A method for positioning a defective chip
By printing cross marks on the chip and establishing a standard image template, and combining the feature matching of the identification code and the cross marks, the problem of low detection efficiency in the existing technology is solved, and the rapid and accurate location of defective chips is achieved, thus improving the detection efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- NANJING OPTICS ROBOT TECH CO LTD
- Filing Date
- 2026-06-09
- Publication Date
- 2026-07-10
Smart Images
Figure CN122373773A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of positioning technology, and in particular to a method for locating defective chips. Background Technology
[0002] Chip manufacturing involves several processes and is highly sensitive to materials, environment, and process parameters. Defects can occur at any stage, therefore, chip testing is necessary after some stages. Chip testing is mainly divided into physical testing and electrical performance testing. Physical testing can be further divided into surface defect detection before packaging and package defect detection after packaging.
[0003] Before being packaged, a chip is a die on a wafer. Due to differences in chip type, function and manufacturing process, the size of the die varies. However, there are at least hundreds or thousands, or even tens of thousands, of dies on a single wafer. Therefore, when a defect is detected in a chip, it is necessary to locate the chip at its specific position on the wafer in order to process the defective chip in the future.
[0004] Existing positioning methods use OCR technology to identify the identification code on the chip, thereby ensuring that each chip can be correctly traced. To ensure detection accuracy, high-precision local inspection usually uses a high-resolution camera with a small field of view. However, for some larger chips, the identification code on the chip may not be displayed or fully displayed in the small field of view. It is necessary to repeatedly move the camera or wafer stage to make the identification code fully enter the camera's field of view in order to locate the chip, which seriously affects the detection efficiency. Summary of the Invention
[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for locating defective chips, so as to solve the technical problem that the prior art requires repeated movement of the camera or wafer stage to ensure that the identification code can be fully entered into the camera's field of view, which seriously affects the detection efficiency.
[0006] As a first aspect of the present invention, a method for locating a defective chip is provided. The chip has an identification code printed on it, and a cross mark is printed at each of the four corners of the chip. A standard image template for each cross mark is established. Before locating the detected defective chip, a camera is controlled to take pictures of the current defective chip at multiple preset teaching positions. Each teaching position corresponds to a teaching field of view. At least one teaching field of view contains the identification code of the current defective chip, and simultaneously, the teaching field of view contains the cross mark of the current defective chip. The method for locating the defective chip includes: Step S1: After the wafer loading is completed, the current defective chip is positioned. The camera is controlled to take pictures of the current defective chip sequentially from the first teaching position point. During the process of taking pictures of the current defective chip, it is determined whether there is an identification code image of the current defective chip within the current field of view of the camera. Step S2: When the identification code image of the current defective chip exists in the current field of view of the camera, the position information of the current defective chip on the wafer is determined directly based on the identification code image of the current defective chip, and the process ends; when the identification code image of the current defective chip does not exist in any of the current field of view of the camera, the current field of view containing the complete cross mark of the current defective chip is found from all the current field of view of the camera. Step S3: Based on the standard image template of the complete cross mark of the current defective chip, determine the teaching field of view to which the complete cross mark belongs, and then calculate the distance between the current position of the camera and the position of the current defective chip identification code; Step S4: Move the wafer or camera according to the distance so that the camera can capture the identification code above the identification code of the current defective chip, and determine the position information of the current defective chip on the wafer based on the captured identification code image. The process ends.
[0007] Furthermore, the step of identifying the current field of view containing the complete crosshair of the current defective chip from all current field of view of the camera also includes: When a complete crosshair mark of the current defective chip is present in the current field of view of the camera, the placement angle of the wafer is corrected according to the deflection angle of the complete crosshair mark in the current field of view. After the wafer's placement angle is corrected, the camera is controlled again to take pictures of the current defective chip sequentially based on multiple teaching position points until a complete cross mark of the current defective chip is present in the current field of view of the camera.
[0008] Furthermore, the calculation of the distance between the current camera position and the current defective chip identifier position also includes: The formula for calculating this distance is as follows: Distance = (pixel coordinates of the complete crosshair within the current field of view - pixel coordinates of the complete crosshair within the teaching field of view) * camera pixel ratio + (coordinates of the first teaching position point - coordinates of the teaching position point corresponding to the teaching field of view to which the complete crosshair belongs).
[0009] Furthermore, the establishment of a standard image template for each cross mark includes: Select a defect-free qualified chip on the wafer, control the camera to capture images of the complete cross mark on the four corners of the qualified chip, remove image noise, and extract one or more features from the edge contour, the ratio of horizontal and vertical line lengths and the sub-pixel grayscale texture of the complete cross mark image. Combine the differential features of the complete cross mark images at each corner to generate four independent and distinguishable standard image templates.
[0010] Further, identifying the current field of view containing the complete crosshair of the current defective chip from all current field of view of the camera includes: By comparing the cross mark within the current field of view of the camera with one or more features of the edge contour, horizontal and vertical line length ratio, and subpixel grayscale texture of the standard image template, if the combined deviation value of all the features involved in the comparison is less than a preset deviation threshold, it is determined that there is a complete cross mark of the current defective chip within the current field of view of the camera.
[0011] Furthermore, the number of the plurality of teaching locations is dynamically adjusted according to the size of the current defective chip. The larger the size of the current defective chip, the more teaching locations there are, and they cover the central area and four edge areas of the current defective chip.
[0012] Furthermore, the multiple teaching locations are arranged at equal intervals.
[0013] Furthermore, the camera takes pictures sequentially from the first teaching position point in either the order of spreading outwards from the center of the current defective chip, or in the order of ascending coordinates of the teaching position points.
[0014] Furthermore, there is a 10%-20% overlap between adjacent teaching field angles.
[0015] Further, determining whether the identification code image of the current defective chip exists within the current field of view of the camera includes: The image within the current field of view is matched with a preset identifier feature library. When the matching degree is greater than a preset threshold, it is determined that the identifier image of the current defective chip exists within the current field of view. Among them, a two-level identification strategy of first coarse matching and then fine matching is adopted for matching; Coarse matching: Extract the global grayscale histogram and outer contour geometric features of the image within the current field of view, as well as the global grayscale histogram and outer contour geometric features of the identifier samples in the identifier feature library. Calculate the coarse matching similarity using the similarity calculation formula, and select identifier samples that meet the coarse matching similarity standard from the identifier feature library to obtain the identifier samples after coarse matching. Fine matching: Extract the depth features of the image within the current field of view and the depth features of the identification code sample after coarse matching for comparison, calculate the spatial distance between the depth features to obtain the fine matching degree, and when the fine matching degree is greater than a preset threshold, determine that there is an identification code image of the current defective chip in the current field of view.
[0016] The defect chip location method provided by this invention has the following advantages: by matching the cross mark printed on the corner of the defect chip, the difference between the current position of the camera and the position of the identification code on the defect chip can be quickly calculated. Then, the camera is positioned above the identification code to take a picture, and the accurate position of the defect chip on the wafer can be obtained by recognizing the information in the identification code image, which shortens the detection time and greatly improves the detection efficiency. Attached Figure Description
[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the following detailed description to explain the invention, but do not constitute a limitation thereof.
[0018] Figure 1 This is a schematic diagram of the chip arrangement on a wafer provided by the present invention.
[0019] Figure 2 This is a comparative schematic diagram showing the wafer before and after correction provided by the present invention.
[0020] Figure 3 A flowchart of the defective chip location method provided by the present invention. Detailed Implementation
[0021] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0022] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.
[0024] like Figure 1 As shown, the chips are arranged on the wafer according to certain row and column rules. OCR identification codes (using numbers 1, 2, 3... as an example) are printed at specific locations on the chips using printing technologies such as laser marking, photolithography, chemical etching, and inkjet printing. The OCR identification codes can contain information such as product model, batch, production date, and product location. A cross mark (or other feature patterns) is printed at each of the four corners of the chip, and a standard image template is established for each cross mark. The standard image template is formed by combining the cross mark with the inherent features of each chip, which aims to distinguish which corner of the chip the cross mark belongs to. Before locating the detected defective chip, the camera is controlled to take pictures of the current defective chip at multiple preset teaching position points. Each teaching position point corresponds to a teaching field of view. At least one teaching field of view contains the identification code of the current defective chip. At the same time, there is at most one cross mark within each teaching field of view (fov), and each cross mark should only appear within one teaching field of view (fov). The pixel coordinates of the cross mark within the teaching field of view (fov) are retained. It's important to note that for positioning on a wafer, the machine uses a camera and XY and R axes. Before positioning begins, a set of starting positions is taught during recipe creation. The number of starting positions corresponds to the number of field-of-view angles required to photograph one die. The first taught position is the top-left corner of the die, and the others are other parts of the die. Each taught position is followed by a taught field of view (FOV). Each FVO contains different content, such as crosshairs and OCR codes on a die. A taught FVO may contain zero, one, or more of these elements. One or two taught FVOs will contain OCR codes.
[0025] This embodiment provides a method for locating defective chips, such as... Figure 3 As shown, the method for locating the defective chip includes: Step S1: After the wafer loading is completed, the current defective chip is positioned. The camera is controlled to take pictures of the current defective chip sequentially from the first teaching position point. During the process of taking pictures of the current defective chip, it is determined whether there is an identification code image of the current defective chip within the current field of view of the camera. Step S2: When the identification code image of the current defective chip is present in the current field of view of the camera, the position information of the current defective chip on the wafer is directly determined based on the identification code image of the current defective chip, and the process ends; when the identification code image of the current defective chip is not present in any of the current field of view of the camera, the current field of view containing the complete cross mark of the current defective chip is found from all the current field of view of the camera; wherein, the captured image is matched with a standard image template, if the match fails, it is determined that the complete cross mark of the current defective chip is not present in the current field of view of the camera, otherwise, it is determined that the complete cross mark of the current defective chip is present in the current field of view of the camera; It should be noted that during wafer loading, due to the poor consistency of loading positions, the image taken at the beginning of the teaching process differs significantly from the actual position on the wafer. Therefore, positioning is necessary to determine the location of the content captured by the camera at the current field of view on the wafer. Before starting the operation, a map file is imported, which stores the arrangement of all dies on the wafer as a two-dimensional array. Ocr(X,Y) usually represents the coordinates of the die on the wafer. However, due to process differences, such as the wafer often discarding a certain range of dies on the outermost ring, or after sorting by a sorting machine, the position of each die is not in its original arrangement. We use WFX / Y to represent the position of the die in the unsorted map, and BFX / Y to represent its position on the actual blue film (different from WFX / Y after sorting, the same as before), and Chip(X,Y) to represent the specific value of Ocr(X,Y).
[0026] It should be noted that after the wafer is loaded, the camera will take a picture at the first teaching position. Due to the position difference, the position of the picture cannot be determined. There may be two situations: one is that the current field of view contains the identification code image of the current defective chip, and the other is that it does not. If the current field of view does not contain the identification code image of the current defective chip, the camera will move to the second teaching position to take a picture, and so on.
[0027] Step S3: Based on the standard image template of the complete cross mark of the current defective chip, determine the teaching field of view to which the complete cross mark belongs, and then calculate the distance between the current position of the camera and the position of the current defective chip identification code, that is, the distance that needs to be moved to photograph the identification code. Specifically, the calculation of the distance between the current position of the camera and the current position of the defective chip identifier also includes: The formula for calculating this distance is as follows: Distance (mm) = (Pixel coordinates of the complete crosshair within the current field of view - Pixel coordinates of the complete crosshair within the teaching field of view) * Camera millimeter-to-pixel ratio + (Coordinates of the first teaching position point - Coordinates of the teaching position point corresponding to the teaching field of view to which the complete crosshair belongs).
[0028] Step S4: Move the wafer or camera according to the distance so that the camera can capture the identification code above the identification code of the current defective chip, and determine the position information of the current defective chip on the wafer based on the captured identification code image. The process ends.
[0029] It should be noted that the wafer stage carries the wafer and moves it this distance to below the camera for image capture. Alternatively, the camera can move this distance to above the identification code for image capture.
[0030] Preferably, the step of identifying the current field of view containing the complete crosshair of the current defective chip from all current field of view of the camera further includes: like Figure 2 As shown, when the camera's current field of view contains a complete cross mark of the current defective chip, the wafer's placement angle is corrected according to the deflection angle of the complete cross mark within the current field of view, so that the wafer is in a horizontal position in the field of view. After the wafer's placement angle is corrected, the camera is controlled again to take pictures of the current defective chip sequentially based on multiple teaching position points until a complete cross mark of the current defective chip is present in the current field of view of the camera.
[0031] It should be noted that since the wafer has moved after correction and the position of the rotation center cannot be confirmed, the camera needs to return to the first teaching position point to take pictures again. If there is no complete cross mark in the current field of view, the camera moves to the subsequent teaching position point to continue taking pictures until there is a complete cross mark of the current defective chip in the current field of view of the camera.
[0032] Preferably, the process of establishing a standard image template for each cross mark includes: Select a defect-free qualified chip on the wafer, control the camera to capture images of the complete cross mark on the four corners of the qualified chip, remove image noise, and extract one or more features from the edge contour, the ratio of horizontal and vertical line lengths and the sub-pixel grayscale texture of the complete cross mark image. Combine the differential features of the complete cross mark images at each corner to generate four independent and distinguishable standard image templates.
[0033] Specifically, ① the edge contour extraction adopts a combination algorithm of Canny operator + subpixel edge fitting. First, the Canny multi-scale gradient detection is used to detect coarse edges and suppress uneven illumination and reflective noise on the wafer surface; then, based on the quadratic polynomial subpixel interpolation algorithm, the discrete pixels of the crosshair edges are fitted at the subpixel level to correct the integer pixel level edge offset error and improve the positioning accuracy of the cross contour edge.
[0034] ② Horizontal and vertical line length ratio coefficient : ; : The effective pixel span of the crosshair; : Effective pixel span of the crosshair; By combining the least squares method to perform straight line fitting on the effective line segments of horizontal and vertical lines, invalid pixels with edge defects and breakpoints are eliminated to ensure the validity of length values and avoid distortion of proportion calculation caused by local occlusion.
[0035] ③ Subpixel grayscale texture: The Local Binary Pattern (LBP) texture operator is used to extract local grayscale texture and edge gradient distribution features around the cross, solidify the micro-process differences of each corner of the cross, and introduce a lightweight convolutional neural network (CNN) shallow feature extraction module. The enhanced extraction of texture details and grayscale abrupt changes is completed only through shallow convolution kernels without increasing the device's computational load. This improves the feature recognition in weak texture and low contrast scenes, constructs a grayscale gradient co-occurrence matrix, and statistically analyzes the texture entropy, contrast, and correlation parameters of the cross region to form a unique texture feature fingerprint.
[0036] Preferably, finding the current field of view containing the complete crosshair of the current defective chip from all current field of view of the camera includes: By comparing the cross mark within the current field of view of the camera with one or more features of the edge contour, horizontal and vertical line length ratio, and subpixel grayscale texture of the standard image template, if the combined deviation value of all the features involved in the comparison is less than a preset deviation threshold, it is determined that there is a complete cross mark of the current defective chip within the current field of view of the camera.
[0037] Specifically, when all three features—edge contour, horizontal and vertical line length ratio, and sub-pixel grayscale texture—are extracted, the weights of the deviation values for each feature are W1=0.5, W2=0.3, and W3=0.2, respectively; among them, the deviation value of edge contour △1≤0.15, the deviation value of horizontal and vertical line length ratio △2≤0.2, and the deviation value of sub-pixel grayscale texture △3≤0.25; the comprehensive deviation value = W1*△1 + W2*△2 + W3*△3. When the comprehensive deviation is ≤0.185, it is determined that there is a complete cross mark of the current defective chip in the current field of view of the camera.
[0038] When only one or any two features are extracted for comparison, the extracted one or two features are normalized and adaptively assigned according to the original weights of the features mentioned above. Unextracted features do not participate in the weighted calculation; the comprehensive deviation value is calculated based on the reassigned weights.
[0039] In this embodiment of the invention, the number of the plurality of teaching location points is dynamically adjusted according to the size of the current defective chip. The larger the size of the current defective chip, the more teaching location points there are, and they cover the central area and four edge areas of the current defective chip.
[0040] In this embodiment of the invention, multiple teaching position points are arranged at equal intervals.
[0041] Preferably, the camera takes pictures sequentially from the first teaching position point in the following order: either in the order of spreading outward from the center of the current defective chip, or in the order of increasing coordinates of the teaching position points.
[0042] Preferably, adjacent teaching field angles have a 10%-20% overlap area.
[0043] It should be noted that the cross marks on the four corners of the chip are prone to falling on the boundary of a single teaching field of view; if there is no overlap, the cross marks are easily clipped, becoming incomplete images and unable to be identified as complete cross marks. However, the teaching field of view has a small overlap, which can ensure that the cross marks and identification codes are completely imaged within a single teaching field of view.
[0044] Preferably, determining whether the identification code image of the current defective chip exists within the current field of view of the camera includes: The image within the current field of view is matched with a preset identifier feature library. When the matching degree is greater than a preset threshold, it is determined that the identifier image of the current defective chip exists within the current field of view. Among them, a two-level identification strategy of first coarse matching and then fine matching is adopted for matching; Coarse matching: Extract the global grayscale histogram and outer contour geometric features of the image within the current field of view, as well as the global grayscale histogram and outer contour geometric features of the identifier samples in the identifier feature library. Calculate the coarse matching similarity using the similarity calculation formula, and select identifier samples that meet the coarse matching similarity standard from the identifier feature library to obtain the identifier samples after coarse matching. Specifically, the contour feature vector P of the image within the current field of view: P={p1,p2,...,pn}; the contour feature vector Q of the identifier sample in the identifier feature library: Q={q1,q2,...,qn}; coarse matching similarity The calculation formula is as follows: ; Coarse-match similarity is selected from the identifier feature library. ≥80% of the identification code samples reduce the amount of subsequent fine matching calculations.
[0045] Fine matching: Extract the high-dimensional depth features of the image within the current field of view and the high-dimensional depth features of the identification code sample after coarse matching for comparison, calculate the spatial distance between the high-dimensional depth features to obtain the fine matching degree, and when the fine matching degree is greater than a preset threshold (e.g., 95%), determine that there is an identification code image of the current defective chip in the current field of view.
[0046] Specifically, Euclidean distance is used. Quantifying differences: ; Precision matching The calculation formula is as follows: ; in, Euclidean distance is the distance between high-dimensional deep features. The smaller the value, the higher the similarity and the better the matching of the high-dimensional deep features; m is the total number of dimensions of the deep features; The k-th dimension depth feature value of the image within the current field of view of the camera; The k-th dimension depth feature value of the identifier code sample after coarse matching and filtering; The preset maximum feature distance threshold is preferably 0.3; ≥95% is considered a valid criterion.
[0047] The defect chip location method provided by this invention quickly calculates the difference between the current position of the camera and the position of the identification code on the defect chip by matching the cross mark printed on the corner of the defect chip. Then, the camera is positioned above the identification code to take a picture, and the accurate position of the defect chip on the wafer is obtained by recognizing the information in the identification code image. This shortens the detection time and greatly improves the detection efficiency.
[0048] It is understood that the above embodiments are merely exemplary implementations used to illustrate the principles of the present invention, and the present invention is not limited thereto. For those skilled in the art, various modifications and improvements can be made without departing from the spirit and essence of the present invention, and these modifications and improvements are also considered to be within the scope of protection of the present invention.
Claims
1. A method for locating defective chips, wherein the chips are arranged in a regular pattern on a wafer, and each chip has an identification code printed on it, characterized in that, A cross mark is printed on each of the four corners of the chip, and a standard image template for each cross mark is established. Before locating the detected defective chip, the camera is controlled to take pictures of the defective chip at multiple preset teaching positions. Each teaching position corresponds to a teaching field of view. At least one teaching field of view contains the identification code of the defective chip, and the cross mark of the defective chip is also contained in the teaching field of view. The method for locating the defective chip includes: Step S1: After the wafer loading is completed, the current defective chip is positioned. The camera is controlled to take pictures of the current defective chip sequentially from the first teaching position point. During the process of taking pictures of the current defective chip, it is determined whether there is an identification code image of the current defective chip within the current field of view of the camera. Step S2: When the identification code image of the current defective chip exists in the current field of view of the camera, the position information of the current defective chip on the wafer is determined directly based on the identification code image of the current defective chip, and the process ends; when the identification code image of the current defective chip does not exist in any of the current field of view of the camera, the current field of view containing the complete cross mark of the current defective chip is found from all the current field of view of the camera. Step S3: Based on the standard image template of the complete cross mark of the current defective chip, determine the teaching field of view to which the complete cross mark belongs, and then calculate the distance between the current position of the camera and the position of the current defective chip identification code; Step S4: Move the wafer or camera according to the distance so that the camera can capture the identification code above the identification code of the current defective chip, and determine the position information of the current defective chip on the wafer based on the captured identification code image. The process ends.
2. The method for locating defective chips according to claim 1, characterized in that, The step of identifying the current field of view containing the complete crosshair of the defective chip from all current field of view of the camera includes: When a complete crosshair mark of the current defective chip is present in the current field of view of the camera, the placement angle of the wafer is corrected according to the deflection angle of the complete crosshair mark in the current field of view. After the wafer's placement angle is corrected, the camera is controlled again to take pictures of the current defective chip sequentially based on multiple teaching position points until a complete cross mark of the current defective chip is present in the current field of view of the camera.
3. The method for locating defective chips according to claim 1, characterized in that, The calculation of the distance between the current position of the camera and the current position of the defective chip identifier also includes: The formula for calculating this distance is as follows: Distance = (pixel coordinates of the complete crosshair within the current field of view - pixel coordinates of the complete crosshair within the teaching field of view) * camera pixel ratio + (coordinates of the first teaching position point - coordinates of the teaching position point corresponding to the teaching field of view to which the complete crosshair belongs).
4. The method for locating defective chips according to claim 1 or 2, characterized in that, The establishment of a standard image template for each cross mark includes: Select a defect-free qualified chip on the wafer, control the camera to capture images of the complete cross mark on the four corners of the qualified chip, remove image noise, and extract one or more features from the edge contour, the ratio of horizontal and vertical line lengths and the sub-pixel grayscale texture of the complete cross mark image. Combine the differential features of the complete cross mark images at each corner to generate four independent and distinguishable standard image templates.
5. The method for locating defective chips according to claim 4, characterized in that, The step of identifying the current field of view containing the complete crosshair of the defective chip from all current field of view of the camera includes: By comparing the cross mark within the current field of view of the camera with one or more features of the edge contour, horizontal and vertical line length ratio, and subpixel grayscale texture of the standard image template, if the combined deviation value of all the features involved in the comparison is less than a preset deviation threshold, it is determined that there is a complete cross mark of the current defective chip within the current field of view of the camera.
6. The method for locating defective chips according to claim 1, characterized in that, The number of teaching locations is dynamically adjusted according to the size of the current defective chip. The larger the size of the current defective chip, the more teaching locations there are, and they cover the central area and four edge areas of the current defective chip.
7. The method for locating defective chips according to claim 6, characterized in that, Multiple teaching locations are arranged at equal intervals.
8. The method for locating defective chips according to claim 6 or 7, characterized in that, The camera takes pictures sequentially from the first teaching position point in either the order of spreading outwards from the center of the current defective chip, or in the order of ascending coordinates of the teaching position points.
9. The method for locating defective chips according to claim 1, characterized in that, There is a 10%-20% overlap between adjacent teaching field angles.
10. The method for locating defective chips according to claim 1, characterized in that, The determination of whether the identification code image of the current defective chip exists within the current field of view of the camera includes: The image within the current field of view is matched with a preset identifier feature library. When the matching degree is greater than a preset threshold, it is determined that the identifier image of the current defective chip exists within the current field of view. Among them, a two-level identification strategy of first coarse matching and then fine matching is adopted for matching; Coarse matching: Extract the global grayscale histogram and outer contour geometric features of the image within the current field of view, as well as the global grayscale histogram and outer contour geometric features of the identifier samples in the identifier feature library. Calculate the coarse matching similarity using the similarity calculation formula, and select identifier samples that meet the coarse matching similarity standard from the identifier feature library to obtain the identifier samples after coarse matching. Fine matching: Extract the depth features of the image within the current field of view and the depth features of the identification code sample after coarse matching for comparison, calculate the spatial distance between the depth features to obtain the fine matching degree, and when the fine matching degree is greater than a preset threshold, determine that there is an identification code image of the current defective chip in the current field of view.