A PMI pin mark rejudgment method, device and system compatible with a light surface and a dot matrix pattern pad

Abstract

The application discloses a PMI needle mark rejudgment method, device and system compatible with light surface and dot matrix pattern PAD. The application belongs to the technical field of wafer detection. The method comprises the following steps: acquiring a PMI source picture to be analyzed, and performing image preprocessing on the PMI source picture to be analyzed; distinguishing the type of wafer PAD in the PMI source picture to be analyzed; if it is a light surface PAD, locating the light surface PAD region; if it is a dot matrix pattern PAD, calling a pre-stored dot matrix PAD template, and searching for the dot matrix pattern PAD region through template matching; identifying a shadow defect, and calculating the distance parameter of the shadow defect from the PAD frame; determining a needle mark rejudgment result based on the distance parameter and a preset distance threshold; when the rejudgment result is abnormal, sending a pause instruction to a host computer to control a Prober device to pause work, and outputting and pushing a rejudgment prompt and a picture comparison result. The application can simultaneously adapt to needle mark rejudgment of light surface PAD and dot matrix pattern PAD, and solves the problems of PAD positioning deviation and rejudgment misjudgment caused by dot matrix texture interference.

Description

Technical Field

[0001] This application belongs to the field of wafer inspection technology, specifically relating to a PMI needle mark re-judgment method, device and system compatible with both smooth and dot matrix pattern PADs. Background Technology

[0002] In the semiconductor wafer testing process, the Prober probe equipment uses the Prober Indentation Inspection (PMI) tool to check the quality of the probe indentations on the wafer pad surface, identifying the positional relationship between shadow defects within the pad area and the pad edge, thereby determining whether the indentation is acceptable. Existing indentation re-inspection solutions typically develop recognition algorithms for glossy pads, capable of directly locating the glossy pad area and calculating the distance to shadow defects.

[0003] With the iteration of semiconductor products, some wafers adopt a dot matrix pattern PAD structure. The PAD surface is covered with dot matrix textures. Traditional algorithms that directly identify PAD areas are easily interfered with by the dot matrix textures, making it difficult to accurately define the effective area of ​​the PAD. If the PAD area positioning is inaccurate, it will directly cause the calculation of the distance parameter of the shadow defect to be distorted, resulting in misjudgment of PMI needle mark re-judgment, and problems such as failure to judge good products or missed detection of defects.

[0004] Currently, there is a lack of offline PMI needle mark verification methods that can simultaneously support both glossy PADs and dot matrix pattern PADs. Some solutions only support single PAD type recognition, and for dot matrix PADs, manual image verification is required, increasing labor costs and resulting in low verification efficiency, which is difficult to meet the real-time offline verification requirements of Prober equipment on the production line. Summary of the Invention

[0005] This application provides a PMI needle mark re-judgment method, apparatus, and system compatible with glossy and dot matrix pattern PADs. The purpose is to solve the technical problems that existing PMI needle mark re-judgment schemes are difficult to simultaneously recognize glossy PADs and dot matrix pattern PADs. Dot matrix PADs are easily affected by texture interference, causing regional positioning deviations, which leads to re-judgment errors, the need for manual verification, and low production line re-judgment efficiency. The application aims to achieve automated offline needle mark re-judgment of both types of PADs.

[0006] In a first aspect, embodiments of this application provide a PMI (Personal Instrument Detection) method for re-judging pin marks on PADs compatible with both glossy and dot-matrix patterned PADs, the method comprising: S1, monitor the PMI failure image storage directory corresponding to the Prober device. When a new image file is added to the directory, obtain the PMI source image to be parsed and perform image preprocessing on the PMI source image to be parsed. S2, distinguish the type of wafer PAD in the PMI source image to be analyzed, the type includes smooth PAD and dot matrix pattern PAD; S3, if it is a glossy PAD, then locate the glossy PAD area; if it is a dot matrix pattern PAD, then call the pre-stored dot matrix PAD template and obtain the dot matrix pattern PAD area through template matching. S4. Identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. S5, determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold; S6, when the re-judgment result is abnormal, a pause command is sent to the host to control the Prober device to pause the operation, and the re-judgment prompt and image comparison result are output and pushed.

[0007] Furthermore, the step of calling a pre-stored dot matrix PAD template and obtaining the dot matrix pattern PAD area through template matching includes: Read the dot matrix PAD template corresponding to the current product model, and perform image preprocessing on the dot matrix PAD template; The preprocessed dot matrix PAD template is slid-searched on the preprocessed PMI source image to be parsed, and the image matching score corresponding to each search position is calculated to obtain several candidate PAD regions. Based on the matching score, the valid region is selected from the candidate PAD region, and the dot pattern PAD region is defined.

[0008] Furthermore, the sliding search is a top-to-bottom and left-to-right traversal sliding search; When multiple candidate PAD regions with the same maximum matching score are retrieved, the candidate PAD region in the top left corner that was retrieved first is selected.

[0009] Furthermore, the step of filtering valid regions from candidate PAD regions based on matching scores includes: The candidate PAD regions are deduplicated, overlapping regions are removed, and the candidate region with the highest matching score is retained. When the matching score of a candidate region is greater than a preset matching threshold, the candidate region is determined as a dot matrix pattern PAD region; wherein, the preset matching threshold is 0.6.

[0010] Furthermore, the distance parameters include: The area of ​​the shadow defect, the pixel distance of the shadow defect from the top, bottom, left, and right sides of the PAD, and the minimum distance of the shadow defect from the PAD border; Correspondingly, if the minimum distance between the shadow defect and the PAD border is less than the preset distance threshold, the re-judgment result is FAIL; if the minimum distance between the shadow defect and the PAD border is greater than or equal to the preset distance threshold, the re-judgment result is PASS.

[0011] Furthermore, the acquisition of the PMI source image to be parsed described in S1 includes: An independent service program collects PMI failure source images in real time from the ImgSrc directory on the Prober device. Each Prober device corresponds to an independent re-judgment service instance, and each re-judgment service instance listens to the image storage directory of the corresponding Prober device.

[0012] Furthermore, S6 includes: Send the RCMDpause command to the host, and the host will forward the pause command to the Prober device; Push notifications for review and image comparison results include sending a notification message containing image comparison information to a designated recipient.

[0013] Furthermore, the dot matrix PAD templates are pre-classified and stored according to product model and device platform, and different device platforms are configured with corresponding dot matrix PAD template sets.

[0014] Secondly, embodiments of this application provide a PMI needle mark re-judgment device compatible with both glossy and dot-matrix patterned PADs, the device comprising: The image monitoring and preprocessing module is used to monitor the directory where PMI failed images are stored for the Prober device. When a new image file is added to the directory, the PMI source image to be parsed is obtained, and the image preprocessing is performed on the PMI source image to be parsed. The PAD type identification and positioning module is used to distinguish the type of wafer PAD in the PMI source image to be parsed. The types include smooth PAD and dot matrix pattern PAD. If it is a smooth PAD, the smooth PAD area is located. If it is a dot matrix pattern PAD, a pre-stored dot matrix PAD template is called and the dot matrix pattern PAD area is obtained by template matching. The shadow defect calculation module is used to identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. The needle mark re-judgment module is used to determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold. The business response execution module is used to send a pause command to the host to control the Prober device to pause its operation when the re-judgment result is abnormal, and to output and push the re-judgment prompt and image comparison result.

[0015] Thirdly, embodiments of this application provide a PMI pin mark re-judgment system compatible with glossy and dot matrix pattern PADs, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the PMI pin mark re-judgment method compatible with glossy and dot matrix pattern PADs as described above.

[0016] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0017] The technical solution provided in this application acquires the PMI source image to be parsed by monitoring the image directory and performs preprocessing. It distinguishes between two PAD types: glossy PADs and dot matrix pattern PADs, and performs region positioning for each. Then, it calculates the distance parameter between the shadow defect and the PAD border for re-judgment. This solution can simultaneously adapt to needle mark re-judgment scenarios with two different PAD structures. It avoids the PAD positioning deviation problem caused by dot matrix textures, reduces distance parameter calculation distortion, and lowers the risk of re-judgment misjudgment. When re-judgment is abnormal, it automatically issues a pause command and pushes the result, eliminating the manual image review step and improving the automation level and operational efficiency of offline PMI needle mark re-judgment on the production line. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the PMI needle mark re-judgment method for compatible glossy and dot matrix patterned PADs provided in this application embodiment; Figure 2 This is a schematic diagram of the overall business process for PMI image offline review provided in this embodiment; Figure 3 This is a schematic diagram of a microscopic image of a smooth PAD wafer provided in this embodiment; Figure 4 This is a schematic diagram of an example image of a pin mark defect on a smooth PAD provided in this embodiment; Figure 5 This is a schematic diagram of a dot matrix patterned PAD wafer micrograph provided in this embodiment; Figure 6 This is a schematic diagram illustrating the annotation of the FAIL result for the smooth PAD re-judgment provided in this embodiment; Figure 7 This is a schematic diagram illustrating the PASS result annotation of the smooth PAD in this embodiment; Figure 8 This is another schematic diagram illustrating the annotation of the FAIL result for the smooth PAD re-judgment provided in this embodiment; Figure 9 This is a schematic diagram of the AP3000E platform dot matrix PAD template storage directory interface provided in this embodiment; Figure 10 This is a schematic diagram of the OPUS platform dot matrix PAD template storage directory interface provided in this embodiment; Figure 11 This is a schematic diagram illustrating the annotation of the FAIL result of the dot matrix PAD re-judgment on the OPUS platform provided in this embodiment; Figure 12This is a schematic diagram of the OPUS platform dot matrix PAD template file directory interface provided in this embodiment; Figure 13 This is a schematic diagram illustrating the annotation of FAIL results for another OPUS platform dot matrix PAD re-judgment provided in this embodiment; Figure 14 This is a schematic diagram of the AP3000E platform dot matrix PAD re-judgment PASS result provided in this embodiment; Figure 15 This is a schematic diagram of the AP3000E platform dot matrix PAD template file directory interface provided in this embodiment; Figure 16 This is a schematic diagram of the structure of the PMI needle mark detection device that is compatible with both glossy and dot matrix patterned PADs provided in the embodiments of this application. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but may also have additional steps not included in the drawings. The process can correspond to a method, function, procedure, subroutine, subroutine, etc.

[0020] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0021] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0022] The following description, in conjunction with the accompanying drawings, details the PMI needle mark re-judgment method, apparatus, equipment, and medium for compatible glossy and dot matrix patterned PADs provided in this application, through specific embodiments and application scenarios.

[0023] Example 1 Figure 1 This is a flowchart illustrating the PMI needle mark re-judgment method for compatible glossy and dot matrix patterned PADs provided in Embodiment 1 of this application. Figure 1 As shown, the specific steps include the following: S1, monitor the PMI failure image storage directory corresponding to the Prober device. When a new image file is added to the directory, obtain the PMI source image to be parsed and perform image preprocessing on the PMI source image to be parsed. Prober equipment can be a probe station device used in semiconductor testing scenarios to perform wafer probe testing. This device can perform PMI pin mark inspection, acquire corresponding images after the probe presses against the wafer pads, and output the original image of the pin mark failure.

[0024] The directory for storing PMI failure images can be a specified folder path deployed on the host side to receive pin mark inspection failure images returned by Prober devices. Different models of production line devices can be configured with their own independent directory paths, such as E:\CP\ProberPmi\ImgSrc. This directory continuously receives raw image files synchronized from the devices.

[0025] This solution allows the background service program to continuously detect file additions and modifications within the directory by using the operating system's file event subscription mechanism or a timed polling scan. It does not require manual triggering of the scan and can capture newly generated image files in real time.

[0026] The PMI source image to be parsed can be the original image file that was marked as failed after the Prober device completed the PMI pin mark inspection. The file carries complete original image information such as probe indentations, pad textures, and wafer circuits, without secondary image processing such as cropping or filter modification.

[0027] Specifically, after capturing the event of adding a new file to the directory, the file extension and file integrity can be verified. The file writing process can be completed to avoid reading corrupted files. Then, the binary data stream of the newly added image file can be read to complete the file loading and obtain the PMI source image to be parsed.

[0028] Image preprocessing can involve the review service program performing format conversion and decompression operations on the read original image files, converting the special packaged files output by the device into a standard image format that the image algorithm can load and parse, while also performing basic noise reduction to eliminate noise interference caused by device transmission, thus providing qualified input materials for subsequent image recognition operations.

[0029] The re-evaluation service program can continuously monitor the directory where PMI failure images are stored for the Prober device. Once a complete new image file is detected in the directory, the file status is verified, the file is read to obtain the PMI source image to be parsed, and image preprocessing operations are performed on the PMI source image to be parsed.

[0030] S2, distinguish the type of wafer PAD in the PMI source image to be analyzed, the type includes smooth PAD and dot matrix pattern PAD; A wafer pad is a metal pad on the wafer surface used for probe bonding contact. It is the contact point for probe testing. PMI pin mark inspection mainly detects whether the shadow defects left by the probe at this location are compliant.

[0031] A smooth pad can be a pad with a flat metal surface, without a periodic dot matrix texture, and the grayscale of the pad area in the image is uniform without a large number of discrete dot matrix patterns.

[0032] A dot matrix pattern PAD can be a pad structure with a large number of dot matrix textures pre-made on the surface of the pad. The pad area is covered with dense dots, and the image will have a large number of alternating bright and dark dots, which will interfere with ordinary edge recognition algorithms.

[0033] Specifically, the product process configuration information corresponding to the current image can be retrieved, and the pad attribute tags pre-bound to the product model can be used to determine whether the pads in the current image are glossy PADs or dot matrix pattern PADs. Image texture features can also be used to assist in type identification.

[0034] S3, if it is a glossy PAD, then locate the glossy PAD area; if it is a dot matrix pattern PAD, then call the pre-stored dot matrix PAD template and obtain the dot matrix pattern PAD area through template matching. The glossy PAD area can be the entire pixel range occupied by the glossy pad in the image, including the complete four boundaries of the pad. Subsequent defect calculations are all limited to this pixel range.

[0035] Localization can be achieved by using image algorithms to identify the contour boundary of grayscale transitions between the pad metal and the surrounding circuit medium, selecting the complete pad pixel range through contour closure operations, and outputting the coordinates of the pad rectangular boundary.

[0036] The dot matrix PAD template can be a standard reference image that has been pre-made and stored on the local disk. The image content is an image of the standard dot matrix pattern PAD of the product model, which is used to compare similarity with the image to be parsed.

[0037] Specifically, the re-judgment service program accesses the storage path to read the corresponding dot matrix PAD template file based on the product model index corresponding to the currently processed image, loads the template image data into the algorithm memory, and uses it for subsequent matching operations.

[0038] Template matching retrieval can be an image similarity comparison algorithm that uses a template image as a comparison benchmark, traverses within a large source image to be parsed to search for image sub-regions with high similarity, and outputs the matching score for each sub-region.

[0039] For example, if the PAD is identified as a glossy surface, the edge detection algorithm is used to locate the glossy surface PAD area; if the PAD is identified as a dot matrix pattern PAD, the pre-stored dot matrix PAD template is read, the template matching retrieval operation is performed, and the pixel position of the dot matrix pattern PAD area is output.

[0040] S4. Identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. Shadow defects can be dark spots left by probes pressing into the PAD surface. In grayscale images, they appear as a set of pixels with a grayscale value significantly lower than that of the surrounding pad surface. These are the target defects that need to be evaluated in PMI probe mark inspection.

[0041] Image algorithms can be used to segment dark spot defect pixels from the pad background within the predefined PAD pixel area using grayscale thresholding, extracting the defect's contour pixel set. Further, the coordinates of all pixels on the shadow defect contour are extracted, and the pixel distances from the defect pixels to the top, bottom, left, and right borders of the PAD are calculated, resulting in a set of quantized values.

[0042] Among them, the distance parameter can be a set of quantitative values ​​that characterize the positional relationship between the shadow defect and the PAD border, serving as the original calculation basis for subsequent re-judgment.

[0043] The proposed image re-judgment algorithm is limited to performing grayscale segmentation to identify shadow defects within the already located PAD area, traversing the defect contour pixels, and calculating the distance parameters between the shadow defect and each border of the PAD.

[0044] S5, determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold; The preset distance threshold can be a process boundary value pre-configured in the production line process, representing the minimum safe pixel distance between the defect and the pad border. Different product models can be configured with different thresholds.

[0045] Specifically, it can read a preset distance threshold, compare the distance parameter with the threshold, and output a PASS or FAIL result for needle mark re-judgment.

[0046] S6, when the re-judgment result is abnormal, a pause command is sent to the host to control the Prober device to pause operation, and a re-judgment prompt and image comparison result are pushed.

[0047] A pause command can be a control command message that can be parsed and recognized by the host and used to notify the Prober device to stop continuing the probe testing operation.

[0048] The host can be a host machine, which is the upper host of the production line EAP system. It is responsible for the functions of instruction relay and message forwarding, and connects the review service program and the Prober device.

[0049] Then, all the original data and judgment results of this complete review can be printed and recorded in the local logs and business interface. The message push interface can be called to send the images and judgment information to the specified receiving channel, such as email or production line alarm message client.

[0050] Image comparison results can be image annotation files containing the PAD selection area, shadow defect markers, various distance parameters, and corresponding numerical reports.

[0051] When the needle mark re-examination result is determined to be abnormal, the re-examination service program sends a pause command to the host. The host forwards the command to control the Prober device to pause the operation, while outputting the re-examination record locally and pushing the re-examination prompt and image comparison results outward.

[0052] This technical solution fully realizes the entire process of image acquisition, image preprocessing, pad type differentiation, separate positioning of the two types of pads, defect parameter calculation, re-judgment, and abnormal business handling. It can simultaneously adapt to offline re-judgment of PMI pin marks for two different pad structures, namely glossy PADs and dot matrix pattern PADs, solving the recognition and adaptation problem caused by dot matrix texture interference, and realizing automated re-judgment on the production line without the need for manual intervention in the basic judgment process.

[0053] In one embodiment, optionally, the step of calling a pre-stored dot matrix PAD template and retrieving the dot matrix pattern PAD area through template matching includes: Read the dot matrix PAD template corresponding to the current product model, and perform image preprocessing on the dot matrix PAD template; The preprocessed dot matrix PAD template is used for sliding search on the preprocessed PMI source image to be parsed. The image matching score corresponding to each search position is calculated to obtain several candidate PAD regions. Based on the matching score, the effective region is selected from the candidate PAD regions to define the dot matrix pattern PAD region.

[0054] The current product model can be the product model identifier corresponding to the wafer currently being processed. Different products correspond to different dot matrix PAD shapes and dot matrix layouts. Each product model is equipped with corresponding dot matrix PAD template resources.

[0055] Specifically, based on the product model field carried in the current business process, the template storage path can be retrieved, and the dot matrix PAD template image file corresponding to that model can be loaded into the algorithm's memory.

[0056] Image preprocessing for dot matrix PAD templates can perform noise reduction and grayscale normalization on the template image, ensuring that the image format and grayscale range of the template image are consistent with the PMI source image to be parsed, thus eliminating the problem of matching score distortion caused by image format differences.

[0057] Sliding retrieval can be a template matching algorithm that uses the size of the template image as the sliding window size, continuously moving the window position on the PMI source image to be parsed, and completing a similarity comparison operation each time the window moves.

[0058] Image matching score is a quantitative value that represents the similarity between the template image and the sub-images in the current window. The higher the score, the higher the similarity between the image content in the window and the standard dot matrix PAD template.

[0059] Candidate PAD regions can be image sub-regions corresponding to each window position during the sliding search process. All window positions participating in the comparison operation will generate corresponding candidate PAD regions and corresponding matching scores.

[0060] Filtering valid regions can be achieved by business logic filtering and discarding all candidate PAD regions based on their matching scores, eliminating invalid candidate positions, retaining regions that meet the criteria, outputting the pixel coordinates of the valid regions, and using rectangular boundaries to mark the positions of the dot pattern PADs in the source image.

[0061] The re-judgment service program reads the dot matrix PAD template corresponding to the current product model and performs image preprocessing on the template. The preprocessed template is then used to perform a sliding search on the PMI source image to be parsed, and the image matching score is calculated for each position to generate several candidate PAD regions. The candidate PAD regions are then selected as valid regions based on the matching scores to define the dot matrix pattern PAD regions.

[0062] This technical solution breaks down and refines the template matching retrieval process for dot matrix PADs, completing template reading preprocessing, sliding retrieval to generate candidate regions, and filtering to obtain valid PAD regions, forming a complete and implementable template matching execution chain. This provides a specific execution path for the accurate positioning of dot matrix pattern PADs, avoiding the ambiguity caused by directly and generally calling template matching.

[0063] In one embodiment, optionally, the sliding search is a top-to-bottom and left-to-right traversal sliding search; when multiple candidate PAD regions with the same maximum matching score are retrieved, the candidate PAD region at the top left corner position that was retrieved first is selected.

[0064] The top-to-bottom and left-to-right traversal sliding search can be the scanning order of the sliding window. The window starts from the top left corner of the image and moves horizontally from left to right pixel by pixel. After scanning one row, it jumps down and continues to scan the next row from left to right, completely traversing all the positions where windows can be placed in the entire PMI source image to be analyzed, without missing any sub-regions in the image.

[0065] Multiple candidate PAD regions with the same maximum matching score can be caused by multiple image sub-frames and dot matrix PAD templates in the image having the highest similarity. Multiple candidate regions can obtain completely consistent maximum matching scores, which can lead to multiple solutions in the algorithm.

[0066] The candidate PAD region in the top left corner that is first retrieved can be scanned from top to bottom and from left to right. The first window in the image that shows the maximum matching score during the scanning process corresponds to the image region that is relatively closer to the top left corner of the image in the overall image space.

[0067] This technical solution clarifies the scanning and traversal order of sliding search, solves the problem of uncertain algorithm output results caused by multiple regions being assigned the same region, provides definite candidate region selection rules, ensures that the algorithm output results are unique under the same set of image input conditions, eliminates the randomness of algorithm results, and improves the reproducibility of production line review results.

[0068] In one embodiment, optionally, filtering valid regions from candidate PAD regions based on matching scores includes: The candidate PAD regions are deduplicated, overlapping regions are removed, and the candidate region with the highest matching score is retained. When the matching score of a candidate region is greater than a preset matching threshold, the candidate region is determined as a dot matrix pattern PAD region; wherein, the preset matching threshold is 0.6.

[0069] Deduplication can be achieved by comparing the rectangular coordinates of each candidate PAD region with an algorithm to determine whether the pixel ranges of different candidate regions overlap spatially. For candidate regions that overlap spatially, a removal operation is performed, and only one candidate output is retained for each real PAD pad, thus avoiding the identification of multiple sets of bounding box coordinates for the same physical pad.

[0070] Eliminating overlapping regions can be done by identifying overlapping regions when the pixel ranges of two candidate PAD regions intersect, discarding the overlapping candidate positions with relatively lower matching scores, and only retaining the candidates with better quality.

[0071] Retaining the candidate region with the highest matching score can be done by prioritizing the candidate position with the highest similarity score among non-overlapping candidate sets, which represents the sub-image that best matches the standard template.

[0072] The preset matching threshold can be a manually set similarity judgment threshold. In this scheme, it is set to 0.6. The matching score can range from 0 to 1. The closer the score is to 1, the higher the similarity.

[0073] This technical solution adds candidate region deduplication and overlap removal logic to avoid redundant calculations caused by multiple sets of bounding box coordinates output for a single pad. At the same time, it sets a quantization matching threshold to filter out mismatched regions with low similarity, reducing the defect of background circuit textures being incorrectly identified as dot matrix PADs and improving the accuracy of dot matrix PAD region positioning.

[0074] In one embodiment, optionally, the distance parameters include: the area of ​​the shadow defect, the pixel distance of the shadow defect from the top, bottom, left, and right sides of the PAD, and the minimum distance of the shadow defect from the PAD border. Correspondingly, if the minimum distance between the shadow defect and the PAD border is less than the preset distance threshold, the re-judgment result is FAIL; if the minimum distance between the shadow defect and the PAD border is greater than or equal to the preset distance threshold, the re-judgment result is PASS.

[0075] The area of ​​a shadow defect can be the total number of pixels contained within the outline of the shadow defect, used to characterize the size of the probe indentation defect itself.

[0076] The pixel distance of the shadow defect from the top, bottom, left, and right sides of the PAD can be the pixel length of the defect outline pixel from the four borders of the PAD, using pixels as the unit of measurement within the image.

[0077] The minimum distance between the shadow defect and the PAD border can be the minimum of four sets of distance values ​​from the shadow defect to the four borders of the PAD, representing the position of the defect closest to the pad edge.

[0078] FAIL indicates a failed re-inspection, meaning the pin mark defect poses a high risk of touching the pad edge, and the pin mark inspection in this image fails the re-inspection. PASS indicates a passed re-inspection, meaning the shadow defect has sufficient safety margin from the pad edge, and the pin mark re-inspection passes.

[0079] When the distance parameter includes the area of ​​the shadow defect, the pixel distance of the shadow defect from the top, bottom, left and right sides of the PAD, and the minimum distance of the shadow defect from the PAD border, the minimum distance of the shadow defect from the PAD border is compared with a preset distance threshold. If it is less than the threshold, the re-judgment result is FAIL; if it is greater than or equal to the threshold, the re-judgment result is PASS.

[0080] This technical solution fully defines the complete set of distance parameters on which the re-judgment depends, clarifies the executable Boolean judgment logic, transforms the process requirements into numerical comparison rules that the machine can calculate, and outputs standardized PASS / FAIL judgment labels to facilitate the production line system to analyze and identify the re-judgment results.

[0081] In one embodiment, optionally, obtaining the PMI source image to be parsed in S1 includes: A separate service program collects PMI failure source images in real time from the ImgSrc directory on the Prober device. Each Prober device corresponds to an independent review service instance, and each review service instance listens to the image storage directory of the corresponding Prober device.

[0082] An independent service program can be a standalone background acquisition program that is decoupled from the image parsing service. It is dedicated to the acquisition and transmission of images from the device to the host and does not participate in image parsing operations.

[0083] The ImgSrc directory can be a local folder on the Prober device used to store PMI needle mark inspection output images.

[0084] Real-time data collection can be achieved by continuously connecting the collection service to the Prober device. Whenever the device generates a PMI failure source image, the collection service will promptly transmit the image to the image storage directory on the host side.

[0085] The review service instance can be a process instance of the review application running. Data isolation and directory monitoring tasks are independent between different instances.

[0086] The pre-processing of acquiring PMI source images to be parsed by S1 is completed by an independent service program. The independent service program collects PMI failure source images in real time from the ImgSrc directory on the Prober device. On the production line, each Prober device is assigned an independent re-judgment service instance, and each instance only listens to the image storage directory corresponding to its own Prober device.

[0087] This technical solution decouples image acquisition and image parsing operations. At the same time, each device corresponds to an independent service instance, and the tasks of different devices do not interfere with each other. An anomaly of a single device will not affect the re-judgment operations of other probe stations, thereby improving the stability of the entire system and adapting to scenarios where multiple devices operate in parallel on a production line.

[0088] In one embodiment, optionally, S6 includes: Send the RCMDpause command to the host, and the host will forward the pause command to the Prober device; Push notifications for review and image comparison results include sending a notification message containing image comparison information to a designated recipient.

[0089] The host can be a host on the production line, belonging to the host node of the EAP equipment automation platform, and is responsible for relaying equipment commands.

[0090] The RCMDpause command can be a remote equipment control command defined by the production line EAP system. RCMD stands for RemoteCommand, and pause means to pause. This command is used to notify the Prober equipment to stop the current test process.

[0091] After receiving the RCMDpause command from the review service, the host re-encapsulates the message according to the device communication protocol and forwards it to the corresponding Prober device.

[0092] The designated recipient can be a pre-configured email account of an engineer, a production line alarm client account, or other receiving entity.

[0093] The image comparison information prompt message can include the annotated image, various distance parameters, and judgment result text, allowing personnel to directly view the re-judgment details.

[0094] This technical solution clarifies the specific command type and command forwarding link for equipment suspension, and also clarifies the content carried in the alarm push message, ensuring that the equipment can be reliably stopped when an anomaly occurs, and that relevant personnel can obtain complete images and data as soon as possible, facilitating the production line personnel to quickly handle the anomaly.

[0095] In one embodiment, optionally, the dot matrix PAD templates are pre-classified and stored according to product model and device platform, and different device platforms are configured with corresponding dot matrix PAD template sets.

[0096] The device platform can be the hardware platform model of the Prober device, such as the OPUS platform or the AP3000E platform. Different device platforms have differences in the lighting and resolution of image imaging, and the same pad will output different images on different platforms.

[0097] Categorized storage allows you to create multi-level folders under the disk storage path. The first level distinguishes between device platforms, and the second level distinguishes between product models. Template files for different products and platforms are placed in the corresponding directories.

[0098] A dot matrix PAD template collection can be a collection of multiple standard reference template images for a specific product model or device platform.

[0099] Specifically, the dot matrix PAD templates are pre-made before being put into operation and are classified and stored according to two dimensions: product model and device platform; each different device platform is configured with a set of dot matrix PAD templates adapted to its own imaging characteristics.

[0100] This technical solution manages templates from two dimensions: product model and equipment platform. This avoids the template matching failure problem caused by imaging differences between different devices. When calling a template, it can quickly locate the matching template resource, further improving the recognition accuracy of template matching and adapting to the mixed production line conditions of multiple platforms and multiple products.

[0101] To enable those skilled in the art to better understand this solution, this application also provides a preferred embodiment.

[0102] This preferred embodiment discloses an offline re-evaluation scheme for PMI (Probe Mark Inspection) pads compatible with both glossy and dot matrix patterned pads. This scheme is deployed on the production line host computer and interfaces with Prober testing equipment to automate offline re-evaluation of wafer PMI failure images. It can simultaneously process pad detection images for both glossy and dot matrix patterned pads. The overall workflow is detailed in the appendix. Figure 2 (Flowchart of PMI image offline review process).

[0103] System deployment architecture: Each Prober probe device on the production line is configured with an independent PMIMG re-evaluation server application. After completing the PMI needle mark inspection, the Prober device saves the images of the failed PMI sources to the device's local ImgSrc directory. An independent acquisition service program collects PMI failure source images from the ImgSrc directory of each Prober device in real time and synchronously transmits them to the corresponding PMIMG failure image storage directory on the host side, such as E:\CP\ProberPmi\OPUS-201\ImgSrc. Each Prober device has its own dedicated monitoring directory, and the re-evaluation service instance independently monitors file changes in its corresponding directory without interference.

[0104] When the directory detects a newly added image file of a specific format, the server reads the image, performs conversion and decompression on the original image, completes image preprocessing, and initiates the image parsing process. After parsing, business processing is differentiated based on the parsing results: if there are no valid pad markers, the process ends directly; if there are valid markers, the host sends an RCMDpause command to the Prober device and simultaneously pushes an email alert to designated personnel, containing image comparison information, shadow distance from the border, etc. The Prober device returns the RCMD command processing status after receiving the command.

[0105] There are two types of PAD pad structures on the wafer, see appendix. Figure 3 - Appendix Figure 5 One type is the smooth PAD, where the pad surface is flat and has no dot matrix texture; the other type is the dot matrix pattern PAD, where the pad surface is covered with periodic dot matrix texture.

[0106] Implementation steps: Step 1: Image Acquisition and Preprocessing; The review service continuously monitors the directory where PMI failure images are stored on the corresponding Prober device. The acquisition service has already synchronized the PMI failure images from the ImgSrc directory on the Prober device to this directory. Once a new image file is detected in the directory, the service waits for the file to finish writing, reads the original image file, performs conversion and decompression, and converts it into an image format that can be processed by the image algorithm to obtain the preprocessed PMI source image to be parsed.

[0107] Step 2, PAD type differentiation; Based on the product model configuration information corresponding to the current wafer, identify whether the PAD in the PMI source image to be analyzed is a smooth PAD or a dot matrix pattern PAD, and carry out PAD area localization in two branches.

[0108] Step 3, PAD area location; Branch 1: Glossy PAD Localization. For glossy PADs, conventional image edge detection algorithms are directly used to identify the pad outlines and define the pixel area of ​​the glossy PAD. See the attached image for the glossy PAD recognition results. Figure 6 Appendix Figure 7 Appendix Figure 8 .

[0109] Within the selected glossy PAD area, identify shadow defects formed by probe indentations, and calculate the area of ​​the shadow defect, the pixel distance of the shadow from the top, bottom, left, and right edges of the PAD, and the minimum distance of the shadow from the PAD border. Compare the minimum distance with a preset threshold, for example, 10 pixels. If the minimum distance is less than the threshold, the result is FAIL; otherwise, it is PASS.

[0110] Branch Two: PAD Positioning Based on Dot Matrix Patterns. The surface of a PAD with a dense dot matrix texture is susceptible to interference from ordinary edge detection algorithms. Therefore, a template matching method is used to locate the PAD. Dot matrix PAD templates are pre-created for different product models and device platforms. The OPUS platform and AP3000E platform maintain their own template sets. Template files are stored in two-level folders: "Device Platform - Product Model". Refer to the attached template storage interface. Figure 9 Appendix Figure 10 .

[0111] When processing dot matrix pattern PAD images, the corresponding dot matrix PAD template is read based on the current product model and device platform. Image preprocessing is performed on the template to unify grayscale and image format. The template is used as a sliding window, and a top-to-bottom, left-to-right traversal sliding search is performed on the preprocessed PMI source image to be parsed. At each sliding position, the image matching score between the template and the window sub-image is calculated, generating a large number of candidate PAD regions. If multiple candidate regions have the same maximum matching score, the candidate region that appears first in the scanning order and is located in the upper left corner is selected. All candidate PAD regions are deduplicated, eliminating spatially overlapping regions and retaining the candidate region with the highest matching score. When the matching score of a candidate region is greater than 0.6, that region is officially defined as a dot matrix pattern PAD region.

[0112] Reference for OPUS platform dot matrix PAD resolution effect Figure 11 Appendix Figure 13 Reference for AP3000E platform dot matrix PAD resolution effect. Figure 14 The corresponding template file is attached. Figure 12 Appendix Figure 15 .

[0113] After defining the PAD area of ​​the dot matrix pattern, the shadow defects of the probe indentation are identified inside the PAD area. The area of ​​the shadow defect, the distance of the shadow from the four sides of the PAD and the minimum distance of the shadow from the PAD border are calculated. The minimum distance is compared with the process preset threshold (e.g., 15px). If it is less than the threshold, it is judged as FAIL. If it is greater than or equal to the threshold, it is judged as PASS.

[0114] Step 4, handling abnormal business operations; When the reassessment result is FAIL (unclear), the host sends an RCMDpause command to the Prober device to pause the testing operation. Simultaneously, a notification message is sent to designated personnel, including an annotated PMI image, various distance parameters, and the assessment result, completing one offline reassessment process.

[0115] The beneficial effects of this technical solution are as follows: This preferred embodiment is compatible with both glossy PADs and dot matrix pattern PADs. Glossy PADs rely on edge detection for rapid positioning; dot matrix PADs employ a platform- and model-specific template matching scheme, setting a 0.6 matching threshold, prioritizing the top left corner for identical scores, and eliminating overlapping areas in the candidate region to effectively avoid misidentification caused by dot matrix textures. The entire solution enables parallel automated offline review of multiple Prober devices on the production line, automatically stopping and sending alarms for anomalies, reducing the workload of manual image review, and improving the efficiency and reliability of PMI needle mark review on the production line.

[0116] Example 2 The second embodiment of this application provides a PMI needle mark detection device compatible with both glossy and dot matrix patterned PADs. Figure 16 This is a schematic diagram of the structure of the PMI needle mark detection device compatible with glossy and dot matrix patterned PADs provided in the embodiments of this application, as shown below. Figure 16 As shown, the device includes: The image monitoring and preprocessing module 1610 is used to monitor the PMI failure image storage directory corresponding to the Prober device. When a new image file is added to the directory, the PMI source image to be parsed is obtained, and the PMI source image to be parsed is preprocessed. PAD type identification and positioning module 1620 is used to distinguish the type of wafer PAD in the PMI source image to be parsed. The type includes glossy PAD and dot matrix pattern PAD. If it is a glossy PAD, the glossy PAD area is located. If it is a dot matrix pattern PAD, a pre-stored dot matrix PAD template is called and the dot matrix pattern PAD area is obtained by template matching. The shadow defect calculation module 1630 is used to identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. The needle mark re-judgment module 1640 is used to determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold. The business response execution module 1650 is used to send a pause command to the host to control the Prober device to pause operation when the re-judgment result is abnormal, and to output and push the re-judgment prompt and image comparison result.

[0117] The PMI needle mark re-judgment device compatible with glossy and dot matrix pattern PADs provided in this application embodiment can realize the various processes and corresponding beneficial effects of the above embodiments. To avoid repetition, it will not be described again here.

[0118] Example 3 This application also provides a PMI pin mark detection system compatible with both glossy and dot matrix pattern PADs, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the PMI pin mark detection method compatible with both glossy and dot matrix pattern PADs as described above.

[0119] It should be noted that the system provided in this application embodiment can perform the PMI needle mark re-judgment method for compatible glossy and dot matrix pattern PADs as described above, and has corresponding beneficial effects.

[0120] Example 4 This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described PMI needle mark re-judgment method embodiment for compatible glossy and dot matrix pattern PADs, and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0121] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element. Furthermore, it should be noted that the scope of the methods and systems in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.

[0122] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0123] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above, which are merely illustrative and not restrictive. Those skilled in the art, under the guidance of this application, can make many modifications without departing from the spirit and scope of the claims, all of which fall within the protection scope of this application.

[0124] The above description is merely a preferred embodiment and the technical principles employed in this application. This application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that can be made by those skilled in the art will not depart from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include more other equivalent embodiments without departing from the concept of this application, the scope of which is determined by the scope of the claims.

Claims

1. A PMI needle mark re-judgment method compatible with both glossy and dot-matrix patterned PADs, characterized in that, The method includes: S1, monitor the PMI failure image storage directory corresponding to the Prober device. When a new image file is added to the directory, obtain the PMI source image to be parsed and perform image preprocessing on the PMI source image to be parsed. S2, distinguish the type of wafer PAD in the PMI source image to be analyzed, the type includes smooth PAD and dot matrix pattern PAD; S3, if it is a glossy PAD, then locate the glossy PAD area; if it is a dot matrix pattern PAD, then call the pre-stored dot matrix PAD template and obtain the dot matrix pattern PAD area through template matching. S4. Identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. S5, determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold; S6, when the re-judgment result is abnormal, a pause command is sent to the host to control the Prober device to pause the operation, and the re-judgment prompt and image comparison result are output and pushed.

2. The method according to claim 1, characterized in that, The step of calling a pre-stored dot matrix PAD template and obtaining the dot matrix pattern PAD area through template matching includes: Read the dot matrix PAD template corresponding to the current product model, and perform image preprocessing on the dot matrix PAD template; The preprocessed dot matrix PAD template is slid-searched on the preprocessed PMI source image to be parsed, and the image matching score corresponding to each search position is calculated to obtain several candidate PAD regions. Based on the matching score, the valid region is selected from the candidate PAD region, and the dot pattern PAD region is defined.

3. The method according to claim 2, characterized in that, The sliding search is a top-to-bottom and left-to-right traversal sliding search. When multiple candidate PAD regions with the same maximum matching score are retrieved, the candidate PAD region in the top left corner that was retrieved first is selected.

4. The method according to claim 2, characterized in that, The process of filtering valid regions from candidate PAD regions based on matching scores includes: The candidate PAD regions are deduplicated, overlapping regions are removed, and the candidate region with the highest matching score is retained. When the matching score of a candidate region is greater than a preset matching threshold, the candidate region is determined as a dot matrix pattern PAD region; wherein, the preset matching threshold is 0.

6.

5. The method according to claim 1, characterized in that, The distance parameters include: The area of ​​the shadow defect, the pixel distance of the shadow defect from the top, bottom, left, and right sides of the PAD, and the minimum distance of the shadow defect from the PAD border; Correspondingly, if the minimum distance between the shadow defect and the PAD border is less than the preset distance threshold, the re-judgment result is FAIL; if the minimum distance between the shadow defect and the PAD border is greater than or equal to the preset distance threshold, the re-judgment result is PASS.

6. The method according to claim 1, characterized in that, The acquisition of the PMI source image to be parsed described in S1 includes: An independent service program collects PMI failure source images in real time from the ImgSrc directory on the Prober device. Each Prober device corresponds to an independent re-judgment service instance, and each re-judgment service instance listens to the image storage directory of the corresponding Prober device.

7. The method according to claim 1, characterized in that, S6 include: Send the RCMDpause command to the host, and the host will forward the pause command to the Prober device; Push notifications for review and image comparison results include sending a notification message containing image comparison information to a designated recipient.

8. The method according to claim 2, characterized in that, The dot matrix PAD templates are pre-categorized and stored according to product model and device platform, with each device platform configured with a corresponding set of dot matrix PAD templates.

9. A PMI needle mark detection device compatible with both glossy and dot-matrix patterned PADs, characterized in that, include: The image monitoring and preprocessing module is used to monitor the directory where PMI failed images are stored for the Prober device. When a new image file is added to the directory, the PMI source image to be parsed is obtained, and the image preprocessing is performed on the PMI source image to be parsed. The PAD type identification and positioning module is used to distinguish the type of wafer PAD in the PMI source image to be analyzed. The types include smooth PAD and dot matrix pattern PAD. If it is a glossy PAD, then locate the glossy PAD area; If it is a dot matrix pattern PAD, then the pre-stored dot matrix PAD template is called, and the dot matrix pattern PAD area is obtained by template matching and retrieval. The shadow defect calculation module is used to identify shadow defects within the located PAD area and calculate the distance parameters of the shadow defects from each border of the PAD. The needle mark re-judgment module is used to determine the needle mark re-judgment result based on the distance parameter and the preset distance threshold. The business response execution module is used to send a pause command to the host to control the Prober device to pause its operation when the re-judgment result is abnormal, and to output and push the re-judgment prompt and image comparison result.

10. A PMI needle mark detection system compatible with both glossy and dot-matrix patterned PADs, characterized in that, The method includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the PMI pin mark re-judgment method for compatible glossy and dot matrix pattern PADs as described in any one of claims 1-8.

Images