A chip appearance inspection method

Abstract

This invention discloses a chip appearance inspection method, belonging to the field of chip manufacturing. It involves image processing of the chip surface to identify pin marks, obtaining pin mark regions in a pin mark distribution map; setting pin mark arrangement parameters and generating a pin mark template distribution map based on these parameters; comparing the pin mark regions with the pin mark template distribution map, and adding a standard pin mark to the areas on the pin mark distribution map where the pin marks in the two maps do not overlap, forming a processed pin mark distribution image; removing the processed pin mark distribution image from the chip surface image to form a pin mark-removed image; and performing defect identification on the pin mark-removed image. Through these steps, the impact of pin marks on appearance can be avoided without extensive sampling; it has fewer missed detections, strong versatility, and is suitable for pin mark arrangements of different types.

Description

Technical Field

[0001] This invention relates to chip manufacturing, and more particularly to a method for inspecting the appearance of chips. Background Technology

[0002] During wafer testing, probes need to contact the chip solder joints for electrical testing. Due to the pressure of the testing equipment, the probes will pierce the surface of the solder joints, leaving pin marks. When the wafer is manufactured into a chip, visual defect inspection is required, and these pin marks on the chip surface will affect the defect inspection results.

[0003] In existing technologies, the following methods are used to address pin marks on the chip surface:

[0004] (1) The needle mark area is directly shielded from detection or the detection area is large, but this method will cause serious abnormal defects to be missed.

[0005] (2) The pin marks are learned to generate a standard template with pin marks, and the template is used to perform a difference operation with the real-time detection image to obtain an abnormal defect map. However, this method requires the size and position of the pin marks to be stable, which is difficult and has poor versatility;

[0006] (3) Use deep learning, such as YOLO-based target detection, to identify needle marks. Although this method does not require professional algorithm knowledge, it requires the collection of a large number of defect samples to achieve a good detection effect. In actual production, timeliness is often very tight, and it is difficult to collect a large number of defect samples. Summary of the Invention

[0007] In order to overcome the shortcomings of the prior art, one of the objectives of this invention is to provide a chip appearance inspection method with fewer missed detections, strong versatility, and no need to collect a large number of defect samples.

[0008] One of the objectives of this invention is achieved through the following technical solution:

[0009] A chip appearance inspection method includes the following steps:

[0010] Image acquisition: Acquiring images of the chip surface;

[0011] Image processing: The acquired image is processed to grayscale, and the grayscale image is processed by image pyramid downsampling to obtain grayscale images at different scales; morphological black hat and top hat processing is performed on the grayscale images at different scales to accommodate some existing bright and dark pin marks, and grayscale images of morphological results are obtained.

[0012] Image enhancement: The morphological result image is enhanced by a linear transformation based on histogram normalization to obtain an enhanced grayscale image. The enhanced grayscale image is then thresholded to become a binary image. The binary image is then upsampled using an image pyramid to obtain an upsampled grayscale image.

[0013] To generate a needle mark distribution image: the upsampled image is thresholded and segmented to become a binary image; then morphological dilation is performed to obtain the needle mark distribution image.

[0014] Generate a needle mark template distribution map: Set the needle mark arrangement parameters, and generate a needle mark template distribution map based on the needle mark arrangement parameters;

[0015] Image recognition processing: Perform needle mark recognition on the needle mark distribution image to obtain the needle mark region in the needle mark distribution image. Compare the needle mark region with the needle mark template distribution image. Add a standard needle mark to the area on the needle mark distribution image where the needle mark regions on the needle mark distribution image and the needle mark template distribution image do not overlap, to form the processed needle mark distribution image.

[0016] Chip image processing: Remove the processed pin mark distribution image from the chip surface image to form a pin mark removal image, and then identify defects from the pin mark removal image.

[0017] Furthermore, in the step of generating the pin mark template distribution map, the pin mark arrangement parameters are specifically one or more of the following: the center spacing of pin marks, the fluctuation of the pin mark spacing, the total number of pin marks, and the maximum number of missing pin marks.

[0018] Furthermore, in the image processing step, when performing image pyramid downsampling on the grayscale image, from the first... layer to the first The general formula for full downsampling of a layer is:

[0019] ,

[0020] in, These are the pixel coordinates in the image; It is the radius of the convolution kernel, for a given... The core, ; It is a Gaussian convolution kernel, whose weights are determined by a Gaussian function. Let σ be the standard deviation.

[0021] Furthermore, in the image processing steps, the morphological black hat and top hat processing of grayscale images at different scales specifically involves:

[0022] ,

[0023] in, This represents the image obtained after black hat computation. This represents the image obtained after the top-hat operation. Morphological images representing the overlay of black hats and top hats. Represents the downsampled image. The rectangular convolution kernel is set.

[0024] Furthermore, in the image enhancement step, the image of the morphological result undergoes a linear transformation enhancement process based on histogram normalization, specifically as follows:

[0025] Generate a corresponding histogram based on the grayscale information of the grayscale image obtained from the morphological results, and output the maximum grayscale value in the histogram. With minimum gray value ;

[0026] Set the enhanced grayscale image Maximum grayscale value With minimum gray value ;

[0027] Calculate the contrast of the enhanced image ,brightness and the grayscale image of the enhanced result. .

[0028] Furthermore, the contrast of the enhanced image is calculated. ,brightness and the grayscale image of the enhanced result. Specifically:

[0029] ,

[0030] In the formula, A morphological image representing the overlay of a black hat and a top hat.

[0031] Furthermore, in the image enhancement step, thresholding the enhanced grayscale image specifically involves:

[0032] ,

[0033] in, The resulting binarized image, The set segmentation threshold.

[0034] Furthermore, in the image enhancement step, when performing image pyramid upsampling on the binarized image, from the first... layer to the first The formula for full upsampling of a layer is:

[0035] ,

[0036] in, For binarized images, This is the image obtained through upsampling.

[0037] Furthermore, in the step of forming the needle mark distribution image, the calculation formula for the morphological dilation process is as follows:

[0038] ,

[0039] in, This is the image after dilation. It is an upsampled segmented image. It is the position of the structural element The value of .

[0040] Furthermore, in the image recognition processing step, needle mark recognition of the needle mark distribution image specifically involves: ... Abnormal residues are divided by row and / or column, and the center coordinates of the residues in each row and / or column are sorted from smallest to largest. If a row and / or column contains only one residue, the row is directly removed without further needle mark identification. Each residual point is recorded as... The continuity of stitch marks is determined based on their arrangement parameters; specifically... ,

[0041] in, The distance between the centers of the needle marks. This represents the fluctuation in the distance between needle marks. The maximum spacing between the centers of the needle marks; The minimum spacing between the centers of the needle marks; The maximum number of missing stitches is allowed; when both of the above formulas are satisfied, it is determined that the stitches are discontinuous.

[0042] Compared to existing technologies, the chip appearance inspection method of this invention identifies pin marks by processing the image of the chip surface, obtaining pin mark regions in a pin mark distribution map; sets pin mark arrangement parameters, and generates a pin mark template distribution map based on the pin mark arrangement parameters; compares the pin mark regions with the pin mark template distribution map, and adds a standard pin mark to the areas on the pin mark distribution map where the pin mark regions and the pin mark template distribution map do not overlap, forming a processed pin mark distribution image; removes the processed pin mark distribution image from the image on the chip surface, forming a pin mark-removed image, and performs defect identification on the pin mark-removed image. Through the above steps, the impact of pin marks on appearance can be avoided without a large number of samples; it has fewer missed detections, strong versatility, and is suitable for pin marks with different arrangements. Attached Figure Description

[0043] Figure 1 This is a flowchart of the chip appearance inspection method of the present invention;

[0044] Figure 2 It is a grayscale image;

[0045] Figure 3 This is a downsampled grayscale image;

[0046] Figure 4 Images representing morphological results;

[0047] Figure 5 To enhance the image;

[0048] Figure 6 A binarized image;

[0049] Figure 7 This is an upsampled grayscale image;

[0050] Figure 8 Image showing the distribution of needle marks;

[0051] Figure 9 Image showing the distribution of needle mark templates;

[0052] Figure 10 This is the processed image of the needle mark distribution. Detailed Implementation

[0053] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0055] Figure 1 This application discloses a chip appearance inspection method, comprising the following steps:

[0056] Image acquisition: Acquiring images of the chip surface;

[0057] Image processing: The acquired image is processed to grayscale, and the grayscale image is processed by image pyramid downsampling to obtain grayscale images at different scales; morphological black hat and top hat processing is performed on the grayscale images at different scales to accommodate some existing bright and dark pin marks, and grayscale images of morphological results are obtained.

[0058] Image enhancement: The morphological result image is enhanced by a linear transformation based on histogram normalization to obtain an enhanced grayscale image. The enhanced grayscale image is then thresholded to become a binary image. The binary image is then upsampled using an image pyramid to obtain an upsampled grayscale image.

[0059] To generate a needle mark distribution image: the upsampled image is thresholded and segmented to become a binary image; then morphological dilation is performed to obtain the needle mark distribution image.

[0060] Generate a needle mark template distribution map: Set the needle mark arrangement parameters, and generate a needle mark template distribution map based on the needle mark arrangement parameters;

[0061] Image recognition processing: Perform needle mark recognition on the needle mark distribution image to obtain the needle mark region in the needle mark distribution image. Compare the needle mark region with the needle mark template distribution image. Add a standard needle mark to the area on the needle mark distribution image where the needle mark regions on the needle mark distribution image and the needle mark template distribution image do not overlap, to form the processed needle mark distribution image.

[0062] Chip image processing: Remove the processed pin mark distribution image from the chip surface image to form a pin mark removal image, and then identify defects from the pin mark removal image.

[0063] Specifically, in the image acquisition step, the images acquired on the chip surface include pin marks and defects.

[0064] The specific steps of image processing are as follows:

[0065] The image after grayscale processing of the acquired image is as follows: Figure 2 As shown; image pyramid downsampling is performed on the grayscale image to obtain grayscale images at different scales, as shown in the figure. Figure 3 As shown. When performing image pyramid downsampling on a grayscale image, starting from the first... layer to the first The general formula for full downsampling of a layer is:

[0066] ,

[0067] in, These are the pixel coordinates in the image; It is the radius of the convolution kernel, for a given... The core, ; It is a Gaussian convolution kernel, whose weights are determined by a Gaussian function. Let σ be the standard deviation.

[0068] Morphological black-hat and top-hat processing is performed on grayscale images at different scales to obtain grayscale images with morphological results, such as... Figure 4 As shown, the morphological black hat and top hat processing for grayscale images at different scales is specifically as follows:

[0069] ,

[0070] in, This represents the image obtained after black hat computation. This represents the image obtained after the top-hat operation. Morphological images representing the overlay of black hats and top hats. Represents the downsampled image. The rectangular convolution kernel is set.

[0071] The image enhancement steps are as follows:

[0072] The image of the morphological results is enhanced by a linear transformation based on histogram normalization to obtain an enhanced grayscale image, such as... Figure 5 As shown. The image enhancement process based on histogram normalization of the morphological results is specifically performed as follows:

[0073] Generate a corresponding histogram based on the grayscale information of the grayscale image obtained from the morphological results, and output the maximum grayscale value in the histogram. With minimum gray value ;

[0074] Set the enhanced grayscale image Maximum grayscale value With minimum gray value ;

[0075] Calculate the contrast of the enhanced image ,brightness and the grayscale image of the enhanced result. .

[0076] ,

[0077] In the formula, A morphological image representing the overlay of a black hat and a top hat.

[0078] The enhanced grayscale image is then thresholded, resulting in a binarized image, such as... Figure 6 As shown. Thresholding segmentation of the enhanced grayscale image is specifically performed as follows:

[0079] ,

[0080] in, The resulting binarized image, The set segmentation threshold.

[0081] Image pyramid upsampling is performed on the binarized image to obtain an upsampled grayscale image, such as... Figure 7 As shown. When performing image pyramid upsampling on a binarized image, starting from the first... layer to the first The formula for full upsampling of a layer is:

[0082] ,

[0083] in, For binarized images, This is the image obtained through upsampling.

[0084] The specific steps for forming a needle mark distribution image are as follows:

[0085] Thresholding is performed on the upsampled image to create a binarized image; then morphological dilation is applied to obtain the needle mark distribution image, such as... Figure 8 As shown; the calculation formula for morphological dilation is:

[0086] ,

[0087] in, This is the image after dilation. It is an upsampled segmented image. It is the position of the structural element The value of .

[0088] The specific steps for generating the needle mark template distribution map are as follows:

[0089] Based on the distribution pattern of needle marks, the arrangement parameters of the needle marks are set. Specifically, these parameters include any one or more of the following: center-to-center spacing of needle marks, fluctuation in needle mark spacing, total number of needle marks, and maximum allowable number of missing needle marks. The generated needle mark template distribution diagram is shown below. Figure 9 As shown.

[0090] The specific steps of image recognition processing are as follows:

[0091] Needle mark recognition on a needle mark distribution image specifically involves: Abnormal residues are divided by row and / or column, and the center coordinates of the residues in each row and / or column are sorted from smallest to largest. If a row and / or column contains only one residue, the row is directly removed without further needle mark identification. Each residual point is recorded as... The continuity of stitch marks is determined based on their arrangement parameters; specifically... ,

[0092] in, The distance between the centers of the needle marks. This represents the fluctuation in the distance between needle marks. The maximum spacing between the centers of the needle marks; The minimum spacing between the centers of the needle marks; The maximum number of missing needle marks is allowed; when both of the above formulas are satisfied, the needle marks are considered discontinuous. The discontinuous region is the area where the needle mark region and the needle mark template distribution map do not intersect. A standard needle mark is added to the area on the needle mark distribution map where the needle mark region and the needle mark template distribution map do not intersect, forming a processed needle mark distribution image, such as... Figure 10 As shown.

[0093] Compared to existing technologies, the chip appearance inspection method of this invention identifies pin marks by processing the image of the chip surface, obtaining pin mark regions in a pin mark distribution map; sets pin mark arrangement parameters, and generates a pin mark template distribution map based on the pin mark arrangement parameters; compares the pin mark regions with the pin mark template distribution map, and adds a standard pin mark to the areas on the pin mark distribution map where the pin mark regions and the pin mark template distribution map do not overlap, forming a processed pin mark distribution image; removes the processed pin mark distribution image from the image on the chip surface, forming a pin mark-removed image, and performs defect identification on the pin mark-removed image. Through the above steps, the impact of pin marks on appearance can be avoided without a large number of samples; it has fewer missed detections, strong versatility, and is suitable for pin marks with different arrangements.

[0094] The above embodiments merely illustrate several implementation methods of the present invention, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the invention patent. It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of the present invention. These are all equivalent modifications and improvements made to the above embodiments based on the essential technology of the present invention, and all of these fall within the protection scope of the present invention.

Claims

1. A method for inspecting the appearance of a chip, characterized in that, Includes the following steps: Image acquisition: Acquiring images of the chip surface; Image processing: The acquired image is processed to grayscale, and the grayscale image is processed by image pyramid downsampling to obtain grayscale images at different scales; Morphological black hat and top hat processing is performed on grayscale images at different scales to accommodate some of the bright and dark needle marks, resulting in grayscale images with morphological results. Image enhancement: The morphological result image is enhanced by a linear transformation based on histogram normalization to obtain an enhanced grayscale image. The enhanced grayscale image is then thresholded to become a binary image. The binary image is then upsampled using an image pyramid to obtain an upsampled grayscale image. To generate a needle mark distribution image: the upsampled image is thresholded and segmented to become a binarized image; then morphological dilation is performed to obtain the needle mark distribution image. Generate a needle mark template distribution map: Set the arrangement parameters of the needle marks, and generate a needle mark template distribution map based on the arrangement parameters of the needle marks; Image recognition processing: Perform pin mark recognition on the pin mark distribution image to obtain the pin mark region in the pin mark distribution image. Compare the pin mark region with the pin mark template distribution image. Add a standard pin mark to the area on the pin mark distribution image where the pin mark region and the pin mark template distribution image do not overlap, to form the processed pin mark distribution image. Chip image processing: Remove the processed pin mark distribution image from the chip surface image to form a pin mark removal image, and then identify defects from the pin mark removal image.

2. The chip appearance inspection method according to claim 1, characterized in that: In the step of generating the pin mark template distribution map, the pin mark arrangement parameters are specifically one or more of the following: the center spacing of pin marks, the fluctuation of pin mark spacing, the total number of pin marks, and the maximum number of missing pin marks.

3. The chip appearance inspection method according to claim 1, characterized in that: In the image processing step, when performing image pyramid downsampling on the grayscale image, from the first... layer to the first The general formula for full downsampling of a layer is: in, These are the pixel coordinates in the image; It is the radius of the convolution kernel, for a given... The core, ; It is a Gaussian convolution kernel, whose weights are determined by a Gaussian function. Let σ be the standard deviation.

4. The chip appearance inspection method according to claim 1, characterized in that: In the image processing steps, the morphological black hat and top hat processing of grayscale images at different scales specifically involves: in, This represents the image obtained after black hat computation. This represents the image obtained after the top-hat operation. Morphological images representing the overlay of black hats and top hats. Represents the downsampled image. The rectangular convolution kernel is set.

5. The chip appearance inspection method according to claim 1, characterized in that: In the image enhancement step, the image of the morphological result undergoes a linear transformation enhancement process based on histogram normalization, specifically as follows: Generate a corresponding histogram based on the grayscale information of the grayscale image obtained from the morphological results, and output the maximum grayscale value in the histogram. With minimum gray value ; Set the enhanced grayscale image Maximum grayscale value With minimum gray value ; Calculate the contrast of the enhanced image ,brightness and the grayscale image of the enhanced result. .

6. The chip appearance inspection method according to claim 5, characterized in that: Calculate the contrast of the enhanced image ,brightness and the grayscale image of the enhanced result. Specifically: In the formula, A morphological image representing the overlay of a black hat and a top hat.

7. The chip appearance inspection method according to claim 1, characterized in that: In the image enhancement step, thresholding the enhanced grayscale image specifically involves: in, The resulting binarized image, The set segmentation threshold.

8. The chip appearance inspection method according to claim 1, characterized in that: In the image enhancement step, when performing image pyramid upsampling on the binarized image, from the first... layer to the first The formula for full upsampling of a layer is: in, For binarized images, This is the image obtained through upsampling.

9. The chip appearance inspection method according to claim 1, characterized in that: In the step of forming the needle mark distribution image, the calculation formula for the morphological dilation process is as follows: in, This is the image after dilation. It is an upsampled segmented image. It is the position of the structural element The value of .

10. The chip appearance inspection method according to claim 9, characterized in that: In the image recognition processing step, needle mark recognition of the needle mark distribution image specifically involves: ... Abnormal residues are divided by row and / or column, and the center coordinates of the residues in each row and / or column are sorted from smallest to largest. If a row and / or column contains only one residue, the row is directly removed without further needle mark identification. Each residual point is recorded as... The continuity of stitch marks is determined based on their arrangement parameters; specifically... in, The distance between the centers of the needle marks. This represents the fluctuation in the distance between needle marks. The maximum spacing between the centers of the needle marks; The minimum spacing between the centers of the needle marks; The maximum number of missing stitches is allowed; when both of the above formulas are satisfied, it is determined that the stitches are discontinuous.

Images