A wafer AOI detection method and system

By using the European distance algorithm and grayscale measurement processing in wafer AOI detection, the problem of color difference influence during wafer production is solved, and the wafer defect detection rate and detection accuracy are improved.

CN118505700BActive Publication Date: 2024-11-15HUAMAO ZHIXIN INTEGRATED ELECTRONICS (JIANGSU) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410962590.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-18
Publication Date
2024-11-15
Estimated Expiration
2044-07-18

AI Technical Summary

Technical Problem

During the production process, the color difference caused by different processes affects the accuracy of AOI detection, resulting in a low detection rate of wafer defects.

Method used

By acquiring the wafer image and dividing it into sub-regions, the chromatic aberration distance sum is calculated based on the European distance algorithm, standardized processing is performed to obtain grayscale quantities, re-dividing the image range to remove the chromatic aberration effect caused by the stripping process, and calculating the adjacent bare chip's Grayscale difference to obtain defect information.

Benefits of technology

It improves tolerance for wafer color aberration, improves the effective detection rate of wafer defects, and ensures the accuracy and efficiency of detection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118505700B_ABST
    Figure CN118505700B_ABST
Patent Text Reader

Abstract

The present invention provides a wafer AOI detection method and system, the method comprising: obtaining a wafer image; dividing the wafer image into a plurality of sub-areas and calculating the color difference distance and of the bare chips in each sub-area based on the Euclidean distance algorithm; obtaining the grayscale of each sub-area; re-dividing the wafer image based on the grayscale of each sub-area to obtain a plurality of continuous image ranges; in each image range, determining a bare chip as a detection object and calculating the grayscale difference between the bare chip and the adjacent bare chip, obtaining the defect information of the bare chip, and then obtaining the defect information of all bare chips in each image range, and obtaining the defect information of each image range. The wafer AOI detection method provided by the present invention cleverly uses grayscale to replace the existing method of directly using grayscale value for defect judgment, which can effectively remove the influence of color difference caused by process on detection, improve the tolerance to wafer color difference, and improve the effective detection rate of wafer defects.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of semiconductor technology, and in particular to a wafer AOI detection method and system. Background Art

[0002] Wafer refers to the silicon chip used in the production of silicon semiconductor integrated circuits. It is the carrier used in the production of integrated circuits (ICs). Therefore, the quality of the wafer directly affects the yield rate and manufacturing cost of the chip. In the process of making wafers, some wafers will inevitably have defects. In order to monitor the defects of the wafers, detection is required. Automated Optical Inspection (AOI) is a high-speed and high-precision optical image detection system. It uses machine vision as the standard detection technology, which can improve the shortcomings of traditional manual optical instrument detection and realize the detection of wafer defects.

[0003] However, in the actual wafer production process, due to the different processes used, there will be color differences in different areas of the produced wafers. For example, the different processes at the center and edge of the wafer will lead to obvious color differences between the center and edge. In the AOI inspection process, the existence of color difference will affect the judgment of grayscale differences caused by wafer defects, that is, color difference is likely to cause other interference defects, affecting the effective detection rate of wafer defects. Summary of the invention

[0004] The present invention aims to provide a wafer AOI inspection method and system to solve the above-mentioned technical problems, improve the tolerance to wafer color difference, and enhance the effective detection rate of wafer defects.

[0005] In order to solve the above technical problems, the present invention provides a wafer AOI detection method, comprising the following steps:

[0006] Acquire wafer images;

[0007] Divide the wafer image into several sub-regions and calculate the color difference distance and the color difference distance of the die in each sub-region based on the Euclidean distance algorithm;

[0008] The color difference distance of the bare chip in each sub-region is normalized to obtain the grayscale of each sub-region;

[0009] Re-dividing the wafer image based on the grayscale of each sub-region to obtain a number of continuous image ranges, wherein the grayscale difference of the sub-regions within each image range meets a preset condition;

[0010] In each image range, a die is determined as the detection object and the grayscale difference between the die and the adjacent die is calculated to obtain the defect information of the die, and then the defect information of all the die in each image range is obtained to obtain the defect information of each image range.

[0011] In the above scheme, the grayscale of the sub-region can characterize the average color difference between each die in the sub-region. The larger the grayscale, the larger the average color difference between each die in the sub-region, that is, the wafer is processed by multiple processes in the sub-region, and there will be obvious color difference lines in the wafer in the sub-region. At this time, the wafer image can be re-divided by grayscale, and the sub-regions produced by only the same process are divided into the same image range to obtain image range defect information, which can effectively remove the influence of color difference caused by the process on the detection and effectively improve the tolerance for color difference. The sub-regions with color difference lines will be divided into the same image range for centralized detection. By calculating the grayscale difference between adjacent die, the die on the color difference line can be screened out. The defect information of this part of the die can be compared and analyzed with the standard qualified wafers produced on the same production line, and the defect information of the die in the non-color difference line area can be obtained by judging the grayscale difference between adjacent die, thereby completing the defect information of the entire wafer.

[0012] The above solution cleverly uses grayscale to replace the existing method of directly using grayscale values ​​for defect judgment. It can accurately classify each sub-area on the wafer, effectively remove the impact of color difference caused by the process on detection, improve the tolerance to wafer color difference, and improve the effective detection rate of wafer defects.

[0013] It should be noted that the preset conditions here can be adjusted based on the actual detection accuracy.

[0014] Furthermore, before acquiring the wafer image, the wafer is pretreated: a hardened protective layer is formed on the front side of the wafer; the back side of the wafer is treated with a concentrated hydrofluoric acid solvent to reduce the color difference on the back side of the wafer; and the hardened protective layer on the front side of the wafer is removed to obtain a pretreated wafer.

[0015] In the above scheme, in order to avoid the occurrence of partial oxidation on the back side of the wafer in the previous process, the wafer can be cleaned before testing. Concentrated hydrofluoric acid solvent can effectively remove impurities such as silicon nitride or silicon oxide on the back side of the wafer that will cause obvious color difference, eliminate the color difference on the back side of the wafer, and thus ensure the accuracy of the test.

[0016] It should be noted that the hardened protective layer can be composed of ion-doped photoresist, that is, first a layer of photoresist is spin-coated on the front side of the wafer, and then ion doping is injected into the photoresist to harden the photoresist, thereby forming a hardened protective layer to protect the front side of the wafer.

[0017] It should be noted that the concentrated hydrofluoric acid solvent may be a solvent mixed with 49% hydrofluoric acid and 51% water.

[0018] Furthermore, the method of acquiring a wafer image specifically includes: optically scanning the wafer to acquire light intensity grayscale quantities at different positions on the wafer; acquiring bright and dark areas of the wafer based on the light intensity grayscale quantities, and performing light selection and light intensity adjustment to obtain an optimal light flux condition for the wafer; and acquiring a wafer image under the optimal light flux condition.

[0019] Furthermore, the bright and dark areas of the wafer are obtained based on the light intensity grayscale, and the light source is selected and the light intensity is adjusted to obtain the optimal luminous flux condition of the wafer, specifically: the bright and dark areas of the wafer are obtained based on the light intensity grayscale; two light sources located at the center of the bright area and the center of the dark area are selected to illuminate the wafer, and the light intensity grayscale of several bare chips from the center of the wafer to the edge of the wafer is obtained in this case; the light intensity of the two light sources is adjusted based on the light intensity grayscale of the several bare chips to obtain the optimal luminous flux condition of the wafer.

[0020] Since the detection of this method is carried out step by step after obtaining the wafer image, in order to ensure the accuracy of the detection, it is very important to accurately obtain the image. In the process of image acquisition, apart from the influence of the chromatic aberration of the wafer itself, the light conditions of the shooting environment should also be considered to avoid the appearance of shadows on the wafer and the aggravation of the chromatic aberration of the wafer. In the above scheme, the placed wafer is first optically scanned to obtain the light intensity grayscale at different positions of the wafer, and then the bright and dark areas of the wafer are determined and two light sources at the center of the bright area and the center of the dark area are selected to illuminate the wafer. The light intensity values ​​emitted by the two light sources can be superimposed, and the optimal light flux conditions of the wafer can be obtained through amplification and compensation. Acquiring the wafer image under the optimal light flux conditions can ensure the accuracy of subsequent detection.

[0021] It should be noted that the bright area and the dark area can be defined in the range of 0 to 255 according to the light intensity grayscale, which is actually the area corresponding to the minimum value and the maximum value.

[0022] Furthermore, the method of dividing the wafer image into several sub-regions and calculating the sum of the color difference distances of the bare chips in each sub-region based on the Euclidean distance algorithm is specifically as follows: dividing the wafer image into several sub-regions; obtaining the RGB signal values ​​of the pixels corresponding to all the bare chips in the several sub-regions; for each sub-region, calculating the RGB signal mean of all the bare chips based on the RGB signal value; and calculating the RGB signal mean of all the bare chips based on the Euclidean distance algorithm to obtain the sum of the color difference distances of the bare chips in each sub-region.

[0023] Furthermore, the Euclidean distance algorithm is used to calculate the mean values ​​of the RGB signals of all the dies to obtain the sum of the color difference distances of the dies in each sub-region, specifically:

[0024] For a die in each sub-region, the color difference distance S is specifically expressed as:

[0025]

[0026] In the formula, It represents the red channel average of the die and other dies, which is specifically expressed as: , Indicates the average red signal of the die, Represents the average red signal of other dies; Indicates the difference between the average red signal of the die and the average red signal of other dies; Indicates the difference between the mean green signal of the die and the mean green signal of other dies; Indicates the difference between the mean blue signal value of the die and the mean blue signal value of other dies;

[0027] The color difference distances of all the bare chips in each sub-region are added together to obtain the sum of the color difference distances of the bare chips in each sub-region.

[0028] The color difference distance and obtained here are only the quantities defined in the above scheme, which are used to characterize the color difference between each bare chip in each sub-region. This quantity needs to be compared with other sub-regions to have a more substantial effect. It should be noted that the color difference distance and defined in the above scheme transform and design the traditional grayscale value, which is similar to characterizing the color uniformity within the sub-region. The wafer preparation process is basically a surface processing process. For example, the processing of the center area and the edge area of ​​the wafer is basically achieved through rotation, which means that the color difference caused by the process on the wafer basically has a color difference line. Thanks to the wafer preparation process, this scheme can well achieve its technical effect. If the defined color difference distance and is too large, it means that there is a color difference line in the sub-region, so that the sub-regions can be classified to remove the color difference caused by the process.

[0029] Furthermore, the color difference distance of the bare chip in each sub-region is normalized to obtain the grayscale of each sub-region, specifically, for each sub-region, the following steps are specifically performed: take the maximum color difference distance Minimum color difference distance ; Calculate the standardized coefficient, specifically:

[0030]

[0031] in =255, =0;

[0032] Normalize the color difference distance of the die in the sub-region to the interval [0,255] to obtain the grayscale value of the sub-region , the specific calculation formula is:

[0033] or

[0034] In the formula, , is the color difference distance of the bare chip and, is the number of dies.

[0035] Furthermore, the wafer image is re-divided based on the grayscale of each sub-region to obtain several continuous image ranges, and the grayscale difference of the sub-regions within each image range meets the preset conditions, specifically: based on the grayscale of each sub-region, the grayscale difference of adjacent sub-regions is calculated, and two adjacent sub-regions whose grayscale difference is less than the preset condition are divided into the same image range, thereby obtaining several continuous image ranges.

[0036] In the above scheme, by setting preset conditions, the wafer area prepared by the same process can be divided into one image range, which can remove the color difference caused by the process and improve the efficiency of detection.

[0037] Furthermore, in each image range, a bare chip is determined as a detection object and the grayscale difference between the bare chip and an adjacent bare chip is calculated to obtain defect information of the bare chip, and then the defect information of all bare chips in each image range is obtained to obtain the defect information of each image range, specifically: in each image range, a bare chip is determined as a detection object; the grayscale difference between the bare chip and the adjacent bare chip is calculated to obtain at least a first difference and a second difference; the difference between the first difference and the second difference is obtained, and if the difference does not meet a preset threshold, the bare chip is marked to obtain the defect information of the bare chip; a bare chip is re-determined, and the above steps are repeated to obtain the defect information of all bare chips to be marked in each image range, and then the defect information of each image range is obtained.

[0038] In the above scheme, each die has multiple pixels, and each pixel has a grayscale value. In theory, the grayscale value of the same pixel at the same position of adjacent dies is the same. By comparing the grayscale values ​​of multiple adjacent dies, the defect information of each pixel on each die can be obtained.

[0039] The present invention also provides a wafer AOI detection system, comprising: an image acquisition module, used to acquire a wafer image; a region division module, used to divide the wafer image into a plurality of sub-regions; a color difference distance calculation module, used to calculate the color difference distance sum of the bare chips in each sub-region based on the Euclidean distance algorithm; a grayscale quantization module, used to standardize the color difference distance sum of the bare chips in each sub-region to obtain the grayscale of each sub-region; an image range division module, used to re-divide the wafer image based on the grayscale of each sub-region to obtain a plurality of continuous image ranges, and the grayscale difference of the sub-regions in each image range meets a preset condition; a defect detection module, used to determine a bare chip as a detection object in each image range and calculate the grayscale difference between the bare chip and an adjacent bare chip to obtain the defect information of the bare chip, and then obtain the defect information of all the bare chips in each image range to obtain the defect information of each image range.

[0040] The system provided by the above solution is simple to build. It cleverly uses grayscale measurement instead of the existing method of directly using grayscale values ​​for defect judgment. It can accurately classify each sub-area on the wafer, effectively remove the impact of color difference caused by the process on detection, improve the tolerance to wafer color difference, and improve the effective detection rate of wafer defects. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A schematic diagram of a wafer AOI inspection method flow chart provided by one embodiment of the present invention;

[0042] Figure 2 A schematic diagram of the architecture of a wafer AOI inspection system provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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.

[0044] See also Figure 1 , this embodiment provides a wafer AOI detection method, comprising the following steps:

[0045] Step S1: acquiring a wafer image;

[0046] Step S2: Divide the wafer image into several sub-regions and calculate the color difference distance sum of the die in each sub-region based on the Euclidean distance algorithm;

[0047] Step S3: normalize the color difference distance of the bare chip in each sub-region to obtain the grayscale of each sub-region;

[0048] Step S4: re-dividing the wafer image based on the grayscale of each sub-region to obtain a plurality of continuous image ranges, wherein the grayscale difference of the sub-regions within each image range meets a preset condition;

[0049] Step S5: In each image range, a die is determined as a detection object and the grayscale difference between the die and the adjacent die is calculated to obtain defect information of the die, and then the defect information of all the die in each image range is obtained to obtain the defect information of each image range.

[0050] In this embodiment, the grayscale of the sub-region can characterize the average color difference between each die in the sub-region. The larger the grayscale, the larger the average color difference between each die in the sub-region, that is, the wafer is processed by multiple processes in the sub-region, and there will be obvious color difference lines in the wafer in the sub-region. At this time, the wafer image can be re-divided by grayscale, and the sub-regions made by only the same process are divided into the same image range to obtain image range defect information, which can effectively remove the influence of color difference caused by the process on the detection, and effectively improve the tolerance for color difference. The sub-regions with color difference lines will be divided into the same image range for centralized detection. By calculating the grayscale difference between adjacent die, the die on the color difference line can be screened out. The defect information of this part of the die can be compared and analyzed with the standard qualified wafers produced on the same production line, and the defect information of the die in the non-color difference line area can be obtained by judging the grayscale difference between adjacent die, thereby completing the defect information of the entire wafer.

[0051] This embodiment cleverly uses grayscale to replace the existing method of directly using grayscale values ​​for defect judgment, which can accurately classify each sub-area on the wafer, effectively remove the impact of color difference caused by the process on detection, improve the tolerance to wafer color difference, and improve the effective detection rate of wafer defects.

[0052] It should be noted that the preset conditions here can be adjusted based on the actual detection accuracy.

[0053] Furthermore, before acquiring the wafer image, the wafer is pretreated: a hardened protective layer is formed on the front side of the wafer; the back side of the wafer is treated with a concentrated hydrofluoric acid solvent to reduce the color difference on the back side of the wafer; and the hardened protective layer on the front side of the wafer is removed to obtain a pretreated wafer.

[0054] In this embodiment, in order to avoid partial oxidation of the back side of the wafer in the previous process, the wafer can be cleaned before detection. Concentrated hydrofluoric acid solvent can effectively remove impurities such as silicon nitride or silicon oxide generated on the back side of the wafer that will cause obvious color difference, eliminate the color difference on the back side of the wafer, and thus ensure the accuracy of detection.

[0055] It should be noted that the hardened protective layer can be composed of ion-doped photoresist, that is, first a layer of photoresist is spin-coated on the front side of the wafer, and then ion doping is injected into the photoresist to harden the photoresist, thereby forming a hardened protective layer to protect the front side of the wafer.

[0056] It should be noted that the concentrated hydrofluoric acid solvent may be a solvent mixed with 49% hydrofluoric acid and 51% water.

[0057] Furthermore, the method of acquiring a wafer image specifically includes: optically scanning the wafer to acquire light intensity grayscale quantities at different positions on the wafer; acquiring bright and dark areas of the wafer based on the light intensity grayscale quantities, and performing light selection and light intensity adjustment to obtain an optimal light flux condition for the wafer; and acquiring a wafer image under the optimal light flux condition.

[0058] Furthermore, the bright and dark areas of the wafer are obtained based on the light intensity grayscale, and the light source is selected and the light intensity is adjusted to obtain the optimal luminous flux condition of the wafer, specifically: the bright and dark areas of the wafer are obtained based on the light intensity grayscale; two light sources located at the center of the bright area and the center of the dark area are selected to illuminate the wafer, and the light intensity grayscale of several bare chips from the center of the wafer to the edge of the wafer is obtained in this case; the light intensity of the two light sources is adjusted based on the light intensity grayscale of the several bare chips to obtain the optimal luminous flux condition of the wafer.

[0059] Since the detection of this method is carried out step by step after obtaining the wafer image, in order to ensure the accuracy of the detection, the accurate acquisition of the image is very important. In the process of image acquisition, aside from the influence of the chromatic aberration of the wafer itself, the light conditions of the shooting environment should also be considered to avoid the appearance of shadows on the wafer and the aggravation of the chromatic aberration of the wafer. In this embodiment, the placed wafer is first optically scanned, and the light intensity grayscale at different positions of the wafer can be obtained, and then the bright and dark areas of the wafer are determined and two light sources at the center of the bright area and the center of the dark area are selected to illuminate the wafer. The light intensity values ​​emitted by the two light sources can be superimposed, and the optimal light flux conditions of the wafer can be obtained by amplification and compensation. Obtaining the wafer image under the optimal light flux conditions can ensure the accuracy of subsequent detection.

[0060] It should be noted that the bright area and the dark area can be defined in the range of 0 to 255 according to the light intensity grayscale, which is actually the area corresponding to the minimum value and the maximum value.

[0061] Furthermore, the method of dividing the wafer image into several sub-regions and calculating the sum of the color difference distances of the bare chips in each sub-region based on the Euclidean distance algorithm is specifically as follows: dividing the wafer image into several sub-regions; obtaining the RGB signal values ​​of the pixels corresponding to all the bare chips in the several sub-regions; for each sub-region, calculating the RGB signal mean of all the bare chips based on the RGB signal value; and calculating the RGB signal mean of all the bare chips based on the Euclidean distance algorithm to obtain the sum of the color difference distances of the bare chips in each sub-region.

[0062] Furthermore, the Euclidean distance algorithm is used to calculate the mean values ​​of the RGB signals of all the dies to obtain the sum of the color difference distances of the dies in each sub-region, specifically:

[0063] For a die in each sub-region, the color difference distance S is specifically expressed as:

[0064]

[0065] In the formula, It represents the red channel average of the die and other dies, which is specifically expressed as: , Indicates the average red signal of the die, Represents the average red signal of other dies; Indicates the difference between the average red signal of the die and the average red signal of other dies; Indicates the difference between the mean green signal of the die and the mean green signal of other dies; Indicates the difference between the mean blue signal value of the die and the mean blue signal value of other dies;

[0066] The color difference distances of all the bare chips in each sub-region are added together to obtain the sum of the color difference distances of the bare chips in each sub-region.

[0067] The color difference distance and obtained here are only the quantities defined in the above scheme, which are used to characterize the color difference between each bare chip in each sub-region. This quantity needs to be compared with other sub-regions to have a more substantial effect. It should be noted that the color difference distance and defined in this embodiment transform and design the traditional grayscale value, which is similar to characterizing the color uniformity inside the sub-region. The wafer preparation process is basically a surface processing process. For example, the processing of the center area and the edge area of ​​the wafer is basically achieved by rotation, which means that the color difference caused by the process on the wafer basically has a color difference line. Thanks to the wafer preparation process, this embodiment can well achieve its technical effect. If the defined color difference distance and is too large, it means that there is a color difference line in the sub-region, so that the sub-regions can be classified to remove the color difference caused by the process.

[0068] Furthermore, the color difference distance of the bare chip in each sub-region is normalized to obtain the grayscale of each sub-region, specifically, for each sub-region, the following steps are specifically performed: take the maximum color difference distance Minimum color difference distance ; Calculate the standardized coefficient, specifically:

[0069]

[0070] in =255, =0;

[0071] Normalize the color difference distance of the die in the sub-region to the interval [0,255] to obtain the grayscale value of the sub-region , the specific calculation formula is:

[0072] or

[0073] In the formula, , is the color difference distance of the bare chip and, is the number of dies.

[0074] Furthermore, the wafer image is re-divided based on the grayscale of each sub-region to obtain several continuous image ranges, and the grayscale difference of the sub-regions within each image range meets the preset conditions, specifically: based on the grayscale of each sub-region, the grayscale difference of adjacent sub-regions is calculated, and two adjacent sub-regions whose grayscale difference is less than the preset condition are divided into the same image range, thereby obtaining several continuous image ranges.

[0075] In this embodiment, by setting preset conditions, the wafer area prepared by the same process can be divided into one image range, which can remove the color difference caused by the process and improve the efficiency of detection.

[0076] Furthermore, in each image range, a bare chip is determined as a detection object and the grayscale difference between the bare chip and an adjacent bare chip is calculated to obtain defect information of the bare chip, and then the defect information of all bare chips in each image range is obtained to obtain the defect information of each image range, specifically: in each image range, a bare chip is determined as a detection object; the grayscale difference between the bare chip and the adjacent bare chip is calculated to obtain at least a first difference and a second difference; the difference between the first difference and the second difference is obtained, and if the difference does not meet a preset threshold, the bare chip is marked to obtain the defect information of the bare chip; a bare chip is re-determined, and the above steps are repeated to obtain the defect information of all bare chips to be marked in each image range, and then the defect information of each image range is obtained.

[0077] In this embodiment, each die has multiple pixels, and each pixel has a grayscale value. In theory, the grayscale values ​​of the same pixel at the same position of adjacent dies are the same. By comparing the grayscale values ​​of multiple adjacent dies, the defect information of each pixel on each die can be obtained.

[0078] See also Figure 2 The present embodiment also provides a wafer AOI inspection system, including: an image acquisition module, used to acquire a wafer image; a region division module, used to divide the wafer image into a plurality of sub-regions; a color difference distance calculation module, used to calculate the color difference distance sum of the bare die in each sub-region based on the Euclidean distance algorithm; a grayscale quantization module, used to standardize the color difference distance sum of the bare die in each sub-region to obtain the grayscale of each sub-region; an image range division module, used to re-divide the wafer image based on the grayscale of each sub-region to obtain a plurality of continuous image ranges, and the grayscale difference of the sub-regions in each image range meets a preset condition; a defect detection module, used to determine a bare die as a detection object in each image range and calculate the grayscale difference between the bare die and the adjacent bare die to obtain the defect information of the bare die, and then obtain the defect information of all the bare die in each image range to obtain the defect information of each image range.

[0079] The system provided in this embodiment is simple to construct and can well implement the above-mentioned wafer AOI inspection method. It cleverly uses grayscale measurement instead of the existing method of directly using grayscale values ​​for defect judgment. It can accurately classify each sub-area on the wafer, effectively remove the influence of color difference caused by the process on detection, improve the tolerance to wafer color difference, and improve the effective detection rate of wafer defects.

[0080] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A wafer AOI inspection method, characterized in that: The following steps are involved: Acquire wafer images; Divide the wafer image into several sub-regions and calculate the color difference distance and the color difference distance of the die in each sub-region based on the Euclidean distance algorithm; The color difference distance of the bare chip in each sub-region is normalized to obtain the grayscale of each sub-region; Re-dividing the wafer image based on the grayscale of each sub-region to obtain a number of continuous image ranges, wherein the grayscale difference of the sub-regions within each image range meets a preset condition; wherein, judging whether there is a color difference line in each sub-region based on the grayscale of each sub-region and dividing the sub-regions with the color difference line into the same image range, comparing and analyzing the bare die within each image range with the standard wafers produced on the same production line to obtain the defect information of all bare die to be marked in each image range, and then obtaining the defect information of each image range; The sub-areas where no color difference lines exist are divided into the same image range. In each image range, a bare chip is determined as a detection object and the grayscale difference between the bare chip and adjacent bare chips is calculated to obtain defect information of the bare chip, and then the defect information of all bare chips in each image range is obtained to obtain the defect information of each image range; specifically: in each image range, a bare chip is determined as a detection object; the grayscale difference between the bare chip and the adjacent bare chips is calculated to obtain at least a first difference and a second difference; the difference between the first difference and the second difference is obtained, and if the difference does not meet a preset threshold, the bare chip is marked to obtain the defect information of the bare chip; a bare chip is re-determined, and the above steps are repeated to obtain the defect information of all bare chips to be marked in each image range, and then the defect information of each image range is obtained.

2. The wafer AOI detection method according to claim 1, characterized in that: Before acquiring the wafer image, the wafer is preprocessed: forming a hardened protective layer on the front side of the wafer; The back of the wafer is treated with concentrated hydrofluoric acid to reduce the color difference on the back of the wafer; The hardened protective layer on the front side of the wafer is removed to obtain a pre-processed wafer.

3. The wafer AOI detection method according to claim 1, characterized in that: The obtaining of the wafer image specifically includes: Optically scan the wafer to obtain the light intensity grayscale at different locations on the wafer; Based on the light intensity grayscale, the bright and dark areas of the wafer are obtained, and the light is selected and the light intensity is adjusted to obtain the optimal light flux condition for the wafer; The wafer image is acquired under the optimal light flux condition.

4. The wafer AOI detection method according to claim 3, characterized in that: The light intensity grayscale is used to obtain the bright and dark areas of the wafer, and the light source is selected and the light intensity is adjusted to obtain the optimal light flux condition of the wafer, specifically: Obtain the bright and dark areas of the wafer based on the light intensity grayscale; Select two light sources located at the center of the bright area and the center of the dark area to illuminate the wafer, and obtain the light intensity grayscale of several bare chips from the center of the wafer to the edge of the wafer in this case; The light intensity of the two light sources is adjusted based on the light intensity grayscale of a plurality of bare chips to obtain the optimum light flux condition for the wafer.

5. A wafer AOI inspection method according to any one of claims 1 to 4, characterized in that: The wafer image is divided into several sub-areas and the color difference distance and the distance of the bare die in each sub-area are calculated based on the Euclidean distance algorithm, specifically: Dividing the wafer image into a number of sub-regions; Obtaining RGB signal values ​​of pixels corresponding to all dies in several sub-areas; For each sub-region, the RGB signal mean of all dies is calculated based on the RGB signal value; The mean values ​​of the RGB signals of all the dies are calculated based on the Euclidean distance algorithm to obtain the sum of the color difference distances of the dies in each sub-region.

6. The wafer AOI inspection method according to claim 5, characterized in that: The Euclidean distance algorithm is used to calculate the mean values ​​of the RGB signals of all the bare chips to obtain the sum of the color difference distances of the bare chips in each sub-region, specifically: For a die in each sub-region, the color difference distance S is specifically expressed as: In the formula, It represents the red channel average of the die and other dies, which is specifically expressed as: , Indicates the average red signal of the die, Represents the average red signal of other dies; Indicates the difference between the average red signal of the die and the average red signal of other dies; Indicates the difference between the mean green signal of the die and the mean green signal of other dies; Indicates the difference between the mean blue signal value of the die and the mean blue signal value of other dies; The color difference distances of all the bare chips in each sub-region are added together to obtain the sum of the color difference distances of the bare chips in each sub-region.

7. The wafer AOI inspection method according to claim 6, characterized in that: The color difference distance and the grayscale of each sub-region are normalized, and the grayscale of each sub-region is obtained, specifically: For each sub-region, perform the following steps: Take the maximum color difference distance Minimum color difference distance ; Calculate the standardized coefficients, specifically: in =255, =0; Normalize the color difference distance of the die in the sub-region to the interval [0,255] to obtain the grayscale value of the sub-region , the specific calculation formula is: or In the formula, , is the color difference distance of the bare chip and, is the number of dies.

8. The wafer AOI detection method according to claim 7, characterized in that: The wafer image is re-divided based on the grayscale of each sub-region to obtain a plurality of continuous image ranges, and the grayscale difference of the sub-regions within each image range meets the preset conditions, specifically: Based on the grayscale of each sub-region, the grayscale difference of adjacent sub-regions is calculated, and two adjacent sub-regions whose grayscale difference is less than a preset condition are divided into the same image range, thereby obtaining several continuous image ranges.

9. A wafer AOI inspection system, comprising: An image acquisition module, used for acquiring wafer images; A region division module, used for dividing the wafer image into a plurality of sub-regions; A color difference distance calculation module, used for calculating the color difference distance sum of the die in each sub-area based on the Euclidean distance algorithm; A grayscale quantization module is used to standardize the color difference distance of the bare die in each sub-region to obtain the grayscale of each sub-region; An image range division module is used to re-divide the wafer image based on the grayscale of each sub-region, and obtain a plurality of continuous image ranges, wherein the grayscale difference of the sub-regions within each image range meets a preset condition; wherein, according to the grayscale of each sub-region, it is determined whether there is a color difference line in each sub-region and the sub-regions with the color difference line are divided into the same image range, and the bare die within each image range is compared and analyzed with the standard wafers produced by the same production line to obtain the defect information of all bare die to be marked in each image range, and then the defect information of each image range is obtained; The defect detection module is used to divide the sub-areas where there are no color difference lines into the same image range, and in each image range, determine a bare chip as a detection object and calculate the grayscale difference between the bare chip and adjacent bare chips to obtain defect information of the bare chip, and then obtain defect information of all bare chips in each image range to obtain defect information of each image range; specifically: in each image range, determine a bare chip as a detection object; calculate the grayscale difference between the bare chip and the adjacent bare chips to obtain at least a first difference and a second difference; obtain the difference between the first difference and the second difference, if the difference does not meet the preset threshold, mark the bare chip to obtain defect information of the bare chip; re-determine a bare chip, and repeat the above steps to obtain defect information of all bare chips to be marked in each image range, and then obtain defect information of each image range.