A method and apparatus for detecting a pellet polymerization defect

Abstract

The application relates to the technical field of wafer detection, and provides a core particle polymerization defect detection method and equipment, which are used for improving the accuracy of core particle polymerization defect detection. The method designs a unique detection mode of core particle polymerization defects for candidate core particles that are preliminarily detected as normal, carries out area analysis on a core particle region by using a binary image of the candidate core particle under a target light source, obtains a target core particle that may have a core particle polymerization defect, and the core particle polymerization defect is generally caused by core particle misplacement. Therefore, after obtaining a core particle out-of-bound situation of the target core particle in a detection frame in a first detection result, four boundary interest regions in a binary image of the target core particle are adjusted, so that a region where the mislocated core particle is located is accurately obtained, and then whether the target core particle has the core particle polymerization defect is determined according to a boundary core particle area proportion of the boundary interest region, the precision of chip polymerization defect detection is effectively improved, the missed detection rate of defective products is reduced, and the quality of electronic products is ensured.

Description

Technical Field

[0001] This application relates to the field of wafer inspection technology, and provides a method and equipment for detecting chip polymerization defects. Background Technology

[0002] In the current era of explosive growth in artificial intelligence and technology products, integrated circuits and semiconductor chips, as core basic components, directly determine the performance and reliability of the entire electronic product. As electronic products develop towards miniaturization and high performance, chip manufacturing processes are constantly improving, chip sizes are continuously shrinking, and integration levels are becoming increasingly higher.

[0003] Typically, a single wafer is about 6-8 inches in size and contains tens of thousands to hundreds of thousands of LED chips, with some reaching up to a million. These chips are arranged regularly on the wafer. When the wafer structure is distorted, especially under certain external environments or stress, the chips may become misaligned, and multiple chips may aggregate (e.g., stick together, be adjacent, or overlap), forming a "twin cell" structure. This structure can sometimes interfere with normal electron flow or signal conduction, affecting the performance of the chips and consequently the operational stability of electronic products.

[0004] Therefore, the detection of defects in core particle polymerization has become an urgent problem to be solved in the current development of electronic products. Summary of the Invention

[0005] This application provides a core defect detection method and device to improve the accuracy of core defect detection.

[0006] In a first aspect, embodiments of this application provide a method for detecting core defects, including:

[0007] Obtain the first detection result of each of the multiple chips contained in the wafer, and take the chips whose first detection result is initially normal as candidate chips;

[0008] Based on the binary image of the candidate core particle under the target light source, the core particle area of ​​the core particle region is calculated, and when the core particle area meets the preset abnormal conditions, the candidate core particle is taken as the target core particle.

[0009] Based on the detection box of the target core in the first detection result, the core cross-boundary situation is obtained, and based on the core cross-boundary situation, the four boundary interest regions in the binary image of the target core are adjusted; the boundary interest regions are preset according to the core model.

[0010] When the adjusted boundary core area of ​​the boundary interest region meets the preset area ratio, the first detection result of the target core is modified to core aggregation defect.

[0011] The beneficial effects of the above technical solution are as follows: after preliminary inspection of multiple chips contained in the wafer, the chips with the first inspection result being initially normal are used as candidate chips for subsequent fine inspection, filtering out defective chips, thereby improving the efficiency of fine inspection. Furthermore, a unique detection method for chip polymerization defects was designed for the subsequent fine inspection process of candidate chips. Considering the large area of ​​the chip region during chip polymerization, the area of ​​the chip region was analyzed using the binary image of the candidate chip under the target light source to obtain the target chips that may have chip polymerization defects. Since chip polymerization defects are generally caused by chip misalignment, four boundary regions of interest (ROIs) in the binary image that may contain misaligned chips are usually pre-defined according to the chip model. In this way, after obtaining the chip boundary crossing situation based on the detection box of the target chip in the first detection result, the four boundary ROIs in the binary image of the target chip are adjusted according to the chip boundary crossing situation, thereby accurately obtaining the area where the misaligned chip is located. Then, based on the boundary chip area ratio of the boundary ROIs, it is determined whether the target chip has a chip polymerization defect, which effectively improves the accuracy of chip polymerization defect detection, reduces the missed detection rate of defective products, and thus ensures the quality of electronic products.

[0012] Optionally, the step of obtaining the boundary crossing status of the target core particle based on the detection box in the first detection result, and adjusting the four boundary interest regions in the binary image of the target core particle based on the boundary crossing status, includes:

[0013] The coordinates of the center point of the detection box of the target particle in the first detection result are compared with the coordinates of the image origin of the target particle; wherein, the image origin is the top left corner vertex of the image;

[0014] Based on the comparison results, the direction of the core particle crossing the boundary is determined, and based on the direction of the core particle crossing the boundary, the four boundary regions of interest in the binary image of the target core particle are reduced.

[0015] The beneficial effects of the above technical solution are as follows: chip aggregation usually affects the position of the detection box, and the image origin is generally the top left corner vertex. In this way, by comparing the coordinates of the center point of the detection box with the coordinates of the image origin, the direction of the chip going out of bounds can be determined, thereby adjusting the boundary region of interest where the misaligned core particles may be located, and thus ensuring the accuracy of core particle aggregation defect detection.

[0016] Optionally, determining the core particle boundary crossing direction based on the comparison result, and narrowing the four boundary interest regions in the binary image of the target core particle according to the core particle boundary crossing direction, includes:

[0017] When the x-coordinate of the center point is less than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the left boundary of the image. The boundary interest region of the left boundary of the image is deleted, while the boundary interest region of the right boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced.

[0018] When the x-coordinate of the center point is greater than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the right boundary of the image. The boundary interest region of the right boundary of the image is deleted, while the boundary interest region of the left boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced.

[0019] When the ordinate of the center point is less than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the upper boundary of the image. The boundary interest region of the upper boundary of the image is deleted, while the boundary interest region of the lower boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced.

[0020] When the ordinate of the center point is greater than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the lower boundary of the image. The boundary interest region of the lower boundary of the image is deleted, while the boundary interest region of the upper boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced.

[0021] The beneficial effects of the above technical solution are: different boundary region adjustment methods are designed for the specific cross-boundary direction of the chip, thereby ensuring the accuracy of subsequent chip polymerization defect analysis.

[0022] Optionally, after modifying the first detection result of the target core particle to a core particle polymerization defect, the method further includes:

[0023] The second detection result of each of the multiple chips contained in the wafer is obtained, and the first detection result and the second detection result of each chip are fused together to obtain the target detection result;

[0024] The multiple core particles are arranged in an orderly manner according to certain rules, and at least one core particle with a target detection result of core particle polymerization defect is selected. The selected at least one core particle is used as the core particle to be analyzed.

[0025] For each of the particles to be analyzed, the target detection result of the particle to be analyzed is obtained based on the gap between the particle to be analyzed and a plurality of adjacent particles.

[0026] The beneficial effects of the above technical solution are as follows: by aggregating the detection results obtained by multiple detection methods, the detection accuracy of core particle defects is improved. Furthermore, since the analysis of core particle aggregation defects using core particle area may be affected by contamination defects, which may lead to inaccurate detection, the gap between core particles is used to re-judge the detected core particle aggregation defects, thereby further improving the accuracy of core particle aggregation defect detection.

[0027] Optionally, obtaining the target detection result of the core particle to be analyzed based on the gap between the core particle to be analyzed and adjacent core particles includes:

[0028] For each of the adjacent core particles, the following steps are performed: calculate the gap between the core particle to be analyzed and each adjacent core particle;

[0029] When the smallest gap among multiple gaps meets the preset gap condition, the target detection result of the core particle to be analyzed is determined to be a core particle aggregation defect;

[0030] When the smallest gap among multiple gaps does not meet the preset gap condition, the target detection result of the chip to be analyzed is determined to be a normal chip.

[0031] The beneficial effect of the above technical solution is that when using the gap between core particles to re-judge the detected core particle polymerization defects, the accuracy of the re-judgment of core particle polymerization defects is guaranteed.

[0032] Optionally, the binary image is obtained in the following manner:

[0033] From the initial core images of the candidate core particles under multiple light sources, an initial core image under a target light source is selected; wherein, in the initial core image under the target light source, the foreground is gray and the background is white;

[0034] The pixel values ​​of the foreground and background in the selected initial core image are flipped, and the gaps in the flipped initial core image are filled to obtain the target core image of the candidate core under the target light source.

[0035] Based on the grayscale histogram of the target core image, target segmentation thresholds for the foreground and background are obtained; wherein, the target segmentation thresholds maximize the grayscale difference between the foreground and background and minimize the grayscale difference within the foreground and background.

[0036] Based on the segmentation threshold, the target core image is binarized to obtain a binary image of the candidate core.

[0037] The beneficial effects of the above technical solution are as follows: Since the foreground and background colors of the initial core image under the target light source are significantly different, the core and background can be better distinguished. After pixel value flipping and gap filling, the core outline is more prominent. Furthermore, by setting a reasonable foreground and background segmentation threshold, the boundary between the core and the background is further clarified, thereby improving the accuracy of subsequent area analysis.

[0038] Optionally, the step of calculating the core area of ​​the core region based on the binary image of the candidate core under the target light source, and using the candidate core as the target core when the core area meets a preset anomaly condition, includes:

[0039] From the binary image of the candidate core particle under the target light source, extract the connected regions with a pixel value of 255, and use the connected regions as the core particle regions;

[0040] The core area is obtained by counting the number of pixels within the core region;

[0041] When the core particle area is greater than a preset multiple of the preset core particle area, the candidate core particle is taken as the target core particle.

[0042] The beneficial effects of the above technical solution are as follows: when there is a core particle polymerization defect, the area of ​​the core particle is relatively large. Therefore, by comparing the area with that of normal core particles, it can be determined whether the candidate core particle is the target core particle of multiple core particles polymerization.

[0043] Optionally, when the area of ​​the central hole in the binary image is greater than a preset area threshold, the method further includes:

[0044] Extract the first maximum connected region with a pixel value of 255 from the binary image, and flip the pixel values ​​within the first maximum connected region to 0;

[0045] Extract the second maximum connected region from the flipped binary image, set the pixel value within the second maximum connected region to 255, and obtain a new binary image.

[0046] The beneficial effects of the above technical solution are as follows: Since the structures of different chip models may be different, some chips contain light-emitting areas inside. The light-emitting areas are generally close to the color of the background, and the chip area is generally calculated based on the chip outline. In order to be applicable to more chip models, the first maximum connected area with a pixel value of 255 is pixel flipped so that the second maximum connected area represents the chip area of ​​multiple chip models, thus ensuring the consistency of the algorithm.

[0047] Secondly, embodiments of this application provide a core defect detection device, including a processor, a memory, and a communication interface, wherein the communication interface, the memory, and the processor are connected via a bus.

[0048] The communication interface is used to acquire images;

[0049] The memory stores a computer program, and the processor performs the following operations according to the computer program:

[0050] Obtain the first detection result of each of the multiple chips contained in the wafer, and take the chips whose first detection result is initially normal as candidate chips;

[0051] Based on the binary image of the candidate core particle under the target light source, the core particle area of ​​the core particle region is calculated, and when the core particle area meets the preset abnormal conditions, the candidate core particle is taken as the target core particle.

[0052] Based on the detection box of the target core in the first detection result, the core cross-boundary situation is obtained, and based on the core cross-boundary situation, the four boundary interest regions in the binary image of the target core are adjusted; the boundary interest regions are preset according to the core model.

[0053] When the adjusted boundary core area of ​​the boundary interest region meets the preset area ratio, the first detection result of the target core is modified to core aggregation defect.

[0054] Optionally, the processor obtains the boundary crossing status of the target core particle based on the detection box in the first detection result, and adjusts the four boundary interest regions in the binary image of the target core particle according to the boundary crossing status. Specifically, the operation is as follows:

[0055] The coordinates of the center point of the detection box of the target particle in the first detection result are compared with the coordinates of the image origin of the target particle; wherein, the image origin is the top left corner vertex of the image;

[0056] Based on the comparison results, the direction of the core particle crossing the boundary is determined, and based on the direction of the core particle crossing the boundary, the four boundary regions of interest in the binary image of the target core particle are reduced.

[0057] Optionally, the processor determines the core particle boundary crossing direction based on the comparison result, and reduces the four boundary regions of interest in the binary image of the target core particle according to the core particle boundary crossing direction. Specifically, the operation is as follows:

[0058] When the x-coordinate of the center point is less than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the left boundary of the image. The boundary interest region of the left boundary of the image is deleted, while the boundary interest region of the right boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced.

[0059] When the x-coordinate of the center point is greater than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the right boundary of the image. The boundary interest region of the right boundary of the image is deleted, while the boundary interest region of the left boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced.

[0060] When the ordinate of the center point is less than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the upper boundary of the image. The boundary interest region of the upper boundary of the image is deleted, while the boundary interest region of the lower boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced.

[0061] When the ordinate of the center point is greater than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the lower boundary of the image. The boundary interest region of the lower boundary of the image is deleted, while the boundary interest region of the upper boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced.

[0062] Optionally, after modifying the first detection result of the target core particle to a core particle aggregation defect, the processor further executes:

[0063] The second detection result of each of the multiple chips contained in the wafer is obtained, and the first detection result and the second detection result of each chip are fused together to obtain the target detection result;

[0064] The multiple core particles are arranged in an orderly manner according to certain rules, and at least one core particle with a target detection result of core particle polymerization defect is selected. The selected at least one core particle is used as the core particle to be analyzed.

[0065] For each of the particles to be analyzed, the target detection result of the particle to be analyzed is obtained based on the gap between the particle to be analyzed and a plurality of adjacent particles.

[0066] Optionally, the processor obtains the target detection result of the core particle to be analyzed based on the gap between the core particle to be analyzed and adjacent core particles, specifically by:

[0067] For each of the adjacent core particles, the following steps are performed: calculate the gap between the core particle to be analyzed and each adjacent core particle;

[0068] When the smallest gap among multiple gaps meets the preset gap condition, the target detection result of the core particle to be analyzed is determined to be a core particle aggregation defect;

[0069] When the smallest gap among multiple gaps does not meet the preset gap condition, the target detection result of the chip to be analyzed is determined to be a normal chip.

[0070] Optionally, the processor obtains the binary image in the following manner:

[0071] From the initial core images of the candidate core particles under multiple light sources, an initial core image under a target light source is selected; wherein, in the initial core image under the target light source, the foreground is gray and the background is white;

[0072] The pixel values ​​of the foreground and background in the selected initial core image are flipped, and the gaps in the flipped initial core image are filled to obtain the target core image of the candidate core under the target light source.

[0073] Based on the grayscale histogram of the target core image, target segmentation thresholds for the foreground and background are obtained; wherein, the target segmentation thresholds maximize the grayscale difference between the foreground and background and minimize the grayscale difference within the foreground and background.

[0074] Based on the segmentation threshold, the target core image is binarized to obtain a binary image of the candidate core.

[0075] Optionally, the processor calculates the core area of ​​the core region based on the binary image of the candidate core under the target light source, and when the core area meets a preset anomaly condition, the candidate core is used as the target core. Specifically, the operation is as follows:

[0076] From the binary image of the candidate core particle under the target light source, extract the connected regions with a pixel value of 255, and use the connected regions as the core particle regions;

[0077] The core area is obtained by counting the number of pixels within the core region;

[0078] When the core particle area is greater than a preset multiple of the preset core particle area, the candidate core particle is taken as the target core particle.

[0079] Optionally, when the area of ​​the central hole in the binary image is greater than a preset area threshold, the processor further executes:

[0080] Extract the first maximum connected region with a pixel value of 255 from the binary image, and flip the pixel values ​​within the first maximum connected region to 0;

[0081] Extract the second maximum connected region from the flipped binary image, set the pixel value within the second maximum connected region to 255, and obtain a new binary image.

[0082] Thirdly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions that, when executed, can implement the steps of the above-described core-particle polymerization defect detection method.

[0083] The technical effects of any of the implementation methods in the third aspect to the third face can be seen in the technical effects of the corresponding implementation methods in the first aspect and the second face, and will not be repeated here. Attached Figure Description

[0084] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0085] Figure 1 A schematic flowchart illustrating a core polymerization defect detection method provided in an embodiment of this application;

[0086] Figure 2 A schematic flowchart of the preliminary core particle detection method provided in the embodiments of this application;

[0087] Figure 3A This is a schematic diagram of a partial wafer image provided in an embodiment of this application;

[0088] Figure 3B This is a schematic diagram of the first detection result of the core provided in an embodiment of this application;

[0089] Figure 4 A schematic diagram of the binarization process of the core image provided in the embodiments of this application;

[0090] Figure 5A and Figure 5B A schematic diagram showing the foreground and background of the initial core images for two different types of cores;

[0091] Figure 6 This is a schematic diagram illustrating the process of adjusting the boundary region of interest provided in an embodiment of this application;

[0092] Figure 7 This is a schematic diagram illustrating the adjustment effect of the boundary region of interest provided in an embodiment of this application;

[0093] Figure 8 A schematic diagram of the re-judgment process for core polymerization defects provided in an embodiment of this application;

[0094] Figure 9 A schematic diagram of the core-particle gap provided in an embodiment of this application;

[0095] Figure 10A This is a schematic diagram of the core image acquisition results provided in an embodiment of this application;

[0096] Figure 10B This is a schematic diagram of the core particle image detection results provided in an embodiment of this application;

[0097] Figure 11 A schematic diagram of the detection process for core polymerization defects provided in an embodiment of this application;

[0098] Figure 12 This is a structural diagram of a core defect detection device provided in an embodiment of this application. Detailed Implementation

[0099] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments of this application. Obviously, the described embodiments are only some embodiments of the technical solutions of this application, and not all embodiments. Based on the embodiments recorded in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the technical solutions of this application.

[0100] Based on the exemplary embodiments shown in this application, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of this application. Furthermore, although the disclosures in this application are presented by way of one or more exemplary examples, it should be understood that each aspect of these disclosures can constitute a complete technical solution on its own.

[0101] It should be noted that the brief descriptions of terms in this application are only for the convenience of understanding the embodiments described below, and are not intended to limit the embodiments of this application. Unless otherwise stated, these terms should be understood in their ordinary and common meaning.

[0102] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar or related objects or entities and do not necessarily imply a specific order or sequence, unless otherwise indicated. It should be understood that such terms can be used interchangeably where appropriate, for example, to implement the application in a sequence other than those given in the embodiments illustrated or described herein.

[0103] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover but not exclusively include, for example, a product or device that includes a series of components is not necessarily limited to those that are explicitly listed, but may include other components that are not explicitly listed or that are inherent to such product or device.

[0104] As used in this application, the term "module" means any known or subsequently developed hardware, software, firmware, artificial intelligence, fuzzy logic, or combination of hardware and / or software code capable of performing the functions associated with that element.

[0105] Particle aggregation refers to the phenomenon where multiple particles adhere to, are adjacent to, or overlap each other. When the number of aggregated particles is 2, it is generally called a "twin cell"; when the number of aggregated particles is greater than 2, it is generally called a "multicell".

[0106] For the core aggregation phenomenon caused by core misalignment, its morphology is complex and easily affected by background noise. Conventional detection methods (such as single light source imaging and single algorithm detection) are difficult to accurately identify, often leading to false detection or missed detection.

[0107] In view of this, this application provides a chip polymerization defect detection method, which designs a unique detection method for chip polymerization defects. Considering that the area of ​​the chip region is large during chip polymerization and chip overruns are possible, the area of ​​the chip region is analyzed using the binary image of a preliminary normal candidate chip under the target light source to obtain the target chip that may have a chip polymerization defect. Since chip polymerization defects are generally caused by chip misalignment, four boundary regions of interest in the binary image that may contain misaligned chips are usually pre-set according to the chip model. In this way, after obtaining the chip overrun situation based on the detection box of the target chip in the first detection result, the four boundary regions of interest in the binary image of the target chip are adjusted according to the chip overrun situation, thereby accurately obtaining the area where the misaligned chip is located. Then, based on the chip area ratio of the boundary region, it is determined whether the target chip has a chip polymerization defect, which effectively improves the accuracy of chip polymerization defect detection, reduces the false negative rate of defective products, and thus ensures the quality of electronic products.

[0108] On the other hand, in the embodiments of this application, preliminary anomaly detection is performed on the core particles on the wafer during core particle polymerization defect detection. Core particles that initially appear normal in the first detection are used as candidate core particles for subsequent fine detection, filtering out defective core particles, thereby improving the efficiency of fine detection. The entire coarse and fine inspection process realizes the synergy of artificial intelligence (AI) algorithms and traditional visual image processing algorithms, effectively improving the accuracy of core particle polymerization defect detection.

[0109] See Figure 1 This application provides a method for detecting defects in core particle polymerization, which mainly includes the following steps:

[0110] S101: Obtain the first detection result of each of the multiple chips contained in the wafer, and select the chips with the first detection result of being initially normal as candidate chips.

[0111] In some embodiments, considering that a single light source cannot simultaneously highlight different types of defects, such as twin gaps and background dirt, images of the core particles under multiple light sources are acquired during the initial inspection of the core particles.

[0112] like Figure 2 The diagram shown illustrates the preliminary testing process for core particles, which mainly includes the following steps:

[0113] S1011: Acquire partial images of the wafer from multiple perspectives under different light sources.

[0114] Among them, multiple partial wafer images from different perspectives under each light source can be stitched together into a complete wafer image, and each partial wafer image contains multiple complete chips.

[0115] like Figure 3A The image shown is a schematic diagram of a partial image of a wafer, where each gray area represents a chip.

[0116] S1012: Cut out the initial core images of multiple cores from each partial image of the wafer.

[0117] Each core particle has initial core particle images under multiple light sources.

[0118] S1013: For each core particle, stitch together the initial core particle images under multiple light sources to obtain a stitched core particle image.

[0119] Each initial core image under each light source occupies one channel.

[0120] S1014: Input the stitched images of multiple core particles into the AI ​​model to obtain the first detection result for each core particle.

[0121] AI models include, but are not limited to, Transform models, BERT models, YOLO models, and ResNet models.

[0122] In one embodiment, the first detection result includes not only the chip type, but also information such as the chip detection frame and the location of the chip.

[0123] like Figure 3B The image shown is a schematic diagram of the first detection result. Abnormal and normal chips can be circled with detection boxes of different colors. The numbers in the boxes indicate the position of the chip output by the AI ​​model, that is, which chip in the image.

[0124] In some embodiments, a wafer may have a large number of chips. To improve system detection efficiency, the acquired partial images of the wafer may be uploaded to the GPU for chip segmentation, detection, and other processes.

[0125] After preliminary inspection of the multiple dies contained in the wafer, abnormal dies with detected defect types do not require further fine inspection. To avoid missing dies aggregation defects, fine inspection can be performed on normal dies after preliminary inspection. By filtering out defective dies, the detection efficiency of dies aggregation defects can be effectively improved.

[0126] S102: Calculate the core area of ​​the core region based on the binary image of the candidate core particles under the target light source.

[0127] When inspecting core defects, a single light source is usually insufficient to detect all defects. Different light sources can better reveal different defects. Therefore, wafers are photographed under multiple light sources. As a result, each core has initial core images under multiple light sources. Thus, for each core, a target light source that makes the core structure clearest can be selected from multiple light sources, and the initial core image corresponding to the target light source can be used for fine defect inspection.

[0128] In one embodiment, the core area is typically large during core polymerization, so core polymerization defects can be identified by area analysis of the binary image.

[0129] like Figure 4 The diagram shown illustrates the process of obtaining a binary image, which mainly includes the following steps:

[0130] S1021: Select an initial core image under a target light source from the initial core images of candidate cores under multiple light sources.

[0131] In the initial core image under the target light source, the foreground (i.e., the core) is gray and the background is white, meaning that the color difference between the background and the foreground is significant, thus clearly reflecting the structure of the core.

[0132] In some embodiments, different types of cores have different structures, resulting in different foreground structures in the image.

[0133] like Figure 5A and Figure 5B The image shown is a schematic diagram of the foreground and background of the initial core images for two different types of core chips. Figure 5A Because the core particles contain light-emitting areas, the foreground contains holes that match the background color.

[0134] S1022: Flip the pixel values ​​of the foreground and background in the selected initial core image.

[0135] For example, a pixel value of 0 is converted to 255, a pixel value of 1 is converted to 254, a pixel value of 2 is converted to 253, and so on.

[0136] By flipping the pixel values, the core (i.e., foreground) phenomenon in the initial core image can be made bright, while the background becomes dark, which facilitates subsequent image denoising and morphological operations, making the core outline more complete.

[0137] S1023: Fill the gaps in the flipped initial core image to obtain the target core image of the candidate core under the target light source.

[0138] Noise is generally present during image acquisition, and the grayscale of pixel boundaries is large after magnification, which affects the detection results. Therefore, the initial core particles after flipping can be smoothed to make the grayscale between pixels smoother.

[0139] Since there are gaps in the middle of the polymer core particles, which affect the detection effect, the gaps in the middle can be filled in by morphological operations such as dilation and erosion after smoothing the initial core particle image, thereby improving the clarity of the core particle outline.

[0140] S1024: Obtain the target segmentation thresholds for the foreground and background based on the grayscale histogram of the target core image.

[0141] Among them, the target segmentation threshold maximizes the grayscale difference between the foreground and the background, and minimizes the grayscale difference within the foreground and the background.

[0142] In practice, all pixels in the target core image are traversed, the frequency of each gray value is counted, a gray-level histogram of the target core image is generated, and then an initial threshold is given to divide the target core image into a class less than or equal to the initial threshold and a class greater than the initial threshold. The inter-class variance is calculated, and the initial threshold is adjusted to maximize the inter-class variance, thus obtaining the target segmentation threshold for the foreground and background.

[0143] S1025: Based on the segmentation threshold, the target core image is binarized to obtain a binary image of the candidate core. Pixel values ​​greater than the segmentation threshold are set to 255, and pixel values ​​less than or equal to the segmentation threshold are set to 0. The target core image is thus converted into a binary image with pixel values ​​of 0 and 255 using the segmentation threshold.

[0144] Because the foreground and background colors of the initial core image under the target light source are significantly different, the core and background can be better distinguished. After pixel value flipping and gap filling, the core outline is more prominent. Furthermore, by setting a reasonable foreground and background segmentation threshold, the boundary between the core and the background is further clarified, thereby improving the accuracy of subsequent area analysis.

[0145] In some embodiments, for a pixel whose image center is a highlighted area, such as Figure 5A As shown, the central region is a large hole after binarization. Therefore, the hole can be filled in the binary image of such core particles.

[0146] In specific implementation, when the area of ​​the central hole in the binary image is greater than a preset area threshold, the first maximum connected region with a pixel value of 255 is extracted from the binary image, the pixel values ​​in the first maximum connected region are flipped to 0, the second maximum connected region in the flipped binary image is extracted again, and the pixel values ​​in the second maximum connected region are set to 255 to obtain a new binary image.

[0147] Since different chip models may have different structures, some chips contain light-emitting areas, which are generally similar in color to the background. The chip area is usually calculated based on the chip outline. In order to be applicable to more chip models, the first maximum connected region with a pixel value of 255 is pixel flipped so that the second maximum connected region represents the chip area of ​​multiple chip models, thereby ensuring the consistency of the algorithm.

[0148] In binary images, the boundaries between the core and the background are relatively clear. Therefore, the core area can be calculated based on the binary image. Specifically: from the binary image of the candidate core under the target light source, the connected regions with a pixel value of 255 are extracted, and these connected regions are used as the core regions. The number of pixels within the core region is then counted to obtain the core area.

[0149] S103: When the core particle area meets the preset abnormal conditions, the candidate core particle is used as the target core particle.

[0150] When core particle aggregation defects exist, the core particle area is relatively large. Therefore, by comparing the area with that of normal core particles, it can be determined whether the candidate core particle is the target core particle of multiple core particle aggregations. In specific implementation, the width and height of normal core particles can be pre-input to obtain the normal core particle region, and the number of pixels within the normal core particle region can be extracted to obtain the preset core particle area. When the core particle area extracted from the binary image is greater than a preset multiple (such as 1.15) of the preset core particle area, it indicates that the candidate core particle is abnormal, and therefore the candidate core particle is used as the target core particle; otherwise, it indicates that the candidate core particle is a normal core particle.

[0151] S104: Based on the detection box of the target core in the first detection result, obtain the core boundary crossing situation, and adjust the four boundary interest regions in the binary image of the target core according to the core boundary crossing situation.

[0152] When multiple core particles are aggregated, there may be core particles that cross the boundary due to core particle misalignment, resulting in the core particle image containing half a core particle. Therefore, according to the core particle model, boundary regions of interest that the core particles may appear can be set in advance at the top, bottom, left and right boundaries of the binary image. Subsequently, the four boundary regions of interest can be adjusted according to the actual core particle boundary crossing situation.

[0153] Typically, individual core particles are small in size, and two core particles often aggregate during core particle cutting. Therefore, taking a "twin" structure as an example, the process of adjusting the boundary region of interest is as follows: Figure 6 As shown, it mainly includes the following steps:

[0154] S1041: Compare the coordinates of the center point of the detection box of the target core in the first detection result with the coordinates of the origin of the target core image.

[0155] The origin of the image is the top left corner vertex of the image.

[0156] S1042: Determine the direction of core particle boundary crossing based on the comparison results.

[0157] Specifically, when the x-coordinate of the center point is less than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined as the left boundary of the image; when the x-coordinate of the center point is greater than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined as the right boundary of the image; when the y-coordinate of the center point is less than the y-coordinate of the image origin, the direction of the core particle crossing the boundary is determined as the upper boundary of the image; and when the y-coordinate of the center point is greater than the y-coordinate of the image origin, the direction of the core particle crossing the boundary is determined as the lower boundary of the image.

[0158] S1043: Based on the direction of the core particle's boundary crossing, reduce the four boundary regions of interest in the binary image of the target core particle.

[0159] like Figure 7 As shown, the boundary regions of interest (ROIs) are filled with diagonal lines. When the core particle crosses the boundary in the direction of the left edge of the image, the ROI at the left edge is deleted, while the ROI at the right edge remains unchanged, and the width of the ROIs at the top and bottom edges of the image is reduced. When the core particle crosses the boundary in the direction of the right edge of the image, the ROI at the right edge is deleted, while the ROI at the left edge remains unchanged, and the width of the ROIs at the top and bottom edges of the image is reduced. When the core particle crosses the boundary in the direction of the top edge of the image, the ROI at the top edge is deleted, while the ROI at the bottom edge remains unchanged, and the height of the ROIs at the left and right edges of the image is reduced. When the core particle crosses the boundary in the direction of the bottom edge of the image, the ROI at the bottom edge is deleted, while the ROI at the top edge remains unchanged, and the height of the ROIs at the left and right edges of the image is reduced.

[0160] For example, if the original size of the boundary region of interest is 50 pixels wide and 50 pixels high, for left and right boundaries, the width of the boundary region of interest at the top and bottom boundaries becomes 30 pixels wide, and for top and bottom boundaries, the height of the boundary region of interest at the left and right boundaries becomes 30 pixels high.

[0161] Chip aggregation typically affects the position of the detection frame, and the image origin is usually the top left corner vertex. By comparing the coordinates of the center point of the detection frame with the coordinates of the image origin, the direction of chip overflow can be determined. Different boundary area adjustment methods can be designed for the specific overflow direction of the chip, thereby accurately obtaining the position of the misaligned chip and ensuring the accuracy of chip aggregation defect detection.

[0162] S105: When the core area of ​​the adjusted boundary region of interest meets the preset area ratio, the first detection result of the target core is modified to core aggregation defect.

[0163] Gray-scale analysis is performed on each remaining boundary region of interest after adjustment to determine the area of ​​the boundary core particles within that region. When the area of ​​a boundary core particle in a remaining boundary region of interest exceeds 50% of the corresponding boundary region of interest, it indicates that the target core particle does not have a "twin" structure. Therefore, the first detection result of the target core particle is modified to a core particle aggregation defect; otherwise, the result of the target core particle being a normal core particle remains unchanged.

[0164] This application's embodiments employ a unique detection method for chip polymerization defects. Since the chip region is relatively large during polymerization, the area of ​​the chip region is analyzed using a binary image of the candidate chip under a target light source. This identifies target chips that may have polymerization defects. Since chip polymerization defects are generally caused by chip misalignment, boundary regions of interest (ROIs) are pre-defined at the four boundaries (top, bottom, left, and right) based on the chip model. After determining the chip's boundary crossing based on the detection frame in the first detection result, the four ROIs are adjusted to accurately obtain the area where the misaligned chip is located. Furthermore, based on the boundary chip area ratio of the ROIs, the presence of a chip polymerization defect in the target chip is determined, effectively improving the accuracy of chip polymerization defect detection, reducing the false negative rate of defective products, and thus ensuring the quality of electronic products.

[0165] In some embodiments, some core particles have a background with large areas of dirt, such as Figure 3B As shown, these contaminants can affect the detection results of core polymerization. In order to further improve the accuracy of core polymerization defect detection, core polymerization defects can be re-evaluated.

[0166] like Figure 8 The diagram shows the re-judgment process for core particle polymerization defects, which mainly includes the following steps:

[0167] S106: Obtain the second detection results of each of the multiple chips contained in the wafer, and fuse the first and second detection results of each chip to obtain the target detection result.

[0168] Typically, core defect detection includes appearance defect detection and internal defect detection. Appearance defect detection is generally obtained through visual images, which is the first detection result in this application embodiment. Internal defects are generally obtained through electrical testing methods (such as measurement with a test pen), which is recorded as the second detection result. Finally, the target detection result of the core (i.e., whether it is a defective core or a normal core) needs to be integrated with the detection results obtained by multiple methods. This process can be called file merging. By aggregating the detection results obtained by multiple detection methods, the detection accuracy of core defects is improved.

[0169] S107: Multiple core particles are arranged in an orderly manner according to certain rules.

[0170] During wafer manufacturing, the chips are arranged neatly according to certain rules, and the spacing between the chips is fixed. However, during initial inspection, the shooting angle and the jitter of the inspection frame output by the AI ​​model can cause the chips in the pixel space to not be neatly arranged. Figure 3B As shown, it is necessary to sort multiple chips in the local wafer and global wafer space. Each chip after sorting has its own position coordinates, and the adjacent chips can also be obtained.

[0171] S108: Select at least one core particle whose target detection result is a core particle polymerization defect, and use the selected at least one core particle as the core particle to be analyzed.

[0172] To avoid over-detection of core particle polymerization defects, core particles with core particle polymerization defects can be used as core particles to be analyzed for re-judgment.

[0173] S109: For each particle to be analyzed, the target detection result of the particle to be analyzed is obtained based on the gap between the particle to be analyzed and multiple adjacent particles.

[0174] Core aggregation defects are caused by core misalignment. After a core is misaligned, the gap between adjacent cores changes. Therefore, for each core to be analyzed, after obtaining the cores that are adjacent to it in the top, bottom, left and right directions from the sorted cores, the gap between the cores can be used to determine whether the core to be analyzed really has a core aggregation defect.

[0175] Taking a single particle to be analyzed as an example, calculate the gaps between the particle to be analyzed and its adjacent particles on the top, bottom, left, and right sides, such as... Figure 9 As shown, the smallest gap is selected from the four gaps (top, bottom, left, and right). When the smallest gap meets the preset gap condition, the target detection result of the chip to be analyzed is determined to be a chip polymerization defect. When the smallest gap does not meet the preset gap condition, the target detection result of the chip to be analyzed is determined to be a normal chip.

[0176] The minimum gap can be either the longitudinal gap between the upper and lower core particles or the transverse gap between the left and right core particles. Each of the longitudinal and transverse gaps corresponds to a gap threshold. When the minimum gap is a longitudinal gap, the preset gap condition is that the minimum gap is 1 / 2 of the normal upper and lower core particle gap. When the minimum gap is a transverse gap, the preset gap condition is that the minimum gap is 1 / 2 of the normal left and right core particle gap.

[0177] It should be noted that the embodiments of this application do not impose restrictive requirements on the gap threshold in the preset gap conditions, and can be flexibly adjusted according to the actual situation.

[0178] When analyzing core particle polymerization defects by core particle area, the influence of contamination defects may lead to inaccurate detection. Therefore, for the core particles to be analyzed that have detected core particle polymerization defects, the gap between the core particles is used to re-judge the core particle polymerization defects, thereby further improving the accuracy of core particle polymerization defect detection.

[0179] like Figure 10A The image shows initial images of multiple core particles. The detection results for normal and abnormal core particles can be marked in different ways (e.g., using different colored detection boxes). Abnormal core particles with different defect types (e.g., dirt, core particle aggregation, scratches, short circuits, etc.) can also be distinguished in different ways. Figure 10B As shown, the embodiments of this application can accurately detect core particle polymerization defects.

[0180] See Figure 11 The diagram below shows the detection architecture for core polymerization defects provided in this application embodiment. The main process is as follows:

[0181] S1: Acquire local wafer images.

[0182] Because of the small size of the die, a single light source may not be able to reveal the defects of the die. Therefore, a high-definition industrial camera can be used to acquire local wafer images from different perspectives under multiple light sources.

[0183] S2: Perform target detection on the local wafer image to obtain the detection box for each chip.

[0184] S3: The wafer image is cropped according to the detection frame to obtain the chip image.

[0185] S4: For each core image, AI detection algorithm and CV detection algorithm are used to detect defects in the core image respectively.

[0186] The AI ​​detection algorithm includes logic for detecting core particle aggregation defects.

[0187] S5: For each core image, the AI ​​detection results and CV detection results are fused.

[0188] S6: Sort the detected core particle images by row and column.

[0189] S7: Re-evaluate core particles with core particle polymerization defects.

[0190] S8: Output the final detection result based on the re-judgment result.

[0191] When the defect is re-identified as a core-particle polymerization defect, the test result can be output directly without modification. When the defect is re-identified as a non-core-particle polymerization defect, the test result should be modified to normal before outputting.

[0192] Those skilled in the art will understand that various aspects of this application can be implemented as a system, method, or program product. Therefore, various aspects of this application can be specifically implemented in the following forms: a completely hardware implementation, a completely software implementation (including firmware, microcode, etc.), or a combination of hardware and software implementations, collectively referred to herein as a "circuit," "module," or "system."

[0193] Based on the same technical concept, this application provides a core particle defect detection device that can implement the steps of the above-described core particle polymerization defect detection method and achieve the same technical effect.

[0194] See Figure 12 The gateway device includes a processor 1201, a memory 1202 and a communication interface 1203, which are connected via a bus 1204.

[0195] Communication interface 1203 is used to acquire images;

[0196] The memory 1202 stores a computer program, and the processor 1201 executes the program according to the computer program. Figure 1 Steps for detecting defects in the polymerization of core particles.

[0197] It should be noted that, Figure 12 As an example only, and not shown, a chip defect detection device may also include components such as memory, power supply, crystal oscillator, and camera.

[0198] In this embodiment, the memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and programs required for running instant messaging functions; the data storage area may store various instant messaging information and operation instruction sets. The memory may be volatile memory, such as random-access memory (RAM); it may also be non-volatile memory, such as read-only memory, flash memory, hard disk drive (HDD), or solid-state drive (SSD); or it may be any other medium capable of carrying or storing a desired computer program having an instruction or data structure form and accessible by a computer, but is not limited thereto. The memory may be a combination of the above-mentioned memories. The processor may include one or more central processing units (CPUs), GPUs, or digital processing units, etc.

[0199] This application does not limit the specific connection medium between the communication interface, memory, and processor. In this application, the bus between the communication interface, memory, and processor is depicted with a thick line in the figure. The connection methods between other components are only for illustrative purposes and are not intended to be limiting. Buses can be divided into address buses, data buses, control buses, etc. For ease of description, only one thick line is used to describe them in the figure, but this does not mean that there is only one bus or one type of bus.

[0200] This application also provides a computer-readable storage medium for storing instructions that, when executed, can perform the steps of any of the core polymerization defect detection methods described in the foregoing embodiments.

[0201] This application also provides a computer program product for storing a computer program that performs the steps of any of the core polymerization defect detection methods described in the foregoing embodiments.

[0202] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0203] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0204] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0205] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0206] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for detecting defects in core particle polymerization, characterized in that, The method includes: Obtain the first detection result of each of the multiple chips contained in the wafer, and take the chips whose first detection result is initially normal as candidate chips; Based on the binary image of the candidate core particle under the target light source, the core particle area of ​​the core particle region is calculated, and when the core particle area meets the preset abnormal conditions, the candidate core particle is taken as the target core particle. The coordinates of the center point of the detection box of the target particle in the first detection result are compared with the coordinates of the image origin of the target particle; wherein, the image origin is the top left corner vertex of the image; Based on the comparison results, the direction of the core particle crossing the boundary is determined, and based on the direction of the core particle crossing the boundary, the four boundary regions of interest in the binary image of the target core particle are reduced; the boundary regions of interest are preset according to the core particle model. When the adjusted boundary core area of ​​the boundary interest region meets the preset area ratio, the first detection result of the target core is modified to core aggregation defect.

2. The method as described in claim 1, characterized in that, The step of determining the core particle boundary crossing direction based on the comparison results, and narrowing the four boundary interest regions in the binary image of the target core particle according to the core particle boundary crossing direction, includes: When the x-coordinate of the center point is less than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the left boundary of the image. The boundary interest region of the left boundary of the image is deleted, while the boundary interest region of the right boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced. When the x-coordinate of the center point is greater than the x-coordinate of the image origin, the direction of the core particle crossing the boundary is determined to be the right boundary of the image. The boundary interest region of the right boundary of the image is deleted, while the boundary interest region of the left boundary of the image remains unchanged. The width of the boundary interest regions of the upper and lower boundaries of the image is reduced. When the ordinate of the center point is less than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the upper boundary of the image. The boundary interest region of the upper boundary of the image is deleted, while the boundary interest region of the lower boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced. When the ordinate of the center point is greater than the ordinate of the image origin, the direction of the core particle crossing the boundary is determined as the lower boundary of the image. The boundary interest region of the lower boundary of the image is deleted, while the boundary interest region of the upper boundary of the image remains unchanged. The height of the boundary interest regions of the left and right boundaries of the image is reduced.

3. The method as described in claim 1, characterized in that, After modifying the first detection result of the target core particle to a core particle polymerization defect, the method further includes: The second detection result of each of the multiple chips contained in the wafer is obtained, and the first detection result and the second detection result of each chip are fused together to obtain the target detection result; The multiple core particles are arranged in an orderly manner according to certain rules, and at least one core particle with a target detection result of core particle polymerization defect is selected. The selected at least one core particle is used as the core particle to be analyzed. For each of the particles to be analyzed, the target detection result of the particle to be analyzed is obtained based on the gap between the particle to be analyzed and a plurality of adjacent particles.

4. The method as described in claim 3, characterized in that, The step of obtaining the target detection result of the core particle to be analyzed based on the gap between the core particle to be analyzed and the adjacent core particles includes: For each of the adjacent core particles, the following steps are performed: calculate the gap between the core particle to be analyzed and each adjacent core particle; When the smallest gap among multiple gaps meets the preset gap condition, the target detection result of the core particle to be analyzed is determined to be a core particle aggregation defect; When the smallest gap among multiple gaps does not meet the preset gap condition, the target detection result of the core particle to be analyzed is determined to be a normal core particle.

5. The method according to any one of claims 1-4, characterized in that, The binary image was obtained in the following way: From the initial core images of the candidate core particles under multiple light sources, an initial core image under a target light source is selected; wherein, in the initial core image under the target light source, the foreground is gray and the background is white; The pixel values ​​of the foreground and background in the selected initial core image are flipped, and the gaps in the flipped initial core image are filled to obtain the target core image of the candidate core under the target light source. Based on the grayscale histogram of the target core image, target segmentation thresholds for the foreground and background are obtained; wherein, the target segmentation thresholds maximize the grayscale difference between the foreground and background and minimize the grayscale difference within the foreground and background. Based on the segmentation threshold, the target core image is binarized to obtain a binary image of the candidate core.

6. The method as described in claim 5, characterized in that, The step of calculating the core area of ​​the core region based on the binary image of the candidate core under the target light source, and selecting the candidate core as the target core when the core area meets a preset anomaly condition, includes: From the binary image of the candidate core particle under the target light source, extract the connected regions with a pixel value of 255, and use the connected regions as the core particle regions; The core area is obtained by counting the number of pixels within the core region; When the core particle area is greater than a preset multiple of the preset core particle area, the candidate core particle is taken as the target core particle.

7. The method as described in claim 5, characterized in that, When the area of ​​the central hole in the binary image is greater than a preset area threshold, the method further includes: Extract the first maximum connected region with a pixel value of 255 from the binary image, and flip the pixel values ​​within the first maximum connected region to 0; Extract the second maximum connected region from the flipped binary image, set the pixel value within the second maximum connected region to 255, and obtain a new binary image.

8. A core defect detection device, characterized in that, It includes a processor, a memory, and a communication interface, wherein the communication interface, the memory, and the processor are connected via a bus; The communication interface is used to acquire images; The memory stores a computer program, and the processor performs the following operations according to the computer program: Obtain the first detection result of each of the multiple chips contained in the wafer, and take the chips whose first detection result is initially normal as candidate chips; Based on the binary image of the candidate core particle under the target light source, the core particle area of ​​the core particle region is calculated, and when the core particle area meets the preset abnormal conditions, the candidate core particle is taken as the target core particle. The coordinates of the center point of the detection box of the target particle in the first detection result are compared with the coordinates of the image origin of the target particle; wherein, the image origin is the top left corner vertex of the image; Based on the comparison results, the direction of the core particle crossing the boundary is determined, and based on the direction of the core particle crossing the boundary, the four boundary regions of interest in the binary image of the target core particle are reduced; the boundary regions of interest are preset according to the core particle model. When the adjusted boundary core area of ​​the boundary interest region meets the preset area ratio, the first detection result of the target core is modified to core aggregation defect.

9. The core defect detection device as described in claim 8, characterized in that, After modifying the first detection result of the target core particle to a core particle aggregation defect, the processor also executes: The second detection result of each of the multiple chips contained in the wafer is obtained, and the first detection result and the second detection result of each chip are fused together to obtain the target detection result; The multiple core particles are arranged in an orderly manner according to certain rules, and at least one core particle with a target detection result of core particle polymerization defect is selected. The selected at least one core particle is used as the core particle to be analyzed. For each of the particles to be analyzed, the target detection result of the particle to be analyzed is obtained based on the gap between the particle to be analyzed and a plurality of adjacent particles.

Images