Semi-solid based semiconductor product dedusting method and dedusting system

Abstract

The application discloses a semi-solid-based semiconductor product dust removal method and a dust removal system, wherein the method comprises the following steps: performing difference regionalization processing on a surface image of a semiconductor product to be dusted in real time to obtain a dust condensation image; the difference regionalization processing is designed to diffuse the difference of outlying pixel points in the surface image, so that the discrete point dust is combined into a region; performing image segmentation on the dust condensation image to obtain a region to be identified; obtaining a dust region according to the gray distribution feature and the contour feature of each region to be identified in the surface image; dynamically determining a dust removal strategy of each dust region based on the use length of a dust removal rod, the size feature of each dust region and the corresponding gray gradient feature; and controlling a functional component matched with the dust removal rod to execute the corresponding dust removal strategy on each dust region, thereby improving the dust removal effect on the semiconductor device.

Description

Technical Field

[0001] This invention relates to the field of semiconductor device dust removal technology, and in particular to a method and system for removing dust from semiconductor products based on semi-solidified solids. Background Technology

[0002] During the manufacturing and use of semiconductor devices, various impurities accumulate on their surfaces. Dust buildup can affect the performance stability of precision semiconductor devices, shorten their lifespan, cause electrical failures, and even lead to malfunctions, ultimately rendering the semiconductor devices unusable. Dust removal from the surfaces of precision semiconductor devices can effectively prevent these problems, ensuring the devices operate at their best and improving the reliability and durability of the overall system.

[0003] However, since existing technologies use fixed dust removal modes to remove dust from semiconductor devices, they may cause secondary pollution to the semiconductor devices, resulting in poor dust removal efficiency.

[0004] Therefore, improving the dust removal effect on semiconductor devices has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a method and system for dust removal of semiconductor products based on semi-solidified solids, in order to solve the technical problem that the existing technology uses a fixed dust removal mode to remove dust from semiconductor devices, which may cause secondary pollution to the semiconductor devices and result in poor dust removal effect.

[0006] To address the aforementioned technical problems, embodiments of the present invention provide a method for dust removal from semiconductor products based on semi-solidified solids.

[0007] During the dust removal process, the surface image of the semiconductor product to be dusted is acquired in real time, and the surface image is processed by differential regionalization to obtain a dust aggregation image. The differential regionalization is designed to diffuse the differences of outlier pixels in the surface image to merge discretely distributed point dust into regions.

[0008] The dust agglomeration image is segmented to obtain various regions to be identified;

[0009] Each dust region is obtained based on the contour features of each region to be identified in the surface image; the dust removal strategy for each dust region is dynamically determined based on the usage time of the semi-solid dust removal rod, the size features of each dust region, and the corresponding grayscale gradient features.

[0010] The dust removal strategy includes a dust removal mode;

[0011] The dust removal modes include a contact dust removal mode and a non-contact dust removal mode. The contact dust removal mode is designed such that the dust removal rod and the surface dust of the semiconductor product are in contact, but the dust removal rod and the semiconductor product are not in contact. The non-contact dust removal mode is designed such that the dust removal rod and the semiconductor product, as well as the surface dust of the semiconductor product, are not in contact.

[0012] The control unit, which is matched with the dust removal bar, executes the corresponding dust removal strategy for each dusty area.

[0013] As one preferred embodiment, the step of performing differential regionization processing on the surface image to obtain a dust aggregation image includes:

[0014] Based on the grayscale difference between each pixel in the surface image and the overall image, and the grayscale difference between each pixel in the surface image and its neighborhood range, a two-level filtering process is performed on all pixels in the surface image to obtain each outlier pixel and each background pixel.

[0015] The similarity of the two outlier pixels within the same region is calculated based on their positional proximity and corresponding grayscale similarity.

[0016] The similarity of belonging to the same region is used as the similarity criterion for the clustering algorithm. The clustering algorithm is used to cluster all the outlier pixels to obtain each point cluster.

[0017] The background pixels are filtered based on the distance between each background pixel and each point cluster to obtain each pixel to be fused;

[0018] Based on the gray values ​​of all the outlier pixels, the gray value of each pixel to be merged is updated by fusion to obtain the fused gray weighted value of each pixel to be merged.

[0019] The grayscale value of each pixel to be fused in the surface image is replaced with the fused grayscale weighted value to obtain a dust agglomeration image.

[0020] As one preferred embodiment, the two-stage screening process is designed as follows:

[0021] The average grayscale value of all pixels in the surface image is used as the semiconductor pixel reference value; the absolute value of the difference between the grayscale value of each pixel in the surface image and the semiconductor pixel reference value is used as the overall grayscale difference value of each pixel in the surface image.

[0022] The overall grayscale difference value is used to perform a first-level comparison and screening of all pixels in the surface image to obtain each singular pixel and each background pixel.

[0023] A corresponding neighborhood window is constructed with each singular pixel as the center, and the neighborhood grayscale difference value of each singular pixel is calculated based on the grayscale difference features of each singular pixel and all pixels in the corresponding neighborhood window;

[0024] The outlier pixels in the surface image are selected by a two-stage comparison using the neighborhood grayscale difference value.

[0025] As one preferred embodiment, the fusion update process is designed as follows:

[0026] The point cluster with the smallest distance to each of the pixels to be merged is taken as the cluster to be merged into each of the pixels to be merged.

[0027] The distance between each pixel to be fused and all outlier pixels in the corresponding cluster to be integrated is normalized to obtain the contribution weight of each outlier pixel in the cluster to be integrated; the gray values ​​of all outlier pixels in the cluster to be integrated for each pixel to be fused are weighted and summed using the contribution weight to obtain the fused gray weight value of each pixel to be fused.

[0028] As one preferred embodiment, obtaining each dust region based on the contour features of each region to be identified in the surface image includes:

[0029] Extract the contour line of each region to be identified, number all pixels on each contour line in a clockwise direction, and calculate the directional change rate of each pixel on each contour line;

[0030] The standard deviation of the directional change rate of all pixels on each contour line is used as the contour complexity index of each contour line; the contour complexity index is used to filter all the regions to be identified to obtain each dust region.

[0031] As one preferred embodiment, the dynamic determination of the dust removal strategy for each dust region based on the usage time of the semi-solidified dust removal bar, the size characteristics of each dust region, and the corresponding grayscale gradient characteristics includes:

[0032] The pixel with the largest difference between the reference value of the semiconductor pixel and the dust region is taken as the dust accumulation point of each dust region.

[0033] The rays connecting the dust accumulation points of each dust region to each pixel on the corresponding contour line are used as dust accumulation lines for each dust region. The dust accumulation index of each dust region is calculated based on the degree of grayscale gradation of the pixels on all dust accumulation lines of each dust region. The dust accumulation index reflects the grayscale gradation characteristics of the corresponding dust region.

[0034] Based on the linear relationship between the dust accumulation index of each dust region and the total number of pixels in each dust region, the dust cleaning urgency index of each dust region is calculated; wherein, the total number of pixels in each dust region reflects the size characteristics of the corresponding dust region;

[0035] The usage time of the semi-solidified dust removal rod and the dust cleaning urgency index of each dust area are input into the mode switching control expression to calculate the non-contact mode switching factor of each dust area.

[0036] The dust removal strategy for each of the dust areas is adjusted using the non-contact mode switching factor.

[0037] As one of the preferred options

[0038] The adjustment of the dust removal strategy for each dust area using the non-contact mode switching factor includes:

[0039] When the non-contact mode switching factor is less than a preset threshold, the dust removal mode of the corresponding dust area is adjusted to the contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level one.

[0040] When the non-contact mode switching factor is greater than or equal to the preset threshold, the dust removal mode of the dust area is adjusted to the non-contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level two.

[0041] As one preferred embodiment, the calculation of the dust accumulation index for each dust region based on the gradient of grayscale values ​​of all pixels on the dust accumulation lines of each dust region includes:

[0042] The difference between the gray value of each pixel on each dust accumulation line and the gray value of the corresponding next pixel is used as the gray gradient value of each pixel on each dust accumulation line.

[0043] The sum of the grayscale gradient values ​​of all pixels on each dust accumulation line is taken as the linear gradient intensity of each dust accumulation line, and the average value of the linear gradient intensity of all dust accumulation lines in each dust region is taken as the dust accumulation index of each dust region.

[0044] As one preferred embodiment, the mode switching control expression is:

[0045]

[0046] in, This is the non-contact dust removal mode switching factor for dusty area j. For the normalization function, Let be the usage time of the semi-solidified dust collector bar, and exp be an exponential function with the natural constant as the base. The urgency index for dust cleaning in dusty area j.

[0047] Another embodiment of the present invention provides a semiconductor product dust removal system based on semi-solidified solids. The system includes a control console, a robotic arm, and a semi-solidified solid dust removal rod. The robotic arm is fixedly mounted in the control area of ​​the control console, and the dust removal rod is placed in the storage area of ​​the control console. A computer device is installed inside the counter of the control console. The computer device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. The processor is communicatively connected to the robotic arm. When the processor executes the computer program, it controls the robotic arm to grasp the semi-solidified solid dust removal rod to realize the semiconductor product dust removal method based on semi-solidified solids as described above.

[0048] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following:

[0049] During the dust removal process, surface images of the semiconductor products to be dusted are acquired in real time. Considering the particulate nature of dust, which may present a discrete, dot-like distribution, to avoid segmenting large, clustered dust areas into discrete smaller dust areas, leading to size recognition errors and subsequent incorrect dust removal strategies, the surface images are first processed by differential regionization to obtain dust aggregation images. Differential regionization is designed to diffuse the differences in outlier pixels in the surface image, merging discretely distributed dust dots into regions, thus improving the accuracy of subsequent semiconductor device segmentation. Image segmentation is then performed on the dust aggregation images to obtain individual regions to be identified. Considering that dust distribution is often irregular and has complex contours, while the contours of components on the semiconductor device surface are relatively simple, each dust region is obtained based on the contour features of each region to be identified in the surface image. The segmentation is based on the usage time of the semi-solid dust removal rod and the size of each dust region. The dust removal strategy for each dust area is dynamically determined based on the characteristics and corresponding grayscale gradient characteristics. The dust removal strategy includes dust removal modes, which include contact dust removal mode and non-contact dust removal mode. The contact dust removal mode is designed so that the dust removal rod contacts the surface dust of the semiconductor product, but the dust removal rod does not contact the semiconductor product itself. The non-contact dust removal mode is designed so that the dust removal rod does not contact the semiconductor product, nor does the surface dust of the semiconductor product. When the non-contact mode switching factor is less than a preset threshold, the dust removal mode for the corresponding dust area is adjusted to contact dust removal mode, and the dust removal intensity for the corresponding dust area is adjusted to level one. When the non-contact mode switching factor is greater than or equal to the preset threshold, the dust removal mode for the dust area is adjusted to non-contact dust removal mode, and the dust removal intensity for the corresponding dust area is adjusted to level two. This targeted real-time adjustment of the dust removal strategy based on the dust distribution on the semiconductor surface improves the dust removal effect on semiconductor devices. Attached Figure Description

[0050] Figure 1 This is a schematic flowchart of a semiconductor product dust removal method based on semi-solidified solids according to one embodiment of the present invention.

[0051] Figure 2 This is a schematic diagram comparing the image segmentation results of a dust agglomeration image and a surface image in one embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram of a semi-solidified dust collector bar in one embodiment of the present invention;

[0053] Figure 4 This is a schematic diagram of dust removal for a semiconductor device in one embodiment of the present invention;

[0054] Figure 5 This is a schematic diagram illustrating the execution of a semiconductor product dust removal method based on semi-solidified solids in one embodiment of the present invention;

[0055] Figure 6 This is a schematic diagram of a dust accumulation line in one embodiment of the present invention;

[0056] Figure 7 This is a schematic diagram illustrating the acquisition of a dust removal strategy in one embodiment of the present invention;

[0057] Figure 8 This is a schematic diagram of a control console in a dust removal system based on semi-solidified solids according to one embodiment of the present invention;

[0058] Figure 9 This is a schematic diagram of the hardware structure of a semiconductor product dust removal system based on semi-solidified solids in one embodiment of the present invention.

[0059] Figure 10 This is a schematic diagram of the structure of a computer device in one embodiment of the present invention;

[0060] Figure 11 This is a schematic diagram of a non-contact dust removal operation performed on the semi-solidified dust removal rod in one embodiment of the present invention.

[0061] Figure 12 This is a schematic diagram of the dust removal head and the semi-solidified solid placement platform in one embodiment of the present invention;

[0062] Figure 13 This is a schematic diagram of a real-time display of a dust interface on a circuit board in one embodiment of the present invention.

[0063] Figure 14 This is a schematic diagram of the automatic installation of the dust removal head and the dust removal head picking up semi-solidified solids in one embodiment of the present invention.

[0064] Figure 15 This is a schematic diagram of a semiconductor dust removal operation performed on a work platform after the semi-solidified solid has been installed, according to one embodiment of the present invention.

[0065] Label Explanation:

[0066] 101. Computer equipment; 102. Robotic arm; 103. Visual inspection module; 104. First moving module; 105. Second moving module. Detailed Implementation

[0067] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0068] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more.

[0069] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. The terms "vertical," "horizontal," "left," "right," "upper," "lower," and similar expressions used herein are for illustrative purposes only and do not indicate or imply that the device or component referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0070] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this specification is for the purpose of describing specific embodiments only and is not intended to limit the invention. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0071] One embodiment of the present invention provides a dust removal method for semiconductor products based on semi-solidified solids. For details, please refer to [link to relevant documentation]. Figure 1 , Figure 1 The diagram shows a flow chart of a semiconductor product dust removal method based on semi-solidified solids, according to one embodiment of the present invention.

[0072] Step S201: During the dust removal process, the surface image of the semiconductor product to be dusted is acquired in real time, and the surface image is processed by differential regionalization to obtain a dust agglomeration image.

[0073] During the dust removal process, the surface image of the semiconductor product to be dusted is acquired in real time. Considering the particulate characteristics of dust, it may present a discrete distribution of dots. In order to avoid the size recognition error caused by the dotted but clustered dust areas, and thus adopt the wrong dust removal strategy, the surface image is first processed by differential regionalization. The gray values ​​of the pixels belonging to the background in the dust area are replaced with gray values ​​similar to those of the surrounding pixels belonging to the dust.

[0074] It should be noted that the differential localization process is designed to diffuse the differences of outlier pixels in the surface image in order to merge discretely distributed point-like dust particles into regions.

[0075] Step S202: Perform image segmentation on the dust agglomeration image to obtain each region to be identified.

[0076] It should be noted that by using differential regionalization processing, the dust particles that are distributed in clusters but are dots in the surface image appear as a large area of ​​dust in the dust agglomeration image. At this time, image segmentation of the dust agglomeration image can completely divide the dust particles that are distributed in clusters, thus avoiding errors in the size identification of the dust area.

[0077] Specifically, an image segmentation algorithm is used to segment the dust agglomeration image to obtain various regions to be identified. For details, please refer to [link to relevant documentation]. Figure 2 , Figure 2 The diagram shows a comparison of image segmentation results between a dust agglomeration image and a surface image in one embodiment of the present invention. Figure 2 The image on the left shows the image segmentation results of the surface image, dividing each point-like but clustered dust particle into separate regions to be identified, region 1 and region 2. Figure 2 The image segmentation result of the dust agglomeration image is shown on the right. The dust particles, which are distributed in clusters but are dotted, are divided into a whole region to be identified.

[0078] It should be noted that, due to the larger size of area 3 to be identified, a non-contact dust removal mode should be used to avoid secondary pollution; while areas 1 and 2 to be identified are smaller and will most likely use a contact dust removal mode to improve dust removal speed. Without differentiated regional processing, area 3 might be mistakenly identified as dust from smaller areas like areas 1 and 2, leading to the use of contact dust removal for these smaller areas. This would cause dust from surrounding smaller areas to adhere to the dust removal rod during contact dust removal of area 1 or 2, resulting in secondary pollution.

[0079] It should be noted that image segmentation algorithms are well-known technologies, and will not be described in detail in this embodiment.

[0080] Step S203: Obtain each dust region based on the contour features of each region to be identified in the surface image, and dynamically determine the dust removal strategy for each dust region based on the usage time of the semi-solid dust removal rod, the size features of each dust region, and the corresponding grayscale gradient features.

[0081] It should be noted that dust distribution is often irregular and has complex outlines, while the outlines of components on the surface of semiconductor devices are relatively simple. Therefore, the complexity of the outlines can be used to accurately identify dust areas.

[0082] Specifically, each dust region is obtained based on the contour features of each region to be identified in the surface image.

[0083] It should be noted that the dust removal strategy includes dust removal modes, which include contact dust removal mode and non-contact dust removal mode. The contact dust removal mode is designed so that the dust removal rod and the surface dust of the semiconductor product come into contact, but the dust removal rod and the semiconductor product do not come into contact. The non-contact dust removal mode is designed so that the dust removal rod and the semiconductor product, as well as the surface dust of the semiconductor product, do not come into contact.

[0084] For details, please see Figure 3 , Figure 3 The diagram shows a semi-solidified dust collector bar in one embodiment of the present invention. For details, please refer to [link / reference]. Figure 4 , Figure 4 The diagram shown is a schematic diagram of dust removal for a semiconductor device in one embodiment of the present invention.

[0085] It should be further noted that contact dust removal mode may cause dust or contaminants to adhere to the surface of the dust removal rod. If not cleaned in time, it may cause secondary pollution in subsequent dust removal processes. By reducing the frequency of contact dust removal, the cleaning cycle of the dust removal rod can be extended, reducing the risk of secondary pollution. Non-contact dust removal mode causes less pollution to the dust removal rod, which can reduce the frequency of cleaning the dust removal rod and thus improve cleaning efficiency.

[0086] Specifically, the dust removal strategy for each dust area is dynamically determined based on the usage time of the semi-solid dust removal rod, the size characteristics of each dust area, and the corresponding grayscale gradient characteristics.

[0087] It should be noted that semi-solid (also known in the industry as viscous semi-solid, see patent applications CN202021203118.1 entitled "A Dust Removal Head" and CN202110693409.6 entitled "Dust Removal Head, Viscous Dust Removal Device, Dust Removal Device and Method") is a substance in a state between solid and liquid, possessing a certain degree of viscosity and fluidity. In air, the surface of the semi-solid gradually dries, and during the drying process, the internal molecular motion intensifies, forming a strong adsorption capacity. This semi-solid is a dust-catching adhesive (full name: dust-catching adhesive), and it is existing technology. The composition and preparation process of the dust-catching adhesive are specifically described in the Chinese invention patent application CN114672277A entitled "Dust-catching adhesive for cleanroom and its preparation method." This semi-solid is common knowledge in the field and is not an improvement of this application; therefore, its composition and preparation process need not be elaborated upon.

[0088] The working principle of the non-contact dust removal mode is as follows: a robotic arm controls a dust removal rod to reach the target position for dust removal. A pre-set distance is maintained between the semi-solidified solid at the bottom of the dust removal rod and the dust. The semi-solidified solid continues to dry in the air, releasing energy during the drying process. The adsorption force generated by the internal molecular motion can adsorb the dust on the surface of the semiconductor product onto the head of the semi-solidified dust removal rod, thereby achieving dust removal from the surface of the semiconductor product. Figure 3 As shown, the semi-solid on the dust collector stick is close to the dust, and the dust can be captured by utilizing the electrostatic adsorption force of the semi-solid. It should be noted that the dust adsorption principle of the dust-collecting adhesive (i.e., the semi-solid) is not an improvement of this application, and is common knowledge.

[0089] like Figure 11 As shown, the robotic arm moves the dust removal rod above the semiconductor device on the operating platform, ensuring that the vertical distance between the semi-condensed solids on the dust removal rod and the surface of the semiconductor device is between 0.1 mm and 0.2 mm. This allows for non-contact dust removal of the semi-condensed solids on the dust removal rod. Figure 12 As shown, each dust collector bar and semi-solidified solid is placed on the platform, and each dust collector bar is of a different size. The controller can control the robotic arm to grab the corresponding size dust collector bar for dust removal according to specific needs, such as... Figure 13 As shown, the device's camera module communicates with the server, sending images of dust on the circuit board. A monitor connected to the server displays the dust interface on the circuit board in real time. Figure 14As shown, the equipment automatically installs dust removal heads of the corresponding size using a robotic arm, and controls the dust removal heads to pick up semi-solidified solids from the adjacent box.

[0090] The working principle of contact dust removal mode is as follows: a robotic arm controls a dust removal rod to reach the target location for dust removal. The semi-solid material at the bottom of the dust removal rod comes into contact with the dust. During the drying process of the semi-solid material, energy is released, and the adsorption force generated by the internal molecular motion can adsorb the dust on the surface of the semiconductor product onto the head of the semi-solid dust removal rod, thereby achieving dust removal from the surface of the semiconductor product. Figure 15 As shown, after the equipment is equipped with dust removal rods and semi-solidified solids, it moves the dust removal rods to the work platform and performs contact dust removal on the surface of each semiconductor device according to the preset cleaning path.

[0091] When the semi-solidified dust removal bar is used for a long time and the size characteristics of the dust area and the corresponding gray scale gradient characteristics are more obvious, a non-contact dust removal mode is adopted to reduce the risk of secondary pollution; when the semi-solidified dust removal bar is used for a short time and the size characteristics of the dust area and the corresponding gray scale gradient characteristics are less obvious, a contact dust removal mode is adopted to improve cleaning efficiency.

[0092] Step S204: Control the functional components that match the dust removal bar to execute the corresponding dust removal strategy for each dust area.

[0093] Specifically, the functional components that control and match the dust removal bar execute corresponding dust removal strategies for each dusty area. As an embodiment of this application, a simulation model is used to conduct simulation experiments on the dust removal strategies for each dusty area, in which the functional components that control and match the dust removal bar execute corresponding dust removal strategies for each dusty area.

[0094] It should be noted that conducting experiments on dust removal strategies through simulation models can verify the feasibility and effectiveness of the strategies before practical application, thereby reducing the risks in actual application.

[0095] The semiconductor product dust removal method in this embodiment performs differential regionalization processing on real-time acquired images of the semiconductor product surface, merging discretely distributed point-like dust particles into regions. An image segmentation algorithm is then used to completely divide the dust regions in the dust aggregation image, ensuring accurate identification of dust regions and avoiding size recognition errors caused by dust cluster distribution. Based on the size characteristics, grayscale gradient characteristics, and usage time of the dust removal rod, a contact or non-contact dust removal mode is dynamically selected. For larger dust areas, contact between the dust removal rod and the surface is avoided to reduce the risk of secondary pollution. The dust removal intensity is adjusted according to the actual situation of the dust area to ensure cleaning effectiveness. This achieves intelligent and refined dust removal strategy, significantly reducing the risk of secondary pollution and improving dust removal efficiency and effect. Before practical application, the feasibility and effectiveness of the dust removal strategy are verified through a simulation model to reduce risks in actual application.

[0096] For details, please see Figure 5 , Figure 5 The diagram illustrates the execution of a semiconductor product dust removal method based on semi-solidified solids in one embodiment of the present invention. Figure 5 In this diagram, 'a' represents the AI ​​system, which executes steps S202 and S203; 'b' represents the camera, which executes step S201; 'c' represents the execution system, which executes step S204; and 'd' represents the chip.

[0097] In one embodiment, in Figure 1 In step S201, the surface image is processed by differential regionalization to obtain a dust aggregation image, including steps S301 to S303:

[0098] Step S301: Based on the grayscale difference between each pixel in the surface image and the overall image, and the grayscale difference between each pixel in the surface image and its neighborhood range, perform a two-level filtering process on all pixels in the surface image to obtain each outlier pixel and each background pixel.

[0099] It should be noted that background pixels refer to pixels in the surface image that belong to the background and are located within the dust area, while outlier pixels refer to pixels in the surface image that belong to the dust and are located within the dust area.

[0100] In this step, the two-stage screening process is designed as follows:

[0101] The average grayscale value of all pixels in the surface image is used as the semiconductor pixel reference value. The absolute value of the difference between the grayscale value of each pixel in the surface image and the semiconductor pixel reference value is used as the overall grayscale difference value of each pixel in the surface image. All pixels in the surface image are compared and filtered using the overall grayscale difference value to obtain each singular pixel and each background pixel. A corresponding neighborhood window is constructed with each singular pixel as the center. The neighborhood grayscale difference value of each singular pixel is calculated based on the grayscale difference features of each singular pixel and all pixels in the corresponding neighborhood window. All singular pixels in the surface image are compared and filtered using the neighborhood grayscale difference value to obtain each outlier pixel.

[0102] It's important to note that the first step involves a primary comparison screening process. This process calculates the difference between each pixel and the overall grayscale reference value, initially identifying all pixels potentially belonging to dust regions as singular pixels, thus reducing the possibility of missed detections. This step uses the overall image grayscale average as a reference value, adapting to different lighting conditions and background variations, avoiding errors caused by fixed thresholds, and quickly separating pixels that might belong to dust regions. The second step involves a secondary comparison screening process that constructs a neighborhood window and calculates the grayscale difference value within that neighborhood. This combines local grayscale features to more accurately determine whether a pixel belongs to a dust region, effectively avoiding misclassifying background pixels as dust pixels and reducing the false detection rate. The primary screening quickly narrows the detection range, while the secondary screening is performed only on singular pixels, avoiding complex calculations across all pixels in the image and improving algorithm efficiency.

[0103] As one embodiment of this application, the average value of the overall grayscale difference of all pixels in the surface image is used as the overall grayscale difference reference value, and all pixels with an overall grayscale difference value greater than or equal to the overall grayscale difference reference value are designated as singular pixels. All singular pixels are binary-classified based on all neighborhood grayscale difference values. The average value of the neighborhood grayscale difference of all singular pixels within each category is used as the neighborhood average grayscale difference. All singular pixels within the category with the largest neighborhood average grayscale difference are designated as outlier pixels, and all singular pixels within the category with the smallest neighborhood average grayscale difference are designated as background pixels.

[0104] Step S302: Calculate the same region similarity between two outlier pixels based on their positional proximity and corresponding grayscale similarity; use the same region similarity as the similarity criterion for the clustering algorithm, and use the clustering algorithm to cluster all outlier pixels to obtain each point cluster.

[0105] It should be noted that considering the spatial distance between outlier pixels based on location proximity allows dust pixels with similar locations to be grouped into the same cluster, avoiding the incorrect merging of scattered dust regions. Similarly, considering the grayscale similarity between outlier pixels allows them to be grouped into the same cluster, avoiding the incorrect merging of dust regions with significant grayscale differences. Combining location proximity and grayscale similarity more accurately reflects the correlation between dust pixels, thus achieving precise clustering of dust regions. This merges scattered pixels belonging to the same dust region into a single cluster, avoiding over-segmentation caused by pixel dispersion, thereby improving the accuracy of subsequent dust size determination.

[0106] As an embodiment of this application, the absolute value of the difference in gray levels between two outlier pixels is taken as the gray level difference value between the two pixels, the sum of the positional distance between the two outlier pixels and the corresponding gray level difference value is taken as the corresponding regional affiliation dissimilarity index, and the calculation result of the exponential function with the natural constant as the base and the negative number of the regional affiliation dissimilarity index as the exponent is taken as the regional affiliation similarity between the two outlier pixels.

[0107] Step S303: Filter background pixels based on the distance between each background pixel and each point cluster to obtain each pixel to be fused; perform fusion update processing on the gray value of each pixel to be fused according to the gray value of all outlier pixels to obtain the fused gray weight value of each pixel to be fused; replace the gray value of each pixel to be fused in the surface image with the fused gray weight value to obtain the dust agglomeration image.

[0108] It should be noted that by calculating the distance between background pixels and dot-like clusters, background pixels close to the dust area can be filtered out, i.e., the pixels to be fused. This filtering mechanism ensures that only background pixels related to the dust area are included in the fusion process, avoiding interference with irrelevant background areas. The dust agglomeration image can better adapt to the dispersed, clustered, or irregular distribution of dust areas, clearly showing the distribution and morphology of dust areas, providing more intuitive support for the assessment of dust size, thereby improving the accuracy of subsequent dust removal strategy determination.

[0109] In this step, the merge update process is designed as follows:

[0110] The point cluster with the smallest distance to each pixel to be merged is taken as the cluster to be merged for each pixel to be merged; the distance between each pixel to be merged and all outlier pixels in the corresponding cluster to be merged is normalized to obtain the contribution weight of each outlier pixel in the cluster to be merged; the gray values ​​of all outlier pixels in the cluster to be merged for each pixel to be merged are weighted and summed using the contribution weight to obtain the fusion gray weight value of each pixel to be merged.

[0111] It should be noted that by using the cluster of points closest to the pixel to be fused as the cluster to be merged, it can be ensured that the pixel to be fused is merged with the most relevant dust area, avoiding the incorrect association of the pixel to be fused with irrelevant dust areas. Normalizing the distance between the pixel to be fused and the outlier pixels in the cluster to be merged can convert the distance information into contribution weights. The closer the outlier pixels are, the greater their contribution to the fused gray value. This can dynamically reflect the degree of influence of outlier pixels on the pixel to be fused, avoiding obvious edges or discontinuous areas in the fused image. The fused dust agglomeration image can more clearly show the characteristics of the dust area, making it easier to count the number, area, distribution and other information of the dust area.

[0112] The semiconductor product dust removal method in this embodiment initially screens outomas by calculating the difference between each pixel and the overall grayscale reference value, reducing the possibility of missed detections. Further screening of outliers is achieved by constructing a neighborhood window and calculating the grayscale difference value of the neighborhood, reducing the false detection rate. This two-stage screening process significantly improves the accuracy of dust detection and reduces missed and false detections. Combining location proximity and grayscale similarity, the similarity between outliers is calculated as a similarity metric for the clustering algorithm, which more accurately reflects the correlation between dust pixels and avoids over-segmentation caused by pixel dispersion. A clustering algorithm is then used to cluster all outliers into point-like clusters, achieving precise division of dust areas. A fusion update process ensures that only background pixels related to the dust area are included in the fusion process, avoiding interference with irrelevant background areas. The grayscale values ​​of the pixels to be fused are replaced with the weighted grayscale values ​​of the fusion process, generating a dust aggregation image that clearly shows the distribution and morphology of the dust area, providing more intuitive support for dust size assessment and thus improving the reliability of dust removal mode determination.

[0113] In one embodiment, in Figure 1 In step S203, each dust region is obtained based on the contour features of each region to be identified in the surface image, including:

[0114] Extract the contour line of each region to be identified in the surface image, number all pixels on each contour line in a clockwise direction, and calculate the direction change rate of each pixel on each contour line; use the standard deviation of the direction change rate of all pixels on each contour line as the contour complexity index of each contour line; use the contour complexity index to filter all regions to be identified to obtain each dust region.

[0115] It should be noted that by mapping the regions to be identified segmented from the dust agglomeration image to the corresponding regions in the surface image, it is possible to analyze the point-like but clustered dust regions as a whole while preserving the original surface image features. Dust regions often have irregular shapes and complex contours. Further calculation of the directional change rate of each pixel on the contour line quantifies the local geometric features of the contour line, reflecting its curvature and trend. The standard deviation of the directional change rate is used as the contour complexity index, which comprehensively reflects the complexity of the entire contour line. Dust regions typically have irregular contours and thus a high contour complexity index, while components on semiconductor surfaces have a lower contour complexity index.

[0116] As an embodiment of this application, the average value of the contour complexity index of all regions to be identified is used as the contour complexity reference threshold, and all regions to be identified corresponding to contour complexity indices greater than or equal to the contour complexity reference threshold are regarded as dust regions.

[0117] In one embodiment, in Figure 1 In step S203, the dust removal strategy for each dust area is dynamically determined based on the usage time of the semi-solidified dust removal rod, the size characteristics of each dust area, and the corresponding grayscale gradient characteristics, including steps S401 to S403:

[0118] Step S401: The pixel with the largest difference from the semiconductor pixel reference value in each dust region is taken as the dust accumulation point of each dust region; the ray obtained by connecting the dust accumulation point of each dust region with each pixel on the corresponding contour line is taken as each dust accumulation line of each dust region; the dust accumulation index of each dust region is calculated based on the degree of grayscale value gradation of the pixels on all dust accumulation lines of each dust region; wherein, the dust accumulation index reflects the grayscale gradation characteristics of the corresponding dust region.

[0119] It should be noted that by selecting the pixel with the largest difference from the semiconductor pixel reference value as the dust accumulation point, the core location of the dust area can be accurately reflected, that is, the area where the dust accumulation is most severe. Based on the gray value change of the pixels on the dust accumulation line, the gray value gradient characteristics of the dust area can be quantified. The dust area usually has a gray value gradient from the center to the edge, while the gray value change of the background area or regular area is more uniform.

[0120] The dust accumulation index for each dust region is calculated based on the gradient of grayscale values ​​of all pixels along the dust accumulation lines in each dust region, including:

[0121] The difference between the grayscale value of each pixel on each dust accumulation line and the grayscale value of the corresponding next pixel is used as the grayscale gradient value of each pixel on each dust accumulation line; the sum of the grayscale gradient values ​​of all pixels on each dust accumulation line is used as the linear gradient intensity of each dust accumulation line; and the average of the linear gradient intensities of all dust accumulation lines in each dust region is used as the dust accumulation index of each dust region.

[0122] It should be noted that the difference between the gray value of each pixel on the dust accumulation line and the gray value of the corresponding next pixel reflects the radial gray value change of the dust area. The dust accumulation line can reflect the distribution of dust from the center to the edge. Through the radially distributed dust accumulation line, the gray value gradient characteristics of the dust area can be captured more comprehensively, which indirectly reflects the thickness of dust accumulation and improves the accuracy of subsequent assessment of the severity of dust accumulation.

[0123] For details, please see Figure 6 , Figure 6 The diagram shown is a schematic representation of a dust accumulation line in one embodiment of the present invention.

[0124] Step S402: Based on the linear relationship between the dust accumulation index of each dust region and the total number of pixels in each dust region, the dust cleaning urgency index of each dust region is calculated; wherein, the total number of pixels in each dust region reflects the size characteristics of the corresponding dust region.

[0125] It should be noted that the dust accumulation index reflects the grayscale gradient characteristics of the dust area, while the total number of pixels in the dust area reflects the size characteristics of the dust area. By combining the dust accumulation index and the size of the dust area, the severity of dust accumulation in the dust area can be comprehensively assessed, thereby improving the accuracy of subsequent dust removal strategy determination.

[0126] Among them, the product of the dust accumulation index of each dust area and the total number of pixels in each dust area can be used as the dust cleaning urgency index of each dust area.

[0127] It should be noted that the Dust Cleaning Urgency Index is used to assess the severity of dust accumulation. By comprehensively considering the thickness and size of the dust accumulation area, it improves the accuracy of assessing the severity of dust accumulation, thereby improving the reliability of determining subsequent dust removal strategies.

[0128] Step S403: Input the usage time of the semi-solidified dust removal rod and the dust cleaning urgency index of each dust area into the mode switching control expression to calculate the non-contact mode switching factor for each dust area; adjust the dust removal strategy for each dust area based on the non-contact mode switching factor.

[0129] It should be noted that each semi-solidified dust removal bar is equipped with a timer, which can record the usage time of the semi-solidified dust removal bar during its use. The surface of semiconductor devices may include multiple dust areas. When using dust removal bars to remove dust from the surface of semiconductor devices, for each dust area, it is necessary to comprehensively consider the characteristics of the dust area and the usage of the dust removal bar. Through the mode switching control expression, the non-contact mode switching factor is automatically calculated and the dust removal strategy is adjusted to reduce manual intervention and reduce the risk of secondary pollution caused by improper operation.

[0130] In this step, the mode switching control expression is:

[0131]

[0132] in, This is the non-contact dust removal mode switching factor for dusty area j. For the normalization function, Let be the usage time of the semi-solidified dust collector bar, and exp be an exponential function with the natural constant as the base. The urgency index for dust cleaning in dusty area j.

[0133] It should be noted that the urgency level of dust cleaning is... The range of values ​​for is from 0 to positive infinity. The value range is specified as 0 to 1. The normalization function restricts the value of the non-contact dust removal mode switching factor to between 0 and 1, effectively preventing gradient explosion and protecting the stability and reliability of the model. When the usage time of the semi-solid dust removal rod is shorter and the severity of dust accumulation is lower, it indicates that less dust is adsorbed on the surface of the semi-solid dust removal rod. In this case, the improvement in cleaning efficiency between contact dust removal and non-contact dust removal is more significant, and contact dust removal mode should be used more often to increase the dust removal rate. The smaller the non-contact dust removal mode switching factor value, the better. Conversely, when the usage time of the semi-solid dust removal rod is longer and the severity of dust accumulation is higher, it indicates that the surface of the semi-solid dust removal rod is more likely to have adsorbed more dust, leading to weakened adsorption capacity. In this case, the improvement in cleaning efficiency between contact dust removal and non-contact dust removal is not significant, and non-contact dust removal mode should be used more often to reduce the risk of secondary pollution. The larger the non-contact dust removal mode switching factor value, the better.

[0134] The semiconductor product dust removal method in this embodiment selects the pixel with the largest difference from the semiconductor pixel reference value as the dust accumulation point, which can accurately reflect the core location of the dust area and achieve precise positioning of the dust accumulation point. By calculating the dust accumulation index, a quantitative assessment of dust accumulation is achieved, reflecting the gray-scale gradient characteristics of the dust area and improving the accuracy of assessing the severity of dust accumulation. In addition, by combining the total number of pixels in the dust area and comprehensively considering the thickness and size of the dust accumulation area, the severity of dust accumulation can be assessed more comprehensively, providing a reliable basis for determining the dust removal strategy. Based on the usage time of the semi-solid dust removal rod and the dust cleaning urgency index, the non-contact mode switching factor is calculated to realize the automatic adjustment of the dust removal strategy. While significantly reducing the risk of secondary pollution, the dust removal efficiency is improved, the cleaning cycle is shortened, and the cleaning cost is reduced.

[0135] In one embodiment, step S403 adjusts the dust removal strategy for each dust area using a non-contact mode switching factor, including:

[0136] When the non-contact mode switching factor is less than the preset threshold, the dust removal mode of the corresponding dust area is adjusted to the contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level one.

[0137] When the non-contact mode switching factor is greater than or equal to the preset threshold, the dust removal mode of the dusty area is adjusted to the non-contact dust removal mode, and the dust removal intensity of the corresponding dusty area is adjusted to level two.

[0138] It should be noted that by combining non-contact dust removal mode with secondary dust removal intensity, the dust removal rate can be increased as much as possible while reducing the risk of secondary pollution.

[0139] It should be noted that the preset threshold is a value preset by the user. In this embodiment, the preset threshold is 0.5. For example, when the usage time of the semi-condensed solid dust collector is 0, the non-contact mode switching factor is 0, the contact dust removal mode is adopted, and the dust removal intensity is adjusted to level one; when the dust cleaning urgency index is 0.5 and the usage time of the semi-condensed solid dust collector is 15, the non-contact mode switching factor is 0.39, the contact dust removal mode is adopted, and the dust removal intensity is adjusted to level one; when the dust cleaning urgency index is 5 and the usage time of the semi-condensed solid dust collector is 15, the non-contact mode switching factor is 0.99, the non-contact dust removal mode is adopted, and the dust removal intensity is adjusted to level two.

[0140] For details, please see Figure 7 , Figure 7 The diagram shown illustrates the acquisition of a dust removal strategy in one embodiment of the present invention.

[0141] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following:

[0142] During the dust removal process, surface images of the semiconductor products to be dusted are acquired in real time. Considering the particulate nature of dust, which may present as discrete point-like distributions, to avoid misidentifying the size of point-like but clustered dust areas and thus adopting incorrect dust removal strategies, the surface images are first processed by differential regionization to obtain dust aggregation images. Differential regionization is designed to diffuse the differences of outlier pixels in the surface image, merging discretely distributed point-like dust into regions, improving the accuracy of subsequent semiconductor device segmentation. Image segmentation is then performed on the dust aggregation images to obtain individual regions to be identified. Considering that dust distribution is often irregular and has complex contours, while the contours of components on the semiconductor device surface are relatively simple, each dust region is obtained based on the contour features of each region to be identified in the surface image. The segmentation is based on the usage time of the semi-solid dust removal rod, the size characteristics of each dust region, and the corresponding... The grayscale gradient feature dynamically determines the dust removal strategy for each dust area. The dust removal strategy includes a dust removal mode, which includes a contact dust removal mode and a non-contact dust removal mode. The contact dust removal mode is designed so that the dust removal rod and the surface dust of the semiconductor product are in contact, but the dust removal rod and the semiconductor product are not in contact. The non-contact dust removal mode is designed so that the dust removal rod and the surface dust of the semiconductor product are not in contact. When the non-contact mode switching factor is less than a preset threshold, the dust removal mode of the corresponding dust area is adjusted to the contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level one. When the non-contact mode switching factor is greater than or equal to the preset threshold, the dust removal mode of the dust area is adjusted to the non-contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level two. This targeted real-time adjustment of the dust removal strategy based on the dust distribution on the semiconductor surface improves the dust removal effect on semiconductor devices.

[0143] One embodiment of the present invention provides a dust removal system for semiconductor products based on semi-solidified solids. For details, please refer to [link to documentation]. Figure 8 , Figure 8 The diagram shows a control console in a semi-solidified dust removal system according to one embodiment of the present invention. The dust removal system includes a control console, a robotic arm, and semi-solidified dust removal rods. The robotic arm is fixedly mounted in the control area of ​​the control console, and the dust removal rods are placed in the storage area of ​​the control console. A computer device is installed inside the console counter. The computer device includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. The processor is communicatively connected to the robotic arm. When the processor executes the computer program, it controls the robotic arm to grasp the semi-solidified dust removal rods to realize the semi-solidified semiconductor product dust removal method described above.

[0144] As one embodiment of this application, the system further includes a visual inspection module, a first motion module, and a second motion module. For details, please refer to [link to relevant documentation]. Figure 9 , Figure 9 The diagram shows a hardware structure schematic of a semiconductor product dust removal system based on semi-solidified solids, according to one embodiment of the present invention. Figure 9 The computer device 101 is installed inside the control panel counter. The first output of the computer device 101 is connected to the robot arm 102 through the first moving module 104, and the second output of the computer device 101 is connected to the vision inspection module 103 through the second moving module 105. When the system is working: the computer device 101 controls the second moving module 105 to move the vision inspection module 103 to the target position, acquire the image of the target being tested, detect the dust position, dust area and dust removal strategy (contact or non-contact) in the image, determine the size of the dust removal rod to be used, and then control the first moving module 104 to move the robot arm 102 to the top of the dust removal rod in the corresponding storage area, grab the dust removal rod of the matching size, and then control the robot arm 102 to move to the dust area to perform the dust removal operation.

[0145] For details, please see Figure 10 , Figure 10 The diagram shown is a structural schematic of a computer device according to one embodiment of the present invention. Figure 10 As shown, the computer device of this embodiment includes: at least one processor ( Figure 10 Only one is shown in the diagram), a memory, and a computer program stored in the memory and executable on at least one processor, wherein the processor executes the computer program to implement the steps in any of the above-described embodiments of the anomaly display detection method.

[0146] This computer device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 10 The examples of computer devices are merely examples and do not constitute a limitation on computer devices. Computer devices may include more or fewer components than shown in the illustration, or combinations of certain components, or different components, such as network interfaces, displays, and input devices.

[0147] The processor referred to can be a CPU, but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0148] Memory includes readable storage media, internal memory, etc., wherein internal memory can be the RAM of a computer device, providing an environment for the operation of the operating system and computer-readable instructions stored in the readable storage media. The readable storage media can be the hard drive of a computer device, or in other embodiments, it can be an external storage device of the computer device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card. Furthermore, memory can include both internal storage units and external storage devices of the computer device. Memory is used to store the operating system, applications, bootloader, data, and other programs, such as program code for computer programs. Memory can also be used to temporarily store data that has been output or will be output.

[0149] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for dust removal from semiconductor products based on semi-solidified solids, characterized in that, The method includes: During the dust removal process, the surface image of the semiconductor product to be dusted is acquired in real time, and the surface image is processed by differential regionalization to obtain a dust aggregation image. The differential regionalization is designed to diffuse the differences of outlier pixels in the surface image to merge discretely distributed point dust into regions. The dust agglomeration image is segmented to obtain various regions to be identified; Each dust region is obtained based on the contour features of each region to be identified in the surface image; the dust removal strategy for each dust region is dynamically determined based on the usage time of the semi-solidified dust removal rod, the size features of each dust region, and the corresponding grayscale gradient features. The dust removal strategy includes a dust removal mode; The dust removal modes include a contact dust removal mode and a non-contact dust removal mode. The contact dust removal mode is designed such that the dust removal rod and the surface dust of the semiconductor product are in contact, but the dust removal rod and the semiconductor product are not in contact. The non-contact dust removal mode is designed such that the dust removal rod and the semiconductor product, as well as the surface dust of the semiconductor product, are not in contact. The dust removal bars of the corresponding size are controlled to execute the corresponding dust removal strategy for each of the dust areas; The dust removal strategy for each dust region is dynamically determined based on the usage time of the semi-solidified dust removal bar, the size characteristics of each dust region, and the corresponding grayscale gradient characteristics, including: The average grayscale value of all pixels in the surface image is used as the reference value for the semiconductor pixel. The pixel with the largest difference between the reference value of the semiconductor pixel and the dust region is taken as the dust accumulation point of each dust region. The rays connecting the dust accumulation points of each dust region to each pixel on the corresponding contour line are used as dust accumulation lines for each dust region. The dust accumulation index of each dust region is calculated based on the degree of grayscale gradation of the pixels on all dust accumulation lines of each dust region. The dust accumulation index reflects the grayscale gradation characteristics of the corresponding dust region. Based on the linear relationship between the dust accumulation index of each dust region and the total number of pixels in each dust region, the dust cleaning urgency index of each dust region is calculated; wherein, the total number of pixels in each dust region reflects the size characteristics of the corresponding dust region; The usage time of the semi-solidified dust removal rod and the dust cleaning urgency index of each dust area are input into the mode switching control expression to calculate the non-contact mode switching factor of each dust area. The dust removal strategy for each of the dust areas is adjusted using the non-contact mode switching factor.

2. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 1, characterized in that, The step of performing differential regionization processing on the surface image to obtain a dust aggregation image includes: Based on the grayscale difference between each pixel in the surface image and the overall image, and the grayscale difference between each pixel in the surface image and its neighborhood range, a two-level filtering process is performed on all pixels in the surface image to obtain each outlier pixel and each background pixel. The similarity of the two outlier pixels within the same region is calculated based on their positional proximity and corresponding grayscale similarity. The similarity of belonging to the same region is used as the similarity criterion for the clustering algorithm. The clustering algorithm is used to cluster all the outlier pixels to obtain each point cluster. The background pixels are filtered based on the distance between each background pixel and each point cluster to obtain each pixel to be fused; Based on the gray values ​​of all the outlier pixels, the gray value of each pixel to be merged is updated by fusion to obtain the fused gray weighted value of each pixel to be merged. The grayscale value of each pixel to be fused in the surface image is replaced with the fused grayscale weighted value to obtain a dust agglomeration image.

3. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 2, characterized in that, The two-level screening process is designed as follows: The absolute value of the difference between the gray value of each pixel in the surface image and the reference value of the semiconductor pixel is taken as the overall gray value difference of each pixel in the surface image; The overall grayscale difference value is used to perform a first-level comparison and screening of all pixels in the surface image to obtain each singular pixel and each background pixel. A corresponding neighborhood window is constructed with each singular pixel as the center, and the neighborhood grayscale difference value of each singular pixel is calculated based on the grayscale difference features of each singular pixel and all pixels in the corresponding neighborhood window; The outlier pixels in the surface image are selected by a two-stage comparison using the neighborhood grayscale difference value.

4. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 2, characterized in that, The fusion update process is designed as follows: The point cluster with the smallest distance to each of the pixels to be merged is taken as the cluster to be merged into each of the pixels to be merged. The distance between each pixel to be fused and all outlier pixels in the corresponding cluster to be fused is normalized to obtain the contribution weight of each outlier pixel in the cluster to be fused. The gray values ​​of all outlier pixels in the cluster to be merged for each pixel to be merged are weighted and summed using the contribution weight to obtain the fused gray weight value for each pixel to be merged.

5. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 3, characterized in that, The step of obtaining each dust region based on the contour features of each region to be identified in the surface image includes: Extract the contour line of each region to be identified, number all pixels on each contour line in a clockwise direction, and calculate the directional change rate of each pixel on each contour line; The standard deviation of the directional change rate of all pixels on each contour line is used as the contour complexity index of each contour line; the contour complexity index is used to filter all the regions to be identified to obtain each dust region.

6. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 1, characterized in that, The adjustment of the dust removal strategy for each dust area using the non-contact mode switching factor includes: When the non-contact mode switching factor is less than a preset threshold, the dust removal mode of the corresponding dust area is adjusted to the contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level one. When the non-contact mode switching factor is greater than or equal to the preset threshold, the dust removal mode of the dust area is adjusted to the non-contact dust removal mode, and the dust removal intensity of the corresponding dust area is adjusted to level two.

7. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 1, characterized in that, The calculation of the dust accumulation index for each dust region based on the gradient of grayscale values ​​of all pixels on the dust accumulation lines in each dust region includes: The difference between the gray value of each pixel on each dust accumulation line and the gray value of the corresponding next pixel is used as the gray gradient value of each pixel on each dust accumulation line. The sum of the grayscale gradient values ​​of all pixels on each dust accumulation line is taken as the linear gradient intensity of each dust accumulation line, and the average value of the linear gradient intensity of all dust accumulation lines in each dust region is taken as the dust accumulation index of each dust region.

8. The method for dust removal from semiconductor products based on semi-solidified solids according to claim 1, characterized in that, The mode switching control expression is: in, This is the non-contact dust removal mode switching factor for dusty area j. For the normalization function, Let be the usage time of the semi-solidified dust collector bar, and exp be an exponential function with the natural constant as the base. The urgency index for dust cleaning in dusty area j.

Images