Processing methods and systems, electronic devices, computer-readable storage media and products

Abstract

This application provides a method and system for processing surface defects on silicon wafers, an electronic device, a computer-readable storage medium, and a product. The processing method includes: acquiring image information, whereby the image information characterizes defect features on the silicon wafer surface, and the defect features characterize the peak and valley differences on the silicon wafer surface within a preset unit area; processing the image information to convert it into a surface feature model, where the surface feature model includes grayscale values ​​corresponding to the defect features, with different defect features corresponding to different grayscale values; coloring the surface feature model to generate a color model, where the color model includes a preset set of colors, with different colors corresponding to different grayscale values; and outputting defect identification results based on the color model. In this way, the nanometer-level peak and valley differences on the silicon wafer surface are transformed from discrete numerical values ​​into a visualized color model, thereby solving the problem of not being able to intuitively present defect features and improving the efficiency and accuracy of defect identification.

Description

Technical Field

[0001] This application relates to the field of wafer processing, and more specifically, to a method and system for processing surface defects on silicon wafers, electronic devices, computer-readable storage media, and products. Background Technology

[0002] In the semiconductor manufacturing field, silicon wafers are the core basic material for integrated circuits, and their surface quality directly determines the performance and yield of subsequent chips. As chip manufacturing processes continue to advance towards the micron and nanometer scales, the requirements for the surface processing precision of silicon wafers are constantly increasing. Even slight deviations in front-end processing technology or equipment parameters can lead to undesirable features on the silicon wafer surface. If these are not identified and addressed in a timely manner, they will cause defects in batches of products and significantly increase production costs.

[0003] Currently, the industry primarily relies on metrology equipment to inspect the surface quality of silicon wafers. Its core function is to collect peak and valley data from the nanoscale regions on the front and back sides of the wafer, reflecting surface morphology characteristics numerically. However, this method cannot directly characterize the specific location, shape, and distribution range of defects on the silicon wafer surface, and it lacks an automated feedback mechanism. Measurement data must be manually processed and transmitted to the upstream processing stations and equipment operators, whereby technicians then develop adjustment strategies based on experience. This results in problems such as data transmission delays, untimely adjustment responses, and low accuracy in strategy matching, making it difficult to meet the high-precision, highly automated, and time-sensitive production requirements of semiconductor manufacturing. Summary of the Invention

[0004] The first aspect of this disclosure provides a method for processing surface defects on a silicon wafer. The method includes: acquiring image information, wherein the image information characterizes defect features on the surface of the silicon wafer, and the defect features characterize the difference information between peaks and valleys on the surface of the silicon wafer within a preset unit area; processing the image information to convert it into a surface feature model, wherein the surface feature model includes grayscale values ​​corresponding to the defect features, and different defect features correspond to different grayscale values; performing coloring processing on the surface feature model to generate a coloring model, wherein the coloring model includes a preset number of colors, and different colors correspond to different grayscale values; and outputting defect identification results based on the coloring model.

[0005] In the above scheme, the nanoscale peak-valley differences on the silicon wafer surface can be transformed from discrete values ​​into a visual color model, thereby solving the problem of not being able to intuitively present defect features. In addition, in this visual color model, different defect features are transformed into different gray values ​​and further assigned different colors for classification, thereby realizing the rapid classification of the degree and distribution of defects on the silicon wafer surface, improving the efficiency and accuracy of defect identification, and providing a basis for subsequent process adjustments.

[0006] In one specific embodiment of the first aspect of this disclosure, the step of processing image information to convert it into a surface feature model may include: performing grayscale processing on the image corresponding to the image information, and extracting the portion corresponding to the silicon wafer from the image as an effective image; establishing a mapping relationship between the nanoscale region of the silicon wafer and the pixel region of the effective image based on the size of the silicon wafer and the number of pixels corresponding to the effective image, wherein the proportion of the nanoscale region in the size of the silicon wafer is equal to the proportion of the number of pixels in the pixel region in the number of pixels in the effective image; obtaining a grayscale difference value based on the grayscale value corresponding to each pixel in the pixel region, wherein the grayscale difference value characterizes the degree of grayscale difference between each pixel in the pixel region; and mapping the grayscale difference value corresponding to each pixel region onto the effective image to obtain a surface feature model.

[0007] In the above scheme, redundant information in the image is removed by grayscale processing and effective image extraction to focus on the core area of ​​the silicon wafer and reduce interference from invalid data. In addition, based on the proportional mapping relationship between silicon wafer size and pixel number, the precise correspondence between nanoscale areas and pixel areas is achieved to ensure the quantitative accuracy of defect features. Furthermore, by calculating grayscale difference values ​​and mapping them to effective images, the surface feature model can accurately characterize the actual surface height difference of the silicon wafer, laying a data foundation for subsequent coloring and defect identification, thereby improving the reliability and specificity of the model.

[0008] In one specific embodiment of the first aspect of this disclosure, the step of comparing the gray values ​​corresponding to each pixel in each pixel region to obtain a gray-scale difference value may include: comparing the gray values ​​corresponding to each pixel in each pixel region, and taking the difference between the largest gray value and the smallest gray value as the gray-scale difference value.

[0009] In the above scheme, the difference between the maximum and minimum gray values ​​within a pixel area is directly used as the gray difference value. This can accurately and intuitively reflect the degree of difference in the area corresponding to the pixel area on the silicon wafer surface, thereby matching the precise requirements of nanoscale defect detection. In addition, the calculation logic is simple and direct, with no additional data redundancy, which can ensure detection efficiency to meet the high timeliness requirements of industrial production.

[0010] In another specific embodiment of the first aspect of this disclosure, the step of comparing the gray values ​​corresponding to each pixel in each pixel region to obtain a gray difference value may include: taking the average gray value of each pixel in each pixel region as the gray difference value.

[0011] In the above scheme, the average gray value within the pixel area is used as the gray difference value, which can measure the degree of defect in the area corresponding to the pixel area on the silicon wafer surface as a whole. In addition, this calculation method is simple and efficient, reduces the complexity of data processing, and can improve the speed of model building.

[0012] In another specific embodiment of the first aspect of this disclosure, the step of comparing the gray values ​​corresponding to each pixel in each pixel region to obtain a gray difference value may include: in each pixel region, taking the weighted average of the gray values ​​of each pixel as the gray difference value, wherein the weight of the largest and smallest gray values ​​is greater than the weight of the non-largest and non-smallest gray values.

[0013] In the above scheme, by giving higher weight to the maximum and minimum gray values, the overall level of gray values ​​within the pixel area is taken into account, while highlighting the impact of key extreme values ​​on defect judgment, which can more comprehensively and accurately reflect the characteristics of complex defects.

[0014] In one specific embodiment of the first aspect of this disclosure, the step of coloring a surface feature model to generate a colored model may include: coloring a pixel region based on the grayscale difference value of the pixel region, wherein different colored pixel regions correspond to different grayscale difference values.

[0015] In the above scheme, by establishing the correspondence between grayscale difference values ​​and colors, the abstract grayscale data is transformed into intuitive color labels, so that defect features of different degrees can be quickly distinguished by color, thereby reducing the difficulty of defect identification. In this way, whether it is an automated system or manual review, abnormal areas on the silicon wafer surface can be quickly located to reduce the risk of misjudgment and missed judgment.

[0016] In one specific embodiment of the first aspect of this disclosure, the step of coloring a pixel region based on the grayscale difference value of the pixel region may include: rendering the pixel region as a first color when the grayscale difference value of the pixel region is greater than a first preset value and not greater than a second preset value; and rendering the pixel region as a second color when the grayscale difference value of the pixel region is greater than the second preset value.

[0017] In the above scheme, by dividing the thresholds of the first preset value and the second preset value, the colors corresponding to different grayscale difference levels are clearly defined, and the severity of defects is graded and identified, so as to make the defect level and distribution more intuitive and provide a clear basis for subsequent targeted machine adjustment strategies, thereby improving the accuracy of process adjustment.

[0018] In one specific embodiment of the first aspect of this disclosure, the processing method may further include: generating and issuing instructions for adjusting upstream process parameters based on the defect identification results. Thus, by adding the step of issuing process adjustment instructions on top of defect identification, an automated closed loop of detection, feedback, and machine adjustment can be constructed, thereby solving the problems of disconnection between detection results and upstream processes, and the lag in manual transmission. Furthermore, this solution achieves real-time linkage between defect detection and process optimization, reducing the generation of batches of defective products and improving production efficiency and product yield.

[0019] In one specific embodiment of the first aspect of this disclosure, the step of outputting defect identification results based on a shading model may include: inputting information of the shading model into a trained defect classification model; determining the defect category information corresponding to the shading model based on the defect classification model to determine the defect category corresponding to the silicon wafer, wherein the defect identification result includes the defect category.

[0020] In the above scheme, the coloring model is analyzed by a trained defect classification model to accurately determine the defect category, thereby providing a precise matching basis for subsequent process adjustments and avoiding the risk of blindly adjusting the machine.

[0021] In one specific embodiment of the first aspect of this disclosure, the step of generating and issuing instructions for adjusting upstream process parameters based on defect identification results may include: sending the defect identification results to a manufacturing execution system; the manufacturing execution system matching a pre-stored process adjustment strategy according to the defect identification results; and converting the process adjustment strategy into instructions and issuing them to upstream process equipment to adjust the equipment parameters.

[0022] In the above solution, the stability of cross-system data transmission and the accuracy of instruction execution are ensured by using defect identification results, manufacturing execution system matching strategies, and standardized processes for issuing conversion instructions to equipment. In addition, the solution relies on pre-stored process adjustment strategies to avoid dependence on manual experience, enabling rapid and standardized issuance of machine adjustment instructions, further enhancing the automation level of the detection and machine adjustment closed loop, continuously optimizing front-end process parameters, and ensuring the stability of silicon wafer surface quality.

[0023] The second aspect of this disclosure provides a system for processing surface defects on a silicon wafer. The system includes an acquisition module, a processing module, a coloring module, and an output module. The acquisition module acquires image information, wherein the image information represents defect features on the silicon wafer surface, and the defect features represent the differences in peaks and valleys on the silicon wafer surface within a preset unit area. The processing module processes the image information to convert it into a surface feature model, wherein the surface feature model includes grayscale values ​​corresponding to the defect features, with different defect features corresponding to different grayscale values. The coloring module performs coloring processing on the surface feature model to generate a coloring model, wherein the coloring model includes a preset number of colors, with different colors corresponding to different grayscale values. The output module outputs defect identification results based on the coloring model.

[0024] In the processing system described above, the nanoscale peak-valley differences on the silicon wafer surface can be transformed from discrete numerical values ​​into a visual color model, thereby solving the problem of not being able to intuitively present defect features. In addition, in this visual color model, different defect features are transformed into different grayscale values ​​and further assigned different colors for classification, thereby enabling rapid classification of the degree and distribution of defects on the silicon wafer surface, improving the efficiency and accuracy of defect identification, and providing a basis for subsequent process adjustments.

[0025] In one specific embodiment of the first aspect of this disclosure, the processing system may further include an adjustment module, which is configured to generate and issue instructions for adjusting front-end process parameters based on defect identification results. Thus, by using defect identification results, manufacturing execution system matching strategies, and a standardized process for issuing conversion instructions to equipment, the stability of cross-system data transmission and the accuracy of instruction execution are ensured. Furthermore, this solution relies on pre-stored process adjustment strategies to avoid dependence on manual experience, enabling rapid and standardized issuance of adjustment instructions, further enhancing the automation level of the detection and adjustment closed loop, continuously optimizing front-end process parameters, and ensuring the stability of silicon wafer surface quality.

[0026] A third aspect of this disclosure provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the processing method described in the first aspect.

[0027] This disclosure provides a fourth aspect of an electronic device, which includes a processor and a memory. The memory stores processor-executable instructions, and the processor performs the processing method described in the first aspect above.

[0028] The fifth aspect of this disclosure provides a computer program product comprising a computer program that, when executed by a processor, implements the processing method described in the first aspect above. Attached Figure Description

[0029] Figure 1A This is one of the schematic diagrams of a method for treating surface defects on silicon wafers according to this application.

[0030] Figure 1B To and Figure 1A The flowchart corresponding to the processing method shown.

[0031] Figure 2 This is a flowchart of a method for treating surface defects on silicon wafers according to this application.

[0032] Figure 3A This is the second schematic diagram of a method for treating surface defects on silicon wafers according to this application.

[0033] Figure 3B To and Figure 3A The flowchart corresponding to the processing method shown.

[0034] Figure 4A This is the third schematic diagram of a method for treating surface defects on silicon wafers according to this application.

[0035] Figure 4B To and Figure 4A The flowchart corresponding to the processing method shown.

[0036] Figure 5A This is the fourth schematic diagram of a method for treating surface defects on silicon wafers according to this application.

[0037] Figure 5B To and Figure 5A The flowchart corresponding to the processing method shown.

[0038] Figure 6 is a schematic diagram showing the display state of the silicon wafer surface model obtained based on the processing method of this application.

[0039] Figure 7 The diagram shown is one of the principle block diagrams of a cutting device provided in an embodiment of this application.

[0040] Figure 8 The following is a second schematic diagram of the cutting device provided in one embodiment of this application.

[0041] Figure 9 The diagram shown is a block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0042] The technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments in this specification, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this specification.

[0043] Currently, the industry primarily relies on specialized metrology equipment to inspect the surface quality of silicon wafers. This equipment collects peak and valley data from the nanometer regions on the front and back sides of the wafer, representing the peak-valley differences on the wafer surface in discrete numerical form. However, this inspection method has significant technical limitations. For example, on the one hand, discrete numerical data cannot intuitively present the specific location, shape, and distribution range of defects. Technicians must indirectly infer the defect situation through complex data conversion and manual judgment. This method not only suffers from low inspection efficiency but is also prone to missed or misjudged defects due to data interpretation biases. On the other hand, existing inspection processes lack automated feedback and adjustment mechanisms. Measurement data must be manually processed and transmitted to upstream process stations and equipment. The formulation of adjustment strategies highly depends on the practical experience of technicians, resulting in problems such as data transmission delays, untimely adjustment responses, and low strategy matching accuracy, making it difficult to meet the core production requirements of semiconductor manufacturing: "high precision, high automation, and high timeliness."

[0044] In view of this, the present disclosure provides a method and system for processing surface defects on silicon wafers, an electronic device, a computer-readable storage medium, and a product to at least solve the aforementioned technical problems. Figure 1A and Figure 1B As shown, the method for treating defects on the silicon wafer surface may include the following steps S100 to S400.

[0045] S100, acquire image information, wherein the image information characterizes the defect features of the silicon wafer surface, and the defect features characterize the difference information of peaks and valleys on the surface of the silicon wafer within a preset unit area.

[0046] For example, such as Figure 1B As shown, after acquiring the raw data about the silicon wafer, this data is filtered, and NanoMap images recording the nanoscale peak and valley characteristics of the silicon wafer surface are selected and downloaded. The downloaded NanoMap images are then saved to a designated storage area. If a subsequent request arises to detect defects on the silicon wafer surface, the request will be responded to by transferring these NanoMap images to subsequent processes.

[0047] In step S200, the image information is processed to convert it into a surface feature model. This model includes grayscale values ​​corresponding to defect features, with different defect features corresponding to different grayscale values. For example, after receiving the NanoMap image transmitted in step S100, the image is converted into a grayscale image, directly linking the grayscale values ​​to the peak-valley features of the silicon wafer surface, thus forming a surface feature model that intuitively reflects the nanometer-level peak-valley differences on the silicon wafer surface.

[0048] In step S300, the surface feature model is colored to generate a colored model. This colored model includes a variety of preset colors, each corresponding to a different grayscale value. For example, after obtaining the surface feature model output from step S200, the correspondence between grayscale values ​​and colors is first determined based on the silicon wafer inspection scenario (such as DSPMonitor inspection, Flatness & Nano inspection, etc.). Then, the grayscale values ​​are converted into corresponding colors according to this rule to generate a colored model that can intuitively identify the location, level, and distribution of defects, providing a visual basis for the defect identification results output in the subsequent step S400.

[0049] S400 outputs defect identification results based on the coloring model. For example, after obtaining the coloring model generated in step S300, key defect information in the coloring model can be extracted using an automated identification algorithm. This includes defect location, defect size, and defect level corresponding to different color regions, and then integrated into a structured output result.

[0050] In the processing method shown in steps S100 to S400 above, the nanoscale peak-valley differences on the silicon wafer surface can be transformed from discrete values ​​into a visual color model, thereby solving the problem of not being able to intuitively present defect features. In addition, in this visual color model, different defect features are transformed into different grayscale values ​​and further assigned different colors for division, thereby realizing the rapid division of the degree and distribution of defects on the silicon wafer surface, improving the efficiency and accuracy of defect identification, and providing a basis for subsequent process adjustments.

[0051] In one scenario, combined Figure 2 right Figure 1A and Figure 1BThe overall process is described below. Specifically, a silicon wafer measurement machine with nanometer-level surface morphology detection capabilities scans and inspects the silicon wafer, then outputs a NanoMap image containing information on the differences in peaks (height of the highest protrusion) and valleys (height of the lowest depression) within a preset unit area on the silicon wafer surface. This image is the core raw data of the entire process. The image is then transmitted to a transit network drive for temporary storage. This step serves as an image transition node, temporarily caching the NanoMap image output by the measurement machine to avoid transmission congestion caused by direct connection of the device to subsequent storage stages. Simultaneously, the images can be initially categorized and temporarily stored according to silicon wafer batch and inspection time to ensure orderly image flow. Afterward, the image is transferred to object storage for image backup, thus serving as a distributed storage carrier adapted to the semiconductor industry scenario. In this way, a redundant storage architecture is used to maintain the NanoMap image. The images are backed up securely for a long time. On the other hand, metadata such as the silicon wafer number, inspection batch, and measurement equipment information corresponding to the images are recorded simultaneously to ensure the traceability of the image data. Then, the process enters the application service to realize image download, saving, and inspection requests. For example, it can download the corresponding NanoMap image from object storage based on industrial-grade transmission protocols and save it to a local specified storage directory. Then, it generates a silicon wafer defect inspection request according to the instructions and transmits the inspection instructions and image storage path to the next stage through an HTTP request. At the same time, it undertakes the function of receiving subsequent inspection results. After that, it enters the image preprocessing stage. After receiving the HTTP request from the application service, it retrieves the saved NanoMap image and processes its data to establish the surface feature model and color model (model) mentioned in the previous embodiment. Finally, it enters the inspection service stage to perform image (here, color model) inspection. For example, after receiving the color model output from the image preprocessing, it extracts key information such as the defect location, size, and level corresponding to different color areas in the model through an automated recognition algorithm. After the inspection is completed, the structured inspection results are fed back to the application service through an HTTP response, ultimately forming a closed loop from image output to inspection result feedback.

[0052] In at least one embodiment of this disclosure, such as Figure 3A and Figure 3B As shown, step S200 above, which is the step of processing image information to convert image information into a surface feature model, may include the following steps S210 to S240.

[0053] S210: Perform grayscale processing on the image corresponding to the image information, and extract the portion of the image corresponding to the silicon wafer as the valid image. Convert the image, which may originally contain multiple color channels (such as a NanoMap image), into a single-channel grayscale image, eliminating the interference of color information on the surface features of the silicon wafer. At this time, the grayscale value of each pixel in the image can be directly correlated with the peak and valley height of the silicon wafer surface (for example, the higher the grayscale value, the more convex the surface, and the lower the grayscale value, the more concave the surface). This transforms visual color information into quantifiable surface topography data, making the peak and valley differences on the silicon wafer surface more intuitive and easier for subsequent models to identify.

[0054] S220 establishes a mapping relationship between the nano-regions of the silicon wafer and the pixel regions of the effective image based on the size of the silicon wafer and the number of pixels corresponding to the effective image. The proportion of the nano-regions in the size of the silicon wafer is equal to the proportion of the number of pixels in the pixel regions in the number of pixels in the effective image.

[0055] For example, the effective image obtained after cropping the original image has 630×630 pixels, and the diameter of the silicon wafer is 300mm. If the nanometer region is 2×2mm, the corresponding pixel count is 630 / (300 / 2)≈4.2, rounded down to a pixel region containing 4×4 pixels; or, if the nanometer region is 8×8mm, the corresponding pixel count is 630 / (300 / 8)≈17, rounded down to a pixel region containing 17×17 pixels; or, if the nanometer region is 10×10mm, the corresponding pixel count is 630 / (300 / 10)≈21, rounded down to a pixel region containing 21×21 pixels. Specifically, the effective image extracted in step S210 can be circular, with a corresponding pixel count set to 630×630 pixels. Then, the silicon wafer nano-region to be mapped (i.e., the preset unit area used for defect detection, such as 2×2mm or 10×10mm) is determined. The pixel count of the corresponding pixel region is calculated according to the principle of "equal proportion": taking a 2×2mm silicon wafer nano-region as an example, its proportion in a 300mm silicon wafer diameter is (2 / 300), corresponding to a 630×630 pixel effective image. The pixel count ratio of each pixel region must also be (2 / 300). Calculations show that the number of pixels corresponding to this nanometer region is approximately 4×4 pixels (630×(2 / 300)≈4). Similarly, for an 8×8mm silicon wafer nanometer region, the corresponding pixel region in the effective image is approximately 17×17 pixels (630×(8 / 300)=17), and for a 10×10mm silicon wafer nanometer region, the corresponding pixel region in the effective image is approximately 21×21 pixels (630×(10 / 300)=21). This mapping relationship establishes a precise correspondence between the nanometer-level physical regions on the silicon wafer surface and the image pixel regions, laying the foundation for subsequently converting the grayscale features of the pixel regions into physical parameters of actual defects in the silicon wafer (such as defect size and peak-valley height difference).

[0056] S230: Based on the grayscale values ​​corresponding to each pixel in the pixel region, a grayscale difference value is obtained. This grayscale difference value characterizes the degree of grayscale difference among pixels in the pixel region. This grayscale difference value directly reflects the degree of grayscale fluctuation of each pixel in the corresponding pixel region. Since the pixel grayscale value is strongly correlated with the peak-valley height on the silicon wafer surface, this difference value also indirectly characterizes the degree of peak-valley height difference in the actual area of ​​the corresponding silicon wafer. This provides a core quantitative basis for subsequently determining the defect level of the silicon wafer surface through the grayscale difference value (e.g., the larger the grayscale difference value, the more significant the peak-valley difference in the corresponding silicon wafer area, and the higher the defect level may be).

[0057] In step S240, the grayscale difference values ​​corresponding to each pixel region are mapped onto the effective image to obtain a surface feature model. Through the mapping relationship between location and grayscale difference values, the effective image no longer presents only single grayscale information, but forms a comprehensive model—the surface feature model—that integrates the quantified peak-valley difference data of each region. This model retains the spatial location information of the silicon wafer surface and intuitively reflects the degree of defect characteristics in different regions through grayscale difference values. Subsequently, based directly on the grayscale difference values ​​of each region in the model, defect-free areas, lightly defective areas, and heavily defective areas on the silicon wafer surface can be quickly distinguished, providing data support for the coloring process in step S300 of the aforementioned embodiments.

[0058] In the processing method shown in steps S210 to S240 above, redundant information in the image is removed by grayscale processing and effective image extraction to focus on the core area of ​​the silicon wafer and reduce interference from invalid data. In addition, based on the proportional mapping relationship between silicon wafer size and pixel number, the precise correspondence between nanoscale areas and pixel areas is achieved to ensure the quantitative accuracy of defect features. Furthermore, by calculating grayscale difference values ​​and mapping them to effective images, the surface feature model can accurately characterize the actual surface height difference of the silicon wafer, laying a data foundation for subsequent coloring and defect identification, thereby improving the reliability and specificity of the model.

[0059] In the embodiments of this disclosure, there are no restrictions on the specific calculation method of the grayscale difference value, as long as it can characterize the degree of defect on the silicon wafer surface. Below, several methods for calculating the grayscale difference value are illustrated by example.

[0060] In some embodiments of this disclosure, step S230, which involves comparing the grayscale values ​​of each pixel in each pixel region to obtain a grayscale difference value, may include: comparing the grayscale values ​​of each pixel in each pixel region and taking the difference between the maximum and minimum grayscale values ​​as the grayscale difference value. Specifically, within a single pixel region, the grayscale values ​​of all pixels are extracted one by one (for example, the value range is 0-255, and according to the grayscale processing rules in step S210, the grayscale value is directly related to the peak and valley height of the silicon wafer surface; that is, the higher the grayscale value, the more convex the corresponding silicon wafer surface position, and the lower the grayscale value, the more concave the corresponding surface position); then, all the extracted grayscale values ​​are numerically compared, and the maximum and minimum grayscale values ​​in the region are selected; finally, the difference between these two extreme values ​​is the grayscale difference value of the pixel region.

[0061] Thus, the difference between the maximum and minimum grayscale values ​​within a pixel region is directly used as the grayscale difference value. The fluctuation of grayscale values ​​within the region directly reflects the peak and valley undulations of the corresponding silicon wafer region. That is, the larger the grayscale difference value, the more significant the height difference between the highest protrusion and the lowest depression within the silicon wafer region, and the higher the probability and severity of defects. Conversely, the smaller the grayscale difference value, the smoother the silicon wafer surface, and the closer it is to a defect-free state. This quantification method transforms the abstract "smoothness" of the silicon wafer surface into a concrete numerical indicator, allowing nanometer-level subtle height differences to be clearly presented through grayscale difference values. This can accurately and intuitively reflect the height differences of the region corresponding to the pixel region on the silicon wafer surface. This method achieves a higher level of precision, matching the needs of nanoscale defect detection. Furthermore, compared to other complex grayscale analysis algorithms (such as statistical methods that require calculating multiple parameters like mean, variance, and standard deviation), the "extremum difference" method requires no additional complex calculations. It can complete the calculation simply through basic numerical comparisons and subtraction, significantly reducing the computational load and redundant data generation during data processing. In industrial production scenarios, silicon wafer inspection faces the demands of large-volume, high-paced production. This efficient calculation method can shorten the inspection time for a single silicon wafer, ensuring that the inspection process matches the rhythm of the production line, avoiding slowing down the overall production progress due to inefficient inspection processes, and aligning with the core demand for high efficiency in industrial production.

[0062] In other embodiments of this disclosure, step S230, which involves comparing the grayscale values ​​of each pixel in each pixel region to obtain a grayscale difference value, may include: taking the average grayscale value of each pixel in each pixel region as the grayscale difference value. Specifically, the grayscale values ​​of all pixels in the target pixel region are extracted one by one. These grayscale values, after grayscale processing in step S210, have established a direct correlation with the peak and valley heights of the silicon wafer surface (higher grayscale values ​​correspond to a more convex surface, and lower grayscale values ​​correspond to a more concave surface), and their value range is strictly limited to, for example, 0-255. Subsequently, the extracted pixel grayscale values ​​are summed and then divided by the total number of pixels in the pixel region. The average grayscale value of all pixels in the pixel region is obtained through the basic calculation logic of "sum / number of pixels". Finally, this average value is directly defined as the grayscale difference value of the pixel region, completing the acquisition of the grayscale difference value of a single pixel region. The calculation of the entire image can be completed by traversing all pixel regions using the same logic.

[0063] Therefore, by using the average grayscale value within a pixel region as the grayscale difference value, the degree of defect in the corresponding silicon wafer region can be measured from an overall perspective. In silicon wafer inspection, some defects (such as large-area slight unevenness, regional uniform bulges or depressions) are not determined by the extreme grayscale values ​​of a single local pixel, but rather by the shift in the overall grayscale level within the region. In this case, by calculating the average value, the overall grayscale state of all pixels in the region can be comprehensively reflected. For example, if the average value is significantly higher than the average grayscale value of the normal area on the silicon wafer surface (the preset normal range is 80-120), it indicates that the corresponding silicon wafer region is generally convex, and there may be large-area slight convexity. If the average value of a certain type of defect is significantly lower than the normal range, it indicates that the corresponding area is generally concave, and there may be large-area mild concavity defects. This holistic measurement method can effectively capture regional uniform defects and avoid missing overall uniform defects by focusing only on local extreme values. In addition, this calculation method is simple and efficient, which greatly reduces the data processing time of a single image and reduces the computing load of the server or inspection equipment. This shortens the overall time from image input to surface feature model construction, allowing the model construction speed to better adapt to the high-speed requirements of industrial production lines and avoids the inspection process lagging behind the production rhythm due to excessive data processing complexity.

[0064] In some embodiments of this disclosure, step S230, which involves comparing the grayscale values ​​of each pixel in each pixel region to obtain a grayscale difference value, may include: in each pixel region, using the weighted average of the grayscale values ​​of each pixel as the grayscale difference value, wherein the weights of the largest and smallest grayscale values ​​are greater than the weights of the non-largest and non-smallest grayscale values. Specifically, the grayscale values ​​of all pixels in the target pixel region are extracted one by one. After grayscale processing in step S210, these grayscale values ​​are directly correlated with the peak and valley heights of the silicon wafer surface. Subsequently, all extracted grayscale values ​​are filtered to distinguish the largest and smallest grayscale values ​​in the region, as well as the remaining non-largest and non-smallest grayscale values ​​(i.e., intermediate grayscale values). Then, differentiated weights are assigned to different types of grayscale values, with the core principle being that "the weights of the largest and smallest grayscale values ​​are higher than those of the intermediate grayscale values." For example, the weight of the largest grayscale value can be set to 0.3, the weight of the smallest grayscale value to 0.3, the weight of the largest grayscale value to 0.3, the weight of the smallest grayscale value to 0.5, the weight of the smallest grayscale value to 0.5, the weight of the largest grayscale value to 0.5 ... The smallest grayscale value has a weight of 0.3, and the remaining weight of each intermediate grayscale value is evenly distributed according to the number of values, with the remaining weight being 0.4 (if a 4×4 pixel area contains 14 intermediate grayscale values, then the weight of each intermediate grayscale value is approximately 0.4÷14≈0.029). Finally, the weighted average value is calculated according to the formula "weighted average = (maximum grayscale value × corresponding weight) + (minimum grayscale value × corresponding weight) + (each intermediate grayscale value × corresponding weight)", and the result is directly used as the grayscale difference value of that pixel area. The same logic is used to obtain the grayscale difference value of all pixel areas.

[0065] Thus, by assigning higher weights to the maximum and minimum grayscale values, the weighted average can reflect both the overall grayscale level of the area represented by the intermediate grayscale values ​​(reflecting the overall defect trend) and highlight the influence of extreme values ​​through high weights (capturing local drastic fluctuation defects). This achieves simultaneous capture of both overall and local defect characteristics, avoiding misjudgments or omissions due to single-dimensional analysis. Furthermore, when calculating grayscale difference values ​​using the weighted average method, high-weighted extreme values ​​directly raise or lower the final result, allowing the grayscale difference values ​​to more accurately match the defect classification logic. For example, if the intermediate grayscale values ​​of a pixel area are concentrated in the normal range (e.g., 80-120), but the maximum grayscale value reaches 200 (e.g., corresponding to severe defects), the weighted average will be significantly higher than the normal range under high weights, thus accurately triggering the severe defect judgment. Conversely, if the extreme values ​​do not exceed the threshold, the overall influence of the intermediate grayscale values ​​will dominate the result, ensuring the accuracy of the judgment of mild defects or defect-free areas, and significantly improving the matching degree between the grayscale difference values ​​and the actual defect level. In this way, while taking into account the overall level of grayscale values ​​within the pixel area, the impact of key extreme values ​​on defect judgment is highlighted, which can more comprehensively and accurately reflect the characteristics of complex defects.

[0066] In at least one embodiment of this disclosure, such as Figure 4A and Figure 4B As shown, step S300 above, which is the step of coloring the surface feature model to generate a colored model, may include the following step S310.

[0067] S310 colors pixel regions based on their grayscale difference values, where different colors correspond to different grayscale difference values. Specifically, before the coloring operation, rules for the correspondence between grayscale difference values ​​and colors can be formulated based on industry standards for silicon wafer inspection and actual production needs. These rules can balance the principles of "clear distinguishability" and "compliance with inspection habits." For example, pixel areas with a grayscale difference value of 0-12 (corresponding to flat, defect-free areas on the silicon wafer surface) are colored with a base light gray, preserving the original tone of the image while clearly representing normal areas. Pixel areas with a grayscale difference value of 13-35 (corresponding to minor defects such as small scratches or slight unevenness) are colored with cyan, a color that provides moderate visibility against a light gray background, avoiding excessive visual interference while clearly marking the extent of minor defects. Pixel areas with a grayscale difference value greater than 35 (corresponding to severe defects such as noticeable protrusions, depressions, or cracks) are colored with purple, utilizing the strong visual contrast between purple and light gray or cyan to make severe defect areas easier to identify. After the rules are determined, an image rendering algorithm traverses all pixel areas in the surface feature model, matching the corresponding color parameters based on the grayscale difference value of each area to complete the area-by-area coloring, ultimately generating a visual coloring model that distinguishes the degree of defect by color. It should be noted that the above numerical ranges can be set according to actual needs and are not limited to the values ​​mentioned above.

[0068] Defect features currently exist only as abstract numerical values ​​representing grayscale differences. Inspectors or automated systems must meticulously verify each value range to determine the severity of the defect, which is not only time-consuming but also prone to misjudgment due to interpretation errors. However, after coloring, defects of different degrees are transformed into distinct colors—for example, cyan for mild defects, purple for severe defects—forming clear visual stratification with the light gray of normal areas. This eliminates the need for complex numerical calculations and comparisons; defect levels can be quickly distinguished simply through visual observation. This transforms abstract data, previously reliant on specialized knowledge, into easily understandable information, allowing even less experienced inspectors to quickly learn and significantly lowering the technical barrier and operational difficulty of defect identification. Furthermore, for automated inspection systems, the color information in the coloring model can be directly converted into easily identifiable image feature parameters. The system no longer needs to analyze massive amounts of grayscale values ​​one by one; it can quickly locate cyan, purple, and light gray using color recognition algorithms. The color model can significantly shorten the detection time for abnormal color areas such as purple. Moreover, the stability of color features is much higher than the slight fluctuations in grayscale values, which can reduce the misjudgment caused by numerical interference. Alternatively, for manual review, inspectors can quickly focus on abnormal color areas through the color model and view the defect details by combining the image magnification function. There is no need to check the grayscale values ​​point by point in the entire image, avoiding visual fatigue caused by staring at a monotonous grayscale image for a long time and missing small defects. In this way, whether in automated or manual processes, the color model can help to quickly locate abnormal areas, making defect identification more efficient and accurate. This effectively reduces the risk of miscalculation of silicon wafer yield or defective products flowing into the next stage due to misjudgment or missed judgment in industrial production.

[0069] In at least one embodiment of this disclosure, step S310, namely the step of coloring a pixel region based on the grayscale difference value of the pixel region, may include: rendering the pixel region as a first color when the grayscale difference value of the pixel region is greater than a first preset value and not greater than a second preset value; and rendering the pixel region as a second color when the grayscale difference value of the pixel region is greater than the second preset value, that is, Figure 4B The coloring is controlled and stored using a threshold difference. For example, a first preset value and a second preset value can be used to distinguish between gray, cyan, and purple in the aforementioned embodiments, with the first color corresponding to cyan and the second color corresponding to purple. Thus, by using the threshold division of the first and second preset values, the colors corresponding to different grayscale difference levels are clearly defined, achieving a graded identification of defect severity. This makes the defect level and distribution more intuitive and provides a clear basis for subsequent targeted machine adjustment strategies, thereby improving the accuracy of process adjustments.

[0070] For example, in the detection scenario of a DSP Monitor: when the pixel area is 4×4 pixels, it corresponds to a 2×2mm nanometer-level detection area on the silicon wafer surface. If the grayscale difference value of this pixel area is greater than 20, then the pixels in this pixel area are set to cyan (to identify minor defects); when the pixel area is 17×17 pixels, it corresponds to an 8×8mm nanometer-level detection area on the silicon wafer surface. If the grayscale difference value of this pixel area is greater than 45, then the pixels in this pixel area are set to purple (to identify severe defects).

[0071] For example, in the Flatness & Nano detection scenario: when the pixel area is 4×4 pixels, it corresponds to a 2×2mm nanometer-level detection area on the silicon wafer surface. If the grayscale difference value of this pixel area is greater than 12, then the pixels in this pixel area are set to cyan (to identify minor defects); when the pixel area includes 21×21 pixels, it corresponds to a 10×10mm nanometer-level detection area on the silicon wafer surface. If the grayscale difference value of this pixel area is greater than 35, then the pixels in this pixel area are set to purple (to identify severe defects).

[0072] In the embodiments of this disclosure, after obtaining the defect identification result based on the coloring model, instructions can be generated and issued based on the result for process optimization, as follows.

[0073] In at least one embodiment of this disclosure, such as Figure 5A and Figure 5B As shown, the processing method may also include the following step S500.

[0074] The S500, based on defect identification results, generates and issues instructions for adjusting upstream process parameters. By adding the step of issuing process adjustment instructions on top of defect identification, an automated closed loop of detection, feedback, and machine adjustment can be constructed, thus solving the problems of disconnect between detection results and upstream processes, and the lag in manual transmission. Furthermore, because this solution automatically associates defects with process parameters and automatically issues adjustment instructions, it achieves real-time linkage between defect detection and process optimization. When a certain type of defect is detected, upstream equipment can complete parameter adjustments in a short time, preventing defects from continuing to occur due to unoptimized processes. This significantly reduces the occurrence of batches of defective products, reducing production cost waste caused by rework and scrap of defective products, shortening ineffective time in the production process, and ultimately improving production efficiency and product yield.

[0075] In at least one embodiment of this disclosure, step S500, namely the step of outputting defect identification results based on the shading model, may include: inputting the information of the shading model into a trained defect classification model; determining the defect category information corresponding to the shading model based on the defect classification model to determine the defect category corresponding to the silicon wafer, wherein the defect identification result includes the defect category. Thus, by analyzing the shading model through the trained defect classification model, accurate determination of the defect category is achieved, thereby providing a precise matching basis for subsequent process adjustments and avoiding the risk of blindly adjusting the machine.

[0076] Specifically, such as Figure 6A and Figure 6B As shown, in the foregoing embodiments, after processing the acquired image, the following is obtained: Figure 6A The surface feature model shown only contains grayscale values ​​in its image; after color processing, the surface feature model is obtained... Figure 6B The coloring model shown. The trained defect classification model stores various defect categories, such as... Figure 6B The categories shown are SA1, SA2, SA3, and SA4.

[0077] For example, in category SA1 (which can be called Saw), the image shows scattered, irregular light (cyan) and dark (purple) areas, which are consistent with the saw marks / line marks defects caused by saw wire vibration and uneven feed speed in the silicon wafer cutting process (such as diamond wire saw cutting). These defects are the cutting marks left on the silicon wafer surface. They are line / surface mixed defects introduced by the process (slicing) and will affect the surface flatness of subsequent grinding and polishing processes.

[0078] For example, in category SA2 (which can be called Ring), the defects in the image are concentrated in the ring-shaped area (mainly purple) at the edge of the silicon wafer. This corresponds to ring-shaped defects caused by uneven edge stress and deviations in grinding / polishing parameters during silicon wafer processing (such as uneven edge thickness and ring-shaped scratch clusters). These are localized process defects, which are usually related to the concentricity of the equipment worktable and the control of edge processing pressure.

[0079] For example, in category SA3 (which can be called Center), defects are concentrated in the center of the silicon wafer as dots / small areas (dark dots in the center + light spots around them). These correspond to center positioning deviations during silicon wafer processing and center wear of the polishing pad, resulting in defects in the center area (such as center pits or center particle aggregation). These are single-point / small-scale local defects that may affect the performance of the center area during device fabrication.

[0080] For example, in category SA4 (which can be called Full), the defects in the image cover most of the silicon wafer (large area of ​​purple + irregular texture), which is a global severe defect. This type of defect is usually caused by serious parameter abnormalities in the slicing / grinding process (such as saw wire breakage, polishing fluid contamination, and uneven thickness across the entire area), which will directly lead to the scrapping of the silicon wafer. It is an extreme defect type under process out of control.

[0081] In at least one embodiment of this disclosure, such as Figure 5B As shown, step S500 above, which involves generating and issuing instructions for adjusting upstream process parameters based on the defect identification results, may include: sending the defect identification results to the Manufacturing Execution System (MES); the MES matching pre-stored process adjustment strategies based on the defect identification results; and converting the process adjustment strategies into instructions and issuing them to the upstream process equipment to adjust the equipment parameters. Specifically, the defect identification results are sent to the MES via "DBLink"; the MES matches pre-stored process adjustment strategies based on the defect identification results, and simultaneously synchronizes the defect information and the proposed adjustment strategy with engineers via "Email & WeChat" for timely confirmation; subsequently, the MES issues the matched process adjustment strategy to the Equipment Automation Program (EAP) via a "message middleware," and the EAP converts the process adjustment strategy into standardized instructions conforming to the SECS / GEM protocol and issues them to the upstream processing equipment; after receiving the instructions, the processing equipment adjusts the process formula or equipment status and simultaneously feeds back the execution results to the EAP.

[0082] In this way, by using defect identification results, manufacturing execution system matching strategies, and standardized processes for issuing conversion instructions to equipment, the stability of cross-system data transmission and the accuracy of instruction execution are ensured. In addition, this solution relies on pre-stored process adjustment strategies to avoid dependence on manual experience, enabling rapid and standardized issuance of machine adjustment instructions, further enhancing the automation level of the detection and machine adjustment closed loop, continuously optimizing front-end process parameters, and ensuring the stability of silicon wafer surface quality.

[0083] At least one embodiment of this disclosure provides a system for processing surface defects on a silicon wafer, such as... Figure 7As shown, the processing system includes an acquisition module 10, a processing module 20, a coloring module 30, and an output module 40. The acquisition module 10 is configured to acquire image information, wherein the image information represents defect features on the surface of a silicon wafer, and the defect features represent the differences in peaks and valleys on the surface of the silicon wafer within a preset unit area. The processing module 20 is configured to process the image information to convert it into a surface feature model, wherein the surface feature model includes grayscale values ​​corresponding to defect features, with different defect features corresponding to different grayscale values. The coloring module 30 is configured to color the surface feature model to generate a coloring model, wherein the coloring model includes a preset number of colors, with different colors corresponding to different grayscale values. The output module 40 is configured to output defect recognition results based on the coloring model. This processing system can transform the nanoscale peak-valley differences on the silicon wafer surface from discrete numerical values ​​into a visualized color model, thereby solving the problem of not being able to intuitively present defect features. Furthermore, in this visualized color model, different defect features are converted into different grayscale values ​​and further assigned different colors for classification, thus enabling rapid classification of the degree and distribution of defects on the silicon wafer surface. This improves the efficiency and accuracy of defect identification and provides a basis for subsequent process adjustments. The processing method for detecting defects on the silicon wafer surface corresponding to this system can be found in the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0084] In at least one embodiment of this disclosure, such as Figure 8 As shown, the processing system may further include an adjustment module 50, which is configured to generate and issue instructions for adjusting front-end process parameters based on defect identification results. Thus, through the standardized process of defect identification results, manufacturing execution system matching strategies, and conversion of instructions to equipment, the stability of cross-system data transmission and the accuracy of instruction execution are ensured. Furthermore, this solution relies on pre-stored process adjustment strategies to avoid dependence on manual experience, enabling rapid and standardized issuance of adjustment instructions, further enhancing the automation level of the detection and adjustment closed loop, continuously optimizing front-end process parameters, and ensuring the stability of silicon wafer surface quality. The processing method for detecting silicon wafer surface defects corresponding to this system can be found in the relevant descriptions in the foregoing embodiments, and will not be repeated here.

[0085] At least one embodiment of this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the processing methods described in the above embodiments.

[0086] At least one embodiment of this application provides an electronic device, such as... Figure 9As shown, the electronic device 1 includes a processor 2 (which may be at least one) and a memory 3. The memory 3 is used to store executable instructions (e.g., applications) of the processor 2. The applications stored in the memory 3 may include one or more modules, each corresponding to a set of instructions. The processor 2 is configured to execute instructions to perform the processing methods in the above embodiments.

[0087] Electronic device 1 may also include a power supply component configured to perform power management of electronic device 1, a wired or wireless network interface configured to connect electronic device 1 to a network, and an input / output (I / O) interface. Electronic device 1 may operate on an operating system stored in memory 3, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or similar.

[0088] At least one embodiment of this application provides a computer program product, which includes a computer program that, when executed by a processor, implements the processing methods described in the above embodiments. Specifically, the computer program can be stored in a computer-readable storage medium. When the above processing methods need to be executed, the processor can read the computer program from the computer-readable storage medium and load it into memory for execution, thereby driving relevant hardware devices to complete operations such as control of the cutting process, parameter adjustment, wear compensation (band saw thickness compensation), and thickness monitoring according to the steps, processes, and logic described in the above embodiments.

[0089] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0090] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0091] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0092] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0093] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0094] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program verification codes, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0095] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0096] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0097] It should be understood that the terms "first," "second," etc., mentioned in the embodiments of the present invention are merely for the purpose of more clearly describing the use of the technical solutions in the embodiments of the present invention, and are not intended to limit the scope of protection of the present invention.

[0098] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Any modifications or equivalent substitutions made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for treating surface defects on a silicon wafer, characterized in that, include: Acquire image information, wherein the image information characterizes the defect features of the silicon wafer surface, and the defect features characterize the difference information of peaks and valleys on the surface of the silicon wafer within a preset unit area; The image information is processed to convert it into a surface feature model, wherein the surface feature model includes gray values ​​corresponding to the defect features, and different defect features correspond to different gray values; The surface feature model is colored to generate a colored model, wherein the colored model includes a variety of preset colors, and different colors correspond to different grayscale values; and The defect identification results are output based on the coloring model.

2. The processing method according to claim 1, characterized in that, The process of processing the image information to convert it into a surface feature model includes: The image corresponding to the image information is processed in grayscale, and the portion corresponding to the silicon wafer is extracted from the image as a valid image; Based on the size of the silicon wafer and the number of pixels corresponding to the effective image, a mapping relationship is established between the nano-region of the silicon wafer and the pixel region of the effective image, wherein the proportion of the nano-region in the size of the silicon wafer is equal to the proportion of the number of pixels in the pixel region in the number of pixels in the effective image. Based on the grayscale values ​​corresponding to each pixel in the pixel region, a grayscale difference value is obtained, wherein the grayscale difference value characterizes the degree of grayscale difference among the pixels in the pixel region. The grayscale difference values ​​corresponding to each pixel region are mapped into the effective image to obtain the surface feature model.

3. The processing method according to claim 2, characterized in that, The step of comparing the grayscale values ​​corresponding to each pixel in each of the pixel regions to obtain the grayscale difference value includes: In each pixel region, the grayscale values ​​corresponding to each pixel are compared, and the difference between the largest and smallest grayscale values ​​is taken as the grayscale difference value.

4. The processing method according to claim 2, characterized in that, The step of comparing the grayscale values ​​corresponding to each pixel in each of the pixel regions to obtain the grayscale difference value includes: In each pixel region, the average grayscale value of each pixel is taken as the grayscale difference value.

5. The processing method according to claim 2, characterized in that, The step of comparing the grayscale values ​​corresponding to each pixel in each of the pixel regions to obtain the grayscale difference value includes: In each pixel region, the weighted average of the gray values ​​of each pixel is used as the gray difference value, wherein the weights of the largest and smallest gray values ​​are greater than the weights of the non-largest and non-smallest gray values.

6. The processing method according to claim 2, characterized in that, The step of coloring the surface feature model to generate a colored model includes: The pixel region is colored based on the grayscale difference value of the pixel region, wherein the pixel regions of different colors correspond to different grayscale difference values.

7. The processing method according to claim 2, characterized in that, The step of coloring the pixel region based on the grayscale difference value of the pixel region includes: If the grayscale difference value of the pixel region is greater than a first preset value but not greater than a second preset value, the pixel region is rendered as a first color. If the grayscale difference value of the pixel region is greater than the second preset value, the pixel region is rendered as the second color.

8. The processing method according to any one of claims 1 to 7, characterized in that, Also includes: Based on the defect identification results, instructions for adjusting the upstream process parameters are generated and issued.

9. The processing method according to claim 8, characterized in that, The defect identification results output based on the coloring model include: The information from the coloring model is input into the trained defect classification model; Based on the defect classification model, the defect category information corresponding to the coloring model is determined to determine the defect category corresponding to the silicon wafer, wherein the defect identification result includes the defect category.

10. The processing method according to claim 9, characterized in that, The step of generating and issuing instructions for adjusting upstream process parameters based on the defect identification results includes: The defect identification results are sent to the manufacturing execution system; The manufacturing execution system matches a pre-stored process adjustment strategy based on the defect identification result; The process adjustment strategy is converted into instructions and sent to the upstream process equipment to adjust the parameters of the equipment.

11. A system for processing surface defects on silicon wafers, characterized in that, include: The acquisition module is configured to acquire image information, wherein the image information characterizes the defect features of the silicon wafer surface, and the defect features characterize the difference information of peaks and valleys on the surface of the silicon wafer within a preset unit area; The processing module is configured to process the image information to convert the image information into a surface feature model, wherein the surface feature model includes gray values ​​corresponding to the defect features, and different defect features correspond to different gray values; A coloring module is configured to perform coloring processing on the surface feature model to generate a coloring model, wherein the coloring model includes a variety of preset colors, and different colors correspond to different grayscale values; and The output module is configured to output defect identification results based on the coloring model.

12. The processing system according to claim 11, characterized in that, Also includes: The adjustment module is configured to generate and issue instructions for adjusting the upstream process parameters based on the defect identification results.

13. A computer-readable storage medium having computer-executable instructions stored thereon, characterized in that, When the executable instructions are executed by the processor, they implement the processing method as described in any one of claims 1 to 10.

14. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is used to execute the processing method according to any one of claims 1 to 10.

15. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the processing method as described in any one of claims 1 to 10.

Images