A high-precision BGA chip posture detection method and related device

Abstract

The present invention belongs to the fields of chip positioning technology and visual inspection technology, and discloses a high-precision BGA chip posture detection method and related device. The method includes: preprocessing the original grayscale image of the bottom package of the BGA chip to obtain a pin edge map; numbering the pins of the BGA chip on the pin edge map; calculating the centroid of each pin in the pin edge numbering map and analyzing the shape of the pin edge; extracting the convex hull vertex on the pin centroid point map after shape analysis; matching the pin centroid positioning point map with a pre-established first-class template point set to obtain a rough rotation angle of the BGA chip; and using the rough rotation angle based on the pre-established second-class template point set and the pin centroid point map to calculate the precise rotation angle of the BGA chip, thereby realizing BGA chip posture detection. The technical solution of the present invention can achieve fast and high-precision detection of BGA chip posture.

Description

Technical Field

[0001] The present invention belongs to the field of chip positioning technology and visual inspection technology, and in particular relates to a high-precision BGA chip posture detection method and related devices. Background Art

[0002] As integrated circuits continue to scale, chip packaging formats are becoming increasingly diverse and complex. IC chips are moving towards smaller sizes, finer pins, and tighter pin spacing. The quality of chip placement directly impacts the performance of electronic devices. BGA (ball grid array) packaged components, offering higher integration and superior performance, are increasingly used in electronic circuits. BGA-packaged chips utilize round solder joints instead of extended pins for their input and output terminals. BGA chips are crucial for component placement and are core components on electronic product PCBs. These small, powerful chips command a price significantly higher than single-function resistors and capacitors, as well as some simpler QFP and SOP chips. Therefore, accurate chip positioning after placement is essential to ensure the proper function of electronic products.

[0003] Traditionally, rectangle fitting and template matching techniques are used to inspect the position and posture of BGA components. However, these methods have limitations. The rectangle fitting method is favored for its simplicity and speed, but because it relies on fewer feature points, it may not guarantee detection accuracy and stability. This method may not be suitable for small BGA components or those with irregular arrangements. Template matching can provide more stable detection results to a certain extent, but it is more complex to operate, requires higher computing resources, and its detection accuracy is difficult to guarantee.

[0004] Currently, most BGA component pose detection methods rely on point set registration. However, this method, which uses feature point sets for registration, still suffers from the following issues: Positioning multi-pin chips often requires longer processing time, and obtaining stable feature points for spherical pins is difficult. For BGA components, using the centroid of the contour point set as the feature point set directly can lead to poor stability and accuracy. Extracting the center points of all pins as the matching point set consumes significant resources and time.

[0005] Specifically, such as Figure 1The image shows a raw grayscale image of the bottom package of a BGA chip. The pins used for identification appear as white circular patterns under a symmetrical bar light source, and all pins are arranged in a grid array. The simplest method is to use fitting to estimate pose information. However, when the component is small, the number of feature points available for fitting is small, and the stability and accuracy of the fitting results cannot be guaranteed. While using template matching to detect pose can improve the accuracy and stability of the detection results, it requires consideration of angular rotation and scaling in the X and Y directions. Template production can affect detection accuracy, and the algorithm has a high time complexity, which does not meet the requirements of the placement machine for fast positioning. If traditional feature point set registration is used, although it can improve the resource requirements of template matching to a certain extent, in actual applications, the numerous and large number of BGA pin arrangements make it difficult to obtain stable feature points, and the time-consuming feature point registration is also very difficult. It is also difficult to simultaneously meet the high precision and high efficiency requirements of the placement machine. Summary of the Invention

[0006] In order to overcome the above-mentioned deficiencies of the prior art, the purpose of the present invention is to provide a high-precision BGA chip posture detection method and related devices, which can realize rapid and high-precision detection of the BGA chip posture.

[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0008] A high-precision BGA chip posture detection method includes the following steps:

[0009] Get the original grayscale image of the bottom package of the BGA chip;

[0010] Preprocessing the original grayscale image to obtain a pin edge map;

[0011] Numbering the pins of the BGA chip on the pin edge diagram to obtain a pin edge numbering diagram;

[0012] Calculate the centroid of each pin in the pin edge numbering diagram and analyze the shape of the pin edge to obtain a pin centroid point diagram after shape analysis;

[0013] Extract the convex hull vertices on the pin centroid point diagram after shape analysis to obtain the pin centroid positioning point diagram;

[0014] Matching the pin centroid positioning point map with a pre-established first-class template point set to obtain a rough rotation angle of the BGA chip; the first-class template point set is the vertices of the convex hull of the point set consisting of the ideal pin circle centers;

[0015] According to the pre-established second-type template point set and pin centroid point diagram, the precise rotation angle of the BGA chip is calculated using the rough rotation angle, realizing the BGA chip posture detection. The second-type template point set is a point set composed of several pin circle centers randomly selected on each side of the outermost edge of the BGA chip.

[0016] Preferably, when the original grayscale image is preprocessed to obtain the pin edge map:

[0017] The original grayscale image is subjected to median filtering for noise reduction, and then the chip pins are segmented using OTSU threshold segmentation. The pin edges are then detected using the Canny edge operator to obtain a pin edge map.

[0018] Preferably, when the pins of the BGA chip are numbered on the pin edge diagram to obtain the pin edge numbering diagram:

[0019] The connected domain analysis method based on 8-neighborhood traversal is used to extract all pin edges and number them. When extracting pin edges, a threshold range of pin edge pixel count is set. The rules are as follows:

[0020] P std =round(2*π*R*P e )

[0021] Among them, P std The value is the pixel perimeter of each pin in pixel calibration, P e is the calibrated pixel equivalent, and R is the actual radius size of each pin of the BGA chip.

[0022] Preferably, the centroid of each pin is calculated in the pin edge numbering diagram and the shape of the pin edge is analyzed to obtain the pin centroid point diagram after shape analysis:

[0023] The weighted average method is used to obtain the centroid of each pin of the chip. While calculating the centroid of each pin edge, the shape of the pin edge is analyzed through principal component analysis to obtain the pin centroid point diagram after shape analysis.

[0024] Among them, when analyzing the shape of the pin edge through principal component analysis, the principal component analysis algorithm is used to perform data dimensionality reduction analysis on the numbered pin edges to obtain the first principal component eigenvector and the second principal component eigenvector. The ratio of the second principal component eigenvector to the first principal component eigenvector is set as the non-circularity measure P, the normal pin threshold range is set to 0.8-1, the pin reference mark threshold range is set to 0.45-0.55, and the defective pin threshold range is set to 0-0.45 and 0.55-0.8.

[0025] Preferably, when matching the pin centroid positioning point map with the pre-established first-class template point set to obtain the rough rotation angle of the BGA chip, convex hull vertices are extracted from the pin centroid point map and the point set, and the convex hull vertices and the first-class template point set are decomposed by SVD to obtain the rough rotation angle of the BGA chip, wherein when extracting the convex hull vertices from the pin centroid point map and the point set, the convex hull vertices are extracted by an improved convex hull detection algorithm, and the improved convex hull detection algorithm includes the following process:

[0026] Selection of the reference point: determine the point with the smallest ordinate value in the point set, then find the point with the smallest abscissa value in the neighborhood of the ordinate value, and update this point as the reference point;

[0027] Convex hull detection: Sort all points in the point set according to their polar angles and distances from the reference point and perform boundary judgment to obtain the convex hull;

[0028] Convex hull screening: Convex hull screening is performed based on the judgment condition of whether the angle between any point and two adjacent points is within the π neighborhood. If the angle is within the π neighborhood, the arbitrary point is excluded; if the angle is not within the π neighborhood, the arbitrary point is retained.

[0029] Preferably, when obtaining the precise rotation angle of the BGA chip using the rough rotation angle according to the pre-established second type template point set and the pin centroid point diagram:

[0030] After the second type of template point set is rotated according to the obtained rough rotation angle of the chip, it is matched with the pin centroid point map to obtain the pin centroid corresponding to the outermost periphery of the BGA chip. The centroid is replaced with the fitting circle center through the circle fitting algorithm. Finally, the second type of template point set is matched with the corresponding peripheral pin circle center point set and the ICP point set to obtain the final precise rotation angle of the chip.

[0031] Preferably, the first type of template point set and the second type of template point set are both selected according to the BGA chip data manual.

[0032] The present invention also provides a high-precision BGA chip posture detection system, comprising:

[0033] Image acquisition unit: used to acquire the original grayscale image of the bottom package of the BGA chip;

[0034] Preprocessing unit: used for preprocessing the original grayscale image to obtain a pin edge map;

[0035] Numbering unit: used for numbering the pins of the BGA chip on the pin edge diagram to obtain a pin edge numbering diagram;

[0036] The first calculation unit is used to calculate the centroid of each pin in the pin edge numbering diagram and analyze the shape of the pin edge to obtain a pin centroid point diagram after shape analysis;

[0037] Extraction unit: used to extract convex hull vertices on the pin centroid point diagram after shape analysis to obtain the pin centroid positioning point diagram;

[0038] Matching unit: used to match the pin centroid positioning point map with the pre-established first-type template point set to obtain the rough rotation angle of the BGA chip; the first-type template point set is the vertices of the convex hull of the point set composed of the ideal pin circle centers;

[0039] The second calculation unit is used to calculate the precise rotation angle of the BGA chip based on the pre-established second-type template point set and the pin center point diagram, using the rough rotation angle, thereby realizing the BGA chip posture detection. The second-type template point set is a point set composed of several pin circle centers randomly selected on each side of the outermost edge of the BGA chip.

[0040] The present invention also provides an electronic device, comprising:

[0041] one or more processors;

[0042] a storage device having one or more programs stored thereon;

[0043] When the one or more programs are executed by the one or more processors, the one or more processors implement the high-precision BGA chip posture detection method of the present invention as described above.

[0044] The present invention also provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the high-precision BGA chip posture detection method of the present invention as described above is implemented.

[0045] The present invention has the following beneficial effects:

[0046] In the present invention, two types of template point sets are pre-established for the BGA chip through the chip data manual, the pins are processed and equivalently converted into the pin centroid point set after the principal component analysis of the shape, and the centroid positioning point set composed of convex hull vertices is matched with the first type of template point set to quickly calculate the rough rotation angle of the chip; then the second type of template point set is rotated by a rough angle to obtain the peripheral corresponding pins in the centroid point set, and the corresponding pins that meet the shape analysis are circle-fitted to update the centroid to the center of the circle, and then the point set matching is performed with the second type of template point set to calculate the precise rotation angle of the chip, thereby realizing high-precision detection of the BGA chip posture. The method of the above process of the present invention has universal applicability to BGA chips with different pin arrangements, and has a fast detection speed and high accuracy, thereby overcoming the problems of long calculation time and difficulty in ensuring detection accuracy due to a large number of pins and different pin arrangements. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 The original grayscale image of the bottom package of the BGA chip obtained in an embodiment of the present invention;

[0049] Figure 2 Schematic diagram of the process of BGA chip posture detection according to an embodiment of the present invention;

[0050] Figure 3 The first type of template point set constructed according to the chip data manual in the embodiment of the present invention;

[0051] Figure 4 The second type of template point set constructed according to the chip data manual in the embodiment of the present invention;

[0052] Figure 5 The chip pin centroid positioning point and its pin edge corresponding map extracted in an embodiment of the present invention;

[0053] Figure 6 The extracted peripheral pin feature points and their corresponding pin outlines in the embodiment of the present invention; DETAILED DESCRIPTION

[0054] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present invention and is not intended to limit the present invention.

[0055] This embodiment uses Figure 1The components shown are objects, and this embodiment is described with respect to the problem of fast and high-precision posture detection of BGA components.

[0056] The first aspect of the embodiment of the present invention provides a high-precision BGA chip posture detection method, the method flow is as follows Figure 2 As shown, the following steps are included:

[0057] S1. Pre-establish two types of chip template point sets:

[0058] There are many pin arrangements for BGA chips, and the number of pins ranges from dozens to hundreds. Extracting the center of all pins as matching feature points will increase the time complexity of pose detection. Therefore, a point set matching method is constructed for pins at special positions on the BGA chip. The two pre-established template point sets are selected and established based on the BGA chip data sheet, such as Figure 3 , the first type of template point set is the convex hull vertices of the point set composed of the ideal pin circle centers; Figure 4 The second type of template point set is a set of points consisting of several pin circle centers randomly selected on each side of the chip's outermost periphery, see Figure 6 .

[0059] S2. Obtain the original grayscale image of the BGA chip, the chip pin segmentation map, and the pin edge map.

[0060] The original grayscale image of the bottom of the BGA chip to be inspected is collected by an industrial camera. Due to the presence of noise and the uneven grayscale value of the surface of the white metal spherical pin due to light and oxidation, the original grayscale image is obtained through a preprocessing step to obtain a pin edge image.

[0061] In the preprocessing operation, a 5*5 median filter operator is used for filtering and noise reduction. Then, the OTSU adaptive threshold segmentation algorithm is used to segment the chip pins from the filtered image. The 5*5 Canny edge operator is used to detect the pin edges. The first and second thresholds in edge detection are set to 50 and 120, respectively.

[0062] S3. Number all the pins of the chip on the pin edge diagram

[0063] The connected domain analysis method based on 8-neighborhood traversal is used to extract all pin edges and number the pins, resulting in a pin edge numbering map. To prevent interference from impurities and noise when extracting pin edges, a threshold range for the number of pin edge pixels is set, as follows:

[0064] P std =round(2*π*R*P e )

[0065] The pixel size of the pin in the image to be detected is calculated based on the pixel equivalent of the camera calibration and the actual size of the solder ball in the chip data sheet, P std The value is the pixel perimeter of each pin in pixel calibration, P e is the calibrated pixel equivalent, and R is the actual radius of each chip pin. This approach can achieve universal applicability for pin contour extraction in BGA position detection of different sizes or under different image acquisition conditions, and can also remove tiny noise points and impurity interference points.

[0066] S4. Replace the chip pins with the pin centroid point diagram

[0067] For each chip pin edge after numbering, the center of mass of each chip pin is calculated by weighted average method according to the serial number to obtain the pin center of mass point diagram. While calculating the center of mass of each pin edge, the shape of the pin edge is analyzed by principal component analysis.

[0068] Principal component analysis reduces the dimensionality of the data by calculating the covariance matrix of each contour point set. Solving the covariance matrix yields two eigenvectors: the first and second principal component eigenvectors. The largest eigenvalue is selected as the first principal component, and the relatively smaller eigenvalue as the second principal component. These two principal components capture the most significant variation in the contour point set, indicating that the data changes primarily occur in these two directions.

[0069] According to the characteristics of the principal component analysis algorithm mentioned above, if the two principal components of a triangle differ by more than twice, but if they are approximately circular, the two principal components are considered to be close (the ratio of the second principal component to the first principal component will be greater than 0.8 and less than 1). The pin shape is judged by the change in the two directions in which the pin contour point set changes most significantly in the data.

[0070] The ratio of the second principal component to the first principal component is set as the non-circularity metric P. The threshold range for normal pins is set to 0.8-1, the threshold range for triangular pin reference marks is set to 0.45-0.55, and the threshold range for defective pins is within other value ranges (i.e., 0-0.45 and 0.55-0.8). The centroid of each chip pin is calculated using the weighted average method of the points on each pin edge. This yields a set of chip pin centroid points, and all numbered pin edges are labeled according to the P value.

[0071] S5. Extract the convex hull vertices from the pin centroid to obtain the centroid positioning point map

[0072] Extract the convex hull vertices from the pin centroid point map and the point set, and decompose the convex hull vertices and the first type of template point set through SVD to obtain the rough rotation angle of the chip. The pin centroid point map replaces the chip pins, but the number of pin centroids is large. If all are extracted as feature points, it will not only reduce the accuracy of pin position replacement, but also increase the computational complexity of point set matching. Therefore, the pin centroid positioning point map that can represent the chip centroid point map is first extracted from the pin centroid point map. The purpose is to quickly obtain a rough shape angle of the chip. The convex hull vertices of the pin centroid point map are selected as the pin centroid positioning point map (see Figure 5 ), by using the improved convex hull detection algorithm Graham algorithm.

[0073] The Graham algorithm is an efficient algorithm for calculating the convex hull of a set of points. The convex hull is the smallest convex polygon of a set of points, that is, all points are located inside or on the boundary of the polygon, and all points are stored counterclockwise according to the polygon outside the point set. However, the traditional Graham algorithm is very simple in selecting reference points, and only considers the minimum value of the horizontal coordinate x or the vertical coordinate y of the point. However, in the case of slight angle offset, inappropriate reference points may be selected, which will affect the extraction of matching feature points; at the same time, the traditional Graham algorithm only considers whether it is a convex point when screening convex points, but for the more regularly arranged BGA pin centroid point set, there may be slight angle changes, and more points that are not positioning feature points will be extracted;

[0074] Based on the above two problems, the present invention improves the traditional Graham algorithm for the centroid of the BGA pin array. The improved convex hull detection algorithm includes the following steps:

[0075] Reference point selection: determine the point with the smallest y-value in the point set, then find the point with the smallest x-value in the neighborhood of the y-value, and update that point as the reference point.

[0076] Convex hull detection: Sort all points in the point set according to their polar angles and distances from the reference point and perform boundary judgment to obtain the convex hull;

[0077] Convex hull screening: Convex hull screening is performed based on the judgment condition of whether the angle between any point and two adjacent points is within the π neighborhood. If the angle is within the π neighborhood, the arbitrary point is excluded; if the angle is not within the π neighborhood, the arbitrary point is retained.

[0078] Specifically, in this embodiment, the improvement of reference point selection and the rules of convex hull screening are as follows:

[0079] p s =min{p i →y}&&min{p s →x}∈[p s →y-ε,ps →y+ε]

[0080] θ∈[π-δ,π]

[0081] Among them, p s is the coordinate of the reference point, p i is the coordinate of any point in the pin centroid set, ε is the half-width of the neighborhood selected by the marker point, which is set to 5 pixels in height; θ is the vector angle formed by any two points a and b in the salient point set and the standard point s, and δ is the angular error value of the salient point screening, which is set to 5°.

[0082] S6. Calculate the precise chip rotation angle based on the second type of template point set

[0083] Because the first type of template point set and the S5 pin centroid positioning point map use a small number of BGA chip pins and use the pin centroid (with poor position accuracy), this method can only obtain a rough chip rotation angle, so an additional step of precise angle detection is added.

[0084] After rotating the second-category template point set according to the rough chip rotation angle, it is matched with the pin centroid point map from S4 to obtain the pin centroid points corresponding to the second-category template point set (including the pin edge number and the label based on the P value from S4). Only normal pins meet the requirements. Then, a circle fitting algorithm is used to replace the original pin centroid with the fitting circle center. Finally, the second-category template point set is combined with the corresponding peripheral pin circle center point set through the ICP point set matching algorithm to obtain the final precise chip rotation angle.

[0085] In order to reasonably arrange the computational complexity, improve the pin position accuracy and eliminate defective pins, the circle fitting algorithm is adaptive iterative reweighted least squares, and the pin center of mass is replaced by the pin fitting circle center. The number of fitting iterations k is determined according to the non-circularity measure P. For defective pins, the smaller the P value, the more iterations k are, and the maximum k is 5. The weight function used is the Huber weight function w i (δ):

[0086]

[0087] Among them, the parameter γ is the clipping function, the median of the distance from the point participating in the fitting to the center of the fitting circle in the last cycle is divided by 0.675 and then multiplied by 2, |δ i | is the distance from point i to the center of the current fitted circle.

[0088] The weighted least squares problem for the kth iteration can be expressed as:

[0089] e i =(x i-a) 2 +(y i -b) 2 -r 2

[0090]

[0091] Among them, let the parameters of the circle be (a, b, r), (a, b) are the coordinates of the center of the circle, and r is the radius. i For the residuals of the data points, is the weight of the kth iteration.

[0092] Adaptive iterative circle fitting is only performed on the pin edges that correspond to the second type of template point set and meet the P value range, which reduces the computational complexity and improves the pin position accuracy, making the fitting circle more consistent with the actual situation.

[0093] As can be seen from the above scheme, this embodiment proposes a connected domain analysis algorithm based on principal component analysis for BGA chips with different pin arrangements. This algorithm can greatly reduce the interference of impurity noise and defective pins, thereby obtaining a more stable and realistic spherical pin outline. Furthermore, an adaptive iterative reweighted least squares circle fitting algorithm is proposed for high-precision matching feature point extraction of BGA pin arrays, which can stably obtain a high-precision circle center point set for peripheral pins. This overcomes the problems of BGA chip pose detection, such as the difficulty in obtaining stable feature points, the long calculation time caused by the large number of pins, and the difficulty in ensuring detection accuracy.

[0094] The present invention also provides a high-precision BGA chip posture detection system based on the above-mentioned method, which includes:

[0095] Image acquisition unit: used to acquire the original grayscale image of the bottom package of the BGA chip;

[0096] Preprocessing unit: used for preprocessing the original grayscale image to obtain a pin edge map;

[0097] Numbering unit: used for numbering the pins of the BGA chip on the pin edge diagram to obtain a pin edge numbering diagram;

[0098] The first calculation unit is used to calculate the centroid of each pin in the pin edge numbering diagram and analyze the shape of the pin edge to obtain a pin centroid point diagram after shape analysis;

[0099] Extraction unit: used to extract convex hull vertices on the pin centroid point diagram after shape analysis to obtain the pin centroid positioning point diagram;

[0100] Matching unit: used to match the pin centroid positioning point map with the pre-established first-class template point set to obtain the rough rotation angle of the BGA chip;

[0101] The second calculation unit is used to calculate the precise rotation angle of the BGA chip based on the pre-established second type template point set and the pin centroid point map, using the rough rotation angle, thereby realizing the BGA chip posture detection.

[0102] The embodiments of the present invention also provide corresponding electronic devices and computer-readable storage media for implementing the solutions provided by the embodiments of the present invention.

[0103] The device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes so that the device executes the high-precision BGA chip posture detection method described in any embodiment of the present application.

[0104] The storage medium stores a computer program, wherein when the computer program is executed by the processor, the high-precision BGA chip posture detection method described in any embodiment of the present application is implemented.

[0105] The preferred embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited thereto. After considering the specification and practicing the invention disclosed herein, those skilled in the art will easily think of other embodiments of the present invention. Within the scope of the technical concept of the present invention, the technical solution of the present invention can be subjected to a variety of simple modifications, including combining various specific technical features in any appropriate manner. In order to avoid unnecessary repetition, the present invention will not further describe various possible combinations. However, these simple modifications and combinations should also be regarded as the contents disclosed by the present invention and belong to the scope of protection of the present invention.

Claims

1. A high-precision BGA chip posture detection method, characterized in that: The process includes the following: Get the original grayscale image of the bottom package of the BGA chip; Preprocessing the original grayscale image to obtain a pin edge map; Numbering the pins of the BGA chip on the pin edge diagram to obtain a pin edge numbering diagram; Calculate the centroid of each pin in the pin edge numbering diagram and analyze the shape of the pin edge to obtain a pin centroid point diagram after shape analysis; Extract the convex hull vertices on the pin centroid point diagram after shape analysis to obtain the pin centroid positioning point diagram; Match the pin centroid positioning point map with the pre-established first-category template point set to obtain the rough rotation angle of the BGA chip; The first type of template point set is the vertices of the convex hull of the point set consisting of the centers of the ideal pin circles; According to the pre-established second-type template point set and pin centroid point diagram, the precise rotation angle of the BGA chip is calculated using the rough rotation angle, realizing the BGA chip posture detection. The second-type template point set is a point set composed of several pin circle centers randomly selected on each side of the outermost edge of the BGA chip.

2. A high-precision BGA chip posture detection method according to claim 1, characterized in that: When the original grayscale image is preprocessed to obtain a pin edge map: The original grayscale image is subjected to median filtering for noise reduction, and then the chip pins are segmented using OTSU threshold segmentation. The pin edges are then detected using the Canny edge operator to obtain a pin edge map.

3. A high-precision BGA chip posture detection method according to claim 1, characterized in that: When the pins of the BGA chip are numbered on the pin edge diagram to obtain a pin edge numbering diagram: The connected domain analysis method based on 8-neighborhood traversal is used to extract all pin edges and number them. When extracting pin edges, a threshold range of pin edge pixel count is set. The rules are as follows: P std =round(2*π*R*P e ) Among them, P std The value is the pixel perimeter of each pin in pixel calibration, P e is the calibrated pixel equivalent, and R is the actual radius size of each pin of the BGA chip.

4. A high-precision BGA chip posture detection method according to claim 1, characterized in that: When calculating the centroid of each pin in the pin edge numbering diagram and analyzing the shape of the pin edge, and obtaining the pin centroid point diagram after shape analysis: The weighted average method is used to obtain the centroid of each pin of the chip. While calculating the centroid of each pin edge, the shape of the pin edge is analyzed through principal component analysis to obtain the pin centroid point diagram after shape analysis. Among them, when analyzing the shape of the pin edge through principal component analysis, the principal component analysis algorithm is used to perform data dimensionality reduction analysis on the numbered pin edges to obtain the first principal component eigenvector and the second principal component eigenvector. The ratio of the second principal component eigenvector to the first principal component eigenvector is set as the non-circularity measure P, the normal pin threshold range is set to 0.8-1, the pin reference mark threshold range is set to 0.45-0.55, and the defective pin threshold range is set to 0-0.45 and 0.55-0.

8.

5. A high-precision BGA chip posture detection method according to claim 1, characterized in that: The pin centroid positioning point map is matched with the pre-established first-class template point set to obtain a rough rotation angle of the BGA chip. The convex hull vertices are extracted from the pin centroid point map and the point set. The convex hull vertices and the first-class template point set are decomposed by SVD to obtain a rough rotation angle of the BGA chip. When the convex hull vertices are extracted from the pin centroid point map and the point set, the convex hull vertices are extracted by an improved convex hull detection algorithm. The improved convex hull detection algorithm includes the following steps: Selection of the reference point: determine the point with the smallest ordinate value in the point set, then find the point with the smallest abscissa value in the neighborhood of the ordinate value, and update this point as the reference point; Convex hull detection: Sort all points in the point set according to their polar angles and distances from the reference point and perform boundary judgment to obtain the convex hull; Convex hull screening: Convex hull screening is performed based on the judgment condition of whether the angle between any point and two adjacent points is within the π neighborhood. If the angle is within the π neighborhood, the arbitrary point is excluded; if the angle is not within the π neighborhood, the arbitrary point is retained.

6. A high-precision BGA chip posture detection method according to claim 1, characterized in that: According to the pre-established second type template point set and pin centroid point diagram, the precise rotation angle of the BGA chip is obtained by using the rough rotation angle: After the second type of template point set is rotated according to the obtained rough rotation angle of the chip, it is matched with the pin centroid point map to obtain the pin centroid corresponding to the outermost periphery of the BGA chip. The centroid is replaced with the fitting circle center through the circle fitting algorithm. Finally, the second type of template point set is matched with the corresponding peripheral pin circle center point set and the ICP point set to obtain the final precise rotation angle of the chip.

7. A high-precision BGA chip posture detection method according to claim 1, characterized in that: The first type of template point set and the second type of template point set are both selected according to the BGA chip data manual.

8. A high-precision BGA chip posture detection system, characterized in that: include: Image acquisition unit: used to acquire the original grayscale image of the bottom package of the BGA chip; Preprocessing unit: used for preprocessing the original grayscale image to obtain a pin edge map; Numbering unit: used for numbering the pins of the BGA chip on the pin edge diagram to obtain a pin edge numbering diagram; The first calculation unit is used to calculate the centroid of each pin in the pin edge numbering diagram and analyze the shape of the pin edge to obtain a pin centroid point diagram after shape analysis; Extraction unit: used to extract convex hull vertices on the pin centroid point diagram after shape analysis to obtain the pin centroid positioning point diagram; Matching unit: used to match the pin centroid positioning point map with the pre-established first-class template point set to obtain the rough rotation angle of the BGA chip; The first type of template point set is the vertices of the convex hull of the point set consisting of the centers of the ideal pin circles; The second calculation unit is used to calculate the precise rotation angle of the BGA chip based on the pre-established second-type template point set and the pin center point diagram, using the rough rotation angle, thereby realizing the BGA chip posture detection. The second-type template point set is a point set composed of several pin circle centers randomly selected on each side of the outermost edge of the BGA chip.

9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the high-precision BGA chip posture detection method as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the high-precision BGA chip posture detection method according to any one of claims 1 to 7 is implemented.

Images