussian filter, reordered and stacked from high to low resolution, and an image pyramid is constructed to generate a reduced version image set.

[0018] As a further solution of the present invention, the step of identifying the potential pin connectivity domain is:

[0019] Extracting a single image based on the reduced version image set, converting each pixel in the image into black and white according to the grayscale value to obtain a binary image set;

[0020] Based on the binary image set, an image dilation operator is applied to each image to process the image, thereby increasing the range of white pixels and connecting pins that are visually similar but not physically touching, thereby generating a preliminary pin connectivity area;

[0021] Based on the preliminary pin connectivity area, a depth analysis is performed on the preliminary pin connectivity area to identify all connected white blocks in the image, a unique identifier is assigned to each independent connectivity area, and a potential pin connectivity domain identification result is established.

[0022] As a further solution of the present invention, the steps for obtaining the initial positioning result of the pin area are:

[0023] Based on the potential pin connected domain identification results, perform boundary extraction on each connected domain using the formula:

[0024] A=(x max -x min )×(y max -y min )

[0025] Calculate the area of ​​the circumscribed rectangle A and obtain the circumscribed rectangle set, where x min , x max ,y min ,y max Represent the minimum and maximum coordinates of the connected domain in the x and y directions respectively;

[0026] According to the circumscribed rectangle set, the area screening condition is set using the formula:

[0027] S m =max(A i )

[0028] Filter and output the largest connected region S m , where A i is the area of ​​the circumscribed rectangle of the i-th connected domain;

[0029] The maximum connected domain is merged, and adjacent connected domains are checked. If the circumscribed rectangle intersects and is adjacent to the maximum connected domain, a merge operation is performed to obtain an initial pin area positioning result.

[0030] As a further solution of the present invention, the corresponding area calibration step is:

[0031] Performing upsampling processing based on the original chip image data to increase pixel density and output upsampled image data;

[0032] Based on the initial positioning result of the pin area, the relative coordinates of the pin are converted into the coordinate system of the upsampled image, and accurately mapped to obtain the mapped pin area;

[0033] According to the mapped pin area, image segmentation is performed, gradient information of pixels is analyzed to identify boundaries, the actual position of each pin is calibrated, and a corresponding area calibration result is generated.

[0034] As a further solution of the present invention, the steps for obtaining accurate pin contour information are:

[0035] Based on the corresponding area calibration results, the formula is used:

[0036]

[0037] Calculate the characteristic size H of the region, where x i and y i are the coordinates of points in the region, and is the average value of the coordinates, N is the total number of points,

[0038]

[0039] θ=max(Δφ)

[0040] Calculate the edge curvature C and angle change rate θ, where κ i is the local curvature of each point, Δφ is the change in angle between adjacent points, and the pin shape that matches the standard is screened according to the preset shape standard and size threshold to generate a connected pin shape that has been screened by the size and shape standards;

[0041] For the connected pin shapes that have been screened by size and shape standards, analyze the magnitude and direction of the image gradient, determine and connect the edges, identify the actual outline of each pin, and generate pin outline data;

[0042] The pin outline data is integrated, formatted and matched to industry standard data files, the geometric dimensions and position information of the pins are recorded, and accurate pin outline information is generated.

[0043] As a further solution of the present invention, the step of comparing with the preset center position is:

[0044] Based on the accurate pin contour information, the coordinates of all pixels in each connected domain are extracted to generate a pixel coordinate extraction result;

[0045] According to the pixel point coordinate extraction result, the formula is used:

[0046]

[0047] Calculate the center position Z of the i-th connected component i , get the set of connected domain center positions, where (x m ,y m) is the coordinate of the mth pixel, M is the total number of pixels in the connected component;

[0048] Based on the connected domain center position set, the connected domain center position is compared with the preset center position, the deviation is evaluated and the accuracy of the position is verified, and the comparison result is output.

[0049] As a further solution of the present invention, the steps for obtaining the accurate pin positioning result are:

[0050] Based on the comparison results, identifying connected domains whose deviations exceed an allowable range, and generating a list of connected domains with excessive deviations;

[0051] Eliminate the list of connected domains with deviations exceeding the standard, and only retain the connected domains with deviations within the allowable range, thereby obtaining a filtered set of valid connected domains;

[0052] According to the set of effective connected domains after screening, the formula is adopted:

[0053]

[0054] Calculate the average center position coordinates C of all valid connected regions final , generate accurate pin positioning results, where C valid,i Represents the center position of the i-th effective connected domain, and U is the total number of effective connected domains.

[0055] Compared with the prior art, the advantages and positive effects of the present invention are:

[0056] In the present invention, the original chip image is processed by step-by-step downsampling and Gaussian filtering, which effectively improves the image processing speed when constructing a multi-level image pyramid. The pin connectivity domain is clearly identified by applying binarization processing and pin neighbor connection strategy, which enhances the system's pin area recognition capability. The upsampled image of the original image data size is used for area recalibration, which can perform accurate pin positioning while maintaining high resolution. By accurately extracting pin contour information and effectively excluding non-target connectivity domains, the probability of erroneous positioning is significantly reduced, the chip assembly quality and speed are improved, and significant benefits are brought to high-speed electronic assembly lines. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a system flow chart of the present invention;

[0058] Figure 2 This is a flow chart for obtaining a reduced version image set of the present invention;

[0059] Figure 3 This is a flow chart for identifying potential pin connectivity domains of the present invention;

[0060] Figure 4This is a flow chart for obtaining the initial positioning result of the pin area of ​​the present invention;

[0061] Figure 5 This is a flow chart for the calibration of the corresponding area of ​​the present invention;

[0062] Figure 6 This is a flow chart for obtaining accurate pin outline information of the present invention;

[0063] Figure 7 A flow chart showing the comparison between the present invention and the preset center position;

[0064] Figure 8 This is a flow chart for obtaining accurate pin positioning results of the present invention. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0066] In the description of the present invention, it should be understood that the terms "length," "width," "up," "down," "front," "back," "left," "right," "vertical," "horizontal," "top," "bottom," "inside," "outside," and the like, indicating positions or relationships, are based on the positions or relationships shown in the accompanying drawings and are intended only to facilitate the description of the present invention and simplify the description. They do not indicate or imply that the devices or elements referred to must have a specific orientation, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limiting the present invention. Furthermore, in the description of the present invention, "plurality" means two or more, unless otherwise expressly and specifically defined.

[0067] See also Figure 1 , a QFP chip image pyramid positioning system includes:

[0068] The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling step by step, and applies Gaussian filtering to each level of the image to smooth the image edges and noise, thus constructing a pyramid structure and forming a reduced version of the image set.

[0069] The pin initialization module performs binarization processing on a reduced-size image set, converting the image into black and white format, connecting touching and adjacent pins, and identifying all potential pin connectivity domains. Based on the potential pin connectivity domain identification results, the module sets filtering conditions using the bounding rectangle of the largest connectivity domain, selects and merges the connected domains that match the conditions, and creates the pin area initialization result.

[0070] The pin precise positioning module uses the initial pin region positioning results to mark the corresponding area in the upsampled image of the original chip image data size. Based on the corresponding area marking results, the module further screens the connected pin shapes according to size and shape standards, identifies the actual outline of each pin, and outputs the pin's accurate outline information.

[0071] The interference area exclusion module calculates the center position of each connected domain based on the accurate pin contour information and compares it with the preset center position. According to the comparison results, the connected domains with abnormal deviations are excluded, and the area including only pins is accurately screened to obtain accurate pin positioning results.

[0072] The reduced image set includes multi-resolution images and filtered image sequences; the initial pin area positioning results include the merged connected domain, pin connectivity analysis results and boundary screening parameters; the accurate pin contour information includes pin shape details, positioning accuracy and contour size; the accurate pin positioning results include pin position verification results, deviation analysis records and final connected domain identification records.

[0073] See also Figure 2 , the steps to obtain the reduced version of the image set are:

[0074] Collect the original image data of the chip and perform step-by-step downsampling, reducing the image resolution to half of the original size at each level. The formula is:

[0075]

[0076] Calculate the size S of the i-th level image i , generate difference resolution image data, where S i-1 Represents the size of the previous level image;

[0077] The initial image size S0 is collected to be 1024 pixels (width or height), and the downsampling process is simulated as follows:

[0078] First level downsampling:

[0079]

[0080] Second level downsampling:

[0081]

[0082] The third level of downsampling:

[0083]

[0084] The downsampling process continuously halves the image size until the desired resolution level is reached. Results show that by downsampling, the image data size can be effectively reduced while maintaining the integrity of the image structure, making it suitable for further image processing steps and generating image data of different resolutions.

[0085] For the difference resolution image data, Gaussian filtering is applied using the formula:

[0086]

[0087] Calculate the weight value G(x,y) of the Gaussian filter to obtain the Gaussian filtered image data, where x and y represent the pixel positions of the image and σ represents the standard deviation of the Gaussian filter;

[0088] Consider a 3×3 area, σ = 1.0, and the filter weights of the center pixel (0, 0) and the surrounding points are calculated as follows:

[0089] Center point weight:

[0090]

[0091] Adjacent point weights (such as (1,0)):

[0092]

[0093] Corner weight (such as (1,1)):

[0094]

[0095] Using the corresponding weights, a convolution operation is performed to smooth the pixel value of the center point. The calculated new pixel value is the result of weighting the original pixel value with the surrounding pixel values. The results show that Gaussian filtering reduces image noise and enhances the visual effect of the image by weighting neighboring pixels. It has a significant smoothing and noise reduction effect, generating Gaussian filtered image data.

[0096] Image data is processed based on Gaussian filtering, reordered and stacked from high to low resolution to construct an image pyramid and generate a reduced version of the image set;

[0097] In the process of constructing an image pyramid, the Gaussian filtered image dataset must first be carefully sorted and layered to ensure that the images in each layer are resized at an appropriate reduction ratio to maintain the consistency and integrity of the image content; then each reduced image layer is quality-preserving, including using interpolation methods to compensate for image details that may be lost due to reduced resolution, ensuring that each layer of images is visually seamlessly connected; the layered images are then stacked to form a structured pyramid, with each layer representing a lower resolution version of the original image. The structured stacking not only reduces the amount of data required to process subsequent tasks, but also maintains the efficiency and speed of image processing; finally, through such stacking and layering, the image pyramid formed allows for rapid access to image data at different levels, providing a convenient data structure for various image analysis and machine learning algorithms, and generating a reduced version of the image collection.

[0098] See also Figure 3 , the steps for identifying potential pin connectivity domains are:

[0099] Based on the reduced version of the image set, a single image is extracted, and each pixel in the image is converted into black and white according to the grayscale value to obtain a binary image set;

[0100] A single image is extracted from the reduced image set and binarized according to the grayscale threshold setting. This process determines the grayscale value of each pixel. If the grayscale value is greater than the set threshold, the pixel value is set to 1 (white), otherwise it is set to 0 (black). The threshold is determined based on the grayscale distribution analysis of the entire image. The optimal threshold is selected by statistically analyzing the histogram of the image, so that the main information of the image is retained, while maximizing the distinction between foreground and background, generating a clearly defined binary image set.

[0101] Based on the binary image set, the image dilation operator is applied to each image to increase the range of white pixels, connect the pins that are visually similar but not physically touching, and generate a preliminary pin connectivity area;

[0102] The image dilation operator is applied to process each image in the binary image set, mainly through the dilation operation in mathematical morphology. The dilation operation uses a structural element to increase the boundaries of objects in the image in order to connect pins that are visually close but physically not touching. The operation involves checking the neighborhood of each pixel and setting the positions of white pixels in the neighborhood to white, which can ensure that adjacent pins are visually merged together. Through this process, a preliminary pin connectivity area is generated, which clearly marks all pins in the image that are physically connected due to the dilation operation.

[0103] Based on the preliminary pin connectivity area, perform in-depth analysis on the preliminary pin connectivity area to identify all connected white blocks in the image, assign a unique identifier to each independent connected area, and establish a potential pin connectivity domain identification result;

[0104] To analyze the expanded image, use the formula:

[0105]

[0106] By scanning each pixel in the image and determining whether each pixel is connected to a white pixel, a connected pixel is marked as 1, and a disconnected pixel is marked as 0. This allows all connected white areas in the image to be identified. Each independent connected area is assigned a unique identifier, which is assigned based on line-by-line scanning and background comparison to ensure that each independent white area is correctly identified. This accurately establishes the identification results of potential pin connectivity domains, indicating the precise location and range of the pins in the image, providing important information for subsequent circuit analysis or fault diagnosis.

[0107] See also Figure 4 , the steps to obtain the initial positioning results of the pin area are:

[0108] Based on the potential pin connected domain identification results, perform boundary extraction on each connected domain using the formula:

[0109] A=(x max -x min )×(y max -y min )

[0110] Calculate the area of ​​the circumscribed rectangle A and obtain the circumscribed rectangle set, where x min , x max ,y min ,y max Represent the minimum and maximum coordinates of the connected domain in the x and y directions respectively;

[0111] The minimum value of the horizontal coordinate of a connected region is x min =3, the maximum value is x max =10, the minimum value of the vertical axis is y min =4, the maximum value is y max = 12. Substitute the values ​​into the formula to calculate the area of ​​the circumscribed rectangle:

[0112] A=(10-3)×(12-4)=7×8=56squnits

[0113] x max -x min The width of the connected domain is calculated, and y max -y minThe height is calculated. Multiplying these two lengths yields the 56 squares that give the area of ​​the connected domain's circumscribed rectangle. The area metric is used to determine the size of the connected domain relative to other connected domains. The results indicate that connected domains with this area are potentially significant enough to warrant further consideration in subsequent steps, such as filtering and merging.

[0114] According to the circumscribed rectangle set, set the area screening condition and use the formula:

[0115] S m =max(A i )

[0116] Filter and output the largest connected region S m , where A i is the area of ​​the circumscribed rectangle of the i-th connected domain;

[0117] Get the areas of three connected domains, A1 = 20squnits, A2 = 56squnits, and A3 = 15squnits, and then select the connected domain with the largest area, i.e. S m =max(20, 56, 15) = 56squnits. The result shows that the area of ​​the largest connected domain is 56 square units, which crucially points out the connected domain to be further processed and sets a benchmark for the merging operation.

[0118] Perform a merge operation on the maximum connected domain and check the adjacent connected domains. If the circumscribed rectangle intersects or is adjacent to the maximum connected domain, merge the domains to obtain the initial pin area positioning result.

[0119] After selecting the largest connected domain, the key step in the merge operation is to check whether other connected domains meet the merge conditions. The merge conditions are based on spatial position relationships, especially whether the bounding rectangles of the connected domains intersect or are adjacent. The specific operation includes calculating the coordinates of the bounding rectangle of each connected domain and determining whether there are overlapping or touching boundaries. If the bounding rectangles of two connected domains intersect or border in space, they are considered adjacent and a merge operation is performed. During the merging process, the newly formed connected domain will include the spatial range of all the original connected domains, that is, the adjacent connected domains are included in a larger single connected domain through the union operation. Ensuring that the final pin area has wider connectivity and larger area coverage provides a more solid foundation for subsequent fine operations and optimizes the accuracy of pin positioning.

[0120] See also Figure 5 , the calibration steps for the corresponding area are:

[0121] Based on the original chip image data, upsampling is performed to increase the pixel density and output upsampled image data;

[0122] The original chip image is upsampled to increase image resolution. This process uses bicubic interpolation, which constructs a cubic polynomial to insert new pixels around known pixels and calculates the values ​​of the interpolated points to achieve image magnification. This technology is particularly suitable for technical images with rich details and can maintain high image clarity and smoothness during the magnification process. In addition, bicubic interpolation calculates the weighted average of 16 surrounding pixels. The weight depends on the distance between the pixel and the target point, and closer pixels have higher weights. This can effectively reduce aliasing and blurring during image magnification, thereby obtaining the upsampled image.

[0123] Based on the initial positioning results of the pin area, the relative coordinates of the pin are converted into the coordinate system of the upsampled image and accurately mapped to obtain the mapped pin area;

[0124] Based on the initial positioning results of the pin area, the relative coordinates of the pin are converted to the coordinate system of the upsampled image, and affine transformation is applied to achieve accurate mapping. Affine transformation is a geometric transformation that maintains the linearity and parallelism of the image and is suitable for image rotation, scaling, and translation operations. First, the coordinate position of the pin is obtained according to the initial positioning results of the pin, and then the transformation matrix and displacement vector required for the affine transformation are calculated to ensure the accurate correspondence of the pin position in the high-resolution image, including adjusting the scaling ratio and rotation angle to adapt to different image resolutions, thereby generating the mapped pin area.

[0125] According to the mapped pin area, image segmentation is performed, pixel gradient information is analyzed to identify boundaries, the actual position of each pin is calibrated, and the corresponding area calibration result is generated;

[0126] In the mapped pin area, customized image segmentation is deployed to identify and accurately calibrate the specific location of each pin. The process involves the use of gradient-based edge detection technology. Edge detection determines the outline of objects in the image by analyzing the changes in color intensity between pixels. The Sobel operator is used to calculate the image gradient and highlight the boundaries of the pin area. The Sobel operator enhances the edge signal by performing horizontal and vertical convolution operations on the image. By setting an appropriate gradient threshold, areas with higher gradient values ​​are identified as edges. The threshold is determined based on an analysis of the overall brightness and contrast of the image to ensure that the pin boundaries are accurately identified, thereby generating a calibrated corresponding area.

[0127] See also Figure 6 , the steps to obtain accurate pin outline information are:

[0128] Based on the calibration results of the corresponding area, the formula is used:

[0129]

[0130] Calculate the characteristic size H of the region, where x i and y i are the coordinates of points in the region, and is the average value of the coordinates, N is the total number of points,

[0131]

[0132] and

[0133] θ=max(Δφ)

[0134] Calculate the edge curvature C and angle change rate θ, where κ i is the local curvature of each point, Δφ is the change in angle between adjacent points, and the pin shape that matches the standard is screened according to the preset shape standard and size threshold to generate a connected pin shape that has been screened by the size and shape standards;

[0135] An area contains N = 100 points, and the coordinates of each point are (x i ,y i ), calculate the average coordinates of all points The sum of the squares of the differences between the coordinates of each point and the average value is calculated to be 2500, and substituting it into the formula to calculate H is:

[0136]

[0137] Collect the local curvature κ of each point i The average value is 0.003, and the curvature formula is substituted to calculate C:

[0138]

[0139] If the maximum value of the maximum angle change Δφ between adjacent points is 45°, then θ=45°.

[0140] The relative calculation results reveal the geometric characteristics of the region, and when the screening criteria are met, the connectivity of the region and the shape of the pins can be confirmed.

[0141] For the connected pin shapes that have been screened by size and shape standards, the amplitude and direction of the image gradient are analyzed, the edges are determined and connected, the actual outline of each pin is identified, and the pin outline data is generated;

[0142] Based on the pin shape obtained by screening, the Canny edge detection algorithm is applied to refine the boundary. First, the image is smoothed with a Gaussian filter to eliminate image noise, maintain the basic structure of the image while removing unnecessary details and interference. Then, the gradient amplitude and direction of the image are analyzed to identify potential edge points. Areas with high gradient amplitude are identified as edge point candidates. Dual threshold technology is used to determine the authenticity of the edge. The higher threshold is used to connect the edge, while the lower threshold is used to suppress noise. It can accurately capture the outline of the pin and effectively separate it from the surrounding background. The true edge is confirmed and connected through edge tracking technology to ensure the complete recognition and accurate depiction of the pin shape. The process enhances the visual features of the pin in the image, making it more suitable for subsequent quality inspection and mechanical processing.

[0143] Integrate pin outline data, format and process it to match industry standard data files, record pin geometry and position information, and generate accurate pin outline information;

[0144] Integrate the identified pin contour information and format it into a data file that meets industry standards. In this step, detailed information such as the pin's geometric size, shape, and position data are converted into a universal data format that is easy for machine vision systems to parse, allowing automated equipment on the production line to directly use the data for precise pin assembly and quality control. The process involves multi-dimensional parameter measurement of the pin contour, including key dimensional information such as the length, width, and curvature of each pin, and encoding the parameters into a structured database or file. This database supports fast query and retrieval, so that accurate pin data can be obtained in real time at any step in the production process. The final output data file is not just a record, but a key asset that can directly affect production efficiency and product quality, ensuring the continuity of the entire production process and the consistency of technology.

[0145] See also Figure 7 , and the comparison steps with the preset center position are:

[0146] Based on the accurate pin contour information, the coordinates of all pixels in each connected domain are extracted to generate pixel coordinate extraction results;

[0147] Based on the accurate pin outline information, high-resolution scanning technology is first used to capture images of the object being measured. The captured images are then preprocessed using high-precision image processing techniques, including denoising, contrast enhancement, and edge detection. To ensure accurate pin outline extraction, every pixel in the image is analyzed to distinguish the pin from the background. Next, an algorithm based on the edge detection results is used to accurately calculate the boundary coordinates of each connected domain. This process involves traversing all pixels in the image and determining the geometric center of the edge pixels, thereby determining the geometric center of the entire connected domain. This process requires not only image processing technology but also powerful computing power to process large amounts of data and ensure that the center position of each connected domain is accurately calculated and recorded.

[0148] According to the pixel coordinate extraction results, the formula is used:

[0149]

[0150] Calculate the center position Z of the i-th connected component i , get the set of connected domain center positions, where (x m ,y m ) is the coordinate of the mth pixel, M is the total number of pixels in the connected component;

[0151] A connected domain has 3 pixels with coordinates (1, 2), (3, 4), and (5, 6), then N = 3. The center position Z is calculated according to the formula i for:

[0152]

[0153] It shows that the center of this connected domain is at coordinates (3, 4).

[0154] Based on the connected domain center position set, compare it with the preset center position, evaluate the deviation and verify the accuracy of the position, and output the comparison result;

[0155] In the process of comparing the calculated center position of each connected domain with the preset center position, it is necessary to first list the calculated center position of each connected domain as a series of coordinate points, and then compare the center position coordinates with the pin center position coordinates preset in the design. The comparison method is to calculate the Euclidean distance between the two points. In this calculation process, the distance formula is used. Here, x1, y1 and x2, y2 represent the coordinates of the connected domain center and the preset center, respectively. Each distance value is compared with a pre-set error threshold to determine whether the center position of each connected domain is within the allowable error range. The process also involves statistical analysis to assess the overall degree of deviation in the positions of all connected domains and determine whether manufacturing or design parameters need to be adjusted to ensure correct pin placement and quality control. This not only ensures the accuracy of the center position of each connected domain, but also provides data support for quality control throughout the entire production process.

[0156] See also Figure 8 , the steps to obtain accurate pin positioning results are:

[0157] Based on the comparison results, the connected domains whose deviations exceed the allowable range are identified, and a list of connected domains with deviations exceeding the allowable range is generated;

[0158] The previously obtained connected domain center position data is analyzed in detail. The analysis process includes comparing the actual center coordinates of each connected domain with the preset center coordinates, and calculating the position deviation of each connected domain. The comparison process uses advanced data analysis tools to process the data of each connected domain in turn, record the coordinate differences, and convert the differences into actual deviation values. The deviation value of each connected domain is then compared with the pre-set maximum tolerance deviation, which is set based on historical quality control data and product specification requirements. If the deviation value of a connected domain exceeds this threshold, it is identified as an anomaly and marked as an object that needs to be excluded. This screening process ensures that only those connected domains with deviations within an acceptable range are processed, providing a solid foundation for ensuring the accuracy and reliability of subsequent steps, and generating a list of connected domains with excessive deviations.

[0159] The list of connected domains with excessive deviations is excluded to ensure that only connected domains with deviations within the allowable range are retained, thus obtaining a set of filtered valid connected domains.

[0160] Connected domains marked as exceeding the deviation limit are screened and excluded, including a detailed review of the data for each marked connected domain to ensure that all connected domains exceeding the predetermined deviation threshold are completely removed from the dataset. This process is completed by an automated script that iterates through the entire connected domain dataset, checks the deviation flags for each connected domain, and excludes connected domains that meet the deletion criteria from further analysis. Only those connected domains with deviations within the allowable range are retained, ensuring the accuracy of the final analysis. This screening process not only optimizes data quality but also creates conditions for accurate pin positioning. The final set of valid connected domains includes only those connected domains that are accurately positioned and meet design standards, providing a solid foundation for ensuring product consistency and reliability.

[0161] According to the set of effective connected domains after screening, the formula is used:

[0162]

[0163] Calculate the average center position coordinates C of all valid connected regions final , generate accurate pin positioning results, where C valid,i Represents the center position of the i-th effective connected domain, and U is the total number of effective connected domains;

[0164] There are three valid connected domains, whose center positions are (1, 1), (2, 2) and (3, 3), then M = 3. The final pin positioning coordinate C is calculated according to the formula final for:

[0165]

[0166] The calculation results show that the averaged center position is at coordinate (2, 2), which is the center of all valid connected domains. This result not only demonstrates how to obtain accurate pin positions through averaging, but also reflects the effectiveness of this method in ensuring design alignment and quality control.

[0167] The above are merely preferred embodiments of the present invention and do not limit the present invention in any other form. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention shall still fall within the scope of protection of the technical solution of the present invention.

Claims

1. A QFP chip image pyramid positioning system, characterized in that: The system includes a pyramid construction module, a pin initial positioning module, a pin precise positioning module and an interference area exclusion module; The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling step by step, and applies Gaussian filtering to each level of the image to smooth the image edges and noise, thus constructing a pyramid structure and forming a reduced version of the image set. The pin initial positioning module performs binarization processing based on the reduced version image set, converts the image into black and white format, connects touching and adjacent pins, identifies all potential pin connected domains, sets filtering conditions based on the circumscribed rectangle of the largest connected domain based on the potential pin connected domain identification results, selects and merges the connected domains that match the conditions, and creates a pin area initial positioning result; The pin precise positioning module uses the initial pin area positioning results to mark the corresponding area in the upsampled image of the original chip image data size. Based on the corresponding area marking results, the module further screens the connected pin shapes according to size and shape standards, identifies the actual outline of each pin, and outputs the pin accurate outline information; The interference area exclusion module calculates the center position of each connected domain based on the accurate pin contour information and compares it with the preset center position. According to the comparison result, the connected domains with abnormal deviations are excluded, and the area including only pins is accurately screened to obtain accurate pin positioning results.

2. The QFP chip image pyramid positioning system according to claim 1, characterized in that: The steps for obtaining the reduced version image set are: Collect the original image data of the chip and perform step-by-step downsampling, reducing the image resolution to half of the original size at each level. The formula is: ； Calculate the Level image size , generating difference resolution image data, where Represents the size of the previous level image; Gaussian filtering is applied to the difference resolution image data using the formula: ； Calculate the weight values ​​of the Gaussian filter , get Gaussian filter processed image data, where, represents the pixel position of the image, represents the standard deviation of Gaussian filtering; The image data is processed based on the Gaussian filter, reordered and stacked from high to low resolution, and an image pyramid is constructed to generate a reduced version image set.

3. The QFP chip image pyramid positioning system according to claim 2, wherein: The steps for identifying the potential pin connectivity domain are: Extracting a single image based on the reduced version image set, converting each pixel in the image into black and white according to the grayscale value to obtain a binary image set; Based on the binary image set, an image dilation operator is applied to each image to process the image, thereby increasing the range of white pixels and connecting pins that are visually similar but not physically touching, thereby generating a preliminary pin connectivity area; Based on the preliminary pin connectivity area, a depth analysis is performed on the preliminary pin connectivity area to identify all connected white blocks in the image, a unique identifier is assigned to each independent connectivity area, and a potential pin connectivity domain identification result is established.

4. The QFP chip image pyramid positioning system according to claim 3, characterized in that: The steps for obtaining the initial positioning result of the pin area are: Based on the potential pin connected domain identification results, perform boundary extraction on each connected domain using the formula: ； Calculate the area of ​​the bounding rectangle , get the circumscribed rectangle set, where Represent the connected domains in and minimum and maximum coordinates in direction; According to the circumscribed rectangle set, the area screening condition is set using the formula: ； Filter and output the largest connected area ,in, For the The area of ​​the circumscribed rectangle of a connected domain; The maximum connected domain is merged, and adjacent connected domains are checked. If the circumscribed rectangle intersects and is adjacent to the maximum connected domain, a merge operation is performed to obtain an initial pin area positioning result.

5. The QFP chip image pyramid positioning system according to claim 4, characterized in that: The calibration steps of the corresponding area are: Performing upsampling processing based on the original chip image data to increase pixel density and output upsampled image data; Based on the initial positioning result of the pin area, the relative coordinates of the pin are converted into the coordinate system of the upsampled image, and accurately mapped to obtain the mapped pin area; According to the mapped pin area, image segmentation is performed, pixel gradient information is analyzed to identify boundaries, the actual position of each pin is calibrated, and a corresponding area calibration result is generated.

6. The QFP chip image pyramid positioning system according to claim 5, characterized in that: The steps for obtaining the accurate pin outline information are: Based on the corresponding area calibration results, the formula is used: ； Calculate the characteristic size of the area ,in and are the coordinates of points in the region, and is the mean value of the coordinates, is the total number of points, ； and ； Calculate edge curvature and the rate of change of angle ,in, is the local curvature at each point, It is the change of the angle of adjacent points. The pin shape that matches the standard is selected according to the preset shape standard and size threshold, and the connected pin shape that has been selected by the size and shape standards is generated; For the connected pin shapes that have been screened by size and shape standards, analyze the magnitude and direction of the image gradient, determine and connect the edges, identify the actual outline of each pin, and generate pin outline data; The pin outline data is integrated, formatted and matched to industry standard data files, the geometric dimensions and position information of the pins are recorded, and accurate pin outline information is generated.

7. The QFP chip image pyramid positioning system according to claim 6, characterized in that: The steps of comparing with the preset center position are: Based on the accurate pin contour information, the coordinates of all pixels in each connected domain are extracted to generate a pixel coordinate extraction result; According to the pixel point coordinate extraction result, the formula is used: ； Calculate the The center position of the connected domain , get the set of connected domain center positions, where It is The coordinates of the pixels, is the total number of pixels in the connected domain; Based on the connected domain center position set, the connected domain center position is compared with the preset center position, the deviation is evaluated and the accuracy of the position is verified, and the comparison result is output.

8. The QFP chip image pyramid positioning system according to claim 7, characterized in that: The steps for obtaining the accurate pin positioning result are: Based on the comparison results, identifying connected domains whose deviations exceed an allowable range, and generating a list of connected domains with excessive deviations; Eliminate the list of connected domains with deviations exceeding the standard, and only retain the connected domains with deviations within the allowable range, thereby obtaining a filtered set of valid connected domains; According to the set of effective connected domains after screening, the formula is adopted: ； Calculate the average center position coordinates of all valid connected regions , generating accurate pin positioning results, where Representative The center position of the effective connected domain, is the total number of valid connected domains.

Images