Chip defect detection system based on image recognition technology

Through the chip defect detection system based on image recognition technology, surface defect detection, edge division and point classification analysis are used to solve the problem of low chip detection efficiency in the existing technology, and the accurate detection and effective treatment of chip surface defects are achieved. Improves the efficiency and stability of the chip.

CN118982497AInactive Publication Date: 2024-11-19SHENZHEN ZHIHE TECH CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410884526.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-03
Publication Date
2024-11-19
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art cannot effectively detect chip surface defects, resulting in low detection efficiency and inaccurate chip detection, affecting chip usage efficiency.

Method used

Using a chip defect detection system based on image recognition technology, including a surface defect detection unit, an image edge division detection unit and a point classification detection unit. Through image acquisition, edge division and pixel point analysis, defect detection coefficient, edge blur information and Point parameters, infer whether there are defects on the chip surface.

Benefits of technology

Accurate detection of chip surface defects is achieved, reducing operating efficiency caused by not detecting obvious defects is avoided, and the accuracy and efficiency of chip detection is improved, ensuring the stability and reliability of chips.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982497A_ABST
    Figure CN118982497A_ABST
Patent Text Reader

Abstract

The invention discloses a chip defect detection system based on an image recognition technology, relates to the technical field of chip defect detection, and solves the technical problem that in the prior art, defect detection cannot be further carried out on the surface of a chip through the image recognition technology after obvious defect detection. Specifically, edge division detection is carried out on an analysis picture of a chip, image analysis is carried out on the analysis picture of the chip, chip defect detection is further carried out, and the problems that defect detection cannot be carried out in time when no obvious defect exists, the defect detection is inaccurate, the chip detection efficiency is reduced, and the actual use feasibility of the chip is affected are avoided. Point location classification division detection is carried out on the analysis pictures, the number of the analysis pictures is continuously increased according to the increase of the shooting duration of the analysis pictures, and therefore point location classification detection is carried out on the analysis pictures in a summarizing mode in the increasing process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of chip defect detection, and in particular to a chip defect detection system based on image recognition technology. Background Art

[0002] Chip defect detection is an important part of the chip production process. Its main purpose is to detect various defects on the chip surface or package to ensure the quality and performance of the chip.

[0003] However, in the prior art, it is impossible to detect obvious defects on the chip surface during chip inspection. At the same time, it is impossible to further perform defect detection on the chip surface through image recognition technology after obvious defects are detected. As a result, the chip defect detection efficiency is low, and the chip inspection cannot be completed reasonably and accurately, resulting in reduced chip utilization efficiency.

[0004] In view of the above technical defects, a solution is now proposed. Summary of the invention

[0005] The purpose of the present invention is to solve the above-mentioned problems and to propose a chip defect detection system based on image recognition technology.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A chip defect detection system based on image recognition technology includes a chip defect detection platform, and the chip defect detection platform is connected to:

[0008] The surface defect detection unit is used to collect images of the chip surface and divide the collected images into several analysis pictures. The label o is set, where o is a natural number greater than 1. The analysis pictures are sorted in the order of image collection time. The surface defect detection parameters are collected and substituted into the formula to obtain the chip surface defect detection coefficient. The surface defect detection is inferred to be normal based on the coefficient comparison.

[0009] The image edge division detection unit is used to set the area of ​​the image division area and divide the edge of the analyzed image according to the area of ​​the image division area, and mark the divided area as the edge division area, collect edge blur information and edge fitting information, and infer whether the edge division is qualified according to the information analysis, and process the brightness of the sub-area after it is qualified. After the processing is completed, the reduction trend parameter and the growth trend parameter are obtained, and the edge detection is inferred whether it is normal according to the parameter comparison;

[0010] The point classification and division detection unit is used to divide the pixels according to the real-time area of ​​the analysis image, and the divided pixel array is distributed throughout the area of ​​the corresponding analysis image. The pixel parameters are collected and set as point parameters, and single point data and multiple point data are collected. Point division is performed according to data analysis, and whether the point division detection is qualified is inferred according to the point type.

[0011] As a preferred embodiment of the present invention, the surface defect detection parameters include the numerical value of the printed character width deviation at any moment in the analysis picture sequence where the characters printed on the chip surface are located and the printing track of the set characters, the horizontal offset of the corresponding pins of the same type set at the same position on the chip surface in the analysis picture, and the numerical ratio of the deformation amount at the same position of the packaging shell and the chip body after the chip surface is packaged in the analysis picture.

[0012] As a preferred embodiment of the present invention, if the chip surface defect detection coefficient exceeds the detection coefficient threshold, a surface defect abnormal signal is generated; if the chip surface defect detection coefficient does not exceed the detection coefficient threshold, a surface defect normal signal is generated.

[0013] As a preferred embodiment of the present invention, the edge blur information and edge fitting information are respectively the ratio of the deviation value between the clarity of the edge divided area in the analysis picture and the set clarity of the analysis picture and the deviation value between the clarity of the non-edge divided area and the set clarity of the analysis picture, and the image distance deviation between the overlap of the boundary images of adjacent edge divided areas in the analysis picture and the corresponding position of the actual analysis picture.

[0014] As a preferred embodiment of the present invention, if the edge blur information exceeds the deviation value ratio threshold, or the edge fit information exceeds the image distance deviation threshold, it is determined that the edge segmentation of the analyzed image is qualified;

[0015] If the edge blur information does not exceed the deviation value ratio threshold, and the edge fit information does not exceed the image distance deviation threshold, the edge segmentation of the analyzed image is determined to be unqualified.

[0016] As a preferred embodiment of the present invention, the brightness processing process is as follows:

[0017] The grayscale value of the displayed pixel in the analysis image when the defect-free sub-area in the analysis image is obtained when the image is collected, and the grayscale value range is set. The grayscale value of the displayed pixel in each sub-area in the real-time edge division area is collected, and the grayscale value of the displayed pixel is compared. The sub-area whose grayscale value of the displayed pixel is not within the grayscale value range is darkened, and vice versa, the sub-area whose grayscale value of the displayed pixel is within the grayscale value range is brightened.

[0018] As a preferred embodiment of the present invention, the decreasing trend parameter and the increasing trend parameter are respectively a floating span value of the decreasing trend of the overall average brightness value of the edge division area in the process of analyzing the increase of the picture, and a floating span value of the increasing trend of the ratio of the number of brightness increasing areas to the brightness decreasing areas in any edge division area in the process of analyzing the increase of the picture;

[0019] If the decreasing trend parameter exceeds the decreasing trend floating span threshold, or the increasing trend parameter does not exceed the increasing trend floating span threshold, an edge detection abnormal signal is generated; if the decreasing trend parameter does not exceed the decreasing trend floating span threshold, and the increasing trend parameter exceeds the increasing trend floating span threshold, an edge detection normal signal is generated.

[0020] As a preferred embodiment of the present invention, the single point data and the multiple point data are respectively the maximum floating span value of the point parameters corresponding to the pixel points during the analysis of the picture increase process and the maximum growth peak value of the point parameter deviation value of adjacent pixel points during the analysis of the picture increase process.

[0021] As a preferred implementation mode of the present invention, if a single point data exceeds the parameter maximum floating span value threshold, or multiple point data exceed the parameter deviation maximum growth peak value threshold, the corresponding pixel point in the analysis image is marked as a high floating point; if a single point data does not exceed the parameter maximum floating span value threshold, and multiple point data do not exceed the parameter deviation maximum growth peak value threshold, the corresponding pixel point in the analysis image is marked as a low floating point, and then the detection result is inferred based on whether the real-time pixel point is qualified. If the real-time pixel point is qualified, the point classification and division detection is qualified, otherwise the point classification and division detection is unqualified;

[0022] As a preferred embodiment of the present invention, if the pixel point is a high floating point, and the peak and valley values ​​of the real-time parameter values ​​of the corresponding point parameters are both within the corresponding parameter range, then the point classification and division detection of the corresponding pixel point is judged to be qualified, a point classification and division detection qualified signal is generated and the point classification and division detection qualified signal is sent to the chip defect detection platform; when the peak and valley values ​​of the real-time parameter values ​​of the corresponding point parameters are not both within the corresponding parameter range, then the point classification and division detection of the corresponding pixel point is judged to be unqualified, and the point parameters of the corresponding pixel point are subjected to floating trend control.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] 1. In the present invention, the chip is inspected for surface defects, and the chip surface is inspected for defects through image recognition technology. The chips with obvious defects on the surface are returned to the factory for processing to avoid the chip surface defects causing the chip's operating efficiency to decrease, resulting in the chip having performance abnormalities and being unable to guarantee the real-time operation execution stability; the chip analysis picture is inspected for edge division, and the chip defect detection is further performed by performing image analysis on the chip analysis picture to avoid the failure to perform defect detection in time when there are no obvious defects, resulting in inaccurate defect detection, which reduces the chip detection efficiency and affects the practical feasibility of the chip.

[0025] 2. In the present invention, point classification and division detection is performed on the analysis pictures. As the shooting time of the analysis images increases, the number of corresponding analysis pictures also increases continuously. Therefore, point classification detection is performed on the analysis pictures in the increasing process to infer whether there is abnormal wear on the current chip surface, causing chip surface defects that affect the use efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0027] Figure 1 This is a principle block diagram of Embodiment 1 of the present invention;

[0028] Figure 2 This is a principle block diagram of the second embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] Embodiment 1

[0032] This embodiment performs defect detection on the chip. Figure 1As shown, a chip defect detection system based on image recognition technology includes a chip defect detection platform, wherein the chip defect detection platform is communicatively connected with a surface defect detection unit, an image edge division detection unit, and a point classification division detection unit;

[0033] The chip defect detection platform generates a surface defect detection signal and sends the surface defect detection signal to the surface defect detection unit. After receiving the surface defect detection signal, the surface defect detection unit performs surface defect detection on the chip, detects defects on the chip surface through image recognition technology, and returns the chip with obvious defects on the surface to the factory for processing, so as to avoid the chip surface defects causing the chip's operating efficiency to decrease, resulting in abnormal performance of the chip and the inability to guarantee the real-time operation execution stability;

[0034] Capture images of the chip surface, and divide the captured images into several analysis pictures, set a label o, o is a natural number greater than 1, and sort the analysis pictures according to the image acquisition time sequence when analyzing, obtain the value of the printed character width deviation at any moment between the character printing track in the analysis picture sequence where the printed characters on the chip surface in the analysis picture are located and the printing track of the set characters, and mark the value of the printed character width deviation at any moment between the character printing track in the analysis picture sequence where the printed characters on the chip surface in the analysis picture are located and the printing track of the set characters as YSKo;

[0035] The horizontal offset of the pins corresponding to the same type of pins at the same position on the chip surface in the analysis picture is obtained, and the horizontal offset of the pins corresponding to the same type of pins at the same position on the chip surface in the analysis picture is marked as PYLo. The pin types are input pins, output pins, and bidirectional pins, which have the same appearance but different functions. The offset or missing installation position will affect the operation of the chip;

[0036] The numerical ratio of the deformation amount of the package shell and the chip body at the same position after the chip surface is packaged in the analysis picture is obtained, and the numerical ratio of the deformation amount of the package shell and the chip body at the same position after the chip surface is packaged in the analysis picture is marked as SZBo, mainly to evaluate the anti-deformation ability of the package shell;

[0037] The above collected data are uniformly marked as surface defect detection parameters, and substituted into the formula to obtain the chip surface defect detection coefficient QJ, where the formula is Among them, fqr1, fqr2 and fqr3 are preset proportional coefficients of the numerical value of the printed character width deviation, the corresponding pin horizontality offset and the numerical value ratio of the deformation variable at any time, and e is a natural constant, and β is an error correction factor with a value of 1.09;

[0038] Compare the chip surface defect detection coefficient QJ with the detection coefficient threshold:

[0039] If the chip surface defect detection coefficient QJ exceeds the detection coefficient threshold, the chip surface defect detection is determined to be abnormal, a surface defect abnormality signal is generated and sent to the chip defect detection platform. After the chip defect detection platform receives the surface defect abnormality signal, the chip is returned to the factory for processing;

[0040] If the chip surface defect detection coefficient QJ does not exceed the detection coefficient threshold, the chip surface defect detection is determined to be normal, and a surface defect normal signal is generated and sent to the chip defect detection platform; at the same time, an image edge division detection signal and a point classification division detection signal are generated and sent to the image edge division detection unit and the point classification division detection unit respectively;

[0041] After receiving the image edge division detection signal, the image edge division detection unit performs edge division detection on the chip analysis picture, and further performs chip defect detection by performing image analysis on the chip analysis picture, so as to avoid the inability to perform defect detection in time when there are no obvious defects, resulting in inaccurate defect detection, reducing chip detection efficiency and affecting the actual feasibility of chip use;

[0042] Set the area of ​​the image division region and divide the analysis image into edges according to the area of ​​the image division region, and mark the divided area as the edge division region, obtain the ratio of the deviation value between the edge division region clarity and the analysis image set clarity deviation value and the non-edge division region clarity and the analysis image set clarity deviation value, and at the same time obtain the image distance deviation between the overlap of the boundary images of adjacent edge division regions in the analysis image and the actual analysis image corresponding position, and mark the ratio of the deviation value between the edge division region clarity and the analysis image set clarity deviation value and the non-edge division region clarity and the analysis image set clarity deviation value, and the image distance deviation between the overlap of the boundary images of adjacent edge division regions in the analysis image and the actual analysis image corresponding position as edge blur information and edge fitting information, respectively, and compare them with the deviation value ratio threshold and the image distance deviation threshold, respectively:

[0043] If the ratio of the deviation value between the edge division area clarity and the analysis picture set clarity deviation value and the non-edge division area clarity and the analysis picture set clarity deviation value exceeds the deviation value ratio threshold, or the image distance deviation between the overlapped boundary images of adjacent edge division areas in the analysis picture and the corresponding position of the actual analysis picture exceeds the image distance deviation threshold, then the edge division of the analysis picture is determined to be qualified;

[0044] If the ratio of the deviation value between the edge division area clarity and the analysis picture set clarity deviation value in the analysis picture and the deviation value between the non-edge division area clarity and the analysis picture set clarity deviation value does not exceed the deviation value ratio threshold, and the image distance deviation between the overlapped boundary images of adjacent edge division areas in the analysis picture and the corresponding position of the actual analysis picture does not exceed the image distance deviation threshold, then the edge division of the analysis picture is determined to be unqualified;

[0045] If the edge division is unqualified, the area of ​​the image division area is changed and re-divided. If the edge division is qualified, the edge division area in the analysis image is divided into i sub-areas, i is a natural number greater than 1, and the grayscale value of the displayed pixel in the analysis image when the defect-free sub-area in the analysis image is collected, and the grayscale value range is set, and the grayscale value of the displayed pixel in each sub-area in the real-time edge division area is collected, and the grayscale value of the displayed pixel is compared, and the sub-area whose grayscale value of the displayed pixel is not within the grayscale value range is darkened, and vice versa, the sub-area whose grayscale value of the displayed pixel is within the grayscale value range is brightened;

[0046] After the sub-region is processed, the floating span value of the decreasing trend of the overall average brightness value of the edge division region during the increase of the analysis image and the floating span value of the increasing trend of the ratio of the number of brightness increase regions to the brightness decrease regions in any edge division region during the increase of the analysis image are obtained, and the floating span value of the decreasing trend of the overall average brightness value of the edge division region during the increase of the analysis image and the floating span value of the increasing trend of the ratio of the number of brightness increase regions to the brightness decrease regions in any edge division region during the increase of the analysis image are marked as decreasing trend parameters and increasing trend parameters, respectively, and compared with the decreasing trend floating span threshold and the increasing trend floating span threshold, respectively:

[0047] If the floating span value of the decreasing trend of the overall average brightness value of the edge division area during the analysis image addition process exceeds the decreasing trend floating span threshold, or the floating span value of the increasing trend of the ratio of the number of brightness increase areas to the brightness decrease areas in any edge division area during the analysis image addition process does not exceed the increasing trend floating span threshold, then it is determined that the edge division detection of the analysis image is abnormal, an edge detection abnormality signal is generated and sent to the chip defect detection platform, and after the chip defect detection platform receives the edge detection abnormality signal, it performs chip maintenance on the sub-area corresponding to the analysis image;

[0048] If the floating span value of the decreasing trend of the overall average brightness value of the edge division area during the analysis image addition process does not exceed the decreasing trend floating span threshold, and the floating span value of the increasing trend of the ratio of the number of brightness increasing areas to the brightness decreasing areas in any edge division area during the analysis image addition process exceeds the increasing trend floating span threshold, then it is determined that the edge division detection of the analysis image is normal, an edge detection normal signal is generated, and the edge detection normal signal is sent to the chip defect detection platform;

[0049] After receiving the point classification detection signal, the point classification detection unit performs point classification detection on the analysis image. As the shooting time of the analysis image increases, the number of corresponding analysis images also increases continuously. Therefore, the analysis images are summarized and point classification detection is performed on the analysis images during the increase process to infer whether there is abnormal wear on the current chip surface, causing chip surface defects that affect the use efficiency;

[0050] Pixel points are divided according to the real-time area of ​​the analysis image, and the divided pixel point array is distributed to cover the area corresponding to the analysis image. Parameters of the pixel points are collected and set as point parameters. The maximum floating span value of the point parameters corresponding to the pixel points in the process of increasing the analysis image is obtained. At the same time, the maximum growth peak value of the point parameter deviation value of the adjacent pixel points in the process of increasing the analysis image is obtained. The maximum floating span value of the point parameters corresponding to the pixel points in the process of increasing the analysis image and the maximum growth peak value of the point parameter deviation value of the adjacent pixel points in the process of increasing the analysis image are marked as single point data and multiple point data, respectively, and compared with the parameter maximum floating span value threshold and the parameter deviation maximum growth peak threshold, respectively: the point parameters are expressed as size, color, brightness, etc.;

[0051] If the maximum floating span value of the point parameter corresponding to the pixel point during the analysis of the image increase exceeds the parameter maximum floating span value threshold, or the maximum growth peak value of the point parameter deviation value of adjacent pixel points during the analysis of the image increase exceeds the parameter deviation maximum growth peak value threshold, then the corresponding pixel point in the analysis of the image is marked as a high floating point; if the maximum floating span value of the point parameter corresponding to the pixel point during the analysis of the image increase does not exceed the parameter maximum floating span value threshold, and the maximum growth peak value of the point parameter deviation value of adjacent pixel points during the analysis of the image increase does not exceed the parameter deviation maximum growth peak value threshold, then the corresponding pixel point in the analysis of the image is marked as a low floating point; then the detection result is inferred based on whether the real-time pixel point is qualified. If the real-time pixel point is qualified, the point classification and division detection is qualified, otherwise the point classification and division detection is unqualified;

[0052] If the pixel point is a high floating point, when the peak and valley values ​​of the real-time parameter values ​​of the corresponding point parameters are both within the corresponding parameter range, the point classification and division detection of the corresponding pixel point is judged to be qualified, and a point classification and division detection qualified signal is generated and sent to the chip defect detection platform; if the peak and valley values ​​of the real-time parameter values ​​of the corresponding point parameters are not both within the corresponding parameter range, the point classification and division detection of the corresponding pixel point is judged to be unqualified, and the point parameters of the corresponding pixel point are subjected to floating trend control;

[0053] Embodiment 2

[0054] The previous embodiment performs real-time detection of chip defects, and this embodiment, based on the previous embodiment, performs chip defect detection in combination with chip failure impact analysis. Figure 2 As shown, the chip defect detection platform is communicated with a fault impact analysis unit;

[0055] The chip defect detection platform generates a fault impact analysis signal and sends the fault impact analysis signal to the fault impact analysis unit. After receiving the fault impact analysis signal, the fault impact analysis unit performs a fault impact analysis on the chip after defect detection, and infers whether the fault impact of the chip after the current defect detection is qualified, so as to perform a fault impact analysis on the chip, thereby avoiding the decrease in chip operation feasibility caused by untimely defect detection efficiency;

[0056] The decreasing span value of the chip damage probability after a fault occurs during the chip operation after defect detection is completed is obtained, and at the same time, the delay time increase of the defect in the chip component during the fault generation stage during the chip operation after defect detection is completed is obtained, and the decreasing span value of the chip damage probability after a fault occurs during the chip operation after defect detection is completed and the delay time increase of the defect in the chip component during the fault generation stage during the chip operation after defect detection are marked as damage impact data and delay defect data, respectively, and compared with the probability decreasing span value threshold and the delay time increase threshold, respectively:

[0057] If the decreasing span value of the chip damage probability after a fault occurs during the operation of the chip after the defect detection is completed exceeds the probability decreasing span value threshold, or the increase in the delay time of the defect in the chip component during the fault generation stage during the operation of the chip after the defect detection is completed does not exceed the delay time increase threshold, then it is determined that the defect detection has a small impact on the fault, a low impact signal is generated and sent to the chip defect detection platform, and after the chip defect detection platform receives the low impact signal, the defect detection cycle of the current batch of chips is extended;

[0058] If the decreasing span value of the probability of chip damage after a fault occurs during chip operation after defect detection is completed does not exceed the probability decreasing span value threshold, and the increase in the delay time for defects in components in the chip during the fault generation phase during chip operation after defect detection is completed exceeds the delay time increase threshold, then it is determined that the defect detection has a large impact on the fault, and a high-impact signal is generated and sent to the chip defect detection platform. After receiving the high-impact signal, the chip defect detection platform shortens the defect detection cycle of the current batch of chips and prohibits their use until the defects are rectified.

[0059] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by technicians in this field according to actual conditions;

[0060] When the present invention is in use, the surface defect detection unit collects images of the chip surface and divides the collected images into several analysis images, and the analysis images are sorted in the order of image collection time when analyzing, and surface defect detection parameters are collected, and the chip surface defect detection coefficient is obtained by substituting it into a formula, and whether the surface defect detection is normal is inferred according to the coefficient comparison; the image edge division detection unit sets the area of ​​the image division area and divides the analysis image into edges according to the area of ​​the image division area, and marks the divided area as the edge division area, collects edge blur information and edge fitting information, and infers whether the edge division is qualified according to the information analysis, and processes the brightness of the sub-area after it is qualified, and obtains the reduction trend parameter and the growth trend parameter after the processing is completed, and infers whether the edge detection is normal according to the parameter comparison; the point classification division detection unit divides the pixels according to the real-time area of ​​the analysis image, and the divided pixel array is distributed to cover the area corresponding to the analysis image, and the pixel points are collected and set as point parameters, and single point data and multiple point data are collected, and point division is performed according to data analysis, and whether the point division detection is qualified is inferred according to the point type.

[0061] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A chip defect detection system based on image recognition technology, characterized in that: Including chip defect detection platform, the chip defect detection platform communication connections are: The surface defect detection unit is used to collect images of the chip surface and divide the collected images into several analysis pictures. The label o is set, where o is a natural number greater than 1. The analysis pictures are sorted in the order of image collection time. The surface defect detection parameters are collected and substituted into the formula to obtain the chip surface defect detection coefficient. The surface defect detection is inferred to be normal based on the coefficient comparison. The image edge division detection unit is used to set the area of ​​the image division area and divide the edge of the analyzed image according to the area of ​​the image division area, and mark the divided area as the edge division area, collect edge blur information and edge fitting information, and infer whether the edge division is qualified according to the information analysis, and process the brightness of the sub-area after it is qualified. After the processing is completed, the reduction trend parameter and the growth trend parameter are obtained, and the edge detection is inferred whether it is normal according to the parameter comparison; The point classification and division detection unit is used to divide the pixels according to the real-time area of ​​the analysis image, and the divided pixel array is distributed throughout the area of ​​the corresponding analysis image. The pixel parameters are collected and set as point parameters, and single point data and multiple point data are collected. Point division is performed according to data analysis, and whether the point division detection is qualified is inferred according to the point type.

2. According to claim 1, a chip defect detection system based on image recognition technology is characterized in that: The surface defect detection parameters include the deviation value of the printed character width at any moment in the analysis picture sequence between the character printing track and the set character printing track, the horizontal offset of the corresponding pins of the same type set at the same position on the chip surface in the analysis picture, and the numerical ratio of the deformation amount at the same position of the packaging shell and the chip body after the chip surface is packaged in the analysis picture.

3. The chip defect detection system based on image recognition technology according to claim 2 is characterized in that: If the chip surface defect detection coefficient exceeds the detection coefficient threshold, a surface defect abnormal signal is generated; if the chip surface defect detection coefficient does not exceed the detection coefficient threshold, a surface defect normal signal is generated.

4. The chip defect detection system based on image recognition technology according to claim 1, characterized in that: The edge blur information and edge fitting information are respectively the ratio of the deviation value between the clarity of the edge divided area in the analysis picture and the set clarity of the analysis picture to the deviation value between the clarity of the non-edge divided area and the set clarity of the analysis picture, and the image distance deviation between the overlap of the boundary images of adjacent edge divided areas in the analysis picture and the corresponding position of the actual analysis picture.

5. The chip defect detection system based on image recognition technology according to claim 4 is characterized in that: If the edge blur information exceeds the deviation value ratio threshold, or the edge fit information exceeds the image distance deviation threshold, the edge segmentation of the analyzed image is determined to be qualified; If the edge blur information does not exceed the deviation value ratio threshold, and the edge fit information does not exceed the image distance deviation threshold, the edge segmentation of the analyzed image is determined to be unqualified.

6. The chip defect detection system based on image recognition technology according to claim 5, characterized in that: The brightness processing process is as follows: The grayscale value of the displayed pixel in the analysis image when the defect-free sub-area in the analysis image is obtained when the image is collected, and the grayscale value range is set. The grayscale value of the displayed pixel in each sub-area in the real-time edge division area is collected, and the grayscale value of the displayed pixel is compared. The sub-area whose grayscale value of the displayed pixel is not within the grayscale value range is darkened, and vice versa, the sub-area whose grayscale value of the displayed pixel is within the grayscale value range is brightened.

7. The chip defect detection system based on image recognition technology according to claim 6 is characterized in that: The decreasing trend parameter and the increasing trend parameter are respectively the decreasing trend floating span value of the overall average brightness value of the edge division area in the process of analyzing the increase of the picture, and the increasing trend floating span value of the ratio of the number of brightness increase areas to the brightness decrease areas in any edge division area in the process of analyzing the increase of the picture; If the decreasing trend parameter exceeds the decreasing trend floating span threshold, or the increasing trend parameter does not exceed the increasing trend floating span threshold, an edge detection anomaly signal is generated; If the decreasing trend parameter does not exceed the decreasing trend floating span threshold, and the increasing trend parameter exceeds the increasing trend floating span threshold, an edge detection normal signal is generated.

8. The chip defect detection system based on image recognition technology according to claim 1, characterized in that: The single point data and multiple point data are respectively the maximum floating span value of the point parameters corresponding to the pixel points in the process of increasing the analysis picture and the maximum growth peak value of the point parameter deviation value of adjacent pixel points in the process of increasing the analysis picture.

9. The chip defect detection system based on image recognition technology according to claim 8, characterized in that: If a single point data exceeds the parameter maximum floating span value threshold, or multiple point data exceed the parameter deviation maximum growth peak value threshold, the corresponding pixel in the analysis image will be marked as a high floating point; if a single point data does not exceed the parameter maximum floating span value threshold, and multiple point data do not exceed the parameter deviation maximum growth peak value threshold, the corresponding pixel in the analysis image will be marked as a low floating point, and then the detection result will be inferred based on whether the real-time pixel point is qualified. If the real-time pixel point is qualified, the point classification and division detection is qualified, otherwise the point classification and division detection is unqualified.

10. The chip defect detection system based on image recognition technology according to claim 9, characterized in that: If the pixel point is a high floating point, and the peak and valley values ​​of the corresponding point parameter real-time parameter values ​​are both within the corresponding parameter range, the corresponding pixel point classification and division detection is judged to be qualified, and a point classification and division detection qualified signal is generated and sent to the chip defect detection platform; if the peak and valley values ​​of the corresponding point parameter real-time parameter values ​​are not both within the corresponding parameter range, the corresponding pixel point classification and division detection is judged to be unqualified, and the point parameters of the corresponding pixel point are subjected to floating trend control.

Citation Information

Cited By

  • Self-adaptive brightness classification method and system for sparsely distributed microbeads on gene chip

    CN119600091A

  • Semiconductor chip packaging defect detection method

    CN119985531A

  • Automatic optical detection device for PCB (Printed Circuit Board) defects based on machine vision

    CN120404738A