A wire bonding quality visual inspection method, system, medium and terminal

Abstract

The present application discloses a method, system, terminal, and medium for visual inspection of wire bonding quality, belonging to the field of visual recognition and semiconductor packaging technology. The inspection method comprises the following steps: obtaining an image of the area to be inspected, converting it into a grayscale image, and performing preprocessing; constructing a detection model based on the preprocessed image; performing a preliminary judgment based on the detection model, obtaining a prediction result, and generating a control instruction; and updating the detection model according to the control instruction. The present application provides a method, system, terminal, and medium for visual inspection of wire bonding quality, which solves the problems of low efficiency and insufficient accuracy of traditional manual inspection, realizes efficient and accurate inspection of the wire bonding quality of semiconductor devices, significantly improves the degree of automation of the inspection process, reduces the subjectivity and uncertainty of manual inspection, and also reduces the reliance on highly skilled operators.

Description

Technical Field

[0001] The present application belongs to the field of visual recognition and semiconductor packaging technology, and in particular relates to a wire bonding quality visual inspection method, system, medium and terminal. Background Art

[0002] Wire bonding is a key technology in semiconductor packaging, primarily used to establish electrical connections between the silicon die and the external leads of a semiconductor device. This technology is crucial to ensuring the functionality, reliability, and performance of semiconductor devices. It is widely used to interconnect various microelectronic components, such as semiconductor devices, chips, LEDs, and printed circuit boards.

[0003] Wire bonding challenges during semiconductor packaging include wire droop, wire wander, and coiling. Wire droop, when the wire stretches or sags under tension, can result in inadequate contact and compromised electrical performance. Wire wander, when the wire moves laterally during the bonding process, can cause misalignment and unreliable connections. Furthermore, coiling caused by unintentional excess wire can compromise bond quality and device functionality.

[0004] To improve the overall quality of packaged products, quality control of the product after wire bonding is crucial. Common testing methods include pull testing, ball shear testing, visual inspection, and automated optical inspection (AOI).

[0005] However, these testing methods require appropriate testing tools, equipment, and skilled personnel. Manual testing has an unstable false detection and missed detection rate, while machine testing, such as automated optical inspection (AOI), requires a certain investment and is difficult to guarantee foolproofness for a wide variety of products, especially the various postures that non-rigid wires may assume. This also increases the threshold for skilled personnel. Summary of the Invention

[0006] In order to solve the above-mentioned problems in the prior art, the present application provides a wire bonding quality visual inspection method and system thereof, which uses machine vision and automated image analysis to efficiently and accurately detect the wire bonding quality of semiconductor devices.

[0007] To achieve the above objectives, the specific technical solutions of a wire bonding quality visual inspection method and system of the present application are as follows:

[0008] A wire bonding quality visual inspection method comprises the following steps:

[0009] S1. Obtain an image of the area to be detected, convert it into a grayscale image and perform preprocessing;

[0010] S2. Build a detection model based on the preprocessed image;

[0011] S3. Make a pre-judgment based on the detection model, obtain the prediction results, and generate control instructions;

[0012] S4. Update the detection model according to the control instruction.

[0013] Furthermore, based on the grayscale image, image noise is reduced by blurring and morphological erosion and dilation operations. The morphological dilation is the maximum pixel value in the output domain, and the morphological erosion operation is the minimum pixel value in the output domain. The specific formula is as follows:

[0014]

[0015] Among them, A is a set of grayscale images, B is a structural element that is considered as a subset of A, is the morphological expansion operation symbol, It is the symbol of morphological corrosion operation.

[0016] Furthermore, preprocessing is performed based on the converted grayscale image to obtain the coordinates of two target points of the conductor to be tested in the detection area image, that is, to obtain the position of the conductor;

[0017] Based on the image of the area to be detected and the coordinates of the two target points, the maximum inscribed circle of the target point position is fitted, the radius and the overall routing angle of the wire are calculated, and a set of moving points that meet the conditions is obtained;

[0018] Based on the features of the coordinate points in the set, different scores and correctability rates are assigned to build a detection model and train the detection model.

[0019] Furthermore, the calculation steps for the radius of the fitted maximum inscribed circle are:

[0020] The radius coordinates parallel to the coordinate axis satisfy,

[0021] N j (P1)={(x1+1,y1),(x1+2,y1),...,(x1+j,y1)},

[0022] N k (P1)={(x1-1,y1),(x1-2,y1),...,(x1-k,y1)},

[0023] N l (P1)={(x1,y1+1), (x1,y1+2),..., (x1,y1+l)},

[0024] N m (P1)={(x1, y1-1)(x1, y1-2),..., (x1, y1-m)}, (1)

[0025] Among them, P1 is the center of the circle, its coordinates are (x1, y1), and its grayscale value is V1, N j , N k , N l , N m are the point sets of one of the radii of the possible maximum inscribed circle on the image,

[0026] And the gray value of the point in the point set satisfies,

[0027] V j , V k , V l , V m ∈[V1-range,V1+range] (2)

[0028] Among them, V j , V k , V l , V m is the grayscale value of the coordinate point in each set, and range is the tolerable grayscale error.

[0029] Combining formula (1) and formula (2), the formula for calculating the maximum inscribed circle radius is:

[0030] =min(j,k,l,m),

[0031] Among them, j, k, l, m are N j , N k , N l , N m The number of points in the point set.

[0032] Furthermore, the steps to calculate the overall routing angle of the wire are as follows:

[0033] Based on the preprocessed image, calculate the straight-line distance between the two targets,

[0034] Based on one of the target points, move to another target point with a fixed step and a continuously updated slope.

[0035] Based on the mobile points that meet the conditions during movement,

[0036] Based on the distance between the moving point and the circle, the coordinates of the moving point after a fixed rotation step are calculated and added to the point set to simulate the random posture of the non-rigid wire.

[0037] Furthermore, when the detection model determines that the wire integrity score in the image of the area to be detected reaches the minimum acceptance threshold, it is determined that the wire integrity meets the requirements;

[0038] When the detection model determines that the wire integrity score in the image of the area to be detected is lower than the minimum acceptance threshold, it is determined that the wire integrity does not meet the requirements.

[0039] Furthermore, when the detection model determines that the wire integrity does not meet the requirements, manual judgment is made as to whether the detection results meet expectations.

[0040] If the expectation is not met, update the detection model parameters;

[0041] If the expectations are met, the wire integrity is determined to meet the requirements.

[0042] The present application also provides a wire bonding quality visual inspection system, based on the above method, comprising:

[0043] The acquisition module is used to acquire the image of the area to be detected, convert it into a grayscale image and perform preprocessing;

[0044] Model building module, used to build a detection model based on the preprocessed image;

[0045] The pre-judgment module is used to make pre-judgments based on the detection model, obtain prediction results, and generate control instructions;

[0046] The model callback module is used to update the detection model according to control instructions.

[0047] The present application also provides a computer-readable storage medium having a computer program stored thereon, which implements the steps of the above method when executed by a processor.

[0048] The present application also provides a terminal, including a terminal body and a controller, the controller including a memory, a processor and a computer program stored in the memory and executable on the processor, and the steps of the above method are implemented when the processor executes the program.

[0049] The technical solution includes at least the following technical effects:

[0050] By leveraging machine vision, automated image analysis, and image processing algorithms, this system overcomes the low efficiency and inaccuracy of traditional manual inspection, enabling efficient and accurate quality testing of semiconductor device wire bonds. This significantly increases the automation level of the inspection process, reduces the subjectivity and uncertainty inherent in manual inspection, and reduces the reliance on highly skilled operators. The system is highly adaptable to varying image quality and lighting conditions, demonstrating strong anti-interference capabilities. Integrated into wire bonders, this inspection system reduces both additional equipment investment and labor costs, resulting in significant cost savings.

[0051] It should be understood that the foregoing general description and the following detailed description are merely illustrative and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0053] Figure 1 A schematic diagram of a wire bonding quality visual inspection system provided by an exemplary embodiment of the present application is shown;

[0054] Figure 2 A schematic diagram showing a posture of a non-rigid material wire detected according to an exemplary embodiment of the present application is shown;

[0055] Figure 3 A schematic diagram of prediction results of a simulation detection model provided by an exemplary embodiment of the present application is shown;

[0056] Figure 4 A flow chart of a wire bonding quality visual inspection method provided by an exemplary embodiment of the present application is shown.

[0057] Figure 5 A flow chart for calculating the overall routing angle of a conductor provided by an exemplary embodiment of the present application is shown. DETAILED DESCRIPTION

[0058] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0059] A wire bonding quality visual inspection method and system in an embodiment of the present application mainly detects the position and integrity of the wire after wire bonding. By analyzing and processing the images captured by the onboard visual platform, it is detected whether there are integrity problems such as missing welds and uneven welds in the product after welding.

[0060] The wire bonding quality visual inspection method comprises the following steps:

[0061] S1. Obtain an image of the area to be detected, convert it into a grayscale image and perform preprocessing;

[0062] S2. Build a detection model based on the preprocessed image;

[0063] S3. Make a pre-judgment based on the detection model, obtain the prediction results, and generate control instructions;

[0064] S4. Update the detection model according to the control instruction.

[0065] Figure 1 A schematic diagram of a wire bond quality visual inspection system, provided by an exemplary embodiment of the present application, is shown. Based on a wire bond quality visual inspection method, the system includes an acquisition module, a model construction module, a pre-judgment module, and a model callback module. The acquisition module is used to acquire an image of the area to be inspected, convert it into a grayscale image, and perform pre-processing. The model construction module is used to construct an inspection model based on the pre-processed image. The pre-judgment module is used to perform a pre-judgment based on the inspection model, obtain a prediction result, and generate a control instruction. The model callback module is used to update the inspection model based on the control instruction.

[0066] In one embodiment, a wire bond quality visual inspection system is integrated into the wire bonding machine software. It features two modes: "while-bonding inspection" and "post-bonding inspection." These modes allow for immediate inspection of the bond quality of the smallest unit wire bond, and inspection of all positions after wire bonding. This allows wire bonding and inspection to be performed on the same machine, reducing additional time and capital investment. The system's high level of integration and the comprehensive prior knowledge built into the algorithm significantly reduce operator skill requirements. For products with relatively stable wire positions, an optimal model can be constructed from a standard post-bonding product image, meeting accuracy, speed, and real-time requirements. Furthermore, the system offers high robustness in terms of imaging quality and light quality, as well as strong anti-interference capabilities.

[0067] like Figure 4 FIG. 1 is a flow chart of a method for visually inspecting wire bonding quality, the method comprising:

[0068] Step 401: Obtain an image of the area to be detected and convert it into a grayscale image;

[0069] Step 402 , performing preprocessing based on the converted grayscale image to obtain the coordinates of two target points of the conductor to be tested in the detection area image, that is, obtaining the position of the conductor;

[0070] In the embodiment of the present application, based on the converted grayscale image, the interference of image noise on accuracy is reduced by blurring and morphological corrosion and dilation operations.

[0071] Among them, the morphological expansion is the maximum pixel value in the output area, and the morphological erosion operation is the minimum pixel value in the output area. The formula is:

[0072]

[0073] Among them, A is a set of grayscale images, B is a structural element that is considered as a subset of A, is the morphological expansion operation symbol, It is the symbol of morphological corrosion operation.

[0074] Step 403: Based on the image of the area to be detected and the coordinates of the two target points, the maximum inscribed circle of the target point position is fitted, the radius and the overall routing angle of the wire are calculated, and a set of moving points that meet the conditions is obtained;

[0075] In the embodiment of the present application, the coordinates of two points P1 and P2 of the conductor to be tested in the image of the area to be tested are obtained, where P1 is the first welding point and P2 is the second welding point;

[0076] Based on the image of the area to be detected and the coordinates of two points, a maximum inscribed circle is fitted between the two points, and the radius and its grayscale range are calculated. The method for fitting the maximum inscribed circle is as follows: the adjacent points q of a coordinate point p in a certain direction are calculated. If the pixel values of the adjacent points q are within a set range, the two pixels are considered connected. In this direction, if point q is connected to a third pixel, and the third pixel is connected to point p, points p and q are also considered connected. If there is a path between points p and q, and every point on the path is connected to the next pixel, then the two points p and q are connected.

[0077] The steps to calculate the radius of the fitted maximum inscribed circle are as follows:

[0078] The radius coordinates parallel to the coordinate axis satisfy:

[0079] N j (P1)={(x1+1,y1),(x1+2,y1),...,(x1+j,y1)},

[0080] N k (P1)={(x1-1,y1),(x1-2,y1),...,(x1-k,y1)},

[0081] N l (P1)={(x1,y1+1),(x1,y1+2),...,(x1,y1+l)},

[0082] N m (P1) = {(x1, y 1-1 ), (x1, y 1-2 ),...,(x1,y 1-m )}, (1)

[0083] Among them, P1 is the center of the circle, its coordinates are (x1, y1), and its grayscale value is V1, N j , N k , N l , N m are the point sets of one of the radii of the possible maximum inscribed circle on the image,

[0084] And the gray value of the point in the point set satisfies,

[0085] V j , V k , V l , V m ∈[V1-range,V1+range] (2)

[0086] Among them, V j , V k , V l , V m is the grayscale value of the coordinate point in each set, and range is the tolerable grayscale error.

[0087] Combining formula (1) and formula (2), the formula for calculating the maximum inscribed circle radius is:

[0088] r=min(j,k,l,m),

[0089] Among them, j, k, l, m are N j , N k , N l , N m The number of points in the point set.

[0090] Based on the coordinates of the first solder point and the second solder point and the maximum inscribed circle radius of the fitted first solder point and the second solder point, the overall routing angle of the wire is calculated. The steps of calculating the overall routing angle of the wire are as follows:

[0091] Calculate the straight line l formed by the first and second welding points using the following formula:

[0092] (y-y2) / (y1-y2)=(x-x2) / (x1-x2), (x1-x2)≠0,

[0093] Wherein, x1, y1 are the coordinates of the first welding point P1, and x2, y2 are the coordinates of the second welding point P2;

[0094] The formula for calculating the maximum inscribed circles R1 and R2 at the first and second welding points is:

[0095] (x-x1) 2 +(y-y1) 2 =r1 2 ,

[0096] (x-x2) 2 +(y-y2) 2 =r2 2 ,

[0097] Where x1, y1 are the coordinates of the first welding point P1, x2, y2 are the coordinates of the second welding point P2, r1, r2 are the radii of circles R1, R2;

[0098] Calculate the straight-line distance between two points based on the centers P3 and P4 of the two circles R1 and R2;

[0099] Starting from R3, move the fixed steps in sequence, and continuously update the slope to correct the direction. Calculate the coordinates of each moving point that meets the conditions to form a set N M ;

[0100] Based on the distance between the moving point and the circles R1 and R2, the coordinates of the moving point after rotating by a specific step length are calculated and passed to the set N. M , to simulate the deformation posture of non-rigid materials, such as Figure 3 Schematic diagram of the prediction results of the simulation detection model shown.

[0101] Step 404: Based on the features of the coordinate points in the set, different scores and correctability rates are assigned to construct a detection model and train the detection model;

[0102] In the embodiment of the present application, based on the set N M The characteristics of the mid-coordinate point are assigned a score w and a correctability rate e to construct a model.

[0103] Step 405: determining the integrity of the wire in the image of the area to be inspected based on the inspection model;

[0104] In an embodiment of the present application, during prediction, the difference between the coordinates of the first welding point and the second welding point in the image to be tested and the coordinates of the first welding point and the second welding point in the image to be tested is calculated as an offset, and the detection score is calculated using the offset model.

[0105] Step 406: When the detection model determines that the wire integrity score in the image of the area to be detected reaches the minimum acceptance threshold, it is determined that the wire integrity meets the requirements;

[0106] Step 407: When the detection model determines that the wire integrity score in the image of the area to be detected is lower than the minimum acceptance threshold, it is determined that the wire integrity does not meet the requirements;

[0107] Step 408: When the detection model determines that the wire integrity does not meet the requirements, a manual determination is made as to whether the detection result meets expectations.

[0108] Step 409: If the expectation is not met, update the detection model parameters; if the expectation is met, determine that the wire integrity meets the requirements.

[0109] In the embodiment of the present application, based on the manual judgment of the detection results, N in the detection model M The coordinate point is corrected according to the correctability rate e.

[0110] like Figure 5 The flowchart for calculating the overall conductor routing angle is shown in the figure. The specific steps are as follows:

[0111] Step 501: Calculate the straight-line distance between two targets based on the pre-processed image.

[0112] Step 502: Based on one of the target points, move toward another target point with a fixed step length and a continuously updated slope.

[0113] Step 503: forming a set based on the moving points that meet the conditions during movement.

[0114] Step 504 : Based on the coordinate points in the set, the coordinates of the point after being rotated by a fixed step are calculated and added to the point set to simulate the random posture of the non-rigid wire.

[0115] like Figure 3 The schematic diagram of the simulation model prediction results shown in the figure has a preset tolerable offset angle of 10 degrees. The welding point detection model inputs the image of the area to be detected for prediction. If the prediction result reaches the minimum score value, it is determined that the wire integrity meets the requirements; if the prediction result does not reach the minimum score value, it is determined that the wire integrity does not meet the requirements.

[0116] A computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, the steps of a method for visually inspecting wire bonding quality are implemented.

[0117] A terminal is provided, comprising a terminal body and a controller. The controller comprises a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, steps of a method for visually inspecting wire bonding quality are implemented.

[0118] In summary, a wire bond quality visual inspection method and system can construct an in-algorithm inspection model by inputting at least one pair of solder joints (i.e., the first and second solder joints) during the model building phase. A more refined model can be constructed by inputting at least one standard post-weld product image. The number of training images required for optimal results varies depending on parameters such as trace angle and wire material. During the prediction phase, custom screening of post-weld products can be achieved by controlling two preset parameters: maximum tolerance deviation and maximum tolerance offset angle. The screening results are updated in real time on the display interface, and products that do not meet the preset standards are displayed and an alarm is issued. The operator can handle the alarm results, including stopping and resuming inspection, and updating the model. Updating the model automatically transmits a captured product image at that location back to the algorithm to update the model parameters.

[0119] By leveraging machine vision, automated image analysis, and image processing algorithms, this system overcomes the low efficiency and inaccuracy of traditional manual inspection, enabling efficient and accurate quality testing of semiconductor device wire bonds. This significantly increases the automation level of the inspection process, reduces the subjectivity and uncertainty inherent in manual inspection, and reduces the reliance on highly skilled operators. The system is highly adaptable to varying image quality and lighting conditions, demonstrating strong anti-interference capabilities. Integrated into wire bonders, this inspection system reduces both additional equipment investment and labor costs, resulting in significant cost savings.

[0120] Those skilled in the art will readily appreciate other embodiments of the present invention after considering the specification and practicing the application herein. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not covered by this application. The description and examples are to be considered merely as exemplary, and the true scope and spirit of the present invention are indicated by the appended claims.

[0121] It should be understood that the present application is not limited to the exact structures described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A wire bonding quality visual inspection method, characterized in that: include: S1. Obtain an image of the area to be detected, convert it into a grayscale image and perform preprocessing; S2. Construct a detection model based on the preprocessed image, perform preprocessing based on the converted grayscale image, obtain the coordinates of the two target points of the wire to be tested in the detection area image, that is, obtain the position of the wire, fit the maximum inscribed circle of the target point position based on the image of the detection area and the coordinates of the two target points, calculate the radius and the overall alignment angle of the wire, and form a set of moving points that meet the conditions in the movement. Based on the characteristics of the coordinate points in the set, assign different scores and correction rates to construct a detection model, train the detection model, and calculate the overall alignment angle of the wire in the following steps: calculate the straight-line distance between the two targets based on the preprocessed image, move from one target point to the other target point with a fixed step and a continuously updated slope, form a set of moving points that meet the conditions in the movement, and calculate the coordinates of the moving point after rotating the fixed step based on the distance of the moving point from the circle, and add them to the point set to simulate the random posture of the non-rigid wire; S3. Make a pre-judgment based on the detection model, obtain the prediction results, and generate control instructions; S4. Update the detection model according to the control instruction.

2. The wire bonding quality visual inspection method according to claim 1, wherein: It includes reducing image noise based on grayscale images through blurring and morphological erosion and dilation operations, where morphological dilation is the maximum pixel value in the output domain, and morphological erosion is the minimum pixel value in the output domain. The specific formula is: in, is a set of grayscale images, to be regarded as The structural elements of the subset, is the morphological expansion operation symbol, It is the symbol of morphological corrosion operation.

3. The wire bonding quality visual inspection method according to claim 1, wherein: The calculation steps for the radius of the fitted maximum inscribed circle are: The radius coordinates parallel to the coordinate axis satisfy, （1） in, is the center of the circle, and its coordinates are , whose gray value is , are the point sets of one of the radii of the maximum inscribed circle on the image, And the gray value of the point in the point set satisfies, （2） in, is the grayscale value of the coordinate point in each set, is the tolerable grayscale error, Combining formula (1) and formula (2), the formula for calculating the maximum inscribed circle radius is: in, for The number of points in the point set.

4. The wire bonding quality visual inspection method according to claim 1, wherein: include: When the detection model determines that the wire integrity score in the image of the area to be detected reaches the minimum acceptance threshold, the wire integrity is determined to meet the requirements; When the detection model determines that the wire integrity score in the image of the area to be detected is lower than the minimum acceptance threshold, it is determined that the wire integrity does not meet the requirements.

5. The wire bonding quality visual inspection method according to claim 4, wherein: include: When the detection model determines that the wire integrity does not meet the requirements, manual judgment is made as to whether the detection results meet expectations. If the expectation is not met, update the detection model parameters; If the expectations are met, the wire integrity is determined to meet the requirements.

6. A wire bonding quality visual inspection system, characterized in that: Based on the wire bonding quality visual inspection method according to any one of claims 1 to 5, the system comprises: The acquisition module is used to acquire the image of the area to be detected, convert it into a grayscale image and perform preprocessing; Model building module, used to build a detection model based on the preprocessed image; The pre-judgment module is used to make pre-judgments based on the detection model, obtain prediction results, and generate control instructions; The model callback module is used to update the detection model according to control instructions.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A terminal comprising a terminal body and a controller, characterized in that: The controller includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the steps of the method according to any one of claims 1 to 5 are implemented.

Images