WO2017020559A1 - 一种基于行列直线聚类的多类型bga芯片视觉识别方法 - Google Patents

一种基于行列直线聚类的多类型bga芯片视觉识别方法 Download PDF

Info

Publication number
WO2017020559A1
WO2017020559A1 PCT/CN2016/071119 CN2016071119W WO2017020559A1 WO 2017020559 A1 WO2017020559 A1 WO 2017020559A1 CN 2016071119 W CN2016071119 W CN 2016071119W WO 2017020559 A1 WO2017020559 A1 WO 2017020559A1
Authority
WO
WIPO (PCT)
Prior art keywords
solder ball
bga solder
bga
equivalent
column
Prior art date
Application number
PCT/CN2016/071119
Other languages
English (en)
French (fr)
Inventor
高会军
靳万鑫
杨宪强
于金泳
孙光辉
林伟阳
李湛
Original Assignee
哈尔滨工业大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 哈尔滨工业大学 filed Critical 哈尔滨工业大学
Priority to US15/323,100 priority Critical patent/US9965847B2/en
Priority to JP2016574415A priority patent/JP6598162B2/ja
Publication of WO2017020559A1 publication Critical patent/WO2017020559A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/001—Industrial image inspection using an image reference approach
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/64—Analysis of geometric attributes of convexity or concavity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30141—Printed circuit board [PCB]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30148—Semiconductor; IC; Wafer
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30152—Solder

Definitions

  • the invention relates to a multi-type BGA chip visual recognition method, in particular to a multi-type BGA chip visual recognition method based on row and column linear clustering.
  • BGA (BallGrayArray) packaged chips are widely used in integrated circuits due to their high integration, I/O leads and excellent electrical properties. Due to this feature, BGA chips are being manufactured. It is more prone to defects such as loss of solder balls, excessive or too small solder balls, and displacement of solder balls.
  • SMT printed circuit board automated placement
  • BGA chip high-density I / O solder ball pins and multiple types of solder ball arrangement structure put more stringent requirements on the accuracy and speed of chip identification and detection.
  • the automatic optical inspection (AOI) system is often used to identify, detect and locate chips in the placement process.
  • Chip identification is the basis of placement.
  • AOI optical inspection
  • a standard database of this type of chip is built in the AOI system to provide a standard data reference for the detection and positioning process.
  • the above working process of the AOI system includes two stages of offline training and online inspection:
  • the standard database for BGA chips includes the solder ball distribution matrix (characterizing the distribution pattern of the solder balls), the standard area and roundness of the solder balls, the standard line spacing of the solder balls, and the standard column spacing of the solder balls.
  • the standard parameters of the chip are mostly performed by manual measurement and manual input, so the workload is large, the labor cost is high, and the degree of automation is low.
  • Detection image processing and data extraction of the image of the BGA chip to be mounted, and using the BGA chip standard database as a reference to perform defect detection and precise positioning on the mounted chip.
  • the defects of the BGA chip mainly include the following aspects: missing solder balls, unqualified solder ball diameter or roundness, and adjacent solder ball bridging.
  • the BGA chip has many solder ball pins, small pitch and various solder ball arrangement forms, which puts higher requirements on the reliability and detection speed of the detection and positioning algorithm.
  • the basic steps of the BGA chip in the detection phase are image acquisition, solder ball extraction, solder ball array positioning, and solder ball parameter detection.
  • the solder ball extraction is designed to extract BGA solder balls by image segmentation algorithm.
  • the traditional solder ball extraction method generally uses the global threshold to binarize the image: for example, the literature "Analysis Ball Grid Array defects by using new image technique” uses all pixels.
  • the binarization threshold is calculated by the mean and variance of the gray scale;
  • the A system for automated BGA Inspection uses the Otsu algorithm of adaptive image gray to solve this global threshold;
  • the literature "Automated detection and classification of non-wet" Solder joints” uses a graph based An iterative calculation method like pixel gray value statistics to find this threshold, and so on.
  • the conventional method separates the obtained solder balls. Therefore, when the BGA image collected by the AOI system is unevenly distributed, the above-mentioned binarization method will cause some solder balls to be “over-segmented” and “under-segmented”. Finally, the detection result is wrong; at the same time, the noise introduced by image binarization may also be mistaken for the solder ball, which will interfere with the subsequent positioning of the solder ball array.
  • the ball grid array positioning is intended to determine the angle of rotation of the solder ball array and the position of the center in the image coordinate system.
  • Existing more mature methods (such as the Halcon machine vision algorithm library) generally use a template matching strategy.
  • the standard solder ball array template is first constructed according to the BGA standard database, and then the traversal and iteration are performed according to the adjacent positional relationship of the solder balls to determine the corresponding transformation relationship between the template solder ball array and the BGA chip solder ball array to be tested. .
  • This method involves a large number of iterations and traversals, high algorithm complexity, and long detection time for BGA chips with large-scale solder ball pins, so it is difficult to apply to AOI systems with high real-time requirements.
  • solder ball array positioning algorithm based on the least squares rectangle method can only be applied to a limited number of BGA chips for solder ball arrangement.
  • the BGA chip with sparse solder ball arrangement or irregular solder ball arrangement has poor applicability.
  • the object of the present invention is to solve the following problems of the existing BGA chip identification or detection method: 1) the recognition range is narrow, and only the BGA chip of a specific type or a limited number of solder ball arrangements can be identified, and the solder ball arrangement is sparse.
  • the BGA chip has poor applicability or is not recognized at all; 2) the robustness to illumination and interference is poor, using the global threshold
  • the value method is used for the extraction of the solder ball.
  • the robustness of the non-uniform illumination image is poor, which may lead to under-segmentation or over-segmentation of the solder ball, which ultimately leads to detection errors.
  • the interference introduced by the method will affect the accuracy of subsequent chip positioning.
  • Adopt The template matching algorithm has low flexibility.
  • a standard array template is needed for each BGA chip, and the interference introduced by the solder ball segmentation is poor. 4)
  • the traditional algorithm has high time complexity, and in most cases, the recognition process requires manual BGA.
  • the chip parameters are measured and recorded.
  • a visual recognition method based on row and column linear clustering is proposed for multi-type BGA chips.
  • Step 1 The image of the gray BGA chip collected by the camera is the original image, and the dynamic threshold segmentation is performed to obtain the binary solder ball image, and the morphological opening operation and the closed operation processing are performed on the binary solder ball image.
  • Each connected field on the binary solder ball image is recorded as a binarized BGA solder ball, and then each connected BGA solder ball is connected to the connected domain;
  • Step 2 Extract each of the binarized BGA solder balls obtained by the connected domain obtained in the first step in the corresponding neighborhood of the original image to obtain a full grayscale BGA solder ball, and establish a complete gray.
  • Degree BGA solder ball information list
  • the complete grayscale BGA solder ball information list includes: grayscale pixels included in each complete grayscale BGA solder ball, and the center point position of each complete grayscale BGA solder ball calculated by grayscale pixels. Coordinates, the minimum outer circle diameter corresponding to each full grayscale BGA solder ball, the perimeter and roundness of each full grayscale BGA solder ball; the grayscale pixels included in each solder ball include pixel coordinates and gray values;
  • Step 3 Using the complete grayscale BGA solder ball information list obtained in step 2, create a background image in which the pixel gray values are all 0 and the same size as the original image; and in the background image, each of the original images is correspondingly complete.
  • the gray value at the center point of the gray BGA solder ball becomes the identification number corresponding to the binarized BGA solder ball.
  • the background image at this time is recorded as the BGA solder ball identification image, and each non-BGA solder ball logo image A pixel of 0 gray value is called an equivalent BGA solder ball, and an array of all equivalent BGA solder balls is called an equivalent BGA array;
  • Step 4 using the equivalent BGA solder ball pitch typical value ⁇ obtained in the third step, performing a local analysis on the BGA solder ball identification image to determine the equivalent BGA array coarse deflection angle ⁇ ;
  • the local analysis is divided into a partial analysis for a regular BGA chip and a local analysis for an irregular BGA chip.
  • the regular BGA chip is a BGA chip in which adjacent BGA solder balls are arranged neatly, and the irregular BGA chip is a phase. Adjacent row BGA solder balls into staggered BGA chips;
  • Step 5 using the equivalent BGA ball pitch typical value ⁇ obtained in the third step and the equivalent BGA array rough deflection angle ⁇ obtained in the step 4, performing the equivalent BGA solder balls on the row and column on the BGA solder ball identification image. Straight line clustering to obtain equivalent BGA solder balls for each row, equivalent BGA solder balls for each column, and boundary equivalent BGA solder balls;
  • Step 6 Using the gray value of each equivalent BGA solder ball in the boundary equivalent BGA solder ball obtained in step 5, find the corresponding full gray BGA in the complete gray BGA solder ball information list.
  • the coordinates of the center point position of the solder ball are based on the coordinates of the center point of the complete gray BGA solder ball, and the boundary straight line is fitted, and the deflection angle and the center position of the BGA chip in the original image are solved by the boundary fitting straight line;
  • Step 7 using each row of equivalent BGA solder balls and each column of equivalent BGA solder balls ⁇ identified in step 5, and the complete grayscale BGA solder ball information list obtained in step 2 for each line of equivalent BGA solder balls Fit and the equivalent BGA solder ball ⁇ straight line fit for each column; the average of the fitted line spacing of all adjacent two rows of equivalent BGA solder balls ⁇ is taken as the standard line spacing of the BGA chip solder balls; all adjacent columns will be The average value of the line spacing of the equivalent BGA solder balls is taken as the standard column spacing of the BGA chip solder balls;
  • Step 8 Using the equivalent BGA solder ball ⁇ fitting line obtained in step 7 and the equivalent BGA solder ball ⁇ fitting line of each column, the equivalent BGA solder ball is performed line by line or on the BGA solder ball identification image. Column search, and then get the BGA solder ball distribution matrix; according to the row-by-row or column-by-column search for all equivalent BGA solder balls and the complete gray-scale BGA solder ball information list obtained in step 2 to obtain the BGA solder ball standard diameter, BGA solder balls Standard perimeter and BGA solder ball standard roundness.
  • the BGA chip visual recognition method proposed by the invention does not need to establish a matching template in the process of multi-type BGA chip identification, and has strong flexibility.
  • the method can be used to identify the standard BGA chip samples and establish a standard database.
  • the identification method can be used to extract the parameters of the BGA chip to be detected and detect it, and obtain the BGA.
  • the deflection angle and center position of the solder ball array can be used to extract the parameters of the BGA chip to be detected and detect it, and obtain the BGA.
  • the deflection angle and center position of the solder ball array The method of the invention comprises a solder ball extraction method based on the gray connected domain, which can ensure the integrity of the BGA solder ball extracted under non-uniform illumination to the greatest extent compared with the traditional binary extraction algorithm. In order to ensure the accuracy of the recognition results.
  • the comparison results of the two solder ball extraction methods are shown in Fig. 6 and Fig. 7; 3)
  • the multi-type BGA chip visual recognition method based on row and column linear clustering can identify most BGA chips: regular BGA chip, irregular Type BGA chip, solder ball arrangement sparse BGA chip or other solder ball distribution type BGA chip.
  • the identification method can automatically identify and filter the interference introduced by the solder ball extraction (which is easily mistaken for solder balls), and improve the robustness of the algorithm to background interference. Sexuality ensures the accuracy and accuracy of BGA solder ball array positioning.
  • Figures 8 and 9 representatively show the non-bump ball interference that the algorithm can recognize.
  • the recognition algorithm proposed by the invention can be robust to non-uniform illumination of the acquired image and adaptive to non-ideal illumination intensity.
  • the chip to be tested is shown in Figure 10.
  • the resolution of the acquired image is 840*1160.
  • the effect of different illumination intensity on the running time and detection accuracy of the algorithm is shown in Figure 11;
  • the identification method based on line and column linear clustering proposed by the invention has low computational complexity.
  • the illumination condition is ideal, the resolution of the acquired image is 840*1160.
  • the running time comparison between a commercial algorithm and the recognition algorithm proposed by the present invention is shown in Fig. 12. :(PC platform reference: B960.2.2 GHz dual-core, C++) 7)
  • the identification method based on the line and column linear clustering proposed by the present invention can also be applied to the identification and detection of graphic arrays with different distribution patterns.
  • FIG. 1 is a flow chart of a method for visual recognition of a multi-type BGA chip based on row-column linear clustering according to a first embodiment
  • FIG. 2 is a diagram showing an example of performing dynamic threshold segmentation on an original image, and a morphological opening operation and a closed operation processing procedure according to the first embodiment
  • FIG. 3(a) is a diagram showing an example of local analysis of a regular BGA chip according to Embodiment 3;
  • FIG. 3(b) is a diagram showing an example of local analysis of an irregular type BGA chip according to Embodiment 4;
  • FIG. 4 is a diagram showing an example of linear clustering of row and column BGA solder balls according to a fifth embodiment, wherein Indicates the intercept of the line equation of the equivalent BGA solder ball corresponding to the jth equivalent BGA solder ball coordinate, The intercept representing the linear equation of the equivalent BGA solder ball column corresponding to the jth equivalent BGA solder ball coordinate, The intercept of the line equation of the equivalent BGA solder ball corresponding to the i-th equivalent BGA solder ball coordinate, The intercept representing the linear equation of the equivalent BGA solder ball column corresponding to the i-th equivalent BGA solder ball coordinate, Indicates the intercept of the linear equation of the equivalent BGA solder ball column corresponding to the interference;
  • FIG. 5 is a view showing an example of a BGA chip in which the solder balls are arranged sparsely;
  • Figure 8 is a representative representation of the non-solder ball interference in the upper left corner that the algorithm can recognize
  • Figure 9 is a representative representation of the non-solder ball interference in the lower left corner that the algorithm can recognize
  • Figure 10 is an image of a BGA chip to be detected
  • FIG. 11 is a graph showing effects of different illumination intensities on algorithm running time and detection accuracy for the BGA chip to be tested of FIG. 10; FIG.
  • Figure 12 is a comparison of the running time of a commercial algorithm and the recognition algorithm proposed by the present invention.
  • Embodiment 1 A visual recognition method for a multi-type BGA chip based on row-column linear clustering in the embodiment is specifically implemented according to the following steps:
  • Step 1 The image of the gray BGA chip collected by the camera is the original image, and the dynamic threshold segmentation is performed to obtain the binary solder ball image, and the morphological opening operation and the closed operation processing are performed on the binary solder ball image.
  • Each connected field on the binary solder ball image is recorded as a binarized BGA solder ball, and then each connected BGA solder ball is connected to the connected domain;
  • f(x, y) is the original image
  • g(x, y) is the average filtered image of f(x, y).
  • C is a preset constant.
  • Step 2 Extract each of the binarized BGA solder balls obtained by the connected domain obtained in the first step in the corresponding neighborhood of the original image to obtain a full grayscale BGA solder ball, and establish a complete gray.
  • Degree BGA solder ball information list
  • the complete grayscale BGA solder ball information list includes: grayscale pixels included in each complete grayscale BGA solder ball, and the center point position of each complete grayscale BGA solder ball calculated by grayscale pixels. Coordinates, the minimum outer circle diameter for each full grayscale BGA solder ball, the perimeter of each full grayscale BGA solder ball, and the roundness of each full grayscale BGA solder ball; each full grayscale BGA solder ball contains Grayscale pixels include pixel coordinates and grayscale values;
  • Step 3 Using the complete grayscale BGA solder ball information list obtained in step 2, create a background image in which the pixel gray values are all 0 and the same size as the original image; and in the background image, each of the original images is correspondingly complete.
  • the gray value at the center point of the gray BGA solder ball becomes the identification number corresponding to the binarized BGA solder ball.
  • the background image at this time is recorded as the BGA solder ball identification image, and each non-BGA solder ball logo image A pixel of 0 gray value is called an equivalent BGA solder ball, and an array of all equivalent BGA solder balls is called an equivalent BGA array;
  • Step 4 using the equivalent BGA solder ball pitch typical value ⁇ obtained in the third step, performing a local analysis on the BGA solder ball identification image to determine the equivalent BGA array coarse deflection angle ⁇ ;
  • the local analysis is divided into a partial analysis for a regular BGA chip and a local analysis for an irregular BGA chip.
  • the regular BGA chip is a BGA chip in which adjacent BGA solder balls are arranged neatly, and the irregular BGA chip is a phase. Adjacent row BGA solder balls into staggered BGA chips;
  • Step 5 using the equivalent BGA ball pitch typical value ⁇ obtained in the third step and the equivalent BGA array rough deflection angle ⁇ obtained in the step 4, performing the equivalent BGA solder balls on the row and column on the BGA solder ball identification image. Straight line clustering to obtain equivalent BGA solder balls for each row, equivalent BGA solder balls for each column, and boundary equivalent BGA solder balls;
  • Step 6 Using the gray value of each equivalent BGA solder ball in the boundary equivalent BGA solder ball obtained in step 5, find the corresponding full gray BGA in the complete gray BGA solder ball information list.
  • the coordinates of the center point position of the solder ball are based on the coordinates of the center point of the complete gray BGA solder ball, and the boundary straight line is fitted, and the deflection angle and the center position of the BGA chip in the original image are solved by the boundary fitting straight line;
  • Step 7 using each row of equivalent BGA solder balls and each column of equivalent BGA solder balls ⁇ identified in step 5, and the complete grayscale BGA solder ball information list obtained in step 2 for each line of equivalent BGA solder balls Fit and the equivalent BGA solder ball ⁇ straight line fit for each column; the average of the fitted line spacing of all adjacent two rows of equivalent BGA solder balls ⁇ is taken as the standard line spacing of the BGA chip solder balls; all adjacent columns will be The average value of the equivalent BGA solder ball ⁇ fitting line spacing is taken as BGA chip solder ball standard column spacing;
  • Step 8 Using the equivalent BGA solder ball ⁇ fitting line obtained in step 7 and the equivalent BGA solder ball ⁇ fitting line of each column, the equivalent BGA solder ball is performed line by line or on the BGA solder ball identification image. Column search, and then get the BGA solder ball distribution matrix; according to the row-by-row or column-by-column search for all equivalent BGA solder balls and the complete gray-scale BGA solder ball information list obtained in step 2 to obtain the BGA solder ball standard diameter, BGA solder balls Standard perimeter and BGA solder ball standard roundness.
  • step 2 This embodiment differs from the specific embodiment in that: in step 2, a complete gray-scale BGA solder ball information list is established for step one, and the specific acquisition method of the related data is as follows:
  • Gray-scale pixels included in each complete gray-scale BGA solder ball Each binary BGA solder ball is extracted in the corresponding neighborhood on the original image by the following formula:
  • (x p , y p ) is the horizontal and vertical coordinate value of the pixel to be judged on the original image
  • R is the complete grayscale BGA solder ball pixel set
  • mean[R] represents all pixels in the full grayscale BGA solder ball.
  • the average gray value, (x adj , y adj ) represents a coordinate value of a pixel adjacent to (x p , y p ) eight and already belonging to R,
  • ⁇ are preset constants;
  • f(x p , y p ) is the gray value of the original image at the point (x p , y p );
  • the gray-scale connected domain extraction can extract all the grayscale pixels contained in each BGA solder ball to the greatest extent, and record the finally obtained grayscale BGA solder ball as a complete grayscale BGA solder ball;
  • (x k , y k ) is the kth gray pixel coordinate contained in the i-th full gray BGA solder ball, and N i is the number of gray pixels included in the i-th full gray BGA solder ball, i is a positive integer and k is a positive integer;
  • the area S i of the i-th full gray-scale BGA solder ball is the number of pixels N i included in the complete gray-scale BGA solder ball; the roundness C i of the i-th full gray-scale BGA solder ball is calculated as follows:
  • L i is the perimeter of the ith full grayscale BGA solder ball, ie the number of peripheral pixels of the full grayscale BGA solder ball.
  • Step 41 Select any equivalent BGA solder ball on the BGA solder ball identification image
  • Step 42 Search for [-45°, 45°), [45°, 135°), [135°, 255°) in the range of 3 ⁇ radius centering on the equivalent BGA solder ball selected in step 41. And the equivalent BGA solder balls closest to the equivalent BGA solder balls selected in step 41 in the four directions of [-135°, -45°), if the nearest equivalent BGA solder balls in the four directions do not exist , re-execute step 41;
  • Step 4 For the nearest neighbor equivalent BGA solder balls found in each direction, obtain the rough deflection angle ⁇ j of the equivalent BGA array corresponding to the jth direction according to the following formula.
  • All ⁇ j are averaged as the final equivalent BGA array coarse deflection angle ⁇ , and in the image coordinate system, ⁇ is positive in the clockwise direction.
  • the other steps and parameters are the same as one of the specific embodiments one to two.
  • Step 41 Select any equivalent BGA solder ball on the BGA solder ball identification image
  • Step 42 Searching for [0°, 90°), [90°, 180°), [180°, 270°) and the equivalent BGA solder balls selected in step 41 in the range of 3 ⁇ radius. In the four directions of [-90°, 0°), the nearest equivalent BGA solder ball to the equivalent BGA solder ball in the center, if the nearest equivalent BGA solder balls in the four directions do not exist, re-execute step four One;
  • Step 4 For the nearest neighbor equivalent BGA solder balls found in each direction, obtain the rough deflection angle ⁇ j of the equivalent BGA array corresponding to the jth direction according to the following formula:
  • All ⁇ j are averaged as the final equivalent BGA array coarse deflection angle ⁇ ; in the image coordinate system, ⁇ is positive in the clockwise direction.
  • the other steps and parameters are the same as one of the specific embodiments one to three.
  • This embodiment differs from one of the specific embodiments 1 to 4 in that the equivalent BGA solder ball pitch typical value ⁇ obtained in step 5 in step 5 and the equivalent BGA array coarse deflection angle ⁇ obtained in step four are obtained.
  • the line and column equivalent BGA solder balls are linearly clustered to obtain equivalent BGA solder balls, each column of equivalent BGA solder balls, and boundary equivalent BGA solder balls.
  • the specific process is:
  • (x center , y center ) is the center point position coordinate of the complete gray BGA solder ball
  • b row is the intercept of the line straight line in the image coordinate system y axis
  • b col is the line straight line in the image coordinate system x axis distance
  • Step 52 Determine the equivalent BGA solder ball row and column classification threshold
  • Step 5 Perform row clustering on equivalent BGA solder balls: use the i-th equivalent BGA solder ball coordinates Inversely solve the intercept of the corresponding equivalent BGA solder ball line equation:
  • Step 5 Based on the classification threshold intercept_thresh, all the obtained line intercepts Clustering, ie within the threshold intercept_thresh, equal row intercept Gather into a class that will be associated with a class of line intercepts
  • the corresponding equivalent BGA solder balls are clustered into one row equivalent BGA solder ball cluster, and the row equivalent BGA solder ball clusters are in the same row;
  • Step 5 After the fifth and fourth clusters are clustered, the row equivalent BGA solder ball clusters with the number of elements in the ⁇ are regarded as interferences; the number of equivalent BGA solder ball clusters obtained is BGA chip soldering. Number of rows of balls;
  • Step five or six, column clustering of equivalent BGA solder balls using the i-th equivalent BGA solder ball coordinates Inversely solve the column intercept of the corresponding column equation
  • Step 57 Based on the classification threshold intercept_thresh, all the obtained column intercepts Into the column cluster, that is, within the threshold intercept_thresh range, equal column intercept Gather into a class that will be associated with a class of column intercepts Corresponding equivalent BGA solder balls are clustered into a column of equivalent BGA solder ball clusters, and the equivalent BGA solder ball clusters in this column are in the same column;
  • Step 58 After the clustering of the fifth and fourth steps, the column equivalent BGA solder ball cluster with the number of elements in the ⁇ is regarded as interference to be eliminated; the number of column equivalent BGA solder ball clusters obtained is BGA chip bonding. Number of spheres;
  • Step 5 Extracting the first group of the sorted row equivalent BGA solder ball ⁇ , the last group of the sorted row equivalent BGA solder ball ;; the sorted column equivalent BGA solder ball ⁇ The last set of equivalent BGA solder balls of a set and sorted column is used as the boundary equivalent BGA solder ball.
  • step 5:4 and step 5.7 all the obtained line intercepts are within the threshold intercept_thresh range.
  • Column intercept The specific implementation of clustering is as follows (now only Clustering as an example):
  • Step1 Will Initialized to class 1 ⁇ 1 ;
  • the other steps and parameters are the same as one of the specific embodiments one to four.
  • the corresponding gray level BGA solder ball information is searched for the corresponding full gray BGA solder ball center point position coordinate. Straight line fitting the coordinates of the center point of each column of the complete gray BGA solder ball.
  • This embodiment differs from one of the specific embodiments 1 to 7 in that each row of equivalent BGA solder ball ⁇ fitting straight line and each column of equivalent BGA solder ball ⁇ fit obtained in step VIII is obtained in step VIII.
  • Straight line on the BGA solder ball identification image, the equivalent BGA solder ball is searched row by row or column by column, and then the BGA solder ball distribution matrix is obtained. According to the row-by-row or column-by-column search, all equivalent BGA solder balls and step two are obtained.
  • the complete grayscale BGA solder ball information list is obtained by the BGA solder ball standard diameter, BGA solder ball standard perimeter and BGA solder ball standard roundness.
  • the BGA solder ball standard diameter, the BGA solder ball standard perimeter and the BGA solder ball standard roundness are obtained.
  • the specific process is as follows: the average value of the minimum outer circle diameter corresponding to all the searched equivalent BGA solder balls in the complete grayscale BGA solder ball information list as the BGA solder ball standard diameter; will be in the full grayscale BGA solder ball information list The average of the perimeters corresponding to all the searched equivalent BGA solder balls is taken as the standard perimeter of the BGA solder balls; it will correspond to all the searched equivalent BGA solder balls in the full grayscale BGA solder ball information list. The average of the roundness is taken as the standard roundness of the BGA solder ball.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Electric Connection Of Electric Components To Printed Circuits (AREA)
  • Length Measuring Devices By Optical Means (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

一种基于行列直线聚类的多类型BGA芯片视觉识别方法，涉及BGA芯片视觉识别方法。该方法是要解决现有BGA芯片视觉识别算法识别范围窄、采用模板匹配算法灵活性低、对光照和干扰的鲁棒性差以及时间复杂度高的问题；该方法是通过1对原始图像进行动态阈值分割、形态学以及连通域标记处理；2基于灰度连通域提取BGA焊球并建立完整灰度BGA焊球信息列表；3建立BGA焊球标识图像；4局部分析确定等效BGA阵列粗略偏转角度；5采用行列直线聚类识别每行、每列等效BGA焊球蔟以及边界等效BGA焊球蔟；6边界BGA焊球蔟直线拟合求解原始图像中BGA芯片的精确偏转角度和中心位置；8提取并求解BGA芯片各个标准参数等步骤实现的。该方法应用于BGA芯片视觉识别领域。

Description

一种基于行列直线聚类的多类型BGA芯片视觉识别方法 技术领域
本发明涉及一种多类型BGA芯片视觉识别方法，特别涉及一种基于行列直线聚类的多类型BGA芯片视觉识别方法。
背景技术
BGA(BallGrayArray，球柵阵列)封装形式的芯片由于集成度高、I/O引线多以及电性能优良的特性正广泛应用于集成电路中，也正由于这种特点，BGA芯片在进行生产制造时更容易出现焊球丢失，焊球过大或者过小，焊球移位等缺陷。在印刷电路板自动化贴装(SMT)领域，BGA芯片高密度的I/O焊球引脚以及多类型的焊球排布结构对芯片识别与检测的精度和速度提出了更为严格的要求。
目前常采用自动光学检测(AOI)系统对贴装生产过程中的芯片进行识别、检测与定位。芯片识别是贴装的基础，通过对标准BGA样片进行测量，在AOI系统中建立该类型芯片的标准数据库，进而为检测与定位过程提供标准数据参考。具体而言，AOI系统的上述工作过程包括示教(offlinetraining)和检测(onlineinspection)两个阶段：
示教：通过识别标准BGA样片来建立BGA芯片标准参数数据库。BGA芯片的标准数据库包括焊球分布矩阵(表征焊球的分布样式)，焊球标准面积和圆度，焊球标准行间距以及焊球标准列间距。但是在现有的AOI系统中，芯片的标准参数大都是通过人工测量，手动录入的方式进行的，因此工作量大，人力成本高，自动化程度低。
检测：对采集到的待贴装BGA芯片的图像进行图像处理和数据提取，并以BGA芯片标准数据库为参考，对贴装芯片进行缺陷检测和精确定位。BGA芯片的缺陷主要包含如下几个方面：焊球缺失，焊球直径或圆度不合格，相邻焊球桥接等。BGA芯片焊球引脚多，间距小，焊球排布形式多样等特点对检测定位算法的可靠性和检测速度提出了更高的要求。
BGA芯片在检测阶段的基本步骤为图像采集、焊球提取、焊球阵列定位以及焊球参数检测等。其中，焊球提取旨在通过图像分割算法提取BGA焊球，传统的焊球提取方法一般采用全局阈值对图像进行二值化：例如文献《Analysis Ball Grid Array defects by using new image technique》利用所有像素灰度的均值和方差来计算得到此二值化阈值；文献《A system for automated BGA Inspection》采用了自适应图像灰度的Otsu算法来求解此全局阈值；文献《Automated detection and classification of non-wet solder joints》采用了一种基于图 像像素灰度值统计的迭代计算的方法来求此阈值，等。传统方法分离得到的为二值化焊球，因此，当AOI系统采集到的BGA图像光照分布不均匀时，上述二值化分割方法会使某些焊球产生“过分割”和“欠分割”，最终导致检测结果发生错误；同时，图像二值化引入的噪点也可能会被误认为焊球，对后续的焊球阵列定位产生干扰。
焊球阵列定位旨在确定焊球阵列的旋转角度和中心在图像坐标系下的位置。现有较为成熟的方法(如Halcon机器视觉算法库)一般采用的是模板匹配策略。具体在实现过程中，首先需根据BGA标准数据库构建标准焊球阵列模板，然后根据焊球的相邻位置关系进行遍历与迭代，确定模板焊球阵列和待检测BGA芯片焊球阵列的对应变换关系。这种方法由于涉及到大量的迭代与遍历，算法复杂度高，对具有大规模焊球引脚的BGA芯片检测时间长，因此很难适用于对实时性要求高的AOI系统；同时，由于每检测一种芯片，就要建立一种标准焊球阵列模板，算法灵活性差。因此为了较少上述检测定位过程的时间复杂度，文献《Locating and checking of BGA pins position using gray level》提出采用一种最小二乘矩形方法对BGA焊球阵列进行矩形拟合，显然这种算法不能用适用于外围焊球为非矩形分布的BGA芯片；文献《A system for automated BGA Inspection》中为了避免在进行模板匹配时遍历所有的焊球，选择了特殊位置处的焊球(例如阵列顶角焊球)作为待匹配焊球来减少匹配时间，但是这种方法同样没有克服建立标准模板的繁琐，同时对于焊球分割阶段引入的干扰也不具备鲁棒性。
综上所述，将现有AOI系统在BGA芯片示教、检测过程存在的问题总结如下：
1)示教过程通常需要人工测量BGA样片参数，并手动录入标准参数数据，工作量大，人力成本高。
2)传统采用全局阈值的二值化方法进行焊球提取，对非均匀光照图像鲁棒性差，容易导致焊球欠分割或者过分割，最终造成检测错误。
3)基于最小二乘矩形方法的焊球阵列定位算法只能适用于有限几种焊球排布的BGA芯片，对于焊球排布稀疏、或者焊球排布不规则的BGA芯片适用性差。
4)传统的基于模板匹配方法的焊球阵列定位算法需要对于每种BGA芯片都建立标准阵列模板，且匹配算法时间复杂度高，对焊球分割过程引入的干扰鲁棒性差。
技术问题
本发明的目的是为了解决现有BGA芯片识别或检测方法的如下几方面问题：1)识别范围窄，只能识别特定种类或者有限几种焊球排布的BGA芯片，对于焊球排布稀疏的BGA芯片适用性差，或者根本无法识别；2)对光照和干扰的鲁棒性差，采用全局阈值的二 值化方法进行焊球提取，对非均匀光照图像鲁棒性差，容易导致焊球欠分割或者过分割，最终造成检测错误，同时该方法引入的干扰会影响后续芯片定位的准确性；3)采用模板匹配算法灵活性低，对于每种BGA芯片都需建立标准阵列模板，对焊球分割引入的干扰鲁棒性差；4)传统算法时间复杂度高，而且大部分情况下识别过程需要人工对BGA芯片参数进行测量与录入；而提出的一种基于行列直线聚类的多类型BGA芯片视觉识别方法。
技术解决方案
步骤一、对摄像头采集到的灰度BGA芯片图像，即为原始图像，进行动态阈值分割得到二值焊球图像，并对二值焊球图像进行形态学开运算和闭运算处理，处理后的二值焊球图像上的每个连通域记为一个二值化BGA焊球，然后对每个二值化BGA焊球进行连通域标记；
步骤二、对步骤一得到的经过连通域标记后的每个二值化BGA焊球在原始图像上对应邻域范围内进行灰度连通域提取，获得完整灰度BGA焊球，并建立完整灰度BGA焊球信息列表；
其中，建立的完整灰度BGA焊球信息列表内容包括：每个完整灰度BGA焊球所包含的灰度像素，以及由灰度像素计算得到的每个完整灰度BGA焊球的中心点位置坐标、每个完整灰度BGA焊球对应的最小外包圆直径、每个完整灰度BGA焊球的周长和圆度；每个焊球包含的灰度像素包括像素坐标和灰度值；
步骤三、用步骤二得到的完整灰度BGA焊球信息列表，建立一个像素灰度值均为0且大小与原始图像相同的背景图像；并在背景图像中，将对应原始图像中每个完整灰度BGA焊球中心点位置处的灰度值，变为对应二值化BGA焊球的标识序号，此时的背景图像记为BGA焊球标识图像，BGA焊球标识图像上的每个非0灰度值的像素称为一个等效BGA焊球，所有等效BGA焊球构成的阵列称为等效BGA阵列；
其中，原始图像中有M*N个完整灰度BGA焊球，对应背景图像中就有M*N个等效BGA焊球，等效BGA焊球实质为一个像素，完整灰度BGA焊球与等效BGA焊球一一对应；在BGA焊球标识图像中，计算相邻2个等效BGA焊球的间距Δγ，将此间距作为等效BGA焊球间距典型值；
步骤四、利用步骤三得到的等效BGA焊球间距典型值Δγ，在BGA焊球标识图像上，对等效BGA阵列进行局部分析，确定等效BGA阵列粗略偏转角度Δθ；
其中，局部分析分为针对规则型BGA芯片的局部分析和针对不规则型BGA芯片的局部分析，规则型BGA芯片为相邻行BGA焊球成整齐排列的BGA芯片，不规则型BGA芯片为相邻行BGA焊球成交错排列的BGA芯片；
步骤五、利用步骤三得到的等效BGA焊球间距典型值Δγ以及步骤四得到的等效BGA阵列粗略偏转角度Δθ，在BGA焊球标识图像上，对行和列的等效BGA焊球进行直线聚类得到每行等效BGA焊球蔟、每列等效BGA焊球蔟以及边界等效BGA焊球蔟；
步骤六、利用步骤五得到的边界等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球的中心点位置坐标，根据完整灰度BGA焊球中心点位置坐标，进行边界直线拟合，通过边界拟合直线求解原始图像中BGA芯片的偏转角度和中心位置；
步骤七、利用步骤五识别得到的每行等效BGA焊球蔟和每列等效BGA焊球蔟，以及步骤二得到的完整灰度BGA焊球信息列表进行每行等效BGA焊球蔟直线拟合和每列等效BGA焊球蔟直线拟合；将所有相邻两行等效BGA焊球蔟的拟合直线间距的平均值作为BGA芯片焊球标准行间距；将所有相邻两列等效BGA焊球蔟拟合直线间距的平均值作为BGA芯片焊球标准列间距；
步骤八、利用步骤七得到的每行等效BGA焊球蔟拟合直线和每列等效BGA焊球蔟拟合直线，在BGA焊球标识图像上对等效BGA焊球进行逐行或逐列搜索，进而得到BGA焊球分布矩阵；根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度。
有益效果
1)本发明提出的BGA芯片视觉识别方法在进行多类型BGA芯片识别过程中无需建立匹配模板，灵活性强。在实际AOI系统应用的示教阶段可利用该方法对标准BGA芯片样本进行参数识别，建立标准数据库；对于检测阶段，可利用该识别方法提取待检测BGA芯片的各个参数并进行检测，同时得到BGA焊球阵列的偏转角度和中心位置。2)本发明方法包含了一种基于灰度连通域的焊球提取方法，相较于传统的二值化提取算法，能够最大程度的保证在非均匀光照下提取出的BGA焊球的完整性，进而保证识别结果的准确性。两种焊球提取方法的对比结果如图6和图7所示；3)该基于行列直线聚类的多类型BGA芯片视觉 识别方法能够识别绝大多数的BGA芯片：规则型BGA芯片、不规则型BGA芯片、焊球排布稀疏型的BGA芯片或其它焊球分布类型的BGA芯片。4)由于本发明提出的行列直线聚类方法的特性，本识别方法能够自动识别与滤除因焊球提取而引入的干扰(容易被误认为焊球)，提高了算法对于背景干扰的鲁棒性，保证了BGA焊球阵列定位的准确性和精度。图8和图9代表性的显示了该算法能识别出的非焊球干扰。5)本发明提出的识别算法能够对采集图像的非均匀光照鲁棒，非理想光照强度自适应。在一种典型的实验情况下：待检测芯片如图10所示，采集图像的分辨率为840*1160，不同光照强度对算法运行时间和检测精度的影响曲线如图11所示；6)本发明提出的基于行列直线聚类的识别方法计算复杂度低。在一种典型的实验情况下：光照条件理想，采集图像的分辨率为840*1160，针对不同焊球数的BGA芯片，一种商业算法和本发明提出的识别算法的运行时间对比如图12：(PC平台参：
Figure PCTCN2016071119-appb-000001
B960.2.2GHzdual-core,C++)7)本发明提出的基于行列直线聚类的识别方法也可以推广适用于具有不同分布样式的图形阵列的识别和检测。
附图说明
图1为具体实施方式一提出的一种基于行列直线聚类的多类型BGA芯片视觉识别方法流程图；
图2为具体实施方式一提出的对原始图像进行动态阈值分割，以及形态学开运算和闭运算处理过程的示例图；
图3(a)为具体实施方式三提出的对规则型BGA芯片的局部分析示例图；
图3(b)为具体实施方式四提出的对不规则型BGA芯片的局部分析示例图；
图4为具体实施方式五提出的对行列BGA焊球进行直线聚类的示例图，其中，
Figure PCTCN2016071119-appb-000002
表示第j个等效BGA焊球坐标对应的等效BGA焊球行直线方程的截距，
Figure PCTCN2016071119-appb-000003
表示第j个等效BGA焊球坐标对应的等效BGA焊球列直线方程的截距，
Figure PCTCN2016071119-appb-000004
表示第i个等效BGA焊球坐标对应的等效BGA焊球行直线方程的截距，
Figure PCTCN2016071119-appb-000005
表示第i个等效BGA焊球坐标对应的等效BGA焊球列直线方程的截距，
Figure PCTCN2016071119-appb-000006
表示干扰对应的等效BGA焊球列直线方程的截距；
图5为一种焊球排布较为稀疏的BGA芯片示例图；
图6为传统基于二值化方法的BGA焊球提取效果图；
图7为本发明包含的BGA焊球提取方法效果图；
图8为代表性的显示了该算法能识别出的左上角的非焊球干扰；
图9为代表性的显示了该算法能识别出的左下角的非焊球干扰；
图10为待检测BGA芯片图像；
图11为对于图10的待检测BGA芯片，不同光照强度对算法运行时间和检测精度的影响曲线图；
图12为一种商业算法和本发明提出的识别算法的运行时间对比图。
本发明的实施方式
具体实施方式一：本实施方式的一种基于行列直线聚类的多类型BGA芯片视觉识别方法，具体是按照以下步骤实施的：
步骤一、对摄像头采集到的灰度BGA芯片图像，即为原始图像，进行动态阈值分割得到二值焊球图像，并对二值焊球图像进行形态学开运算和闭运算处理，处理后的二值焊球图像上的每个连通域记为一个二值化BGA焊球，然后对每个二值化BGA焊球进行连通域标记；
其中，原始图像经过动态阈值分割得到二值焊球图像I(x，y)可用如下数学表达式表示：
Figure PCTCN2016071119-appb-000007
式中，f(x，y)原始图像，g(x，y)是f(x，y)经过均值滤波后的图像。C为预设常数。
步骤二、对步骤一得到的经过连通域标记后的每个二值化BGA焊球在原始图像上对应邻域范围内进行灰度连通域提取，获得完整灰度BGA焊球，并建立完整灰度BGA焊球信息列表；
其中，建立的完整灰度BGA焊球信息列表内容包括：每个完整灰度BGA焊球所包含的灰度像素，以及由灰度像素计算得到的每个完整灰度BGA焊球的中心点位置坐标、每个完整灰度BGA焊球对应的最小外包圆直径、每个完整灰度BGA焊球的周长和每个完整灰度BGA焊球的圆度；每个完整灰度BGA焊球包含的灰度像素包括像素坐标和灰度值；
步骤三、用步骤二得到的完整灰度BGA焊球信息列表，建立一个像素灰度值均为0且大小与原始图像相同的背景图像；并在背景图像中，将对应原始图像中每个完整灰度BGA焊球中心点位置处的灰度值，变为对应二值化BGA焊球的标识序号，此时的背景图像记为BGA焊球标识图像，BGA焊球标识图像上的每个非0灰度值的像素称为一个等效BGA焊球，所有等效BGA焊球构成的阵列称为等效BGA阵列；
其中，原始图像中有M*N个完整灰度BGA焊球，对应背景图像中就有M*N个等效BGA焊球，等效BGA焊球实质为一个像素，完整灰度BGA焊球与等效BGA焊球一一对应；在BGA焊球标识图像中，计算相邻2个等效BGA焊球的间距Δγ，将此间距作为等效BGA焊球间距典型值；
步骤四、利用步骤三得到的等效BGA焊球间距典型值Δγ，在BGA焊球标识图像上，对等效BGA阵列进行局部分析，确定等效BGA阵列粗略偏转角度Δθ；
其中，局部分析分为针对规则型BGA芯片的局部分析和针对不规则型BGA芯片的局部分析，规则型BGA芯片为相邻行BGA焊球成整齐排列的BGA芯片，不规则型BGA芯片为相邻行BGA焊球成交错排列的BGA芯片；
步骤五、利用步骤三得到的等效BGA焊球间距典型值Δγ以及步骤四得到的等效BGA阵列粗略偏转角度Δθ，在BGA焊球标识图像上，对行和列的等效BGA焊球进行直线聚类得到每行等效BGA焊球蔟、每列等效BGA焊球蔟以及边界等效BGA焊球蔟；
步骤六、利用步骤五得到的边界等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球的中心点位置坐标，根据完整灰度BGA焊球中心点位置坐标，进行边界直线拟合，通过边界拟合直线求解原始图像中BGA芯片的偏转角度和中心位置；
步骤七、利用步骤五识别得到的每行等效BGA焊球蔟和每列等效BGA焊球蔟，以及步骤二得到的完整灰度BGA焊球信息列表进行每行等效BGA焊球蔟直线拟合和每列等效BGA焊球蔟直线拟合；将所有相邻两行等效BGA焊球蔟的拟合直线间距的平均值作为BGA芯片焊球标准行间距；将所有相邻两列等效BGA焊球蔟拟合直线间距的平均值作为 BGA芯片焊球标准列间距；
步骤八、利用步骤七得到的每行等效BGA焊球蔟拟合直线和每列等效BGA焊球蔟拟合直线，在BGA焊球标识图像上对等效BGA焊球进行逐行或逐列搜索，进而得到BGA焊球分布矩阵；根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度。
具体实施方式二：本实施方式与具体实施方式一不同的是：步骤二中对步骤一建立完整灰度BGA焊球信息列表，相关数据的具体获取方法如下：
(1)每个完整灰度BGA焊球所包含的灰度像素：每个二值化BGA焊球在原始图像上对应邻域范围内进行灰度连通域提取用如下公式表示：
Figure PCTCN2016071119-appb-000008
式中，(xp，yp)为原始图像上待判断的像素的横纵坐标值，R为完整灰度BGA焊球像素集合，mean[R]表示完整灰度BGA焊球中所有像素的平均灰度值，(xadj，yadj)表示与(xp，yp)八邻接且已经属于R的像素的坐标值，
Figure PCTCN2016071119-appb-000009
和Δ为预设常量；f(xp，yp)为原始图像在点(xp，yp)处的灰度值；
灰度连通域提取能够最大程度的提取到每个BGA焊球包含的所有灰度像素，将最终提取得到的灰度BGA焊球记为完整灰度BGA焊球；
(2)第i个完整灰度BGA焊球的中心点位置坐标
Figure PCTCN2016071119-appb-000010
计算公式如下：
Figure PCTCN2016071119-appb-000011
式中：(xk，yk)为第i个完整灰度BGA焊球包含的第k个灰度像素坐标，Ni为第i个完整灰度BGA焊球包含的灰度像素个数，i为正整数，k为正整数；
(3)第i个完整灰度BGA焊球的面积Si为此完整灰度BGA焊球包含的像素个数Ni；第i个完整灰度BGA焊球的圆度Ci计算公式如下：
Figure PCTCN2016071119-appb-000012
式中，Li为第i个完整灰度BGA焊球的周长即此完整灰度BGA焊球的外围像素数。其它步骤及参数与具体实施方式一相同。
具体实施方式三：本实施方式与具体实施方式一或二不同的是：步骤四中针对规则型BGA芯片的局部分析具体过程为图3(a)：
步骤四一、在BGA焊球标识图像上，选择任一个等效BGA焊球；
步骤四二、以步骤四一中选择的等效BGA焊球为中心在3Δγ半径范围内，分别搜索[-45°，45°)、[45°，135°)、[135°，255°)和[-135°，-45°)四个方向范围内，与步骤四一中选择的等效BGA焊球最近的等效BGA焊球，若四个方向的最近等效BGA焊球均不存在，重新执行步骤四一；
步骤四三、对每个方向找到的最近相邻等效BGA焊球，分别按照下式求取第j个方向对应的等效BGA阵列粗略偏转角度Δθj，
Figure PCTCN2016071119-appb-000013
式中，(xcenter，ycenter)为中心的等效BGA焊球的坐标，
Figure PCTCN2016071119-appb-000014
为第j个方向对应的最近等效BGA焊球；同时，±90°和正负号根据四个方向情况进行选择或者舍去；j＝1代表[-45°，45°)范围；j＝2代表[45°，135°)范围：j＝3代表[135°，225°)范围；j＝4代表[-135°，-45°)范围；
将所有Δθj取平均值作为最终的等效BGA阵列粗略偏转角度Δθ，在图像坐标系下，Δθ以顺时针方向为正。其它步骤及参数与具体实施方式一至二之一相同。
具体实施方式四：本实施方式与具体实施方式一至三之一不同的是：步骤四中针对不规则型BGA芯片局部分析的具体过程为图3(b)：
步骤四一、在BGA焊球标识图像上，选择任一个等效BGA焊球；
步骤四二、以步骤四一中选择的等效BGA焊球为中心在3Δγ半径范围内，分别搜索[0°，90°)、[90°，180°)、[180°，270°)和[-90°，0°)的四个方向范围内，与中心的等效BGA焊球最近的等效BGA焊球，若四个方向的最近等效BGA焊球均不存在，重新执行步骤四一；
步骤四三、对每个方向找到的最近相邻等效BGA焊球，分别按照下式求取第j个方向对应的等效BGA阵列粗略偏转角度Δθj，：
Figure PCTCN2016071119-appb-000015
式中，(xcenter，ycenter)为中心的等效BGA焊球的坐标，
Figure PCTCN2016071119-appb-000016
为第j个方向对应的最近等效BGA焊球，同时，上式中的±90°和正负号根据四个方向情况进行 选择或者舍去；j＝1代表[0°，90°)范围；j＝2代表[90°，180°)范围：j＝3代表[180°，270°)范围；j＝4代表[-90°，0°)范围；
将所有Δθj取平均值作为最终的等效BGA阵列粗略偏转角度Δθ；在图像坐标系下，Δθ以顺时针方向为正。其它步骤及参数与具体实施方式一至三之一相同。
具体实施方式五：本实施方式与具体实施方式一至四之一不同的是：步骤五中利用步骤三得到的等效BGA焊球间距典型值Δγ以及步骤四得到的等效BGA阵列粗略偏转角度Δθ，在BGA焊球标识图像上，对行和列的等效BGA焊球进行直线聚类得到每行等效BGA焊球蔟、每列等效BGA焊球蔟以及边界等效BGA焊球蔟的具体过程为：
步骤五一、根据等效BGA阵列粗略偏转角度Δθ，确定等效BGA焊球的行直线方程表达式ycenter＝tan(Δθ)xcenter+brow和等效BGA焊球的列直线方程表达式ycenter＝tan(Δθ+90°)(xcenter-bcol)；
其中，(xcenter，ycenter)为完整灰度BGA焊球的中心点位置坐标，brow为行直线在图像坐标系y轴的截距，bcol为列直线在图像坐标系x轴的截距；
步骤五二、确定等效BGA焊球行和列分类阈值
Figure PCTCN2016071119-appb-000017
步骤五三、对等效BGA焊球进行行聚类：利用第i个等效BGA焊球坐标
Figure PCTCN2016071119-appb-000018
反解出对应的等效BGA焊球行直线方程的截距：
Figure PCTCN2016071119-appb-000019
步骤五四、基于分类阈值intercept_thresh，对所有得到的行截距
Figure PCTCN2016071119-appb-000020
进行聚类，即在阈值intercept_thresh范围内，相等的行截距
Figure PCTCN2016071119-appb-000021
聚为一类，将与一类行截距
Figure PCTCN2016071119-appb-000022
对应的等效BGA焊球聚为一个行等效BGA焊球簇，且行等效BGA焊球簇在同一行；
步骤五五、将步骤五四聚类后，蔟内元素个数为1的行等效BGA焊球簇视为干扰予以剔除；得到的行等效BGA焊球簇的个数即为BGA芯片焊球行数；
然后对所有行等效BGA焊球簇按照蔟内对应的平均行截距的升序进行排序，得到的排序后的行等效BGA焊球蔟则是以等效BGA阵列的行序号由小到大进行排列的；
步骤五六、对等效BGA焊球进行列聚类：利用第i个等效BGA焊球坐标
Figure PCTCN2016071119-appb-000023
反解出对应的列直线方程的列截距
Figure PCTCN2016071119-appb-000024
Figure PCTCN2016071119-appb-000025
步骤五七、基于分类阈值intercept_thresh，对所有得到的列截距
Figure PCTCN2016071119-appb-000026
进列聚类，即在阈值intercept_thresh范围内，相等的列截距
Figure PCTCN2016071119-appb-000027
聚为一类，将与一类列截距
Figure PCTCN2016071119-appb-000028
对应的等效BGA焊球聚为一个列等效BGA焊球簇，且此列等效BGA焊球簇在同一列；
步骤五八、将步骤五四聚类后，蔟内元素个数为1的列等效BGA焊球簇视为干扰予以剔除；得到的列等效BGA焊球簇的个数即为BGA芯片焊球列数；
然后对所有列等效BGA焊球簇按照蔟内对应的平均列截距的升序进行排序，得到的排序后的列等效BGA焊球蔟则是以等效BGA阵列的列序号由小到大进行排列的；
步骤五九、提取经过排序后的行等效BGA焊球蔟的第一组、经过排序后的行等效BGA焊球蔟的最后一组；经过排序后的列等效BGA焊球蔟的第一组和经过排序后的列等效BGA焊球蔟的最后一组作为边界等效BGA焊球蔟。
步骤五四和步骤五七中，在阈值intercept_thresh范围内，对所有得到的行截距
Figure PCTCN2016071119-appb-000029
和列截距
Figure PCTCN2016071119-appb-000030
进行聚类的具体实施方式如下(现仅以
Figure PCTCN2016071119-appb-000031
的聚类为例)：
Step1:将
Figure PCTCN2016071119-appb-000032
初始化为第1类Ω1；
Step2:如果
Figure PCTCN2016071119-appb-000033
i＝2,3…与第j(j＝1,2,3…k)类Ωj的聚类中心
Figure PCTCN2016071119-appb-000034
满足
Figure PCTCN2016071119-appb-000035
则将
Figure PCTCN2016071119-appb-000036
归为j类，并重新计算
Figure PCTCN2016071119-appb-000037
否则，将
Figure PCTCN2016071119-appb-000038
初始化为k+1类；其中，类Ωj的聚类中心
Figure PCTCN2016071119-appb-000039
是Ωj内所有截距的平均值；
Step3:i＝i+1，执行步骤Step2。其它步骤及参数与具体实施方式一至四之一相同。
具体实施方式六：本实施方式与具体实施方式一至五之一不同的是：步骤七中利用步骤二得到的完整灰度BGA焊球信息列表进行每行等效BGA焊球蔟直线拟合的过程如下：
利用每行等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球中心点位置坐标，将每行完整灰度BGA焊球中心点位置坐标进行直线拟合。其它步骤及参数与具体实施方式一至五之一相同。
具体实施方式七：本实施方式与具体实施方式一至六之一不同的是：步骤七中利用 步骤二得到的完整灰度BGA焊球信息列表进行每列等效BGA焊球蔟直线拟合的过程如下：
利用每列等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球中心点位置坐标，将每列完整灰度BGA焊球中心点位置坐标进行直线拟合。其它步骤及参数与具体实施方式一至六之一相同。
具体实施方式八：本实施方式与具体实施方式一至七之一不同的是：步骤八中利用步骤七得到的每行等效BGA焊球蔟拟合直线和每列等效BGA焊球蔟拟合直线，在BGA焊球标识图像上对等效BGA焊球进行逐行或逐列搜索，进而得到BGA焊球分布矩阵，根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度的具体过程为：
求解每行等效BGA焊球蔟拟合直线方程和每列等效BGA焊球蔟拟合直线方程的交点，并以该交点为中心在BGA标识图像上进行半径为Δγ/2局部搜索；如果搜索到某一等效BGA焊球，则该等效BGA焊球对应的BGA标识矩阵的所在行列位置的值置为1，否则该交点位置对应的BGA标识矩阵的所在行列位置的值置为0；
根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度的具体过程如下：在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的最小外包圆直径的平均值作为BGA焊球标准直径；将在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的周长的平均值作为BGA焊球标准周长；将在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的的圆度的平均值作为BGA焊球标准圆度。
其它步骤及参数与具体实施方式一至七之一相同。

Claims (8)

  1. 一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于一种基于行列直线聚类的多类型BGA芯片视觉识别方法具体是按照以下步骤进行的：
    步骤一、对摄像头采集到的灰度BGA芯片图像，即为原始图像，进行动态阈值分割得到二值焊球图像，并对二值焊球图像进行形态学开运算和闭运算处理，处理后的二值焊球图像上的每个连通域记为一个二值化BGA焊球，然后对每个二值化BGA焊球进行连通域标记；
    步骤二、对步骤一得到的经过连通域标记后的每个二值化BGA焊球在原始图像上对应邻域范围内进行灰度连通域提取，获得完整灰度BGA焊球，并建立完整灰度BGA焊球信息列表；
    其中，建立的完整灰度BGA焊球信息列表内容包括：每个完整灰度BGA焊球所包含的灰度像素，以及由灰度像素计算得到的每个完整灰度BGA焊球的中心点位置坐标、每个完整灰度BGA焊球对应的最小外包圆直径、每个完整灰度BGA焊球的周长和圆度；每个焊球包含的灰度像素包括像素坐标和灰度值；
    步骤三、用步骤二得到的完整灰度BGA焊球信息列表，建立一个像素灰度值均为0且大小与原始图像相同的背景图像；并在背景图像中，将对应原始图像中每个完整灰度BGA焊球中心点位置处的灰度值，变为对应二值化BGA焊球的标识序号，此时的背景图像记为BGA焊球标识图像，BGA焊球标识图像上的每个非0灰度值的像素称为一个等效BGA焊球，所有等效BGA焊球构成的阵列称为等效BGA阵列；
    其中，原始图像中有M*N个完整灰度BGA焊球，对应背景图像中就有M*N个等效BGA焊球，等效BGA焊球实质为一个像素，完整灰度BGA焊球与等效BGA焊球一一对应；在BGA焊球标识图像中，计算相邻2个等效BGA焊球的间距Δγ，将此间距作为等效BGA焊球间距典型值；
    步骤四、利用步骤三得到的等效BGA焊球间距典型值Δγ，在BGA焊球标识图像上，对等效BGA阵列进行局部分析，确定等效BGA阵列粗略偏转角度Δθ；
    其中，局部分析分为针对规则型BGA芯片的局部分析和针对不规则型BGA芯片的局部分析，规则型BGA芯片为相邻行BGA焊球成整齐排列的BGA芯片，不规则型BGA芯片为相邻行BGA焊球成交错排列的BGA芯片；
    步骤五、利用步骤三得到的等效BGA焊球间距典型值Δγ以及步骤四得到的等效BGA阵 列粗略偏转角度Δθ，在BGA焊球标识图像上，对行和列的等效BGA焊球进行直线聚类得到每行等效BGA焊球蔟、每列等效BGA焊球蔟以及边界等效BGA焊球蔟；
    步骤六、利用步骤五得到的边界等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球的中心点位置坐标，根据完整灰度BGA焊球中心点位置坐标，进行边界直线拟合，通过边界拟合直线求解原始图像中BGA芯片的偏转角度和中心位置；
    步骤七、利用步骤五识别得到的每行等效BGA焊球蔟和每列等效BGA焊球蔟，以及步骤二得到的完整灰度BGA焊球信息列表进行每行等效BGA焊球蔟直线拟合和每列等效BGA焊球蔟直线拟合；将所有相邻两行等效BGA焊球蔟的拟合直线间距的平均值作为BGA芯片焊球标准行间距；将所有相邻两列等效BGA焊球蔟拟合直线间距的平均值作为BGA芯片焊球标准列间距；
    步骤八、利用步骤七得到的每行等效BGA焊球蔟拟合直线和每列等效BGA焊球蔟拟合直线，在BGA焊球标识图像上对等效BGA焊球进行逐行或逐列搜索，进而得到BGA焊球分布矩阵；根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度。
  2. 根据权利要求1所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤二中对步骤一建立的完整灰度BGA焊球信息列表中的数据具体为：
    (1)每个完整灰度BGA焊球所包含的灰度像素：每个二值化BGA焊球在原始图像上对应邻域范围内进行灰度连通域提取用如下公式表示：
    Figure PCTCN2016071119-appb-100001
    式中，(xp，yp)为原始图像上待判断的像素的横纵坐标值，R为完整灰度BGA焊球像素集合，mean[R]表示完整灰度BGA焊球中所有像素的平均灰度值，(xadj，yadj)表示与(xp，yp)八邻接且已经属于R的像素的坐标值，
    Figure PCTCN2016071119-appb-100002
    和Δ为预设常量；f(xp，yp)为原始图像在点(xp，yp)处的灰度值；
    (2)每个完整灰度BGA焊球的中心点位置坐标：第i个完整灰度BGA焊球的中心点位置坐 标
    Figure PCTCN2016071119-appb-100003
    计算公式如下：
    Figure PCTCN2016071119-appb-100004
    式中：(xk，yk)为第i个完整灰度BGA焊球包含的第k个灰度像素坐标，Ni为第i个完整灰度BGA焊球包含的灰度像素个数，i为正整数，k为正整数；
    (3)每个完整灰度BGA焊球的圆度等参数：第i个完整灰度BGA焊球的面积Si.为完整灰度BGA焊球包含的像素个数Ni；第i个完整灰度BGA焊球的圆度Ci .计算公式如下：
    Figure PCTCN2016071119-appb-100005
    式中，Li为第i个完整灰度BGA焊球周长即完整灰度BGA焊球的外围像素数。
  3. 根据权利要求2所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤四中针对规则型BGA芯片的局部分析具体过程为：
    步骤四一、在BGA焊球标识图像上，选择任一个等效BGA焊球；
    步骤四二、以步骤四一中选择的等效BGA焊球为中心在3Δγ半径范围内，分别搜索[-45°，45°)、[45°，135°)、[135°，225°)和[-135°，-45°)四个方向范围内，与步骤四一中选择的等效BGA焊球最近的等效BGA焊球，若四个方向的最近等效BGA焊球均不存在，重新执行步骤四一；
    步骤四三、对每个方向找到的最近相邻等效BGA焊球，分别按照下式求取第j个方向对应的等效BGA阵列粗略偏转角度Δθj，
    Figure PCTCN2016071119-appb-100006
    式中，(xcenter，ycenter)为中心的等效BGA焊球的坐标，
    Figure PCTCN2016071119-appb-100007
    为第j个方向对应的最近等效BGA焊球；同时，±90°和正负号根据四个方向情况进行选择或者舍去；j＝1代表[-45°，45°)范围；j＝2代表[45°，135°)范围：j＝3代表[135°，225°)范围；j＝4代表[-135°，-45°)范围；
    将所有Δθj取平均值作为最终的等效BGA阵列粗略偏转角度Δθ，在图像坐标系下，Δθ以顺时针方向为正。
  4. 根据权利要求3所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤四中针对不规则型BGA芯片局部分析的具体过程为：
    步骤四一、在BGA焊球标识图像上，选择任一个等效BGA焊球；
    步骤四二、以步骤四一中选择的等效BGA焊球为中心在3Δγ半径范围内，分别搜索[0°，90°)、[90°，180°)、[180°，270°)和[-90°，0°)的四个方向范围内，与中心的等效BGA焊球最近的等效BGA焊球，若四个方向的最近等效BGA焊球均不存在，重新执行步骤四一；
    步骤四三、对每个方向找到的最近相邻等效BGA焊球，分别按照下式求取第j个方向对应的等效BGA阵列粗略偏转角度Δθj，：
    Figure PCTCN2016071119-appb-100008
    式中，(xcenter，ycenter)为中心的等效BGA焊球的坐标，
    Figure PCTCN2016071119-appb-100009
    为第j个方向对应的最近等效BGA焊球，同时，上式中的±90°和正负号根据四个方向情况进行选择或 者舍去；j＝1代表[0°，90°)范围；j＝2代表[90°，180°)范围：j＝3代表[180°，270°)范围；j＝4代表[-90°，0°)范围；
    将所有Δθj取平均值作为最终的等效BGA阵列粗略偏转角度Δθ；在图像坐标系下，Δθ以顺时针方向为正。
  5. 根据权利要求4所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤五中利用步骤三得到的等效BGA焊球间距典型值Δγ以及步骤四得到的等效BGA阵列粗略偏转角度Δθ，在BGA焊球标识图像上，对行和列的等效BGA焊球进行直线聚类得到每行等效BGA焊球蔟、每列等效BGA焊球蔟以及边界等效BGA焊球蔟的具体过程：
    步骤五一、根据等效BGA阵列粗略偏转角度Δθ，确定等效BGA焊球的行直线方程表达式ycenter＝tan(Δθ)xcenter+brow和等效BGA焊球的列直线方程表达式ycenter＝tan(Δθ+90°)(xcenter-bcol)；
    其中，(xcenter，ycenter)为完整灰度BGA焊球的中心点位置坐标，brow为行直线在图像坐标系y轴的截距，bcol为列直线在图像坐标系x轴的截距；
    步骤五二、确定等效BGA焊球行和列分类阈值
    Figure PCTCN2016071119-appb-100010
    步骤五三、对等效BGA焊球进行行聚类：
    利用第i个等效BGA焊球坐标
    Figure PCTCN2016071119-appb-100011
    反解出对应的等效BGA焊球行直线方程的截距：
    Figure PCTCN2016071119-appb-100012
    步骤五四、基于分类阈值intercept_thresh，对所有得到的行截距
    Figure PCTCN2016071119-appb-100013
    进行聚类，即在阈值intercept_thresh范围内，相等的行截距
    Figure PCTCN2016071119-appb-100014
    聚为一类，将与一类行截距
    Figure PCTCN2016071119-appb-100015
    对应的等效BGA焊球聚为一个行等效BGA焊球簇，且行等效BGA焊球簇在同一行；
    步骤五五、将步骤五四聚类后，蔟内元素个数为1的行等效BGA焊球簇视为干扰予以剔除；得到的行等效BGA焊球簇的个数即为BGA芯片焊球行数；
    然后对所有行等效BGA焊球簇按照蔟内对应的平均行截距的升序进行排序，得到的排序后的行等效BGA焊球蔟则是以等效BGA阵列的行序号由小到大进行排列的；
    步骤五六、对等效BGA焊球进行列聚类：利用第i个等效BGA焊球坐标
    Figure PCTCN2016071119-appb-100016
    反解出对应的列直线方程的列截距
    Figure PCTCN2016071119-appb-100017
    Figure PCTCN2016071119-appb-100018
    步骤五七、基于分类阈值intercept_thresh，对所有得到的列截距
    Figure PCTCN2016071119-appb-100019
    进列聚类，即在阈值intercept_thresh范围内，相等的列截距
    Figure PCTCN2016071119-appb-100020
    聚为一类，将与一类列截距
    Figure PCTCN2016071119-appb-100021
    对应的等效BGA焊球聚为一个列等效BGA焊球簇，且此列等效BGA焊球簇在同一列；
    步骤五八、将步骤五四聚类后，蔟内元素个数为1的列等效BGA焊球簇视为干扰予以剔除；得到的列等效BGA焊球簇的个数即为BGA芯片焊球列数；
    然后对所有列等效BGA焊球簇按照蔟内对应的平均列截距的升序进行排序，得到的排序后的列等效BGA焊球蔟则是以等效BGA阵列的列序号由小到大进行排列的；
    步骤五九、提取经过排序后的行等效BGA焊球蔟的第一组、经过排序后的行等效BGA焊球蔟的最后一组、经过排序后的列等效BGA焊球蔟的第一组和经过排序后的列等效BGA焊球蔟的最后一组作为边界等效BGA焊球蔟。
  6. 根据权利要求5所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤七中利用步骤二得到的完整灰度BGA焊球信息列表进行每行等效BGA焊球蔟直 线拟合的过程如下：
    利用每行等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球中心点位置坐标，将每行完整灰度BGA焊球中心点位置坐标进行直线拟合。
  7. 根据权利要求6所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤七中利用步骤二得到的完整灰度BGA焊球信息列表进行每列等效BGA焊球蔟直线拟合的过程如下：
    利用每列等效BGA焊球蔟中的每个等效BGA焊球在标识图像上的灰度值，在完整灰度BGA焊球信息列表中查找对应的完整灰度BGA焊球中心点位置坐标，将每列完整灰度BGA焊球中心点位置坐标进行直线拟合。
  8. 根据权利要求7所述一种基于行列直线聚类的多类型BGA芯片视觉识别方法，其特征在于：步骤八中利用步骤七得到的每行等效BGA焊球蔟拟合直线和每列等效BGA焊球蔟拟合直线，在BGA焊球标识图像上对等效BGA焊球进行逐行或逐列搜索，进而得到BGA焊球分布矩阵，根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度的具体过程为：
    依次求解每行等效BGA焊球蔟拟合直线方程和每列等效BGA焊球蔟拟合直线方程的交点，并以该交点为中心在BGA标识图像上进行半径为Δγ/2局部搜索；如果搜索到某一等效BGA焊球，则BGA标识矩阵的对应行列位置的值置为1，否则BGA标识矩阵的对应行列位置的值置为0；
    根据逐行或逐列搜索得到的所有等效BGA焊球以及步骤二得到的完整灰度BGA焊球信息列表求解得到BGA焊球标准直径、BGA焊球标准周长和BGA焊球标准圆度的具体过程如下：在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的最小外包圆直径的平均值作为BGA焊球标准直径；将在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的周长的平均值作为BGA焊球标准周长；将在完整灰度BGA焊球信息列表中与所有搜索到的等效BGA焊球对应的的圆度的平均值作为BGA焊球标准圆度。
PCT/CN2016/071119 2015-08-05 2016-01-16 一种基于行列直线聚类的多类型bga芯片视觉识别方法 WO2017020559A1 (zh)

Priority Applications (2)

Application Number Priority Date Filing Date Title
US15/323,100 US9965847B2 (en) 2015-08-05 2016-01-16 Multi-type BGA chip visual recognition method using line based clustering
JP2016574415A JP6598162B2 (ja) 2015-08-05 2016-01-16 線形クラスタリングに基づくマルチタイプのｂｇａチップの視覚識別方法

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201510474956.XA CN105005997B (zh) 2015-08-05 2015-08-05 一种基于行列直线聚类的多类型bga芯片视觉识别方法
CN201510474956X 2015-08-05

Publications (1)

Publication Number Publication Date
WO2017020559A1 true WO2017020559A1 (zh) 2017-02-09

Family

ID=54378654

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2016/071119 WO2017020559A1 (zh) 2015-08-05 2016-01-16 一种基于行列直线聚类的多类型bga芯片视觉识别方法

Country Status (4)

Country Link
US (1) US9965847B2 (zh)
JP (1) JP6598162B2 (zh)
CN (1) CN105005997B (zh)
WO (1) WO2017020559A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111144478A (zh) * 2019-12-25 2020-05-12 电子科技大学 一种穿帮镜头的自动检测方法
CN113836489A (zh) * 2021-09-24 2021-12-24 英特尔产品（成都）有限公司 球栅阵列的缺陷分析方法与设备
CN114387223A (zh) * 2021-12-22 2022-04-22 广东正业科技股份有限公司 一种芯片缺陷视觉检测方法及设备

Families Citing this family (23)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP6599672B2 (ja) * 2015-07-17 2019-10-30 日本電産サンキョー株式会社 文字切り出し装置、文字認識装置、および文字切り出し方法
CN105005997B (zh) * 2015-08-05 2017-11-14 哈尔滨工业大学 一种基于行列直线聚类的多类型bga芯片视觉识别方法
CN107507182B (zh) * 2017-09-25 2019-10-25 电子科技大学 一种基于x射线图像的bga焊球提取方法
US10242951B1 (en) 2017-11-30 2019-03-26 International Business Machines Corporation Optical electronic-chip identification writer using dummy C4 bumps
CN108537772B (zh) * 2018-02-08 2020-06-23 杭州蓝雪科技有限公司 贴片电阻正导体印刷缺陷的视觉检测方法
CN108876871B (zh) * 2018-06-15 2022-10-04 广东数相智能科技有限公司 基于圆拟合的图像处理方法、装置与计算机可读存储介质
CN109285181B (zh) * 2018-09-06 2020-06-23 百度在线网络技术（北京）有限公司 用于识别图像的方法和装置
CN110162290B (zh) * 2019-05-28 2022-06-14 易诚高科（大连）科技有限公司 一种针对OLED屏DeMURA数据的压缩方法
CN111462242B (zh) * 2020-03-11 2021-06-18 哈尔滨工业大学 一种基于改进的可变形部件模型的矩形引脚芯片定位方法
CN111257296B (zh) * 2020-03-20 2023-04-11 京东方科技集团股份有限公司 一种检测生物芯片样本的方法、装置及存储介质
CN112419224B (zh) * 2020-07-17 2021-08-27 宁波智能装备研究院有限公司 一种球形引脚芯片定位方法及系统
CN112129780A (zh) * 2020-09-24 2020-12-25 哈尔滨工业大学 Bga芯片焊点质量红外无损检测方法
CN112184715B (zh) * 2020-11-10 2022-07-19 武汉工程大学 一种bga图像的焊点理论中心计算方法
CN112734777B (zh) * 2021-01-26 2022-10-11 中国人民解放军国防科技大学 一种基于簇形状边界闭包聚类的图像分割方法及系统
CN113362290B (zh) * 2021-05-25 2023-02-10 同济大学 点阵平面共线特征快速识别方法、存储设备及装置
CN113945188B (zh) * 2021-09-18 2023-08-08 番禺得意精密电子工业有限公司 分析连接器焊接面在回流焊过程中翘曲的方法及系统
CN114091620B (zh) * 2021-12-01 2022-06-03 常州市宏发纵横新材料科技股份有限公司 一种模板匹配检测方法、计算机设备及存储介质
CN114170191B (zh) * 2021-12-10 2024-08-02 湖北磁创电子科技有限公司 一种网络变压器引脚自动检测机检测图片的处理方法
CN114441555A (zh) * 2022-01-14 2022-05-06 上海世禹精密机械有限公司 自动化焊球阵列封装植球检测系统
CN116298824B (zh) * 2023-05-10 2023-09-15 深圳和美精艺半导体科技股份有限公司 Ic封装基板的测试方法及系统
CN116721105B (zh) * 2023-08-11 2023-10-20 山东淼珠生物科技有限公司 基于人工智能的爆珠生产异常在线检测方法
CN117274246B (zh) * 2023-11-17 2024-02-20 深圳市大族封测科技股份有限公司 一种焊盘识别方法、计算机设备以及存储介质
CN118552539A (zh) * 2024-07-29 2024-08-27 三峡金沙江云川水电开发有限公司 一种基于图像识别技术的座环法兰面垫块识别方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6151406A (en) * 1997-10-09 2000-11-21 Cognex Corporation Method and apparatus for locating ball grid array packages from two-dimensional image data
CN103745475A (zh) * 2014-01-22 2014-04-23 哈尔滨工业大学 一种用于球形引脚元件的检测与定位方法
CN105005997A (zh) * 2015-08-05 2015-10-28 哈尔滨工业大学 一种基于行列直线聚类的多类型bga芯片视觉识别方法
CN105066892A (zh) * 2015-08-05 2015-11-18 哈尔滨工业大学 一种基于直线聚类分析的bga元件检测与定位方法
CN105184770A (zh) * 2015-08-05 2015-12-23 哈尔滨工业大学 一种用于球栅阵列引脚芯片的焊球定位及其参数识别方法

Family Cites Families (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6956963B2 (en) * 1998-07-08 2005-10-18 Ismeca Europe Semiconductor Sa Imaging for a machine-vision system
US6177682B1 (en) * 1998-10-21 2001-01-23 Novacam Tyechnologies Inc. Inspection of ball grid arrays (BGA) by using shadow images of the solder balls
US8306311B2 (en) * 2010-04-14 2012-11-06 Oracle International Corporation Method and system for automated ball-grid array void quantification
JP5574379B2 (ja) * 2010-10-13 2014-08-20 富士機械製造株式会社 画像処理装置及び画像処理方法
CN103761534B (zh) * 2014-01-22 2017-03-01 哈尔滨工业大学 一种用于qfp元件视觉定位的检测方法
US9704232B2 (en) * 2014-03-18 2017-07-11 Arizona Board of Regents of behalf of Arizona State University Stereo vision measurement system and method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6151406A (en) * 1997-10-09 2000-11-21 Cognex Corporation Method and apparatus for locating ball grid array packages from two-dimensional image data
CN103745475A (zh) * 2014-01-22 2014-04-23 哈尔滨工业大学 一种用于球形引脚元件的检测与定位方法
CN105005997A (zh) * 2015-08-05 2015-10-28 哈尔滨工业大学 一种基于行列直线聚类的多类型bga芯片视觉识别方法
CN105066892A (zh) * 2015-08-05 2015-11-18 哈尔滨工业大学 一种基于直线聚类分析的bga元件检测与定位方法
CN105184770A (zh) * 2015-08-05 2015-12-23 哈尔滨工业大学 一种用于球栅阵列引脚芯片的焊球定位及其参数识别方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
CHEN, LIANG ET AL.: "Vision Detection and Locating Algorithm of BGA Package Based on Hough Transform and Point Pattern Matching", ELECTRONICS PROCESS TECHNOLOGY, vol. 29, no. 6, 30 November 2008 (2008-11-30) *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111144478A (zh) * 2019-12-25 2020-05-12 电子科技大学 一种穿帮镜头的自动检测方法
CN111144478B (zh) * 2019-12-25 2022-06-14 电子科技大学 一种穿帮镜头的自动检测方法
CN113836489A (zh) * 2021-09-24 2021-12-24 英特尔产品（成都）有限公司 球栅阵列的缺陷分析方法与设备
CN114387223A (zh) * 2021-12-22 2022-04-22 广东正业科技股份有限公司 一种芯片缺陷视觉检测方法及设备
CN114387223B (zh) * 2021-12-22 2024-04-26 广东正业科技股份有限公司 一种芯片缺陷视觉检测方法及设备

Also Published As

Publication number Publication date
US20170193649A1 (en) 2017-07-06
US9965847B2 (en) 2018-05-08
CN105005997B (zh) 2017-11-14
JP6598162B2 (ja) 2019-10-30
JP2018522293A (ja) 2018-08-09
CN105005997A (zh) 2015-10-28

Similar Documents

Publication Publication Date Title
WO2017020559A1 (zh) 一种基于行列直线聚类的多类型bga芯片视觉识别方法
CN105066892B (zh) 一种基于直线聚类分析的bga元件检测与定位方法
CN105184770B (zh) 一种用于球栅阵列引脚芯片的焊球定位及其参数识别方法
WO2017181724A1 (zh) 电子元件漏件检测方法和系统
Gao et al. A line-based-clustering approach for ball grid array component inspection in surface-mount technology
CN108520514B (zh) 基于计算机视觉的印刷电路板电子元器一致性检测方法
CN113870257B (zh) 印刷电路板缺陷检测分类方法、装置及计算机储存介质
CN106501272B (zh) 机器视觉焊锡定位检测系统
CN110415296B (zh) 一种有阴影光照下矩形状电器件的定位方法
CN109829911B (zh) 一种基于轮廓超差算法的pcb板表面检测方法
CN109449093A (zh) 晶圆检测方法
Liu et al. A novel industrial chip parameters identification method based on cascaded region segmentation for surface-mount equipment
WO2017107529A1 (zh) 一种并排二极管的定位方法及装置
CN116168218A (zh) 一种基于图像识别技术的电路板故障诊断方法
CN110426395B (zh) 一种太阳能el电池硅片表面检测方法及装置
CN116309518A (zh) 一种基于计算机视觉的pcb电路板检测方法及系统
CN113705564B (zh) 一种指针式仪表识别读数方法
CN113393447B (zh) 基于深度学习的针尖正位度检测方法及系统
CN118501177A (zh) 一种化成箔的外观缺陷检测方法及系统
CN112419225B (zh) 一种基于引脚分割的sop型芯片检测方法及系统
CN113744252A (zh) 用于标记和检测缺陷的方法、设备、存储介质和程序产品
CN111508017B (zh) 一种弱对比度定位标记中心的方法和系统
CN113192061A (zh) Led封装外观检测图像的提取方法、装置、电子设备和存储介质
CN108898584B (zh) 一种基于图像分析的全自动贴面电容装焊极性判别方法
CN115375679B (zh) 一种缺陷芯片寻边寻点定位方法及装置

Legal Events

Date Code Title Description
ENP Entry into the national phase

Ref document number: 2016574415

Country of ref document: JP

Kind code of ref document: A

WWE Wipo information: entry into national phase

Ref document number: 15323100

Country of ref document: US

121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 16832049

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 16832049

Country of ref document: EP

Kind code of ref document: A1