A linear detection method and detection equipment for semiconductor discrete components

Abstract

The present invention discloses a linear detection method for semiconductor discrete components. Mean filtering is respectively performed on the square of the discrete component image and the discrete component image with a preset window to obtain filtered images. Then, a local standard deviation image is calculated based on the standard deviation calculation formula, the first filtered image, and the second filtered image, and a local bonding wire image is extracted based on the local standard deviation image. Since the parameters w and h are different, the goal of suppressing the gradient value in a certain direction is achieved, with higher flexibility. Then, the solder ball center coordinates are obtained from the discrete component image. Finally, it can be determined whether the semiconductor discrete component is NG according to the preset NG standard, the set S of bonding wire coordinates of the local bonding wire image, and the solder ball center coordinates. Among them, since the interference of irrelevant gradients has been filtered out in the extracted local bonding wire image, the distance between the solder ball center coordinates can be calculated simply and quickly to determine whether the semiconductor discrete component is NG, which takes a short time and occupies less resources.

Description

Technical Field

[0001] The present invention relates to the field of semiconductor chip production and manufacturing, and particularly to a method and device for detecting the linear type of semiconductor discrete components. Background Art

[0002] The linear type detection of semiconductor discrete components is an important step in the process of semiconductor chip production and manufacturing. At present, the linear type detection of semiconductor discrete components all adopts the means of manual visual inspection. This method has a low automation level, and it is easy for employees to feel tired after long-term work, and the detection results are easily affected by subjective factors such as emotions, resulting in a decrease in the detection accuracy.

[0003] Therefore, in the prior art, some technical personnel proposed to use the Laplace edge detection operator to perform edge detection on the linear type of the semiconductor to obtain an edge image with enhanced edges, and then train it through deep learning or neural network methods. Finally, a training model capable of recognizing the above-mentioned edge image is obtained to realize the automation of linear type detection.

[0004] There are two disadvantages in the above process. One is that the edge image obtained by using the Laplace edge detection operator in the prior art has gradient value interference, resulting in low flexibility of the overall process; the other is that using deep learning or neural network methods takes a long time and occupies a lot of resources.

[0005] In summary, the detection means in the prior art has low flexibility and takes a long time. Summary of the Invention

[0006] The purpose of the present invention is to provide a method and device for detecting the linear type of semiconductor discrete components to solve the problems of low flexibility and long time consumption of the current detection means.

[0007] To achieve this purpose, the present invention adopts the following technical solutions:

[0008] A method for detecting the linear type of semiconductor discrete components includes:

[0009] Preset a filtering window for mean filtering, and the width and height of the filtering window are not equal;

[0010] Obtain a discrete component image I(x, y) of the semiconductor discrete component;

[0011] Perform mean filtering on the square of the discrete component image I(x, y) to obtain a first filtered image E(I 2 )

[0012] Perform mean filtering on the discrete component image I(x, y) to obtain a second filtered image E(I);

[0013] According to the standard deviation calculation formula, E(I 2 ) and E(I), the local standard deviation image I σ is calculated;

[0014] The local wire bonding image I σ is extracted from I WireContours ;

[0015] The center coordinates (x ball , y ball ) of the solder ball in I(x, y) are calculated;

[0016] It is determined whether the distances between the pixels of I WireContours and the solder ball center coordinates (x ball , y ball ) are all within a preset distance range;

[0017] If so, it is determined that the semiconductor discrete component is a normal workpiece;

[0018] If not, it is determined that the semiconductor discrete component is an NG component.

[0019] Optionally, the step of extracting the local wire bonding image I σ from I WireContours specifically includes:

[0020] Perform threshold segmentation on I σ to obtain the gray-scale change region image I m ;

[0021] Perform threshold segmentation on I(x, y) to obtain the pin image I PinMask at the pins of the semiconductor discrete component;

[0022] Let I m minus I PinMask to obtain an intermediate image, and screen out the preliminary wire bonding image I WireMask from the intermediate image, and after refinement, obtain I WireContours .

[0023] Optionally, the step of performing threshold segmentation on I σ to obtain the gray-scale change region image I m specifically includes:

[0024] Assign pixel values to the pixels (x, y) of I σ in sequence to obtain the gray-scale change region image I m ;

[0025] The method of assigning pixel values to the pixels (x, y) of I σ in sequence specifically includes:

[0026] Successively determine I σ for each pixel point (x, y) to see if its pixel value is greater than the sum of the pixel value of the corresponding pixel point (x, y) in the mean image and a preset first pixel value threshold;

[0027] If so, assign the pixel value of the pixel point (x, y) to 255;

[0028] If not, assign the pixel value of the pixel point (x, y) to 0.

[0029] Optionally, the step of: Let I m subtract I PinMask , obtain an intermediate image, and screen out a preliminary wire bonding image I WireMask , and after refinement, obtain I WireContours , including:

[0030] Judge whether the area in the intermediate image is within a preset area threshold;

[0031] If not, judge that the intermediate image is an interference image;

[0032] If so, determine that the intermediate image is an image to be determined; and screen out a wire bonding image from the image to be determined to obtain I WireMask .

[0033] Optionally, the step of: calculating the solder ball center coordinates (x ball , y ball ) of I(x, y), specifically including:

[0034] Successively use the Laplace operator and the Sobel operator to perform edge detection on the discrete component image I(x, y) to obtain a solder ball edge image I conv ;

[0035] Perform threshold segmentation on the solder ball edge image I conv to obtain a solder ball segmentation image I binary , and screen out a solder ball image I binary that meets the preset solder ball standard from the solder ball segmentation image I ball ;

[0036] Calculate the solder ball center coordinates (x ball ) of the screened solder ball image I ball , y ball ).

[0037] Optionally, the step of: successively use the Laplace operator and the Sobel operator to perform edge detection on the discrete component image I(x, y) to obtain a solder ball edge image I conv, specifically including:

[0038] Convolve the discrete component image I(x,y) through a preset Laplace kernel k Laplace to obtain a convolved image I c1 ;

[0039] Convolve the convolved image I Sobel through a preset Sobel kernel k c1 to obtain the solder ball edge image I conv .

[0040] Optionally, the step: determining whether the distance between each pixel of I WireContours and the solder ball center coordinates (x ball , y ball ) is within a preset distance range specifically includes:

[0041] Determine the maximum difference d WireContours on the y-axis between the pixel points in I ball and the solder ball center coordinates (x ball ) to see if it is greater than a preset first NG value T1; yMax If so, determine the semiconductor discrete component as a collapsed wire NG product;

[0042] If not, determine the semiconductor discrete component as a non-collapsed wire NG product.

[0043]

[0044] Optionally, after the step: extracting a local bonding wire image I σ from I WireContours , it further includes:

[0045] Determine whether the shortest distance value d WireContours between the pixel points in I PinMask and I Min is less than a preset second NG value T2;

[0046] If so, determine the semiconductor discrete component as a wire swing NG product;

[0047] If not, determine the semiconductor discrete component as a non-wire swing NG product.

[0048] A detection device, characterized in that it includes a memory and a processor, the memory stores a control program that can run on the processor, and when the control program is executed by the processor, it implements the semiconductor discrete component wire type detection method as described above.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The semiconductor discrete component linear detection method and detection device provided by the present invention respectively perform mean filtering on the square of the discrete component image and the discrete component image with a preset window to obtain a first filtered image and a second filtered image; then calculate a local standard deviation image according to the standard deviation calculation formula, the first filtered image and the second filtered image, and extract a local bonding wire image based on the local standard deviation image. Among them, since the parameters w and h are different, the goal of suppressing the gradient value in a certain direction is achieved, and it has higher flexibility; then obtain the solder ball center coordinates from the discrete component image, and finally, it can be determined whether the semiconductor discrete component is NG according to the preset NG standard, the set S of bonding wire coordinates of the local bonding wire image, and the solder ball center coordinates; among them, since the local bonding wire image has filtered out the interference of irrelevant gradients, it can simply and quickly calculate the distance from the solder ball center coordinates to determine whether the semiconductor discrete component is NG, which takes a short time and occupies less resources; in summary, the semiconductor discrete component linear detection method and detection device of the present invention have the advantages of high flexibility and short time consumption. Description of the Drawings

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained according to these drawings.

[0052] The structures, ratios, sizes, etc. shown in the drawings of this specification are only used to cooperate with the content disclosed in the specification for those who are familiar with this technology to understand and read, and are not used to limit the limited conditions under which the present invention can be implemented. Therefore, they do not have a substantial technical meaning. Any modification of the structure, change of the proportional relationship, or adjustment of the size, without affecting the effects that the present invention can produce and the purposes that can be achieved, should still fall within the scope that the technical content disclosed by the present invention can cover.

[0053] Figure 1 It is a schematic diagram of the overall process of the semiconductor discrete component linear detection method provided by the embodiment of the present invention;

[0054] Figure 2 It is a schematic diagram of the discrete component image of the embodiment of the present invention. Detailed Embodiments

[0055] In order to make the objectives, features, and advantages of the present invention more obvious and understandable, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the embodiments described below are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0056] In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "upper", "lower", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation to the present invention. It should be noted that when a component is considered to be "connected" to another component, it can be directly connected to the other component or there may be an intermediate component present.

[0057] The following further illustrates the technical solutions of the present invention with reference to the accompanying drawings and specific embodiments.

[0058] Please refer to Figures 1 to 2 , Figure 1 which is a schematic diagram of the overall process of the semiconductor discrete component line detection method provided by the embodiment of the present invention, Figure 2 and

[0059] Embodiment 1

[0060] This embodiment provides a semiconductor discrete component line detection method, which is mainly applied to the scenario of detecting the line type of semiconductor discrete components. By optimizing the image processing method, it has higher flexibility, shorter time consumption, and less resource occupation.

[0061] As Figure 1 shown, the semiconductor discrete component line detection method of this embodiment includes:

[0062] S100. Preset a filtering window for mean filtering, where the width w and height h of the filtering window are not equal;

[0063] S200. Obtain the discrete component image I(x, y) of the semiconductor discrete component; among them, the discrete component image I(x, y) is obtained by photographing the semiconductor discrete component with X-ray; among them, the discrete component image I(x, y) is as Figure 2As shown, it includes a wire bonding part (wire), a pin part (pin), and a solder ball part (ball). The separate acquisition of the wire bonding part (wire) and the solder ball part (ball) is mainly achieved through the following steps for determining whether it is a NG workpiece in step S800;

[0064] S300. Square the discrete component image I(x, y) and perform mean wave filtering with a preset window (w, h) to obtain the first filtered image E(I 2 ), where the parameter w is not equal to the parameter h;

[0065] S400. Perform mean wave filtering on the discrete component image I(x, y) with the window (w, h) to obtain the second filtered image E(I);

[0066] S500. Calculate the local standard deviation image I 2 according to the standard deviation calculation formula, the first filtered image E(I σ );

[0067] S600. Extract the local wire bonding image I σ from the local standard deviation image I WireContours ;

[0068] S700. Calculate the solder ball center coordinates (x ball , y ball ) of the discrete component image I(x, y);

[0069] S800. Determine whether the distance between the local wire bonding image I WireContours and the solder ball center coordinates (x ball , y ball ) meets the preset NG standard;

[0070] S810. If so, determine that the semiconductor discrete component is a normal workpiece;

[0071] S820. If not, determine that the semiconductor discrete component is a NG part.

[0072] It should be noted that the standard deviation calculation formula in step S500 is

[0073] Standard Deviation

[0074] where x i is the initial set, N is the number of elements in the set, and μ is the average value of the set;

[0075] In step S300, square the discrete component image I with (w, h) as the window 2(x, y) performs mean filtering to obtain the first filtered image E(I 2 );This first filtered image E(I 2 ) can be regarded as the

[0076] In step S400, the discrete component image I(x, y) is subjected to mean filtering with (w, h) as the window to obtain the second filtered image E(I); this second filtered image E(I) can be regarded as μ in formula (1);

[0077] Therefore, after calculating the first filtered image E(I 2 ) in step S300 and the second filtered image E(I) in step S400, the local standard deviation image I σ can be quickly calculated by the formula:

[0078]

[0079] It should be understood that by calculating the local standard deviation image I σ in the above manner, the local standard deviation image I σ and the second filtered image E(I) can be quickly calculated, where E(I) is also regarded as the mean image I μ , significantly improving the processing efficiency of the semiconductor discrete component line type detection method.

[0080] Specifically, the semiconductor discrete component line type detection method of this embodiment performs mean filtering on the square of the discrete component image and the discrete component image respectively with a preset window to obtain the first filtered image and the second filtered image; then, according to the standard deviation calculation formula, the first filtered image and the second filtered image, the local standard deviation image is calculated, and the local wire bonding image is obtained based on the local standard deviation image. Among them, since the parameters w and h are different, the goal of suppressing the gradient value in a certain direction is achieved, with higher flexibility; then, the solder ball center coordinates are obtained from the discrete component image, and finally, it can be determined whether the semiconductor discrete component is NG according to the preset NG standard, the solder wire coordinate set S of the local wire bonding image, and the solder ball center coordinates; among them, since the local wire bonding image has extracted and filtered out the interference of irrelevant gradients, it can simply and quickly calculate the distance from the solder ball center coordinates to determine whether the semiconductor discrete component is NG, with short time consumption and less resource occupation; in summary, this semiconductor discrete component line type detection method and detection device have the advantages of high flexibility and short time consumption.

[0081] Further, step S600: Extract the local wire bonding image I σ from the local standard deviation image I WireContours , specifically including:

[0082] S610. Perform threshold segmentation on the local standard deviation image I σ to obtain the grayscale change region image I m ;

[0083] S620. Perform threshold segmentation on the discrete component image I(x, y) to obtain the pin image I at the pins of the semiconductor discrete components PinMask ; Among them, the second threshold segmentation is performed using the second threshold t2, so as to separate the pin image I PinMask from the discrete component image I(x, y), where 50 ≤ t2 ≤ 150;

[0084] S630. Let the grayscale change region image I m subtract the pin image I PinMask to obtain an intermediate image, and screen out the preliminary wire bonding image I WireMask from the intermediate image, and after refinement, obtain the local wire bonding image I WireContours ; The grayscale change region image I m mainly includes the wire bonding part, the pin part and the interference part. After subtracting the pin image I PinMask , the obtained intermediate image includes the wire bonding part and the interference part; then, the interference part is removed from the intermediate image through constraints such as area and position, that is, the preliminary wire bonding image I WireMask is obtained; Among them, in this step, a topological refinement algorithm is used to repeatedly perform iterative calculations starting from the boundary, and the boundary of the graph is uniformly peeled off layer by layer until the innermost one-dimensional skeleton remains, so as to obtain the local wire bonding image I WireContours .

[0085] Specifically, the second filtered image E(I) is the mean image I μ ; That is, there is the following formula:

[0086]

[0087] Step S610: Perform the first threshold segmentation on the local standard deviation image I σ to obtain the grayscale change region image I m , specifically including:

[0088] S611. Sequentially judge whether the product of the pixel value of each pixel point (x, y) of I σ and the preset segmentation threshold is less than the difference between the pixel value of the corresponding pixel point (x, y) in the mean image and the pixel value of the corresponding pixel point (x, y) in the discrete component image; The specific judgment formula is as follows: (4);

[0090] Among them, t1 is the preset first threshold parameter, and 0.1 ≤ t1 ≤ 0.3;

[0091] S612. If so, assign the pixel point (x, y) a value of 255, that is, determine that the pixel point (x, y) is a point with a drastic gray-scale change.

[0092] S613. If not, assign the pixel point (x, y) a value of 0.

[0093] S614. After all pixel points are assigned values, obtain the gray-scale change region image I. m .

[0094] Exemplarily, step S630: Let the gray-scale change region image I m subtract the pin image I PinMask , obtain multiple unconnected intermediate images, and screen out the preliminary wire bonding image I WireMask , and after refinement, obtain I WireContours , including:

[0095] S631. Determine whether the area of the intermediate image is within a preset area threshold.

[0096] S632. If not, determine the intermediate image as an interference image.

[0097] S633. If so, determine the intermediate image as an image to be determined; and screen out the wire bonding image from the images to be determined, and obtain I WireMask ; among them, screening out the local wire bonding image I WireContours is mainly through position judgment. For example, by judging whether the pixel coordinates of the image to be determined are located in the target interference region. If not in the interference region, determine the image to be determined as the local wire bonding image.

[0098] Furthermore, step S700: Extract the solder ball center coordinates (x ball , y ball ) from the discrete component image I(x, y), specifically including:

[0099] S710. Perform edge detection on the discrete component image I(x, y) successively using the Laplace operator and the Sobel operator to obtain the solder ball edge image I conv ;

[0100] S720. Perform the first threshold segmentation on the solder ball edge image I conv to obtain the solder ball segmentation image I binary , and screen out the solder ball image I binary that meets the preset solder ball standard from the solder ball segmentation image I ball ; among them, the preset solder ball standard refers to screening out the solder ball image I binary that meets the standard from the solder ball segmentation image I according to the area and position constraint conditions of the solder ball.ball ; among them, the threshold is 100 ≤ t3 ≤ 240;

[0101] S730. Calculate the solder ball image I ball of the center coordinates (x ball , y ball ) of the solder ball.

[0102] Specifically, step S710: Perform edge detection on the discrete component image I(x, y) successively using the Laplace operator and the Sobel operator to obtain the solder ball edge image I conv , specifically including:

[0103] S711. Convolve the discrete component image I(x, y) through a preset Laplace kernel k Laplace to obtain a convolution image I c1 ; among them, as Figure 2 can be seen, the solder ball is located at the end of the bonding wire, and the gray value is less than that of the bonding wire. Therefore, after detecting with the above Laplace operator, the gradient amplitude of the solder ball is greater than that of the bonding wire, so as to quickly find the solder ball; in this embodiment, the convolution kernel h of the Laplace kernel k Laplace can be defined as:

[0104]

[0105] It can also be defined as

[0106]

[0107] S712. Convolve the convolution image I Sobel through a preset Sobel kernel k c1 to obtain the solder ball edge image I conv . It should be noted that the Sobel algorithm usually includes an operator for horizontal edge detection and an operator for vertical edge detection; for example, the following h1 is the Sobel horizontal edge detection operator and h2 is the Sobel vertical edge detection operator;

[0108]

[0109]

[0110] In this embodiment, the Sobel vertical edge detection operator h2 is used as the Sobel kernel k Sobel to perform a convolution operation on the convolution image I c1 , which can eliminate the interference of the gradient information of the bonding wire part and improve the flexibility of this method.

[0111] Further, step S800: determining the local welding line image I WireContours The center coordinates of the solder ball (x ball ,y ball ) meets the preset NG standard, including:

[0112] S810, judging the local welding line image I WireContours The pixel point in the image and the center coordinate of the solder ball (x ball ,y ball ) on the y-axis. yMax Is it greater than the preset first NG value T1, where T1 can be 13;

[0113] S811, if yes, the semiconductor discrete component is determined to be a line collapse NG product;

[0114] S812: If not, the semiconductor discrete component is determined to be a non-collapse NG product.

[0115] Further, step SS600: from I σ The local welding line image I is extracted from WireContours After that, it also includes:

[0116] Step S900: Determination I WireContours The pixels in I PinMask The shortest distance value d Min Is it less than the preset second NG value T2; T2 can be selected as 11;

[0117] S901, if yes, the semiconductor discrete component is determined to be a line swing NG product;

[0118] S902: If not, the semiconductor discrete component is determined to be a non-linear NG product.

[0119] In summary, the semiconductor discrete component line shape detection method provided in this embodiment can simply and quickly calculate the distance between the semiconductor discrete component and the center coordinate of the solder ball to determine whether the semiconductor discrete component is NG. It is time-saving and resource-saving, and has the advantages of high flexibility and short time consumption.

[0120] Embodiment 2

[0121] The detection device provided in this embodiment includes a memory and a processor, wherein the memory stores a control program that can be run on the processor, and when the control program is executed by the processor, the semiconductor discrete component line type detection method in Embodiment 1 is implemented. Embodiment 1 describes the specific steps and technical effects of the semiconductor discrete component line type detection method, and the detection device of this embodiment refers to this method and also has its technical effects.

[0122] In summary, the detection device provided in this embodiment can simply and quickly calculate the distance from the center coordinates of the solder balls to determine whether the semiconductor discrete component is NG, which takes a short time and occupies few resources, and has the advantages of high flexibility and short time consumption.

[0123] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A linear detection method for semiconductor discrete components, characterized in that, Including: Pre-set a filtering window for mean filtering, where the width and height of the filtering window are not equal; Obtain the discrete component image I(x, y) of the semiconductor discrete component; Perform mean filtering on the square of I(x, y) to obtain the first filtered image E(I 2 ); Perform mean filtering on I(x, y) to obtain the second filtered image E(I); According to the standard deviation calculation formula, E(I 2 ) and E(I), the local standard deviation image I σ is calculated; Extract the local wire bonding image I σ from I WireContours ; Calculate the solder ball center coordinates (x ball , y ball ) of I(x, y); Determine I WireContours Whether the distance between each pixel of and the center coordinates (x ball , y ball ) of the solder ball is within a preset distance range; If so, determine that the semiconductor discrete component is a normal workpiece; If not, determine that the semiconductor discrete component is an NG component.

2. The linear detection method of semiconductor discrete components according to claim 1, characterized in that, The steps: Extract from I σ to obtain a partial wire bonding image I WireContours , specifically including: For I σ perform threshold segmentation to obtain the grayscale change region image I m ; Perform threshold segmentation on I(x, y) to obtain the pin image I at the pin of the semiconductor discrete component PinMask ; Let I m subtract I PinMask to obtain an intermediate image, and screen out a preliminary wire bonding image I WireMask from the intermediate image, and after refinement, obtain a local wire bonding image I WireContours .

3. The linear detection method of semiconductor discrete components according to claim 2, wherein The steps: For I σ Perform threshold segmentation to obtain the grayscale change region image I m , specifically including: Assign pixel values to each pixel point (x, y) of I σ in sequence to obtain the grayscale change region image I m ; The method of sequentially assigning pixel values to each pixel point (x, y) of I σ specifically includes: Judge I sequentially σ whether the product of the pixel value of each pixel point (x, y) in and a preset segmentation threshold is less than the difference between the pixel value of the corresponding pixel point (x, y) in the mean image and the pixel value of the corresponding pixel point (x, y) in the discrete component image; If so, assign the pixel value of the pixel point (x, y) to be 255; If not, assign the pixel value of the pixel point (x, y) to be 0.

4. The linear detection method of semiconductor discrete components according to claim 2, characterized in that, The steps are as follows: Let I m subtract I PinMask , to obtain an intermediate image, and screen out a preliminary wire bonding image I WireMask from the intermediate image. After refinement, a local wire bonding image I WireContours is obtained, including: Judge whether the area of the intermediate image is within a preset area threshold; If not, judge that the intermediate image is an interference image; If so, determine that the intermediate image is an image to be determined; and screen out the wire bonding images from the images to be determined to obtain a preliminary wire bonding image I WireMask .

5. The linear detection method of the semiconductor discrete element according to claim 1, characterized in that The said steps: calculating the solder ball center coordinates (x ball , y ball ) of I(x, y), specifically including: Perform edge detection on the discrete component image I(x, y) successively using the Laplace operator and the Sobel operator to obtain the solder ball edge image I conv ; Perform threshold segmentation on the solder ball edge image I conv to obtain a solder ball segmented image I binary , and screen out the solder ball image I binary that meets the preset solder ball standard from the solder ball segmented image I ball ; Calculate the solder ball image I selected by screening ball of the solder ball center coordinates (x ball , y ball ).

6. The linear detection method of the semiconductor discrete component according to claim 5, characterized in that, The steps: perform edge detection on the discrete component image I(x, y) successively using the Laplace operator and the Sobel operator to obtain the solder ball edge image I conv , which specifically includes: By means of a preset Laplace kernel k Laplace perform convolution on the discrete element image I(x, y) to obtain a convolution image I c1 ; Through the preset Sobel kernel k Sobel perform convolution on the convolution image I c1 to obtain the solder ball edge image I conv .

7. The linear detection method of semiconductor discrete components according to claim 1, characterized in that The above steps: Determine I WireContours Whether the distance between each pixel of and the center coordinates (x ball , y ball ) of the solder ball is within a preset distance range, specifically including: Determination I WireContours The pixel points in it and the center coordinates (x ball , y ball ) of the solder ball, the maximum difference d yMax on the y-axis, whether it is greater than the preset first NG value T1; If so, determine that the semiconductor discrete component is a collapsed line NG product; If not, determine that the semiconductor discrete component is a non-collapsed line NG product.

8. The linear detection method of the semiconductor discrete component according to claim 2, characterized in that, The steps: Extract from I σ to obtain a partial wire bonding image I WireContours After that, it further includes: Determination I WireContours The pixel points in PinMask and the shortest distance value d Min Is it less than the preset second NG value T2; If so, determine that the semiconductor discrete component is a line swing NG product; If not, determine that the semiconductor discrete component is a non-line swing NG product.

9. A detection device, characterized in that, Including a memory and a processor, the memory stores a control program that can run on the processor, and when the control program is executed by the processor, it implements the semiconductor discrete component line type detection method according to any one of claims 1 to 8.

Images