QFP chip image pyramid positioning system

Abstract

The invention relates to the technical field of image analysis, in particular to a QFP chip image pyramid positioning system which comprises a pyramid construction module, a pin initial positioning module, a pin accurate positioning module and an interference area elimination module. According to the method, an original chip image is processed through step-by-step down-sampling and Gaussian filtering, the image processing speed is effectively improved when a multi-layer image pyramid is constructed, a pin connected domain is clearly identified by applying binarization processing and a pin neighbor connection strategy, the pin area identification capability of the system is enhanced, and the identification accuracy of the system is improved. Region recalibration is carried out by using an up-sampling image of an original image data size, accurate pin positioning can be executed while high resolution is maintained, the probability of wrong positioning is remarkably reduced by accurately extracting pin contour information and effectively eliminating non-target connected domains, the chip assembling quality is improved, the chip assembling speed is increased, and the chip assembling efficiency is improved. And remarkable benefits are brought to a high-speed electronic assembly line.

Description

Technical Field

[0001] The 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 that involves extracting useful information from images. It includes multiple sub-fields such as image segmentation, edge detection, motion analysis, depth and shape recovery, and color feature determination. It can identify and classify objects in images, determine positions and sizes, and analyze motion and interaction information. Image analysis is often used in monitoring, medical imaging, remote sensing, industrial visual inspection and other fields, emphasizing the extraction of as much quantitative and qualitative information as possible from image data.

[0003] Among them, the QFP chip image pyramid positioning system mainly uses image pyramid technology to accurately 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 electronic assembly and quality inspection. Its main uses include automatically detecting the accuracy of chip placement, ensuring the assembly quality of electronic devices, and quickly identifying and correcting chip position errors on the production line, which plays an important role in improving production efficiency and product quality.

[0004] Existing technologies are obviously insufficient in image processing speed and accuracy, especially when it comes to accurately locating the tiny pins of QFP chips. Relying on a single-resolution image processing strategy, it is difficult to quickly and accurately locate chip pins in complex or low-contrast environments. In addition, traditional technologies require high computing resources when processing large amounts of data, which can easily form bottlenecks on the production line. The lack of efficient pin connectivity domain identification and error elimination mechanisms also increases the risk of mislocation, leading to reduced production efficiency and yield, which in turn affects the functionality and reliability of electronic equipment and increases production and maintenance costs. Summary of the invention

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

[0006] In order to achieve the above object, the present invention adopts the following technical solution: A QFP chip image pyramid positioning system comprises:

[0007] The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling level by level, and applies Gaussian filtering to each level of the image to smooth the image edge and noise, constructs a pyramid structure, and forms 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 the touching and adjacent pins, identifies all potential pin connected domains, sets screening conditions through the circumscribed rectangle of the largest connected domain according to the potential pin connected domain identification results, selects the connected domains that match the conditions and merges them, and creates the pin area initial positioning result;

[0009] The pin accurate positioning module uses the initial positioning result of the pin area to mark the corresponding area in the upsampled image of the original chip image data size, and according to the marking result of the corresponding area, screens the connected pin shapes again by size and shape standards, identifies the actual contour of each pin, and outputs the accurate contour information of the pin;

[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 domain with abnormal deviation is 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, using the formula:

[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 filter processed 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, an image pyramid is constructed, and a reduced version image set is generated.

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

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

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

[0021] According to 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, the boundary extraction is performed 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 domain 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 process is performed to obtain an initial pin region positioning result.

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

[0031] Based on the original chip image data, up-sampling processing is performed to increase pixel density and output up-sampled 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 adopted:

[0036]

[0037] Calculate the characteristic size H of the region, where x i and i are the coordinates of the 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 the angle change rate θ, where κ i is the local curvature of each point, Δφ is the change in angles of adjacent points, and the pin shapes matching the standards are screened according to the preset shape standards and size thresholds to generate connected pin shapes screened by size and shape standards;

[0041] For the connected pin shapes screened by the size and shape standards, the magnitude and direction of the image gradient are analyzed, the edges are determined and connected, the actual contour of each pin is identified, and the pin contour data is generated;

[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 domain i , and obtain 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 domain;

[0048] Based on the connected domain center position set, a comparison is made with a preset center position, the deviation is evaluated and the accuracy of the position is verified, and a 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 result, identifying the connected domains whose deviations exceed the allowable range, and generating a list of connected domains with excessive deviations;

[0051] Excluding the list of connected domains with deviations exceeding the standard, and determining to retain only connected domains with deviations within the allowable range, and obtaining a set of filtered 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 valid connected domain, and U is the total number of valid 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, the image processing speed is effectively improved when constructing a multi-level image pyramid, the pin connectivity domain is clearly identified by applying binarization processing and pin neighbor connection strategy, the pin area recognition capability of the system is enhanced, and the area is recalibrated by using an upsampled image of the original image data size, which can perform accurate pin positioning while maintaining high resolution, and the probability of erroneous positioning is significantly reduced by accurately extracting pin contour information and effectively excluding non-target connectivity domains, thereby improving the chip assembly quality and speed, and bringing significant benefits 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 of obtaining a reduced version image set of the present invention;

[0059] Figure 3 It is a flow chart for identifying potential pin connection 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 It is a flow chart of the calibration of the corresponding area of ​​the present invention;

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

[0063] Figure 7 is a flow chart of comparison between the present invention and a 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 solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with 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 indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

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

[0068] The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling level by level, and applies Gaussian filtering to each level of the image to smooth the image edge and noise, constructs a pyramid structure, and forms a reduced version of the image set;

[0069] The pin initial positioning module performs binarization processing based on the reduced version of the image set, converts the image into black and white format, connects the touching and adjacent pins, identifies all potential pin connectivity domains, sets the screening conditions through the circumscribed rectangle of the largest connectivity domain based on the potential pin connectivity domain identification results, selects the connectivity domains that match the conditions and merges them, and creates the pin area initial positioning results;

[0070] The pin precise positioning module uses the initial positioning results of the pin area to mark the corresponding area in the upsampled image of the original chip image data size. According to the marking results of the corresponding area, the connected pin shapes are screened again by size and shape standards, the actual contour of each pin is identified, and the accurate contour information of the pin is output;

[0071] The interference area exclusion module calculates the center position of each connected domain based on the accurate contour information of the pin, 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 the pins is accurately screened to obtain accurate pin positioning results.

[0072] The reduced version image set includes multi-resolution images and filtered image sequences; the initial positioning results of the pin area 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, using the formula:

[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 step-by-step 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 step-by-step downsampling process continuously halves the image size until the desired resolution level is reached. The results show that step-by-step downsampling can effectively reduce the amount of image data 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 filter processed 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 central pixel (0, 0) and the surrounding points are calculated as follows:

[0089] Center point weight:

[0090]

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

[0092]

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

[0094]

[0095] The corresponding weights are used to perform convolution operations to smooth the pixel value of the center point. The calculated new pixel value is the result of weighting the original pixel value and 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 and generates image data processed by Gaussian filtering.

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

[0097] In the process of constructing an image pyramid, the Gaussian filtered image data set 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; each reduced image layer is then subjected to quality preservation processing, including the use of 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 quick access to image data at different levels, provides a convenient data structure for various image analysis and machine learning algorithms, and generates a reduced version of the image collection.

[0098] See also Figure 3 , the steps to identify 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 gray value to obtain a binary image set;

[0100] A single image is extracted from the reduced version of the 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 the foreground and the background, generating a clearly defined binary image set.

[0101] Based on the binary image set, the image dilation operator is applied to process each image. By increasing the range of white pixels, the pins that are visually similar but not physically touching are connected to generate a preliminary pin connection 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 not physically 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 the 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, identify all connected white blocks in the image, assign a unique identifier to each independent connectivity area, and establish a potential pin connectivity domain identification result;

[0104] The expanded image is analyzed using the formula:

[0105]

[0106] By scanning each pixel in the image and determining whether each pixel is connected to the white pixel, if connected it is marked as 1, and if not connected it is marked as 0. All connected white blocks in the image can be identified. Each independent connected area is assigned a unique identifier. The assignment of identifiers is based on line-by-line scanning and background comparison to ensure that each independent white block is correctly identified. The identification results of the potential pin connectivity domain are accurately established. The results show the precise position 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 domain 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. The 56squnits obtained by multiplying these two lengths is the area of ​​the circumscribed rectangle of the connected domain. The area metric is used to judge the size of the connected domain relative to other connected domains. The results show that the connected domain with this area has enough potential importance to further consider its processing in subsequent steps, such as screening 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 domain S m , where A i is the area of ​​the circumscribed rectangle of the i-th connected domain;

[0117] The areas of the three connected domains are obtained as A1 = 20squnits, A2 = 56squnits, and A3 = 15squnits, and then the connected domain with the largest area is selected, 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] The maximum connected domain is merged, and the adjacent connected domains are checked. If the circumscribed rectangle intersects and is adjacent to the maximum connected domain, the merging operation is performed to obtain the initial positioning result of the pin area.

[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 a 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 of the corresponding area are:

[0121] Based on the original chip image data, up-sampling processing is performed to increase the pixel density and output up-sampled image data;

[0122] The original chip image is upsampled to improve the image resolution. This process uses the bicubic interpolation method to construct a cubic polynomial to insert new pixels around known pixels and calculate the value of the inserted 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 the surrounding 16 pixels. The weight depends on the distance between the pixel and the target point. The closer the pixel, the higher the weight. This can effectively reduce the aliasing and blurring problems during image magnification, thereby obtaining an upsampled image.

[0123] 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;

[0124] 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 the affine transformation is applied to achieve precise mapping. The 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 precise 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 a mapped pin area.

[0125] According to the mapped pin area, image segmentation is performed, the gradient information of pixels is analyzed to identify the boundary, 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 position 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 color intensity changes between pixels. The Sobel operator is used to calculate the image gradient and highlight the boundary of the pin area. The Sobel operator enhances the edge signal by performing horizontal and vertical convolution operations on the image. By setting a suitable gradient threshold, the area with higher gradient value is identified as the edge. The threshold is determined based on the analysis of the overall brightness and contrast of the image to ensure that the boundary of the pin is accurately identified, and then the corresponding calibration area is generated.

[0127] See also Figure 6 , the steps to obtain accurate pin contour 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 i are the coordinates of the 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 the angle change rate θ, where κ i is the local curvature of each point, Δφ is the change in angles of adjacent points, and the pin shapes matching the standards are screened according to the preset shape standards and size thresholds to generate connected pin shapes screened by size and shape standards;

[0135] A region 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 difference 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 of is 0.003, and substituting it into the curvature formula to calculate C is:

[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 magnitude and direction of the image gradient are analyzed, the edges are determined and connected, the actual contour of each pin is identified, and the pin contour data is generated;

[0142] Based on the pin shape obtained by screening, the Canny edge detection algorithm is used to refine the boundary. First, the image is smoothed by a Gaussian filter to eliminate image noise, maintain the basic structure of the image and remove 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. The 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. The outline of the pin can be accurately captured and effectively separated from the surrounding background. The real 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 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 encodes 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 contour information, we first use high-resolution scanning technology to collect images of the object to be tested, and then pre-process the collected images using high-precision image processing technology, including denoising, contrast enhancement, and edge detection. In order to ensure the accurate extraction of the pin contour, each pixel in the image is analyzed to distinguish the pin from the background. After that, an algorithm based on the edge detection results is used to accurately calculate the boundary coordinates of each connected domain. The process involves traversing all the pixels in the image, determining the geometric center of the edge pixels, and then determining the geometric center of the entire connected domain. The process requires not only the support of image processing technology, but also powerful computing power to process large amounts of data to ensure that each center position of the 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 domain i , and obtain 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 domain;

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

[0152]

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

[0154] Based on the connected domain center position set, compare 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 first necessary to 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 Among them, x1, y1 and x2, y2 represent the coordinates of the center of the connected domain and the preset center respectively. Each distance value is compared with the preset error threshold to determine whether the center position of each connected domain is within the allowable error range. In addition, the process also involves statistical analysis to evaluate the overall degree of deviation of the positions of all connected domains and determine whether the manufacturing or design parameters need to be adjusted to ensure the correct position and quality control of the pins, which not only ensures the accuracy of the center position of each connected domain, but also provides data support for the quality control of 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 excessive deviations is generated;

[0158] The previously obtained data on the center positions of connected domains 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 preset 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 is 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, thereby obtaining a set of filtered valid connected domains;

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

[0161] According to the screened effective connected domain set, 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 calculated according to the formula is 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. The results not only show how to obtain accurate pin positions through averaging, but also reflect the effectiveness of this method in ensuring design alignment and quality control.

[0167] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A QFP chip image pyramid positioning system, characterized in that: The system comprises: The pyramid construction module uses the original chip image data, reduces the image resolution by downsampling level by level, and applies Gaussian filtering to each level of the image to smooth the image edge and noise, constructs a pyramid structure, and forms 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 the touching and adjacent pins, identifies all potential pin connected domains, sets screening conditions through the circumscribed rectangle of the largest connected domain according to the potential pin connected domain identification results, selects the connected domains that match the conditions and merges them, and creates the pin area initial positioning result; The pin accurate positioning module uses the initial positioning result of the pin area to mark the corresponding area in the upsampled image of the original chip image data size, and according to the marking result of the corresponding area, screens the connected pin shapes again by size and shape standards, identifies the actual contour of each pin, and outputs the accurate contour information of the pin; 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 domain with abnormal deviation is 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 of 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, using the formula: 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: Calculate the weight value G(x,y) of the Gaussian filter to obtain the Gaussian filter processed image data, where x and y represent the pixel positions of the image and σ represents the standard deviation of the Gaussian filter; The image data is processed based on the Gaussian filter, reordered and stacked from high to low resolution, an image pyramid is constructed, and a reduced version image set is generated.

3. The QFP chip image pyramid positioning system according to claim 2, characterized in that: The steps of identifying the potential pin connectivity domain are: Based on the reduced version image set, extract a single image, convert each pixel in the image into black and white according to the grayscale value, and obtain a binary image set; Based on the binary image set, an image dilation operator is applied to process each image, and a preliminary pin connection area is generated by connecting pins that are visually similar but physically not in contact by increasing the range of white pixels; According to 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, the boundary extraction is performed on each connected domain using the formula: A=(x max -x min )×(and max -and min ) 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; According to the circumscribed rectangle set, the area screening condition is set using the formula: S m =max(A i ) Filter and output the largest connected domain S m , where A i is the area of ​​the circumscribed rectangle of the i-th 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 process is performed to obtain an initial pin region 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: Based on the original chip image data, up-sampling processing is performed to increase pixel density and output up-sampled 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, 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.

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 as follows: Based on the corresponding area calibration results, the formula is adopted: Calculate the characteristic size H of the region, where x i and i are the coordinates of the points in the region, and is the average value of the coordinates, N is the total number of points, and θ=max(Δφ) Calculate the edge curvature C and the angle change rate θ, where κ i is the local curvature of each point, Δφ is the change in angles of adjacent points, and the pin shapes matching the standards are screened according to the preset shape standards and size thresholds to generate connected pin shapes screened by size and shape standards; For the connected pin shapes screened by the size and shape standards, the magnitude and direction of the image gradient are analyzed, the edges are determined and connected, the actual contour of each pin is identified, and the pin contour data is generated; 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 center position Z of the i-th connected domain i , and obtain 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 domain; Based on the connected domain center position set, a comparison is made with a preset center position, the deviation is evaluated and the accuracy of the position is verified, and a 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 result, identifying the connected domains whose deviations exceed the allowable range, and generating a list of connected domains with excessive deviations; Excluding the list of connected domains with deviations exceeding the standard, and determining to retain only connected domains with deviations within the allowable range, and obtaining a set of filtered valid connected domains; According to the set of effective connected domains after screening, the formula is adopted: 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 valid connected domain, and U is the total number of valid connected domains.

Images