Lead bonding welding quality online evaluation method and system

Abstract

The invention belongs to the field of circuit packaging, and particularly discloses a lead bonding welding quality online evaluation method and system, and the method comprises the steps: creating a standard lead template, then collecting a detection image of a to-be-detected workpiece, processing the detection image to obtain the center coordinate of each to-be-detected lead, and obtaining the center coordinate of each to-be-detected lead; calling a standard lead template according to the center coordinates of the to-be-tested leads to match the to-be-tested leads in sequence so as to obtain the contours of the to-be-tested leads; and obtaining the minimum surrounding rectangle of each to-be-tested lead according to the center coordinate and the contour, comparing the minimum surrounding rectangle with the length, the width and the area of the minimum surrounding rectangle of a standard lead, and carrying out weighted calculation to output a quality evaluation score, so as to realize the online evaluation of the lead bonding welding quality. Whether all the leads are detected or not can be accurately judged, meanwhile, whether the leads are too long, too short, excessive welding, insufficient welding, bending and other problems or not can be accurately reflected, and then comprehensive judgment of the overall quality of the leads is achieved.

Description

Technical Field

[0001] The present application belongs to the field of circuit packaging, and more specifically, to an online evaluation method and system for wire bonding welding quality. Background Art

[0002] In the field of high-frequency, high-speed, high-power and high-temperature integrated circuit packaging, wire bonding technology is of great significance. It is mainly used to connect the interface of the chip and the interface of the substrate through high-purity wires. The quality of wire bonding will directly affect the performance of electronic components. Reasonable detection and evaluation of the quality of wire bonding is of great significance in actual production. The lead length is generally 15μm to 250μm. Traditionally, the main way to detect the quality of wire bonding is to observe manually using a microscope. The inspector carefully checks each lead through a microscope to ensure that the position, shape and size of the lead meet the standards. This detection method is inefficient and low in precision, and there is interference from human subjective factors, and the reliability is low. At the same time, there are many leads on electronic components, and the distribution is relatively complex. The traditional manual detection method is time-consuming and labor-intensive, and it is difficult to reasonably evaluate all leads. It is also difficult to have standardized standards for evaluation methods and indicators. Therefore, in the context of increasingly widespread and important applications of wire bonding technology, it is necessary to provide a new wire bonding welding quality inspection and evaluation method that can overcome the related pain points of traditional manual inspection and achieve the purpose of inspection and evaluation with high efficiency, high precision, high reliability and standardized evaluation standards.

[0003] At present, some new detection methods have appeared, such as detection methods based on point cloud deep learning technology, detection methods based on multi-channel information fusion, etc. CN118396961A discloses a SOP type chip pin defect detection method and system based on point cloud deep learning, by pre-processing the point cloud image, extracting the corresponding pin information, and using the pre-built defect detection model, obtaining the defect detection result of the SOP type chip, and realizing the three-dimensional detection of the SOP type chip pin. However, this method only focuses on the detection of pin-related defects, and does not mention whether the lead wire is too long, too short, more welding, less welding, etc., and the overall quality of the lead wire cannot be comprehensively judged. CN117745723A discloses a chip lead wire bonding quality detection method, system and storage medium, proposes an image processing method for multi-channel information fusion, and establishes a mathematical model for chip lead wire bonding quality detection for different quality defects of chip leads. However, this method ignores the detection of whether the lead wire exists and how many leads there are, and only focuses on the quality judgment of the pin and the lead wire alignment, and cannot guarantee whether all the leads in the figure can be detected. Summary of the invention

[0004] In view of the defects of the prior art, the present application provides a method and system for online evaluation of wire bonding welding quality, aiming to solve the problem that the existing lead evaluation method cannot comprehensively judge the overall quality of the lead and cannot ensure that all leads are detected.

[0005] According to one aspect of the present application, a method for online evaluation of wire bonding quality is provided, specifically comprising:

[0006] S1 creates a standard lead template;

[0007] S2: collecting a detection image of the workpiece to be tested, processing the detection image to obtain the center coordinates of each lead to be tested, and then calling a standard lead template according to the center coordinates of each lead to be tested to match them in sequence, so as to obtain the contour of each lead to be tested;

[0008] S3 obtains the minimum surrounding rectangle of each lead to be tested according to the center coordinates and contour obtained in step S2, and compares and weights the minimum surrounding rectangle with the length, width and area of ​​the minimum surrounding rectangle of the standard lead to output a quality evaluation score, thereby realizing online evaluation of wire bonding welding quality.

[0009] Through the above technical scheme conceived by the present application, compared with the prior art, the present application can effectively determine whether all the leads are detected by drawing the outlines of all the leads to be tested, and at the same time, by comparing and weightedly calculating the sizes of the minimum surrounding rectangles of the leads to be tested and the standard leads, the overall quality of the leads to be tested can be comprehensively determined.

[0010] As a further preferred embodiment, step S1 is specifically as follows:

[0011] S11: acquiring a standard image of a standard lead, and then preprocessing the standard image to obtain template information;

[0012] S12 draws the outline of the standard lead to obtain outline information based on the edge information and grayscale information of the preprocessed standard image;

[0013] S13 saves the template information and outline information of the standard lead.

[0014] As a further preferred embodiment, in step S12, a plane projection invariant deformable model or an anisotropic scaling shape model is used to draw the outline of the standard lead.

[0015] As a further preferred embodiment, in step S2, when the lead to be tested is rotationally distorted or partially blocked in the detection image, a plane projection invariant deformable model is used for template matching; when the lead to be tested is scaled and deformed, an anisotropic scaling shape model is used for template matching.

[0016] As a further preferred embodiment, in step S3, the quality evaluation score is calculated as follows:

[0017] Score [i] =0.7×M L[i] +0.2×M W[i] +0.1×M A[i]

[0018] In the formula, Score [i] is the quality evaluation score of the i-th lead, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i] is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas:

[0019] M [i] =min{S,S [i]} / max{S,S [i]}

[0020] Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the minimum surrounding rectangle of the i-th lead to be tested.

[0021] According to another aspect of the present application, there is provided a wire bonding quality online evaluation system, the system comprising a standard module, a wire detection module and a quality evaluation module, wherein:

[0022] The standard module is used to create a standard lead template;

[0023] The lead detection module is used to collect the detection image of the workpiece to be tested, process the detection image to obtain the center coordinates of each lead to be tested, and then call the standard lead template according to the center coordinates of each lead to be tested to match them in turn to obtain the outline of each lead to be tested; the quality evaluation module is used to obtain the minimum surrounding rectangle of each lead to be tested according to the center coordinates and the outline, and compare it with the length, width and area of ​​the minimum surrounding rectangle of the standard lead and perform weighted calculation to output the quality evaluation score, thereby realizing online evaluation of lead bonding welding quality.

[0024] As a further preference, the specific process of the standard module creating a standard lead template is: acquiring a standard image of the standard lead, then preprocessing the standard image to obtain template information, then drawing the outline of the standard lead based on the edge information and grayscale information of the preprocessed standard image to obtain outline information, and finally saving the template information and outline information of the standard lead.

[0025] As a further preference, the standard module uses a plane projection invariant deformable model or an anisotropic scaling shape model to draw the outline of the standard lead.

[0026] As a further preferred embodiment, when the lead to be tested is rotationally distorted or partially occluded in the detection image, the lead detection module uses a plane projection invariant deformable model for template matching; when the lead to be tested is scaled and deformed, the lead detection module uses an anisotropic scaling shape model for template matching.

[0027] As a further preferred method, the quality evaluation score in the quality evaluation module is calculated as follows:

[0028] Score [i] =0.7×M L[i] +0.2×M W[i] +0.1×M A[i]

[0029] In the formula, Score [i] is the quality evaluation score of the i-th lead, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i] is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas:

[0030] M [i] =min{S,S [i]} / max{S,S [i]}

[0031] Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the minimum surrounding rectangle of the i-th lead to be tested.

[0032] In general, the above technical solutions conceived by this application have the following technical advantages compared with the prior art:

[0033] 1. This application matches the lead to be tested with the standard lead template and obtains its outline, which can accurately determine whether all leads are detected. At the same time, this application uses the length, width and area of ​​the minimum surrounding rectangle of the lead to be tested and the standard lead for price comparison and weighted calculation to output the quality evaluation score, which can accurately reflect whether the lead is too long, too short, too many welds, too few welds, bent, etc., thereby achieving a comprehensive judgment of the overall quality of the lead;

[0034] 2. In particular, the present application uses a plane projection invariant deformable model or anisotropic scaling shape model when creating a standard lead template and matching the lead to be tested, which can effectively improve the lead detection accuracy;

[0035] 3. In addition, this application also optimizes the calculation method of the quality evaluation score. By setting the score calculation formula and performing weighted calculation on the scores of length, width and area, the detection accuracy of the overall quality of the lead can be further improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a flow chart of a method for online evaluation of wire bonding quality provided in an embodiment of the present application;

[0037] Figure 2 It is the outline of the standard lead manually drawn after the standard lead is selected in the embodiment of the present application;

[0038] Figure 3 is a detection image of the workpiece to be detected in the embodiment of the present application;

[0039] Figure 4 is the profile and quality evaluation score of all the leads to be tested in the workpiece to be tested in the embodiment of the present application;

[0040] Figure 5 It is an application schematic diagram of the wire bonding welding quality online evaluation system provided in an embodiment of the present application.

[0041] Throughout the drawings, the same reference numerals are used to denote the same elements or structures, wherein:

[0042] 1- movable workbench, 2- workpiece to be tested, 3- split light source, 4- industrial camera, 5- microscope, 6- online evaluation system for wire bonding welding quality. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] like Figure 1 As shown, the present application provides a method for online evaluation of wire bonding welding quality, specifically:

[0045] S1 creates a standard lead template;

[0046] S2 collects the detection image of the workpiece to be tested, processes the detection image to obtain the center coordinates of each lead to be tested, and then calls the standard lead template according to the center coordinates of each lead to be tested to match them in turn to obtain the center coordinates and contours of all leads to be tested. By drawing the contour, it can be intuitively judged whether all leads to be tested in the detection image are detected, thereby ensuring that all leads to be tested are detected. When calling the template, parameters including but not limited to the scaling factor can be adjusted to match different circuit boards. After adjusting the parameters, a preliminary test can be performed, and the accuracy of the test (whether all leads are detected and how the test effect is) can be judged through test pictures or real-time monitoring. If the detection accuracy meets the requirements, it means that the parameters are appropriate, and then all leads will continue to be detected. If the detection accuracy does not meet the requirements, it will return to continue to adjust the parameters until the detection accuracy meets the requirements. In actual applications, the detection accuracy can be adjusted as needed;

[0047] S3 obtains the minimum surrounding rectangle of each lead to be tested according to the center coordinates and contour obtained in step S2, and compares and weights the minimum surrounding rectangle with the length, width and area of ​​the standard lead to output a quality evaluation score. If the lead is too long or too short, bent, missing, too many welds or too few welds, the quality evaluation score will be reduced, thereby realizing online evaluation of wire bonding welding quality.

[0048] The present application can effectively determine whether all leads are detected by drawing the outlines of all leads to be tested, and can comprehensively determine the overall quality of the leads to be tested by comparing and weighting the sizes of the minimum surrounding rectangles of the leads to be tested and the standard leads.

[0049] Further, step S1 is specifically as follows:

[0050] S11 collects a standard image of a workpiece with a standard lead through a microscope and an industrial camera, and then preprocesses the standard image to obtain template information. The template drawing method can adopt the grayscale matching model based on NCC, the feature-based matching model, and the shape-based matching model. The template information includes grayscale removal, edge information, etc.

[0051] S12 draws the outline of the standard lead based on the edge information and grayscale information of the preprocessed standard image to obtain outline information. The present application uses a manual method to draw the outline, and a program can also be used to automatically draw the outline.

[0052] S13 saves the template information and contour information of the standard lead. The contour information is saved mainly for subsequent detection of the lead to be tested, for visualizing the final result, and for judging whether all the lead to be tested are detected.

[0053] Furthermore, in step S12 and step S2, since the lead styles on different types of circuit boards may be different, in order to achieve the purpose of having a good detection effect on the leads on different types of circuit boards, a plane projection invariant deformable model or anisotropic scaling shape model is used to draw the outline of the standard lead. The above two models are both existing mathematical methods. This application applies them to lead detection, which can detect all the leads in the image and output the center coordinates of each lead, effectively solving the problem of poor lead detection effect of traditional template matching algorithms.

[0054] Template matching is to search for one or several areas corresponding to the template image in the entire image. Considering that the lead to be tested presents the overall characteristics of light-dark-light in the image, using the template matching algorithm to detect the lead to be tested can achieve better detection results.

[0055] The plane projection invariant deformable model algorithm is mainly used to detect the projection images of planar objects at different viewing angles. It utilizes the properties of the projection transformation of planar objects. Even if the object undergoes rotation, scaling, translation and other transformations, its projection on the image plane still maintains a certain geometric relationship. At the same time, the model mainly focuses on the geometric shape of the object and is less sensitive to changes in illumination. The anisotropic scaling shape model algorithm is mainly used to detect objects with specific shape and orientation characteristics, and can capture the shape and orientation information of the object. It establishes a statistical model by learning the shape and orientation information of the object, which can describe the shape, orientation and range of variation of the object, and has a certain robustness to transformations such as scaling, rotation, and translation of the object.

[0056] Among them, the key parameters of the above two template matching algorithms are image pyramid level, model scaling factor, minimum score threshold, maximum overlap, etc. The image pyramid is to generate a series of images using the original image. The original image is the first layer of the pyramid level and has the highest resolution. The higher the level, the smaller the image and the lower the image resolution. In the image processing algorithm, multi-resolution operations such as image pyramid can avoid being trapped in local points and enhance the robustness of the model. Selecting a suitable image pyramid level can greatly improve the search efficiency and search accuracy and obtain the best matching results. The model scaling factor is used to set the upper limit and lower limit of the maximum size of the lead that can be detected, because the size of the template lead is not exactly the same as the actual size of the lead detected, and the size of the lead in the actual wire bonding process is large and small; the minimum score threshold is used to exclude those matching results whose algorithm scores are lower than the threshold. Therefore, the reasonable setting of the model scaling factor and the minimum score threshold can avoid the occurrence of false detection and missed detection of the lead to the greatest extent and improve the detection accuracy.

[0057] In step S2, the standard lead outline is transformed to each detected lead using affine transformation to display all detection results. Specifically, after the lead is detected, the lead center coordinates are output, and then the affine transformation is calculated based on the point correspondence and the two corresponding angles, that is, the transformation consisting of rotation and translation. The transformation matrix HomMat2D consists of two parts, the rotation matrix R and the translation vector t:

[0058]

[0059] The standard lead template outline is transformed from the upper left corner to the center coordinates of each lead to display the detected lead results. The coordinates and angles of the standard lead template outline at the center of the upper left corner are input (Row1, Column1, Angle1). After the following matrix calculation, the coordinates and angles of the standard lead template transformed to the center of each lead are output as (Row2, Column2, Angle2).

[0060]

[0061] Preferably, in step S2, the detection mode can be flexibly adjusted according to the lead type or actual working conditions during the detection process. When the lead to be tested is rotationally distorted or partially occluded in the detection image, a plane projection invariant deformable model is used for template matching; when the lead to be tested is scaled and deformed, an anisotropic scaling shape model is used for template matching.

[0062] Further, in step S3, in order to determine the geometric dimensions of each lead to be tested, the outline of the standard lead and the outline of each lead to be tested detected are calculated to obtain the smallest surrounding rectangle in any direction, that is, the rectangle with the smallest area among all rectangles containing the standard lead or the lead to be tested (the smallest surrounding rectangle can be obtained by the method in the prior art). The calculation of the rectangle is based on the center coordinates of the regional pixels, that is, the center coordinates of the standard lead or the lead to be tested, so it is necessary to obtain the smallest surrounding rectangle based on the center coordinates and the outline.

[0063] The quality evaluation score is calculated as follows:

[0064] Score [i] =0.7×M L[i] +0.2×M W[i] +0.1×M A[i]

[0065] In the formula, Score [i] is the quality evaluation score of the i-th lead, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i]is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas:

[0066] M [i] =min{S,S [i]} / max{S,S [i]}

[0067] Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the smallest surrounding rectangle of the i-th lead to be tested, and the above-mentioned sizes are length, width or area respectively.

[0068] According to another aspect of the present application, there is provided a wire bonding quality online evaluation system, the system comprising a standard module, a wire detection module and a quality evaluation module, wherein:

[0069] The standard module is used to create a standard lead template, including collecting a standard image of a standard lead, and then preprocessing the standard image to obtain template information. The template drawing method can adopt an NCC-based grayscale matching model, a feature-based matching model, or a shape-based matching model. Then, based on the edge information and grayscale information of the preprocessed standard image, the outline of the standard lead is drawn to obtain outline information, and finally the template information and outline information of the standard lead are saved.

[0070] The lead detection module is used to collect the detection image of the workpiece to be tested, and call the standard lead template to match the leads to be tested in the detection image in turn to obtain the center coordinates and contours of all the leads to be tested. Template matching is to search for one or several areas corresponding to the template image in the entire image. Considering that the leads to be tested show the characteristics of light-dark-light in the image as a whole, using the template matching algorithm to detect the leads to be tested can achieve better detection results;

[0071] The quality evaluation module is used to obtain the minimum surrounding rectangle of each lead to be tested according to the center coordinates and contour, and compare and weightedly calculate the minimum surrounding rectangle with the length, width and area of ​​the minimum surrounding rectangle of the standard lead to output the quality evaluation score, thereby realizing the online evaluation of the wire bonding welding quality.

[0072] Further, if Figure 5 As shown, the standard module and the lead detection module both use image acquisition units to collect standard images and images to be tested, specifically including a microscope 5, an industrial camera 4 and a split light source 3. The standard workpiece or the workpiece to be tested 2 is fixed on a movable workbench 1. The displacement of the standard workpiece or the workpiece to be tested 2 is controlled by the movable workbench 1, and the lens is aimed at the position where the lead exists to collect the standard image and the image to be tested, and send them to the lead bonding welding quality online evaluation system 6 to display the detection results and scores.

[0073] Furthermore, the lead styles on different types of circuit boards may be different. In order to achieve the goal of having a good detection effect on the leads on different types of circuit boards, a plane projection invariant deformable model or anisotropic scaling shape model is used to draw the outline of the standard lead. The above two models are both existing mathematical methods. This application applies them to lead detection, which can detect all leads in the image and output the center coordinates of each lead; use affine transformation to transform the standard lead outline to each detected lead to display all detection results.

[0074] After the lead is detected, the center coordinates of the lead are output, and then the affine transformation is calculated based on the point correspondence and the two corresponding angles, that is, the transformation consisting of rotation and translation. The transformation matrix HomMat2D consists of two parts, the rotation matrix R and the translation vector t:

[0075]

[0076] The standard lead template outline is transformed from the upper left corner to the center coordinates of each lead to display the detected lead results. When the standard lead template outline is in the upper left corner, the center coordinates and angles are passed in (Row1, Column1, Angle1). After the following matrix calculation, the standard lead template is transformed to the coordinates and angles of the center of each lead and are output as (Row2, Column2, Angle2).

[0077]

[0078] Preferably, in step S2, the detection mode can be flexibly adjusted according to the lead type or actual working conditions during the detection process. When the lead to be tested is rotationally distorted or partially occluded in the detection image, a plane projection invariant deformable model is used for template matching; when the lead to be tested is scaled and deformed, an anisotropic scaling shape model is used for template matching.

[0079] Furthermore, in order to determine the geometric dimensions of each lead to be tested, the minimum surrounding rectangle in any direction is calculated for the outline of the standard lead and the outline of each lead to be tested, that is, the rectangle with the smallest area among all rectangles containing the standard lead or the lead to be tested (the calculation method of the minimum surrounding rectangle is a mathematical calculation method in the prior art). The calculation of the rectangle is based on the center coordinates of the regional pixels, that is, the center coordinates of the standard lead or the lead to be tested, so it is necessary to obtain the minimum surrounding rectangle based on the center coordinates and the outline.

[0080] The quality evaluation score is calculated as follows:

[0081] Score [i] =0.7×M L[i] +0.2×MW[i] +0.1×M A[i]

[0082] In the formula, Score [i] is the quality evaluation score of the i-th lead, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i] is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas:

[0083] M [i] =min{S,S [i]} / max{S,S [i]}

[0084] Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the smallest surrounding rectangle of the i-th lead to be tested, and the above-mentioned sizes are length, width or area respectively.

[0085] The technical solution provided by this application is further described below based on specific embodiments.

[0086] S1 creates a standard lead template, such as Figure 2 As shown, the red one is the outline of the standard lead;

[0087] S2 collects the inspection image of the workpiece to be inspected, such as Figure 3 As shown, the standard lead template is called to match the lead wires to be tested in the detection image in sequence to obtain the center coordinates and contours of all the lead wires to be tested;

[0088] S3 obtains the minimum surrounding rectangle of each lead to be tested according to the center coordinates and contour obtained in step S2, and compares the minimum surrounding rectangle with the length, width and area of ​​the minimum surrounding rectangle of the standard lead and performs weighted calculation to output a quality evaluation score, such as Figure 4 As shown, the red color is the outline of the lead to be tested, and the numbers above each lead to be tested are the lead number and quality evaluation score, for example, 1:78 means that the quality evaluation score of the first lead to be tested is 78, and the quality evaluation score is set to be above 60 to be qualified.

[0089] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for online evaluation of wire bonding quality, characterized in that: The online evaluation method is as follows: S1 creates a standard lead template; S2: collecting a detection image of the workpiece to be tested, processing the detection image to obtain the center coordinates of each lead to be tested, and then calling a standard lead template according to the center coordinates of each lead to be tested to match them in sequence, so as to obtain the contour of each lead to be tested; S3 obtains the minimum surrounding rectangle of each lead to be tested according to the center coordinates and contour obtained in step S2, and compares and weights the minimum surrounding rectangle with the length, width and area of ​​the minimum surrounding rectangle of the standard lead to output a quality evaluation score, thereby realizing online evaluation of wire bonding welding quality.

2. The online evaluation method for wire bonding quality according to claim 1, characterized in that: Step S1 is specifically as follows: S11: acquiring a standard image of a standard lead, and then preprocessing the standard image to obtain template information; S12 draws the outline of the standard lead to obtain outline information based on the edge information and grayscale information of the preprocessed standard image; S13 saves the template information and outline information of the standard lead.

3. The online evaluation method for wire bonding quality according to claim 2, characterized in that: In step S12, the outline of the standard lead is drawn using a plane projection invariant deformable model or an anisotropic scaling shape model.

4. The online evaluation method for wire bonding quality according to claim 1, characterized in that: In step S2, when the lead to be tested is rotationally distorted or partially blocked in the detection image, a plane projection invariant deformable model is used for template matching; when the lead to be tested is scaled and deformed, an anisotropic scaling shape model is used for template matching.

5. The method for online evaluation of wire bonding quality according to claim 1, wherein: In step S3, the quality evaluation score is calculated as follows: Score [i] =0.7×M L[i] +0.2×M W[i] +0.1×M A[i] In the formula, Score [i] is the quality evaluation score of the i-th lead to be tested, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i] is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas: M [i] =min{S,S [i] } / max{S,S [i] } Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the minimum surrounding rectangle of the i-th lead to be tested.

6. A wire bonding quality online evaluation system, characterized in that: The online evaluation system includes a standard module, a lead detection module and a quality evaluation module, wherein: The standard module is used to create a standard lead template; The lead detection module is used to collect the detection image of the workpiece to be tested, process the detection image to obtain the center coordinates of each lead to be tested, and then call the standard lead template according to the center coordinates of each lead to be tested to match them in turn to obtain the outline of each lead to be tested; the quality evaluation module is used to obtain the minimum surrounding rectangle of each lead to be tested according to the center coordinates and the outline, and compare it with the length, width and area of ​​the minimum surrounding rectangle of the standard lead and perform weighted calculation to output the quality evaluation score, thereby realizing online evaluation of lead bonding welding quality.

7. The wire bonding quality online evaluation system according to claim 6, characterized in that: The specific process of the standard module creating a standard lead template is: collecting a standard image of the standard lead, then preprocessing the standard image to obtain template information, then drawing the outline of the standard lead based on the edge information and grayscale information of the preprocessed standard image to obtain outline information, and finally saving the template information and outline information of the standard lead.

8. The wire bonding quality online evaluation system according to claim 7, characterized in that: The standard module uses a planar projection invariant deformable model or an anisotropic scaling shape model to draw the outline of the standard lead.

9. The wire bonding quality online evaluation system according to claim 6, characterized in that: When the lead to be tested is rotationally distorted or partially blocked in the detection image, the lead detection module uses a plane projection invariant deformable model for template matching; when the lead to be tested is scaled and deformed, the lead detection module uses an anisotropic scaling shape model for template matching.

10. The wire bonding quality online evaluation system according to claim 6, characterized in that: The calculation method of the quality evaluation score in the quality evaluation module is: Score [i] =0.7×M L[i] +0.2×M W[i] +0.1×M A[i] In the formula, Score [i] is the quality evaluation score of the i-th lead to be tested, M L[i] is the length fraction of the smallest surrounding rectangle of the i-th lead to be tested, M W[i] M is the width fraction of the smallest surrounding rectangle of the i-th lead to be tested. A[i] is the area fraction of the smallest surrounding rectangle of the i-th lead to be tested. The length fraction, width fraction and area fraction are calculated using the following formulas: M [i] =min{S,S [i] } / max{S,S [i] } Where M [i] is the size fraction, S is the size of the minimum surrounding rectangle of the standard lead, S [i] is the size of the minimum surrounding rectangle of the i-th lead to be tested.

Images