Method of identifying a wafer

Abstract

The application provides a method for identifying a wafer, comprising the following steps. A mark frame is obtained from a wafer detection picture. A gray scale index corresponding to the mark frame is calculated based on respective gray scale values of a plurality of pixels included in the mark frame. The gray scale index represents a proportion of pixels with a gray scale value greater than a specified value. Whether a trace pattern in the mark frame is a scratch or a grain boundary is determined based on the gray scale index.

Description

Technical Field

[0001] This invention relates to a wafer analysis method, and more particularly to a method for identifying wafer scratches and grain boundaries. Background Technology

[0002] Wafer scratches and grain boundaries are common anomalies in semiconductor fabrication plants. Severe scratches can render an entire batch unusable. Current inspection equipment cannot accurately distinguish between grain boundaries and scratches, classifying all as scratches. This leads to unnecessary rework time and costs during subsequent rough or fine polishing due to grain boundaries. Summary of the Invention

[0003] This invention provides a method for identifying wafers, which can accurately distinguish between grain boundaries and scratches.

[0004] The method for identifying a wafer according to the present invention includes: obtaining a marker frame from a wafer detection image; calculating a grayscale index corresponding to the marker frame based on the grayscale values ​​corresponding to each of the multiple pixels included in the marker frame, wherein the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value; and determining, based on the grayscale index, whether the trace pattern in the marker frame is a scratch or a grain boundary.

[0005] According to an embodiment of the present invention, the marker box has a corresponding confidence index and a category label. After obtaining the marker box from the wafer inspection image, the method further includes: if the confidence index is greater than or equal to the confidence threshold, the category label is determined to be correct; and if the confidence index is less than the confidence threshold, the grayscale index corresponding to the marker box is calculated and a threshold is set to determine whether the trace pattern is a scratch or a grain boundary.

[0006] According to an embodiment of the present invention, the step of determining whether a trace pattern is a scratch or a grain boundary includes: determining whether a grayscale index is greater than or equal to a threshold; if the grayscale index is greater than or equal to the threshold, determining that the trace pattern is a scratch; if the grayscale index is less than the threshold, identifying the trace pattern in a marker box and calculating the slope difference index of the trace pattern; and determining whether the trace pattern is a scratch or a grain boundary based on the grayscale index and the slope difference index.

[0007] According to an embodiment of the present invention, the step of calculating the slope difference index of a trace pattern includes: extracting a first endpoint, a second endpoint, and a plurality of target points from the trace pattern; calculating the standard slope of the first endpoint and the second endpoint; calculating a plurality of comparison slopes based on any two points among the target points; and obtaining the slope difference index based on the standard slope and the comparison slope.

[0008] According to an embodiment of the present invention, in the step of obtaining the slope difference index based on the standard slope and the comparison slope, the slope difference index is calculated using the following formula:

[0009]

[0010] Where n is the number of alignment slopes, Si is the i-th alignment slope (i = 1 to n), Sct is the slope difference index, and Sstd is the standard slope.

[0011] According to an embodiment of the present invention, the step of determining whether a trace pattern is a scratch or a grain boundary based on a grayscale index and a slope difference index includes: obtaining a comprehensive index based on the grayscale index and the slope difference index; if the comprehensive index is greater than or equal to a preset value, determining that the trace pattern in the marking box is a scratch; and if the comprehensive index is less than the preset value, determining that the trace pattern in the marking box is a grain boundary.

[0012] According to an embodiment of the present invention, the comprehensive index is obtained by the following formula: C = W1 × G + W2 × (1 / Sct), where C is the comprehensive index, G is the grayscale index, Sct is the slope difference index, and W1 and W2 are weight values.

[0013] The method for identifying a wafer according to the present invention includes: obtaining a marker frame from a wafer detection image; calculating a grayscale index corresponding to the marker frame based on the grayscale values ​​corresponding to the multiple pixels included in the marker frame, wherein the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value; identifying a trace pattern in the marker frame and calculating a slope difference index of the trace pattern; and determining whether the trace pattern is a scratch or a grain boundary based on the grayscale index and the slope difference index.

[0014] Based on the above, the present invention utilizes the different characteristics of grain boundaries and scratches to establish an identification module, which can accurately distinguish between grain boundaries and scratches, thereby reducing the problem of grain boundaries being misjudged as scratches and causing additional rework. Attached Figure Description

[0015] Figure 1 This is a block diagram of a wafer analysis system according to an embodiment of the present invention;

[0016] Figure 2 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention;

[0017] Figure 3 This is a schematic diagram of a wafer inspection image according to an embodiment of the present invention;

[0018] Figure 4 This is a schematic diagram of multiple candidate boxes according to an embodiment of the present invention;

[0019] Figure 5This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention;

[0020] Figure 6A and Figure 6B This is a grayscale distribution curve diagram according to an embodiment of the present invention;

[0021] Figure 7 This is a schematic diagram of a trace pattern according to an embodiment of the present invention;

[0022] Figure 8 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention;

[0023] Figure 9 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention;

[0024] Figure 10 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention.

[0025] Explanation of reference numerals in the attached figures

[0026] 100: Chip Analysis System

[0027] 110: Measuring Instruments

[0028] 120: Analytical device

[0029] 121: Processor

[0030] 122: Memory

[0031] 300, 400: Chip inspection images

[0032] 310, 700: Marker boxes

[0033] 710: Trace Pattern

[0034] c001~c006: Candidate boxes

[0035] C: Confidence Index

[0036] M: Category tag

[0037] P1~P5: Target Points

[0038] PS: First endpoint

[0039] PE: Second endpoint

[0040] S205~S215: Steps for identifying wafer scratches and grain boundaries

[0041] S505~S535: Steps for identifying wafer scratches and grain boundaries

[0042] S805~S840: Steps for identifying wafer scratches and grain boundaries

[0043] S905~S950: Steps for identifying wafer scratches and grain boundaries

[0044] S1005~S1020: Steps for identifying wafer scratches and grain boundaries Detailed Implementation

[0045] Figure 1 This is a block diagram of a wafer analysis system according to an embodiment of the present invention. The wafer analysis system 100 includes a measuring instrument 110 and an analysis device 120. The measuring instrument 110 can transmit data to the analysis device 120 via wired or wireless transmission.

[0046] Measuring instrument 110 is, for example, an Automated Optical Inspection (AOI) instrument. AOI instruments are high-speed, high-precision optical image inspection systems that include measuring lens technology, optical illumination technology, positioning measurement technology, electronic circuit testing technology, image processing technology, and automation technology applications. They utilize machine vision as the standard inspection technology.

[0047] The analysis device 120 is an electronic device with computing capabilities, which can be implemented using a personal computer, laptop computer, tablet computer, smartphone, or any device with computing capabilities; the present invention is not limited thereto. The analysis device 120 includes a processor 121 and a memory 122.

[0048] The processor 121 may be, for example, a central processing unit (CPU), a physical processing unit (PPU), a programmable microprocessor, an embedded control chip, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), or other similar devices.

[0049] The memory 122 is, for example, any type of fixed or removable random access memory (RAM), read-only memory (ROM), flash memory, hard disk, or other similar device or combination thereof. The memory 122 comprises one or more code segments that, after being installed, will be executed by the processor 121.

[0050] Figure 2 This is a flowchart of a method for identifying chip scratches according to an embodiment of the present invention. Please refer to... Figure 1 and Figure 2 In step S205, bounding boxes are obtained from the wafer inspection image. In this embodiment, the processor 121 selects bounding boxes in the wafer inspection image to obtain one or more candidate bounding boxes. For example, the memory 122 includes a Labellimg tool, which detects candidate bounding boxes with corresponding confidence indices and category labels. The confidence indices are, for example, IOU (Intersection over Union) values, and the category labels are grain boundaries or scratches.

[0051] Figure 3 This is a schematic diagram illustrating the model judgment of a wafer inspection image according to an embodiment of the present invention. Please refer to... Figure 3 The model judgment obtained after the wafer inspection image 300 is judged by the detection model includes bounding boxes 310. Each bounding box 310 has a corresponding confidence index C and a category label M. During model training, the IOU value represents, for example, the difference between the real data and the bounding box judgment results. During model testing, the IOU value represents the intersection / union of the bounding box features and the features trained by the model. For example, the model judgment disclosed herein can set an IOU threshold of 0.7, where candidate boxes with an IOU value < 0.7 are directly ignored and not processed, and only candidate boxes with an IOU value ≥ 0.7 are processed subsequently. Here, the IOU threshold of 0.7 is only for illustrative purposes; in other embodiments, the IOU value of the model judgment as the threshold, and the corresponding processing and judgment mechanisms, can be adjusted according to requirements, and are not limited thereto.

[0052] If the detection model of the wafer inspection image 300 determines that the number of detected candidate boxes is 1, the detected candidate box is directly used as the marker box. If the number of detected candidate boxes is greater than 1, at least one candidate box is selected as the marker box based on the confidence index. For example, if the number of candidate boxes is greater than 1, the overlap rate between every two candidate boxes is determined, and at least one candidate box is selected as the marker box based on the confidence index and the overlap rate.

[0053] For example, multiple candidate boxes are sorted from highest to lowest confidence index. The candidate box with the highest confidence index is then selected as the first labeled box. Next, candidate boxes with an overlap rate greater than a preset value (e.g., 60%) with the first labeled box are deleted, retaining those with an overlap rate less than 60%. Then, from the retained candidate boxes, the candidate box with the highest confidence index is selected as the second labeled box. Then, candidate boxes with an overlap rate greater than 60% with the second labeled box are deleted, retaining those with an overlap rate less than 60%. Then, from the retained candidate boxes, the candidate box with the highest confidence index is selected as the third labeled box. This process is repeated until the overlap rate between the remaining candidate boxes and the selected labeled box is no greater than 0.6.

[0054] Figure 4 This is a schematic diagram of multiple candidate boxes according to an embodiment of the present invention. Figure 4 The wafer inspection image 400 shown detects six candidate boxes, namely candidate boxes c001 to c006, with confidence indices of 0.95, 0.9, 0.9, 0.8, 0.7, and 0.7, respectively. First, candidate box c001, corresponding to a confidence index of 0.95, is taken as the first labeled box. Here, it is assumed that the overlap rate between candidate boxes c002 and c005 and the first labeled box (candidate box c001) is greater than 60%. Therefore, among the remaining five candidate boxes, those with an overlap rate of less than 60% with the first labeled box (candidate box c001), namely candidate boxes c003, c004, and c006, are retained.

[0055] Next, among candidate boxes c003, c004, and c006, candidate box c003, corresponding to a confidence index of 0.9, is selected as the second labeled box. Here, the overlap rate between candidate boxes 004 and 006 and the second labeled box (candidate box 003) is greater than 60%. Therefore, candidate boxes 004 and 006 are removed, and no further labeled boxes are selected.

[0056] After determining the marker box, in step S210, the analysis device 120 calculates the grayscale index corresponding to the marker box based on the grayscale values ​​of the multiple pixels included in the marker box. The grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value. For example, if the specified value is set to 212.5, the proportion of pixels with grayscale values ​​greater than 212.5 in the marker box is calculated as the grayscale index.

[0057] Then, in step S215, based on the grayscale index, it is determined whether the trace pattern in the marked box is a scratch or a grain boundary. For example, a threshold is set to determine whether the trace pattern is a scratch or a grain boundary.

[0058] Let me give another example to illustrate this. Figure 5This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention. In this embodiment, a confidence threshold can be set for the first stage of judgment, and a threshold can be set for the second stage of judgment. First, in step S505, a marker box is obtained from the wafer inspection image. Next, in step S510, it is determined whether the confidence index of the marker box is greater than or equal to the confidence threshold. If the confidence index is greater than or equal to the confidence threshold, as shown in step S515, the trace pattern in the marker box is determined to be a scratch or a grain boundary based on the category label.

[0059] For example, with Figure 3 Taking the wafer inspection image 300 as an example, its category label M is divided into two categories: scratches and grain boundaries. Assuming that the category label M of the marker box 310 in the wafer inspection image 300 is a scratch, since the confidence index C (99%) is greater than or equal to the confidence threshold (e.g., 70%), it indicates that the model's judgment is correct, and the trace pattern in the marker box 310 is determined to be a scratch based on the category label. This process continues, using the confidence threshold as the first-stage judgment.

[0060] If the confidence index C is less than the confidence threshold, the category label M in the label box will not be used for judgment. Instead, the subsequent steps S520 to S535 / S530 will be executed.

[0061] That is, if the confidence index C is less than the confidence threshold, the grayscale index is calculated in step S520. Specifically, the analysis device 120 calculates the grayscale index corresponding to the marker box based on the grayscale values ​​of the multiple pixels included in the marker box, where the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value.

[0062] In detail, Figure 6A and Figure 6B This is a grayscale distribution curve diagram according to an embodiment of the present invention. Figure 6A The proportion of grayscale values ​​distributed for scratches. Figure 6B This represents the grayscale distribution ratio of grain boundaries. Since scratches are generally located on the surface of the wafer, they appear as a higher grayscale value (white), while grain boundaries are not on the surface and therefore have a lower grayscale value. Figure 6A and Figure 6B It can be seen that, for grayscale indicators corresponding to scratches, the proportion of pixels with higher grayscale values, such as those with grayscale values ​​greater than 212.5, is higher than that corresponding to grain boundaries. Therefore, the above... Figure 5 In the illustrated embodiment, a specified value can be pre-set to 212.5 to calculate the grayscale index.

[0063] return Figure 5A threshold for a grayscale index is set. After obtaining the grayscale index, in step S525, it is determined whether the grayscale index is greater than or equal to the set threshold. When the grayscale index is greater than or equal to the set threshold, in step S530, the mark pattern in the marking box is determined to be a scratch. When the grayscale index is less than the set threshold, in step S535, the mark pattern in the marking box is determined to be a grain boundary.

[0064] In another embodiment, after using the confidence index for the first-stage judgment, a comprehensive index can be calculated for the second-stage judgment. For details, please refer to the following description. Figure 8 The process.

[0065] Alternatively, in another embodiment, a three-stage judgment can be performed: first, a confidence index is used for the first stage of judgment; then, a grayscale index is used for the second stage of judgment; and finally, a comprehensive index is calculated for the third stage of judgment. For detailed explanation, please refer to the following description. Figure 9 The process.

[0066] The comprehensive index is calculated based on the grayscale index and the slope difference index. The specific calculation method of the slope difference index is as follows. After the analysis device 120 identifies the trace pattern in the marking box, it extracts the first endpoint, the second endpoint, and multiple target points from the trace pattern. Then, it calculates the standard slope of the first endpoint and the second endpoint. Then, it calculates multiple comparison slopes based on any two points among the target points. After that, it obtains the slope difference index based on the standard slope and the comparison slope. For example, the slope difference index is calculated using the following formula (1). In formula (1), n ​​is the number of comparison slopes, Si is the i-th comparison slope, i = 1 to n, Sct is the slope difference index, and Sstd is the standard slope.

[0067]

[0068] The following example illustrates how to calculate the slope difference index.

[0069] Figure 7 This is a schematic diagram of a trace pattern 710 according to an embodiment of the present invention. Please refer to... Figure 7 The marker frame 700 includes a trace pattern 710. The analysis device 120 locates the trace pattern 710 based on grayscale values ​​and obtains the coordinates (X_start, Y_start) of the first endpoint PS of the trace pattern 710 located on the boundary of the marker frame 700, and the coordinates (X_end, Y_end) of the second endpoint PE. Furthermore, it extracts the coordinates (X1, Y1) to (X5, Y5) of multiple target points P1 to P5 from the trace pattern 710.

[0070] Next, the slope between the first endpoint PS and the second endpoint PE is calculated as the standard slope Sstd = (Y_start - Y_end) / (X_start - X_end). Furthermore, the slope between any two target points is calculated as the comparison slope. Here, the slopes S1 to S4 between target point P1 and target points P2 to P5 are calculated; the slopes S5 to S7 between target point P2 and target points P3 to P5 are calculated; the slopes S8 to S9 between target point P3 and target points P4 to P5 are calculated; and the slope S10 between target point P4 and target point P5 is calculated. These slopes S1 to S10 are used as the comparison slope.

[0071] Then, the slope difference index SCT is calculated based on the above formula (1).

[0072] Sct=[(S1-Sstd) / Sstd+(S2-Sstd) / Sstd+…

[0073] +(S9-Sstd) / Sstd+(S10-Sstd) / Sstd] / n.

[0074] The comprehensive index can be obtained through the following formula (2). Where C is the comprehensive index, G is the grayscale index, Sct is the slope difference index, and W1 and W2 are the weight values.

[0075] C=W1×G+W2×(1 / Sct) (2)

[0076] Figure 8 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention. In this embodiment, a confidence threshold is first set for the first stage of judgment, and a preset value is set for comparison with a comprehensive index in the second stage. Steps S805 to S815 of this embodiment are... Figure 5 Steps S505 to S515 are similar and will not be repeated here.

[0077] If the confidence index is less than the confidence threshold, in step S820, the analysis device 120 calculates the grayscale index and the slope difference index. The calculation of the grayscale index can be referred to... Figure 5 Step S520. The slope difference index can be calculated using the formula (1) above. Next, in step S825, the analysis device 120 calculates the comprehensive index based on the grayscale index and the slope difference index. The comprehensive index can be calculated using the formula (2) above.

[0078] After obtaining the comprehensive index, in step S830, the analysis device 120 determines whether the comprehensive index is greater than or equal to a preset value. If the comprehensive index is greater than or equal to the preset value, in step S840, the trace pattern in the marking box is determined to be a scratch. If the comprehensive index is less than the preset value, in step S835, the trace pattern in the marking box is determined to be a grain boundary.

[0079] Figure 9 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention. In this embodiment, a three-stage judgment is performed: first, a confidence threshold is set for the first stage of judgment; second, a threshold is set for comparison with grayscale indicators; and third, a preset value is set for comparison with comprehensive indicators. Steps S905 to S915 of this embodiment are... Figure 5 Steps S505 to S515 are similar, and steps S920 to S925 are the same. Figure 5 Steps S520 to S525 are similar and will not be repeated here.

[0080] exist Figure 9 In this embodiment, when the grayscale index is greater than or equal to the set threshold, it is determined to be a scratch in step S950. When the grayscale index is less than the set threshold, a third stage of judgment is performed. That is, in step S930, the slope difference index is calculated. The analysis device 120 identifies the trace pattern in the marking box and calculates the slope difference index of the trace pattern using formula (1). Next, in step S935, the analysis device 120 calculates the comprehensive index. For example, the comprehensive index is calculated using formula (2).

[0081] After obtaining the comprehensive index, in step S940, the analysis device 120 determines whether the comprehensive index is greater than or equal to a preset value. If the comprehensive index is greater than or equal to the preset value, in step S950, the analysis device 120 determines that the trace pattern in the marking box is a scratch. If the comprehensive index is less than the preset value, in step S945, the analysis device 120 determines that the trace pattern in the marking box is a grain boundary.

[0082] In detail, the method for setting the aforementioned confidence threshold, preset value, specified value, and threshold can be as follows: Take an array of detection images and their corresponding real data, input these detection images into the analysis device 120 for model training, then compare the training results of the model with the real data to calculate the accuracy rate, and select an appropriate and satisfactory accuracy rate and its corresponding value as the confidence threshold or preset value. For example, the following explanation uses the threshold setting for grayscale indicators. For instance, when the set threshold is 15%, the accuracy rate is 90%. When the set threshold is less than 15% or greater than 15%, the accuracy rate is 60%. A 60% accuracy rate does not meet the requirements of the judgment model, but a 90% accuracy rate does. Therefore, the threshold for this judgment model is set to 15%. This method can be applied to the aforementioned confidence threshold, preset value, specified value, threshold, weight value, or any value that needs to be preset. However, this is only an example; it can be set according to different designs and requirements, and is not limited to this method.

[0083] Figure 10 This is a flowchart of a method for identifying wafer scratches and grain boundaries according to an embodiment of the present invention. This embodiment uses grayscale and slope difference indices for judgment. Please refer to... Figure 10 In step S1005, a bounding box is obtained from the wafer detection image. In step S1010, based on the grayscale values ​​of the pixels included in the bounding box, a grayscale index corresponding to the bounding box is calculated, where the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value. The calculation of the grayscale index can be referred to... Figure 5 Step S520. Furthermore, in step S1015, the trace pattern is identified in the marking box, and the slope difference index of the trace pattern is calculated. The slope difference index can be calculated using the formula (1) above.

[0084] Subsequently, in step S1020, based on the grayscale index and slope difference index, the trace pattern is determined to be either a scratch or a grain boundary. For example, a comprehensive index can be obtained based on formula (2), and a preset value can be set. If the comprehensive index is greater than or equal to the preset value, the trace pattern in the marked box is determined to be a scratch. If the comprehensive index is less than the preset value, the trace pattern in the marked box is determined to be a grain boundary. It is worth noting that the order of steps S1010 and S1015 can be adjusted as needed, or they can be completed in the same step. This invention is not limited to this.

[0085] In summary, this invention utilizes the different characteristics of grain boundaries and scratches to establish an identification module, which can accurately distinguish between grain boundaries and scratches, thereby reducing the problem of grain boundaries being misjudged as scratches and causing additional rework.

Claims

1. A method for identifying a chip, characterized in that, include: A marker box is obtained from a wafer inspection image, wherein the marker box contains multiple pixels and a trace pattern; Based on the grayscale values ​​of the multiple pixels included in the marker box, a grayscale index corresponding to the marker box is calculated, wherein the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value. Determine whether the grayscale index is greater than or equal to the threshold; If the grayscale index is greater than or equal to the threshold, the mark pattern in the mark box is determined to be a scratch; as well as If the grayscale index is less than the threshold, determine whether the trace pattern in the marker box is a grain boundary.

2. The method for identifying a chip according to claim 1, characterized in that, The marker frame has a corresponding confidence index and a category marker, wherein the category marker includes scratches and grain boundaries. After obtaining the marker frame from the wafer inspection image, it also includes: If the confidence index is greater than or equal to the confidence threshold, the category label is determined to be correct; and If the confidence index is less than the confidence threshold, then after calculating the grayscale index, it is determined whether the grayscale index is greater than or equal to the threshold.

3. The method for identifying a chip according to claim 1, characterized in that, If the grayscale index is less than the threshold, the step of determining whether the trace pattern is a grain boundary includes: Identify the trace pattern within the marked frame and calculate the slope difference index of the trace pattern; and Based on the grayscale index and the slope difference index, it is determined whether the trace pattern is a grain boundary. The step of calculating the slope difference index of the trace pattern includes: First endpoints and second endpoints are taken from both ends of the trace pattern, and multiple target points are taken between the first endpoints and the second endpoints of the trace pattern. Calculate the standard slopes of the first endpoint and the second endpoint; Calculate multiple alignment slopes based on any two points among the target points; and The slope difference index is obtained based on the standard slope and the comparison slope.

4. The method for identifying a chip according to claim 3, characterized in that, The steps for obtaining the slope difference index based on the standard slope and the comparison slope include: The slope difference index is calculated using the following formula: Where n is the number of alignment slopes, Si is the i-th alignment slope, i = 1 to n, Sct is the slope difference index, and Sstd is the standard slope.

5. The method for identifying a chip according to claim 3, characterized in that, The step of determining whether the trace pattern is a grain boundary based on the grayscale index and the slope difference index includes: A comprehensive index is obtained based on the grayscale index and the slope difference index. If the comprehensive index is greater than or equal to a preset value, the mark pattern in the marking box is determined to be a scratch; and If the comprehensive index is less than the preset value, the trace pattern in the marker box is determined to be a grain boundary.

6. The method for identifying a chip according to claim 5, characterized in that, The comprehensive index is obtained through the following formula: C = W1 × G + W2 × (1 / Sct), Wherein, C is the comprehensive index, G is the grayscale index, Sct is the slope difference index, and W1 and W2 are weight values.

7. A method for identifying a chip, characterized in that, include: Obtaining bounding boxes from wafer inspection images; Based on the grayscale values ​​of the multiple pixels included in the marker box, a grayscale index corresponding to the marker box is calculated, wherein the grayscale index represents the proportion of pixels with grayscale values ​​greater than a specified value. Identify the trace pattern within the marked frame and calculate the slope difference index of the trace pattern; and Based on the grayscale index and the slope difference index, the trace pattern is determined to be either a scratch or a grain boundary. The step of calculating the slope difference index of the trace pattern includes: First endpoints and second endpoints are taken from both ends of the trace pattern, and multiple target points are taken between the first endpoints and the second endpoints of the trace pattern. Calculate the standard slopes of the first endpoint and the second endpoint; Calculate multiple alignment slopes based on any two points among the target points; and The slope difference index is obtained based on the standard slope and the comparison slope.

8. The method for identifying a chip according to claim 7, characterized in that, The marker frame has a corresponding confidence index and a category marker, wherein the category marker includes scratches and grain boundaries. After obtaining the marker frame from the wafer inspection image, it also includes: If the confidence index is greater than or equal to the confidence threshold, the category label is determined to be correct; and If the confidence index is less than the confidence threshold, the grayscale index corresponding to the marked box is calculated and a threshold is set to determine whether the trace pattern is a scratch or a grain boundary.

9. The method for identifying a chip according to claim 7, characterized in that, The steps for obtaining the slope difference index based on the standard slope and the comparison slope include: The slope difference index is calculated using the following formula: Where n is the number of alignment slopes, Si is the i-th alignment slope, i = 1 to n, Sct is the slope difference index, and Sstd is the standard slope.

10. The method for identifying a chip according to claim 7, characterized in that, The step of determining whether the trace pattern is a scratch or a grain boundary based on the grayscale index and the slope difference index includes: A comprehensive index is obtained based on the grayscale index and the slope difference index. If the comprehensive index is greater than or equal to a preset value, the mark pattern in the marking box is determined to be a scratch; and If the comprehensive index is less than the preset value, the trace pattern in the marker box is determined to be a grain boundary.

11. The method for identifying a chip according to claim 10, characterized in that, The comprehensive index is obtained through the following formula: C = W1 xG + W2 × (1 / Sct), Wherein, C is the comprehensive index, G is the grayscale index, Sct is the slope difference index, and W1 and W2 are weight values.

Images