A method and system for dynamic detection of crystal growth

Abstract

The application discloses a kind of dynamic detection method and system of crystal growth, and the present application relates to the technical field of crystal detection, corresponding crystal growth scene is determined based on the scene identification of each crystal growth area, according to the morphology of multiple crystal characteristics, corresponding time node and corresponding crystal growth scene determine crystal growth dynamic graph, guarantee the accuracy of crystal growth dynamic graph.Therefore, based on the identification of abnormal area of crystal growth dynamic graph to determine the abnormal part of crystal characteristics, according to the abnormal part of the crystal characteristics to determine the abnormal growth parameter of the crystal characteristics in the crucible according to the abnormal part of the crystal characteristics;Based on the traversal of the abnormal part of the crystal characteristics to determine abnormal change amount, and according to abnormal change amount and the spatial position of the abnormal part of the crystal characteristics trigger the local compensation measure of the crystal characteristics of the crucible, guarantee the accuracy of the local compensation measure of the crystal characteristics of the crucible, realize the dynamic detection and autonomous optimization of the crystal characteristics.

Description

Technical Field

[0001] The present invention relates to the technical field of crystal detection, and in particular to a dynamic detection method and system for crystal growth. Background Art

[0002] With the development of science and technology, crystals are widely used as semiconductor components in electronic chips. Crystals gradually grow in a crucible. In existing technologies, the crystals are placed in a crucible, which gradually cultivates the crystals during operation. After the crystals are cultivated, they are photographed to capture the final state of the crystal characteristics. This method does not consider the state of the crystal characteristics at different time points, ignores abnormal locations of the crystal characteristics, and cannot achieve dynamic detection and autonomous optimization of the crystal characteristics. Summary of the Invention

[0003] The purpose of the present invention is to overcome the deficiencies of the prior art, and the present invention provides a method and system for dynamic detection of crystal growth.

[0004] In a first aspect, the present invention provides a method for dynamic detection of crystal growth, comprising:

[0005] The crystal is in the crucible, multiple internal images of the crucible are collected in a preset time period, and the crystal growth area in each internal image is marked;

[0006] Determining multiple crystal features based on crystal identification of each crystal growth region, wherein the multiple crystal features correspond to different time nodes;

[0007] Determine the corresponding crystal growth scene based on scene recognition of each crystal growth area, and determine the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes and corresponding crystal growth scenes;

[0008] In the crystal growth dynamic graph, the abnormal portion of the crystal feature is determined based on the identification of the abnormal region of the crystal growth dynamic graph, and the abnormal growth parameter of the crystal feature in the crucible is determined based on the tracing back of the abnormal portion of the crystal feature;

[0009] Based on the traversal of the abnormal part of the crystal feature, the abnormal change amount is determined, and according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature, the crucible triggers the local compensation measure of the crystal feature to trigger the autonomous optimization of the abnormal part of the crystal feature.

[0010] Preferably, the crystal is in a crucible, and collecting multiple internal images of the crucible in a preset time period and marking the crystal growth area in each internal image includes:

[0011] The crystal is placed on the positioning component of the crucible and dynamically grows in the crucible. At this time, the crucible is monitored in real time and the interior of the crucible is triggered to be photographed within a preset time period to determine multiple internal images of the crucible within the preset time period;

[0012] In each internal image, the spatial position of the crystal is determined based on detection of the internal image, a first sub-region is determined based on the spatial position of the crystal and the growth stage of the crystal, and a second sub-region is determined based on the spatial position of the crystal and the internal spatial morphology of the crucible;

[0013] In each internal image, a crystal growth region in each internal image is determined based on a mapping relationship among the first subregion, the second subregion, and the growth region, so as to mark the crystal growth region in each internal image.

[0014] Preferably, the multiple crystal features are determined based on the crystal identification of each crystal growth region, and the multiple crystal features correspond to different time nodes, including:

[0015] In each crystal growth region, the corresponding crystal outer contour is determined based on the traversal of the crystal growth region, and the corresponding crystal recognition mode is determined according to the crystal outer contour and the crystal model, and the crystal recognition mode includes a ring detection mode, a local detection mode and an overall detection mode;

[0016] triggering crystal recognition of a crystal growth region along the crystal recognition pattern to output a plurality of sub-crystal features;

[0017] The corresponding crystal feature is determined by synthesizing multiple sub-crystal features and images corresponding to the crystal growth area. At this time, each crystal feature is compared with an internal image, and the shooting time of the internal image is used as the time node corresponding to the crystal feature.

[0018] Preferably, the determining of the corresponding crystal growth scene based on scene recognition of each crystal growth area, and determining the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes and corresponding crystal growth scenes, include:

[0019] In each crystal growth area, the regional position of the crystal growth area is collected, and the corresponding scene area is determined based on the regional position of the crystal growth area, the morphology of the crystal growth area and the position of the crystal characteristics. The corresponding crystal growth scene is determined based on the scene recognition of the scene area.

[0020] Preferably, the method of determining the corresponding crystal growth scene based on scene recognition of each crystal growth area, and determining the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes and corresponding crystal growth scenes, further includes:

[0021] Determine a dynamic change graph of crystal features based on the morphology of multiple crystal features and corresponding time nodes; determine an environmental change graph of crystal features based on the morphology of multiple crystal features and corresponding crystal growth scenarios;

[0022] The crystal growth dynamics graph is determined based on the synthesis of the dynamics change graph of the crystal characteristics and the environmental change graph of the crystal characteristics.

[0023] Preferably, in the crystal growth dynamic graph, determining the abnormal portion of the crystal feature based on identification of an abnormal region in the crystal growth dynamic graph, and determining the abnormal growth parameter of the crystal feature in the crucible by tracing back the abnormal portion of the crystal feature, includes:

[0024] The crystal growth dynamic graph is monitored in real time, a plurality of abnormal sub-regions are determined according to the traversal of the crystal growth dynamic graph, and the abnormal region of the crystal growth dynamic graph is determined according to the synthesis of the regional positions and regional morphologies of the plurality of abnormal sub-regions.

[0025] Preferably, the method further comprises: determining an abnormal portion of a crystal feature based on identification of an abnormal region in the crystal growth dynamic graph, and determining abnormal growth parameters of the crystal feature in the crucible based on tracing back the abnormal portion of the crystal feature.

[0026] Identifying an abnormal region of the crystal growth dynamic graph, determining an abnormal portion of the crystal characteristic based on the identification of the abnormal region of the crystal growth dynamic graph, and determining corresponding abnormal information based on the detection of the abnormal portion;

[0027] The abnormal information corresponding to the abnormal part is determined based on the position, corresponding morphology and corresponding internal image of the abnormal part. The growth stage and corresponding time node of the crystal are determined based on the tracing of the abnormal information corresponding to the abnormal part. The abnormal growth parameters of the crystal characteristics in the crucible are determined based on the growth stage of the crystal, the corresponding time node and the crucible database.

[0028] Preferably, the determining of the abnormal variation based on the traversal of the abnormal portion of the crystal feature, and triggering the crucible to take local compensation measures for the crystal feature according to the abnormal variation and the spatial position of the abnormal portion of the crystal feature to trigger autonomous optimization of the abnormal portion of the crystal feature, includes:

[0029] Collecting the abnormal part of the crystal feature, determining the theoretical morphology of the part of the crystal feature based on matching the abnormal part of the crystal feature with a crucible database, and determining the abnormal change amount based on comparison between the abnormal part of the crystal feature and the theoretical morphology of the part of the crystal feature;

[0030] The abnormal type of the crystal feature in different dimensions is determined based on the abnormal change amount and the abnormal location of the crystal feature. The first compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the spatial position of the abnormal location. The second compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the current morphology of the crystal.

[0031] Preferably, the determining of the abnormal variation based on the traversal of the abnormal portion of the crystal feature, and triggering the crucible to take local compensation measures for the crystal feature according to the abnormal variation and the spatial position of the abnormal portion of the crystal feature to trigger autonomous optimization of the abnormal portion of the crystal feature, further includes:

[0032] In the crucible, a local compensation measure of the crucible for the crystal feature is determined based on the first compensation measure, the second compensation measure and the compensation mapping relationship, the working parameters of the crucible are adjusted based on the execution of the local compensation measure, and the position of the abnormal part of the crystal feature is targetedly adjusted to achieve autonomous optimization of the abnormal part of the crystal feature until the abnormal part of the crystal feature is adjusted to a normal form.

[0033] In a second aspect, the present invention further provides a dynamic detection system for crystal growth, which is applied to the above-mentioned dynamic detection method for crystal growth, and the dynamic detection system for crystal growth comprises:

[0034] An internal image module is used to collect multiple internal images of the crucible in a preset time period when the crystal is in the crucible, and mark the crystal growth area in each internal image;

[0035] A crystal feature module is used to determine multiple crystal features based on crystal identification of each crystal growth region, where the multiple crystal features correspond to different time nodes;

[0036] A crystal growth dynamic graph module is used to determine the corresponding crystal growth scene based on scene recognition of each crystal growth area, and to determine the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes;

[0037] An abnormal location module is used to determine the abnormal location of the crystal feature based on the identification of the abnormal area in the crystal growth dynamic diagram, and to determine the abnormal growth parameters of the crystal feature in the crucible based on the tracing of the abnormal location of the crystal feature;

[0038] The local compensation module is used to determine the abnormal change amount based on the traversal of the abnormal part of the crystal feature, and trigger the crucible to take local compensation measures for the crystal feature according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature, so as to trigger the autonomous optimization of the abnormal part of the crystal feature.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] In the present invention, through the dynamic detection method of crystal growth, the corresponding crystal growth scene is determined based on the scene recognition of each crystal growth area, and the crystal growth dynamic graph is determined according to the morphology of multiple crystal features, the corresponding time nodes and the corresponding crystal growth scene. It is compatible with the overall consideration of the morphology of multiple crystal features, the corresponding time nodes and the corresponding crystal growth scene, ensures the accuracy of the crystal growth dynamic graph, and presents the status of the crystal features at different time nodes.

[0041] In the crystal growth dynamic graph, the abnormal part of the crystal feature is determined based on the identification of the abnormal area of ​​the crystal growth dynamic graph, and the abnormal growth parameters of the crystal feature in the crucible are determined based on the tracing of the abnormal part of the crystal feature; the abnormal change amount is determined based on the traversal of the abnormal part of the crystal feature, and the crucible's local compensation measures for the crystal feature are triggered according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature to trigger the autonomous optimization of the abnormal part of the crystal feature. The abnormal part of the crystal feature is introduced to ensure the accuracy of the crucible's local compensation measures for the crystal feature, and realize the dynamic detection and autonomous optimization of the crystal feature. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 1 is a schematic flow chart of a dynamic detection method for crystal growth in an embodiment of the present invention;

[0043] Figure 2 It is a flow chart of step S11 in the present invention;

[0044] Figure 3 It is a schematic flow chart of step S12 in the present invention;

[0045] Figure 4 It is a flow chart of step S13 in the present invention;

[0046] Figure 5 It is a schematic flow chart of step S14 in the present invention;

[0047] Figure 6 It is a schematic flow chart of step S15 in the present invention;

[0048] Figure 7 It is a structural schematic diagram of a dynamic detection system for crystal growth in an embodiment of the present invention.

[0049] Description of reference numerals:

[0050] 21. Internal image module; 22. Crystal feature module; 23. Crystal growth dynamic diagram module; 24. Abnormal part module; 25. Local compensation module. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present invention will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present invention.

[0052] Reference Figures 1 to 6 A dynamic detection method for crystal growth is applied to dynamic detection scenarios of crystal growth; the dynamic detection method for crystal growth includes:

[0053] Step S11: The crystal is in the crucible, multiple internal images of the crucible are collected in a preset time period, and the crystal growth area in each internal image is marked;

[0054] Step S12: determining a plurality of crystal features based on crystal identification of each crystal growth region, wherein the plurality of crystal features correspond to different time nodes;

[0055] Step S13: determining a corresponding crystal growth scene based on scene recognition of each crystal growth area, and determining a crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes;

[0056] Step S14: in the crystal growth dynamic graph, determining the abnormal location of the crystal feature based on the identification of the abnormal area of ​​the crystal growth dynamic graph, and determining the abnormal growth parameters of the crystal feature in the crucible by tracing back the abnormal location of the crystal feature;

[0057] Step S15: Determine the abnormal variation based on the traversal of the abnormal part of the crystal feature, and trigger the crucible to take local compensation measures for the crystal feature according to the abnormal variation and the spatial position of the abnormal part of the crystal feature to trigger autonomous optimization of the abnormal part of the crystal feature.

[0058] Reference Figure 2 In step S11, the crystal is in the crucible, a plurality of internal images of the crucible in a preset time period are collected, and the crystal growth area in each internal image is marked;

[0059] In the specific implementation process of the present invention, the specific steps are:

[0060] S111: The crystal is placed on the positioning component of the crucible and dynamically grows in the crucible. At this time, the crucible is monitored in real time, and internal photography of the crucible is triggered within a preset time period to determine multiple internal images of the crucible within the preset time period.

[0061] S112: In each internal image, determine the spatial position of the crystal based on detection of the internal image, determine a first sub-region based on the spatial position of the crystal and the growth stage of the crystal, and determine a second sub-region based on the spatial position of the crystal and the internal spatial shape of the crucible;

[0062] S113: In each internal image, determine the crystal growth region in each internal image based on the mapping relationship among the first sub-region, the second sub-region, and the growth region, so as to mark the crystal growth region in each internal image.

[0063] In an embodiment of the present application, a crystal is placed on a positioning component of a crucible and dynamically grows in the crucible. At this time, the crucible is monitored in real time, and internal shooting of the crucible is triggered within a preset time period to determine multiple internal images of the crucible within the preset time period, thereby introducing multiple internal images of the crucible within the preset time period.

[0064] At this time, the crucible is usually designed with special positioning components, such as brackets, grooves or clamping devices, to ensure that the crystal maintains a stable position and posture during the growth process. These positioning components are customized according to the shape, size and growth characteristics of the crystal. At this time, the operator places the pre-processed crystal (single crystal seed, polycrystalline powder or other forms) on the positioning components; avoid damaging the crystal during the placement process and ensure good contact between it and the positioning components.

[0065] The crucible is placed in a controllable growth environment, such as a heating furnace, atmosphere control box, etc. These devices adjust growth conditions such as temperature, atmosphere, and pressure to meet the growth requirements of the crystal. Under the set growth conditions, the crystal begins to grow dynamically, and its morphology, structure, and properties change over time.

[0066] Monitoring equipment (such as high-definition cameras, infrared thermal imagers, etc.) is installed outside or inside the crucible to monitor the growth status of the crystal in real time; the position, angle and focal length of the monitoring equipment are adjusted according to the shape of the crucible and the growth characteristics of the crystal; according to the growth rate of the crystal and the required monitoring accuracy, a preset time period (such as every few minutes, hours, etc.) is set to trigger shooting; the selection of the shooting time interval needs to balance the real-time nature of monitoring and the amount of data storage; when the preset time period is reached, the monitoring equipment automatically triggers shooting, captures the image of the inside of the crucible, and transmits the image in real time to a computer or data storage device for subsequent analysis and processing.

[0067] Furthermore, in each internal image, the spatial position of the crystal is determined based on the detection of the internal image, the first sub-region is determined according to the spatial position of the crystal and the growth stage of the crystal, and the second sub-region is determined according to the spatial position of the crystal and the internal spatial morphology of the crucible, which is compatible with the overall consideration of the spatial position of the crystal and the internal spatial morphology of the crucible and ensures the accuracy of the second sub-region.

[0068] At this point, the collected internal images are processed using image processing software or algorithms, such as grayscale conversion, filtering and denoising, and edge detection, to enhance the visibility and clarity of the crystals in the image. Through methods such as template matching, shape analysis, or machine learning, the identified features are compared with preset crystal models or samples in the database to identify the features of the crystals in the image (such as edges, contours, textures, etc.), and then determine the spatial position of the crystals. The determined spatial position of the crystals is recorded in the form of coordinates, which represent the position of the center, edge, or specific point of the crystal for subsequent analysis and processing.

[0069] Based on the spatial position of the crystal and known growth conditions (such as temperature, time, atmosphere, etc.), combined with theoretical knowledge of crystal growth and historical data, the current growth stage of the crystal (such as the initial stage, stable stage, end stage, etc.) is identified; according to the identified growth stage and the spatial position of the crystal, a sub-region related to the current growth stage (the first sub-region) is divided in the image. The first sub-region covers the main growth part of the crystal, or the region related to the specific growth characteristics of the crystal (such as morphological changes, defect generation, etc.).

[0070] CAD models, 3D scanning, or image processing technology are used to obtain information about the internal spatial morphology of the crucible, including its shape, size, internal structure, etc.; combined with the spatial position of the crystal and the internal spatial morphology of the crucible, another sub-region related to the crystal growth environment (the second sub-region) is divided in the image. The second sub-region takes into account the influence of the crucible shape on crystal growth (such as temperature gradient, atmosphere distribution, etc.), or the area related to the interaction between the crystal and the crucible (such as heat conduction, material exchange, etc.).

[0071] Specifically, image processing software is used to process the collected internal images to enhance the visibility of the crystals; edge detection and shape analysis are used to identify the outline and position of the crystals in the image; and the identified crystal positions are recorded in the form of coordinates, which represent the position of the crystal center for subsequent analysis and processing.

[0072] Determine the first sub-region based on the spatial position and growth stage of the crystal: combine known growth conditions (such as temperature curve, time nodes, etc.) and crystal growth characteristics (such as melting point, crystallization rate, etc.) to identify the current crystal growth stage (such as molten state, beginning of crystallization, stable growth, etc.); divide a sub-region related to the current growth stage in the image (the first sub-region). The first sub-region covers the part where the crystal begins to crystallize and grows stably, and is used to monitor the morphological changes and growth rate of the crystal.

[0073] The second subregion is determined based on the spatial position of the crystal and the internal spatial morphology of the crucible. The CAD model is used to obtain information about the internal spatial morphology of the crucible, including its shape, size, and internal structure. Combining the spatial position of the crystal and the morphology of the crucible, another subregion related to the crystal growth environment (the second subregion) is delineated within the image. This second subregion takes into account the influence of the crucible's shape on the temperature gradient and atmosphere distribution, as well as the heat conduction and material exchange processes between the crystal and the crucible. By determining the spatial position of the crystal and delineating subregions related to the growth stage and crucible morphology, the crystal growth process can be more accurately analyzed, providing strong support for subsequent feature extraction, anomaly detection, and optimization measures.

[0074] Therefore, in each internal image, the crystal growth area in each internal image is determined based on the mapping relationship between the first sub-area, the second sub-area and the growth area to mark the crystal growth area in each internal image, which is compatible with the overall consideration of the mapping relationship between the first sub-area, the second sub-area and the growth area, and ensures the accuracy of the crystal growth area in each internal image.

[0075] At this time, the growth area usually refers to the part of the crystal that is growing or has grown, which is identified by the crystal's morphology, color, texture and other characteristics. Based on the information of the first sub-area (related to the crystal growth stage) and the second sub-area (related to the internal space morphology of the crucible), as well as the known crystal growth characteristics and crucible design parameters, a mapping relationship is constructed. This mapping relationship describes how to determine the position and range of the crystal growth area based on the position and characteristics of the first sub-area and the second sub-area.

[0076] In each internal image, a mapping relationship is applied based on the positions and features of the first sub-region and the second sub-region previously determined. This can be accomplished by further processing and analysis of the image, such as image registration, feature extraction, and pattern recognition. By applying the mapping relationship, the crystal growth region in each internal image is identified, and the mapping relationship is continuously iteratively adjusted and optimized to obtain more accurate growth region identification results.

[0077] After the crystal growth area is determined, it is visually marked on the image by drawing a bounding box, contour line, color fill or other visual elements on the image; the marked growth area is stored together with the corresponding internal image for subsequent analysis and processing, and a database or file system is established to store these images and related information.

[0078] Specifically, the definition of the growth area is clearly defined, that is, the part of the crystal that is growing or has grown, and the growth area is identified by the morphological changes and color differences of the crystal. Then, based on the previously determined first sub-area (covering the part where the crystal begins to crystallize and stably grows) and the second sub-area (taking into account the influence of the crucible shape on the temperature gradient and atmosphere distribution), combined with the known crystal growth characteristics and crucible design parameters, a mapping relationship is constructed. This mapping relationship describes how to determine the position and range of the crystal growth area based on the position and characteristics of the first sub-area and the second sub-area.

[0079] In each internal image, the image was further processed and analyzed based on the mapping relationship, including image registration to ensure spatial consistency between different images, feature extraction to identify morphological changes and color differences in the crystals, and pattern recognition to apply the mapping relationship and identify the growth area; through this process, the growth area of ​​the crystal was successfully identified in each internal image.

[0080] Finally, the growth areas are marked on each internal image using color filling or bounding boxes. These markings enable more intuitive observation and analysis of the crystal growth process. These marked growth areas are stored together with the corresponding internal images in a database for subsequent analysis and processing. This database will be used to monitor key parameters such as crystal growth rate, morphological changes, and defect generation, and provide data support for optimizing crystal growth conditions.

[0081] In some embodiments of the present application, a growth region matching table is collected, and the growth region matching table is shown in Table 1:

[0082] Table 1 Growth area matching table

[0083] Image number First sub-region characteristics Second sub-region characteristics Growing area 001 Circular, center position Center of the bottom of the crucible Central circular area 002 Long, left-leaning Left side wall of crucible Left long area ... ... ... ... 010 Irregular shape, upper part Upper corner of the crucible Upper irregular area

[0084] refer to Figure 3 In step S12, a plurality of crystal features are determined based on crystal identification of each crystal growth region, and the plurality of crystal features correspond to different time nodes;

[0085] In the specific implementation process of the present invention, the specific steps are:

[0086] S121: In each crystal growth region, determining a corresponding crystal outer contour based on traversal of the crystal growth region, and determining a corresponding crystal recognition mode according to the crystal outer contour and the crystal model, the crystal recognition mode including a ring detection mode, a local detection mode, and an overall detection mode;

[0087] S122: triggering crystal recognition of the crystal growth region along the crystal recognition pattern to output a plurality of sub-crystal features;

[0088] S123: Determine the corresponding crystal feature based on the synthesis of multiple sub-crystal features and images corresponding to the crystal growth area. At this time, each crystal feature is compared with an internal image, and the shooting time of the internal image is used as the time node corresponding to the crystal feature.

[0089] In an embodiment of the present application, in each crystal growth area, the corresponding crystal outer contour is determined based on the traversal of the crystal growth area, and the corresponding crystal identification mode is determined according to the crystal outer contour and the crystal model. The crystal identification mode includes an annular detection mode, a local detection mode and an overall detection mode, which is compatible with the overall consideration of the crystal outer contour and the crystal model, and ensures the accuracy of the corresponding crystal identification mode.

[0090] At this time, traversing the crystal growth area means checking each pixel or small area in the image one by one to determine which areas belong to the crystal growth area. This is usually achieved using image processing techniques such as threshold segmentation, region growing, edge detection, etc. to separate the crystal area from the background. At this time, use image processing software or algorithms to load the internal image; apply appropriate preprocessing steps (such as denoising, contrast enhancement) to improve image quality; use threshold segmentation or edge detection algorithms to identify the boundary between the crystal and the background; and determine the complete crystal growth area through region growing or connected domain analysis.

[0091] After the crystal growth region is determined, the next step is to extract the outer contour of the crystal. The outer contour is the boundary line between the crystal and its surroundings, which provides important information about the crystal's shape and size. At this point, a contour extraction algorithm (such as the Canny edge detector) is applied based on the crystal growth region. The outermost contour is identified and extracted as the outer contour of the crystal. The extracted contour is smoothed to reduce noise and irregularities.

[0092] Crystal recognition mode refers to a specific strategy or method used to detect and analyze crystal features. The most appropriate recognition mode is selected based on the crystal's outer contour and model (i.e., crystal type, specifications, etc.). Ring detection mode: Suitable for crystals with ring-shaped or ring-like structures; this mode focuses on detecting ring features of the crystal, such as inner and outer diameters, ring width, etc.; Local detection mode: Suitable for situations where certain specific areas or features within the crystal are more important; this mode focuses on detecting local areas within the crystal, such as defects, texture changes, etc.; Overall detection mode: Suitable for situations where a comprehensive assessment of the crystal's overall shape, size, and texture is required; this mode comprehensively analyzes the crystal's overall characteristics. Optionally, the most appropriate recognition mode is selected based on the crystal's model and outer contour features; the corresponding image processing algorithm or machine learning model is configured to adapt to the selected mode; and preprocessing steps and parameter settings are prepared for subsequent crystal recognition and analysis.

[0093] Furthermore, crystal recognition of the crystal growth region is triggered along the crystal recognition pattern to output multiple sub-crystal features, thereby introducing the output of multiple sub-crystal features.

[0094] At this point, the appropriate crystal recognition mode is determined based on the outer contour and model of the crystal, and the corresponding image processing algorithm or machine learning model is selected and configured to adapt to the selected recognition mode, which involves adjusting the parameters of the algorithm, selecting the appropriate feature extraction method, etc.; at this point, according to the previously determined recognition mode (such as ring detection mode, local detection mode or overall detection mode), the corresponding image processing algorithm or machine learning model is selected; the algorithm parameters, such as threshold, filter size, feature selection, etc., are configured to adapt to the specific crystal recognition task.

[0095] After selecting and configuring a suitable crystal recognition pattern, the pattern is applied to the crystal growth area, and the pixels or small areas within the growth area are analyzed one by one to extract the characteristics of the crystal. At this point, the image data of the crystal growth area is input into the selected image processing algorithm or machine learning model. The algorithm or model will traverse the pixels or small areas within the growth area and extract the characteristics of the crystal according to the selected recognition pattern. The extracted characteristics include the size, shape, texture, color, defects, etc. of the crystal.

[0096] After applying the crystal recognition mode, multiple sub-crystal features will be obtained. These features reflect the characteristics of the crystal in different aspects, such as geometric shape, surface texture, color distribution, etc. These features will be used for subsequent crystal analysis and quality control; the extracted sub-crystal features will be sorted and classified for subsequent analysis and processing; the features will be further pre-processed, such as normalization, standardization or feature selection, to improve the effectiveness and accuracy of the features; the sub-crystal features will be output for use in subsequent steps.

[0097] Therefore, the corresponding crystal features are determined based on the synthesis of multiple sub-crystal features and images corresponding to the crystal growth area. At this time, each crystal feature is compared with an internal image, and the shooting time of the internal image is used as the time node corresponding to the crystal feature, which is compatible with the overall consideration of the synthesis of multiple sub-crystal features and images corresponding to the crystal growth area, ensuring the accuracy of the corresponding crystal features.

[0098] At this point, in step S122, multiple sub-crystal features have been extracted from the crystal growth region; in this step, the multiple sub-crystal features are synthesized or integrated into a complete crystal feature description, which can be achieved through techniques such as feature fusion, feature selection or feature dimensionality reduction; at this point, the extracted sub-crystal features are sorted to ensure that they correspond one-to-one to the corresponding crystal growth regions; feature fusion techniques (such as series, parallel or weighted summation) are applied to integrate multiple sub-features into a comprehensive feature vector; if the feature vector dimension is too high, consider using feature selection or feature dimensionality reduction methods (such as PCA, LDA) to reduce redundant information and improve the effectiveness of the features.

[0099] After the complete crystal feature description is synthesized, it is compared with the corresponding internal image to ensure the accuracy and consistency of the features. At this time, the synthesized crystal feature vector is compared with the pixel data of the internal image or the extracted image features. A similarity measurement method (such as cosine similarity and Euclidean distance) is used to calculate the degree of match between the feature vector and the image data. Based on the degree of match, the feature vector is adjusted or corrected to ensure its consistency with the internal image.

[0100] Recording the internal image capture time as the time node corresponding to the crystal feature helps understand the temporal evolution of the crystal during growth and provides a time reference for subsequent growth condition optimization and quality control. At this time, the internal image capture time is obtained from the image metadata or capture record. The capture time is associated with the corresponding crystal feature vector and recorded in the dataset. The crystal feature dataset containing the time node is output for subsequent analysis and processing.

[0101] Specifically, during the growth process, the size, shape, texture, and potential defects of the crystals are monitored to ensure the quality and performance of the final product. Multiple sub-features are extracted from images of the crystal growth area, including the crystal's diameter (150μm), height (50μm), texture contrast (0.75), and uniformity (0.90), as well as two tiny defects detected. These sub-features are then integrated into a comprehensive feature vector to reduce the dimensionality of the feature vector and improve computational efficiency.

[0102] The synthesized feature vector is compared with the corresponding internal image, and the cosine similarity between the feature vector and the image data is calculated; based on the similarity result, the feature vector is fine-tuned to ensure its consistency with the internal image; at the same time, the shooting time of the internal image is obtained from the image metadata, that is, a specific time point in the crystal growth process; the shooting time is associated with the corresponding crystal feature vector and recorded in a data set; the crystal feature data set containing the time node is output, which can be used for subsequent growth condition optimization, quality control and performance analysis; through this process, multiple crystal features can be accurately synthesized, compared and recorded, providing important data support for subsequent crystal growth research.

[0103] In some embodiments of the present application, a crystal matching table is collected, and the crystal matching table is shown in Table 2:

[0104] Table 2 Crystal matching table

[0105] Crystal ID Eigenvector (example) Internal Image ID Shooting time C001 Size: 150μm, Shape: Regular, Texture: 0.75, Defects: 2 IMG001 2023-04-01 C002 Size: 145μm, Shape: Slightly Irregular, Texture: 0.80, Defects: 1 IMG002 2023-04-02

[0106] At this point, assume the following weight distribution: Size: 0.4 (most important); Shape: 0.3 (second most important); Texture: 0.2 (less important); Defects: 0.1 (least important, but still influential);

[0107] For crystal C001, its comprehensive score is calculated as follows:

[0108] Size score: 150μm * 0.4 = 60; Shape score: Regularity 1.0 (assumed to be 1.0 because it is the optimal value) * 0.3 = 0.3; Texture score: 0.75 * 0.2 = 0.15; Defect score: 2 * (-0.1) = -0.2 (assumed that the greater the number of defects, the lower the score); Comprehensive score: 60 + 0.3 + 0.15 - 0.2 = 60.25; Therefore, the comprehensive score of crystal C001 is 60.25, and the corresponding time node is 2023-04-01; In this way, a comprehensive score is calculated for each crystal and recorded in the matching table together with the corresponding time node to provide data support for subsequent crystal growth analysis, quality control, and performance evaluation.

[0109] refer to Figure 4 In step S13, the corresponding crystal growth scene is determined based on the scene recognition of each crystal growth area, and the crystal growth dynamic diagram is determined according to the morphology of multiple crystal features, the corresponding time nodes and the corresponding crystal growth scene;

[0110] In the specific implementation process of the present invention, the specific steps are:

[0111] S131: In each crystal growth region, collecting the regional position of the crystal growth region, determining a corresponding scene region based on the regional position of the crystal growth region, the morphology of the crystal growth region, and the position of the crystal feature, and determining a corresponding crystal growth scene based on scene recognition of the scene region;

[0112] S132: Determine a dynamic change graph of the crystal features based on the morphologies of the multiple crystal features and corresponding time nodes; determine an environmental change graph of the crystal features based on the morphologies of the multiple crystal features and corresponding crystal growth scenarios;

[0113] S133: Determine a crystal growth dynamic graph based on a synthesis of a dynamic change graph of crystal characteristics and an environmental change graph of crystal characteristics.

[0114] In an embodiment of the present application, in each crystal growth area, the regional position of the crystal growth area is collected, and the corresponding scene area is determined based on the regional position of the crystal growth area, the morphology of the crystal growth area, and the position of the crystal characteristics. The corresponding crystal growth scene is determined based on the scene recognition of the scene area, which is compatible with the overall consideration of the regional position of the crystal growth area, the morphology of the crystal growth area, and the position of the crystal characteristics, thereby ensuring the accuracy of the corresponding scene area.

[0115] At this time, obtaining the precise location information of the crystal growth area can be achieved by using equipment such as the global positioning system (GPS), laser rangefinder, optical microscope or electron microscope. The goal is to determine the exact location of the crystal growth area in the laboratory, production line or nature, as well as the relative positional relationship between them; at this time, use GPS or laser rangefinder to determine the geographical location of the crystal growth area in the laboratory or production line; use an optical microscope or electron microscope to obtain an image of the crystal growth area, and extract the location information through image processing technology; record the location information, such as coordinates, distance or relative position relationship, for subsequent analysis.

[0116] After obtaining the regional position of the crystal growth area, the corresponding scene area is determined by combining the morphology of the crystal growth area (such as shape, size, boundary, etc.) and the position of the crystal features (such as defects, impurities, texture, etc.). This usually involves a detailed analysis of the image or data of the crystal growth area to identify regions with similar features and classify them as the same scene area; at this time, image processing technology (such as edge detection, texture analysis) is used to extract the morphological features of the crystal growth area; based on the position information of the crystal features, the crystal growth area is divided into different sub-regions, and regions with similar features are identified; combined with the regional position and morphological features, similar regions are classified as the same scene area, and a unique identifier is assigned to each scene area.

[0117] After determining the corresponding scene areas, scene recognition technology is used to determine the crystal growth scene corresponding to each scene area, which usually involves using a machine learning model or a deep learning algorithm to analyze the characteristics of the scene area and match it with the known growth scenes. At this time, images or data of known crystal growth scenes are collected, and a training set of the machine learning model or deep learning algorithm is constructed. The trained model or algorithm is used to analyze the characteristics of the scene area and predict the crystal growth scene corresponding to each scene area. Based on the prediction results, a crystal growth scene label is assigned to each scene area, such as "high-temperature melt growth", "vapor deposition growth", etc.

[0118] Furthermore, a dynamic change graph of crystal features is determined based on the morphologies of multiple crystal features and corresponding time nodes; an environmental change graph of crystal features is determined based on the morphologies of multiple crystal features and corresponding crystal growth scenarios, which is compatible with the overall consideration of the morphologies of multiple crystal features and corresponding time nodes, ensuring the accuracy of the dynamic change graph of crystal features.

[0119] At this time, based on the morphology of multiple crystal features (such as size, shape, texture, etc.) and their corresponding time nodes, a dynamic change diagram of the crystal features can be drawn. This can be done by collecting crystal feature data at a series of time points and displaying these data in a graphical way to observe and analyze how the crystal features change over time.

[0120] At this point, during the crystal growth process, morphological data of crystal characteristics are collected regularly or continuously, which can be achieved through automated measuring equipment, microscopic observation or image processing software; ensure that the exact time node is recorded each time data is collected so that data analysis and graph drawing can be performed in chronological order later; organize the collected data to ensure that the data format is consistent, and standardize data with different characteristics for comparison in the graph; use chart software or data visualization tools to draw dynamic change graphs based on time nodes and crystal characteristic data; common chart types include line charts, scatter charts or area charts, etc.

[0121] Alternatively, assume that you are studying the growth process of a crystal; regularly collect data on the size and shape of the crystals, and record the time node of each measurement; on the first day, the average size of the crystals is measured to be 50 microns, and the shape is relatively regular; on the third day, the crystal size grows to 75 microns, and the shape is still regular; on the fifth day, the crystal size reaches 100 microns, but the shape begins to become irregular; based on these data, draw a line graph showing the change of crystal size over time, and a schematic diagram showing the change of crystal shape from regular to irregular, so as to intuitively understand the dynamic changes of crystals during the growth process.

[0122] Based on the morphologies of multiple crystal features and their corresponding crystal growth scenarios, the environmental change map of the crystal features is inferred and drawn. This usually involves analyzing the relationship between the crystal feature morphology and the growth environment, such as temperature, pressure, atmosphere composition, etc., and using these relationships to predict or draw an environmental change map; at this time, understand the known relationship between different crystal feature morphologies and the growth environment; during the crystal growth process, collect environmental parameter data related to the characteristic morphology changes, including data provided by equipment such as temperature sensors, pressure gauges or gas analyzers; correlate the collected environmental parameter data with the crystal feature morphology data to find the correlation or causal relationship between them; based on the analysis results, use charting software or data visualization tools to draw an environmental change map to show how the environmental parameters change over time and the relationship between them and the changes in crystal feature morphology.

[0123] Specifically, it is known that high temperature and nitrogen atmosphere are conducive to crystal growth, while oxygen contamination leads to a decline in crystal quality; on the first day, the temperature in the growth chamber was measured to be 1000°C, the nitrogen concentration was 99%, and the oxygen concentration was less than 1ppm; at this time, the crystal size grew rapidly and the shape was regular; on the third day, the temperature was maintained at 1000°C, the nitrogen concentration dropped slightly but was still above 98%, and the oxygen concentration was still very low; the crystal size continued to grow, but the shape began to show slight irregularity; on the fifth day, due to equipment failure, the temperature dropped to 900°C, the nitrogen concentration dropped to 95%, and the oxygen concentration rose to 5ppm; at this time, the crystal size growth slowed down and the shape became obviously irregular; based on these data, a line graph showing the changes in temperature, nitrogen concentration, and oxygen concentration over time was drawn, and these changes were correlated with the changes in crystal size and shape to show the impact of environmental changes during crystal growth on the characteristic morphology of the crystal.

[0124] Therefore, the crystal growth dynamic graph is determined based on the synthesis of the dynamic change graph of crystal characteristics and the environmental change graph of crystal characteristics, which is compatible with the overall consideration of the synthesis of the dynamic change graph of crystal characteristics and the environmental change graph of crystal characteristics, ensuring the accuracy of the crystal growth dynamic graph. At the same time, it is compatible with the overall consideration of the morphology of multiple crystal characteristics, the corresponding time nodes and the corresponding crystal growth scenes, ensuring the accuracy of the crystal growth dynamic graph and presenting the status of the crystal characteristics at different time nodes.

[0125] At this point, the dynamic change graph of crystal characteristics and the environmental change graph of crystal characteristics are combined to generate a comprehensive crystal growth dynamic graph, which shows how crystal characteristics (such as size, shape, etc.) change over time, and the relationship between these changes and the growth environment (such as temperature, pressure, atmosphere, etc.); at this point, ensure that the dynamic change graph of crystal characteristics and the environmental change graph of crystal characteristics are aligned at the time node, which means that each data point in the two graphs corresponds to the same time point; integrate the data in the two graphs into a common data framework, and perform data merging and cleaning operations by using data processing software or programming languages ​​(such as Python, R, etc.).

[0126] Use a data visualization tool (such as Excel, Tableau, or Matplotlib) to create a new chart to display the integrated data. Use the crystal feature dynamics graph as the bottom layer, and use a line chart or area chart to show how the features change over time. Use the environmental variation graph of the crystal features as the top layer, using different colors, line types, or icons to represent different environmental parameters. Ensure that the two layers are aligned on the timeline so that the relationship between feature changes and environmental changes can be clearly seen. To make the crystal growth dynamics graph more interactive and user-friendly, add interactive elements such as mouse hover to display detailed information, timeline controls (play, pause, fast forward, rewind, etc.), or data filtering functions. These interactive elements can be implemented using interactive data visualization libraries (such as Plotly or Bokeh).

[0127] After generating a crystal growth graph, review the data and visualizations to ensure they accurately reflect changes in the crystal characteristics and environment. If you find any errors or inconsistencies, correct the data or adjust the visualization settings.

[0128] Specifically, assume that a silicon crystal growth process for the semiconductor industry is being studied, and a dynamic change graph of crystal characteristics and an environmental change graph of crystal characteristics have been generated according to the previous steps; the dynamic change graph of crystal characteristics shows how the size and shape of the silicon crystal change over time; for example, on the first day, the crystal size is small and the shape is regular; on the third day, the crystal size increases and the shape remains regular; on the fifth day, the crystal size continues to increase but the shape begins to become irregular; the environmental change graph of crystal characteristics shows how the temperature and nitrogen concentration in the growth chamber change over time; for example, on the first and third days, the temperature is maintained at 1200°C and the nitrogen concentration is above 99%; on the fifth day, the temperature drops to 1100°C and the nitrogen concentration drops to 95%.

[0129] Now, combine these two charts into a crystal growth dynamics chart: make sure that the data points in both charts correspond to the same time points (day 1, day 3, day 5); create a new chart with the dynamic changes in crystal size as the bottom layer (represented by a line chart) and the environmental changes in temperature and nitrogen concentration as the top layer (represented by different colors and line types); add the function of displaying detailed information on mouse hover to the chart, so that when you hover over a data point, the specific values ​​of crystal size, temperature, and nitrogen concentration at that time point can be displayed; check the data points and visualization effects in the chart to ensure that they accurately reflect the characteristic changes and environmental changes during the growth of silicon crystals; the final generated crystal growth dynamics chart will provide a comprehensive perspective to observe and understand the dynamic changes of silicon crystals during growth and their relationship with the growth environment, so as to better control the crystal growth process and improve the quality and performance of crystals.

[0130] In some embodiments of the present application, a crystal time node matching table is collected, and the crystal time node matching table is shown in Table 3:

[0131] Table 3 Crystal time node matching table

[0132] Time Node Crystal size changes Shape changes Temperature changes Nitrogen concentration changes T1 Increase rule Elevated Increase T2 Stablize rule Stablize Stablize T3 Increase irregular reduce reduce

[0133] In this example, we see that at time node T1, the increase in crystal size matches the increase in temperature and the increase in nitrogen concentration; while at time node T3, the increase in crystal size is accompanied by a decrease in temperature and a decrease in nitrogen concentration, but the shape becomes irregular, indicating that other unknown factors are also affecting crystal growth.

[0134] refer to Figure 5 In step S14, in the crystal growth dynamic graph, the abnormal part of the crystal feature is determined based on the identification of the abnormal area of ​​the crystal growth dynamic graph, and the abnormal growth parameters of the crystal feature in the crucible are determined based on the tracing of the abnormal part of the crystal feature;

[0135] In the specific implementation process of the present invention, the specific steps are:

[0136] S141: real-time monitoring of a crystal growth dynamic graph, determining a plurality of abnormal sub-regions according to traversal of the crystal growth dynamic graph, and determining an abnormal region of the crystal growth dynamic graph according to a synthesis of the regional positions and regional morphologies of the plurality of abnormal sub-regions;

[0137] S142: Identifying an abnormal region of the crystal growth dynamic graph, determining an abnormal portion of the crystal characteristic based on the identification of the abnormal region of the crystal growth dynamic graph, and determining corresponding abnormal information based on the detection of the abnormal portion;

[0138] S143: Determine the abnormal information corresponding to the abnormal part based on the position, corresponding morphology and corresponding internal image of the abnormal part, determine the growth stage and corresponding time node of the crystal based on the tracing of the abnormal information corresponding to the abnormal part, and determine the abnormal growth parameters of the crystal characteristics in the crucible based on the growth stage of the crystal, the corresponding time node and the crucible database.

[0139] In an embodiment of the present application, the crystal growth dynamic graph is monitored in real time, a plurality of sub-abnormal regions are determined based on the traversal of the crystal growth dynamic graph, and the abnormal region of the crystal growth dynamic graph is determined based on the synthesis of the regional positions and regional morphologies of the plurality of sub-abnormal regions. This is compatible with the overall consideration of the synthesis of the regional positions and regional morphologies of the plurality of sub-abnormal regions, thereby ensuring the accuracy of the abnormal region of the crystal growth dynamic graph.

[0140] At this time, the crystal growth dynamic diagram is monitored in real time. The monitoring equipment should have high resolution and high sensitivity to accurately capture subtle changes in the crystal; the monitoring software should have real-time data processing and analysis capabilities, and be able to automatically identify abnormal data and issue alarms; the monitoring process should be continuous to ensure that no abnormal events are missed.

[0141] On the basis of real-time monitoring, the crystal growth dynamic graph is traversed, and each area in the graph is checked one by one to find sub-regions that do not conform to the normal growth pattern. These sub-abnormal areas are manifested as sudden changes in crystal size, irregular shapes, abnormal internal structures, etc. At this time, the traversal process should be detailed and comprehensive to ensure that no potential abnormal areas are missed; when identifying abnormal areas, professional knowledge, historical data and experimental experience should be combined for judgment; optionally, image processing technology and machine learning algorithms can be used to assist in identifying abnormal areas.

[0142] After identifying multiple sub-abnormal regions, the regional positions and regional morphologies of these regions are synthesized to form a comprehensive abnormal region map. The abnormal region map will clearly show all abnormal regions that appear during the crystal growth process, providing strong support for subsequent abnormal analysis and processing. At this time, the synthesis process should consider factors such as the shape, size, position and mutual relationship of the abnormal regions. Use data visualization tools to assist in the synthesis of the abnormal region map so as to more intuitively display the abnormal information.

[0143] Specifically, suppose you are monitoring the growth process of a silicon crystal used in the semiconductor industry. When monitoring the crystal growth dynamic diagram in real time, you discover the following anomalies: at a certain stage of crystal growth, the crystal size in a certain area suddenly increases, with a significant size difference from the surrounding area; the crystal shape in another area becomes irregular, with obvious bending and distortion; and some tiny cracks and defects are found in the internal structure of the crystal.

[0144] Based on these abnormal phenomena, multiple sub-abnormal regions were identified; then, data visualization tools were used to synthesize the regional positions and regional morphologies of these regions to form a comprehensive abnormal region map. This abnormal region map clearly shows all abnormal regions that appear during the crystal growth process, including size abnormal regions, shape abnormal regions and internal defect regions; based on this abnormal region map, the causes of the abnormalities were further analyzed, such as excessive temperature gradients and improper atmosphere control, and corresponding measures were taken to adjust and optimize them to improve the quality and performance of the crystal.

[0145] Furthermore, the abnormal areas of the crystal growth dynamic graph are identified, the abnormal parts of the crystal characteristics are determined based on the identification of the abnormal areas of the crystal growth dynamic graph, and the corresponding abnormal information is determined based on the detection of the abnormal parts, which is compatible with the overall consideration of the identification of the abnormal areas of the crystal growth dynamic graph and ensures the accuracy of the abnormal parts of the crystal characteristics.

[0146] At this time, the abnormal area of ​​the crystal growth dynamic map synthesized in step S141 is more deeply identified; the goal is to accurately identify the specific characteristics of the abnormal area so that the abnormal part can be accurately determined in subsequent steps; at this time, the identification process should be based on professional crystal growth knowledge and experience, combined with image processing and data analysis technology; the morphology of the abnormal area is observed, including shape, edge, texture and other features; the comparison between the abnormal area and the surrounding normal area is considered to highlight the abnormal features; optionally, machine learning algorithms or pattern recognition technology are used to assist the identification process to improve accuracy and efficiency.

[0147] After identifying the abnormal area, the specific location of the abnormal part on the crystal is determined based on the structural characteristics and growth patterns of the crystal. At this time, the specific location of the abnormal part in the crystal structure is inferred based on the morphology and position information of the abnormal area. The accuracy of the inference is verified using high-precision measurement equipment and experimental methods.

[0148] After determining the abnormal part, the abnormal part is inspected and analyzed in detail to determine its corresponding abnormal information, including the type, degree, cause and impact of the abnormality on the crystal quality and performance. At this time, professional detection equipment and experimental methods are used to conduct detailed inspection and analysis of the abnormal part. The morphology, structure, chemical composition and other aspects of the abnormal part are considered to infer the type and degree of the abnormality. The relationship between the abnormal part and the crystal growth environment is analyzed to infer the cause. The impact of the abnormality on the crystal quality and performance is evaluated so that appropriate measures can be taken to deal with and improve it later.

[0149] Specifically, in step S141, a crystal growth dynamic map containing an abnormal area has been synthesized; it is found that a certain area in the crystal growth dynamic map has obvious color changes and morphological abnormalities; the color of this area is darker than the surrounding area, and the shape is irregular and the edges are blurred; based on the understanding of the crystal structure, it is inferred that the abnormal area is located at a defect position in the crystal, such as near a grain boundary or inclusion; through further measurement and analysis, the specific position of the abnormal part in the crystal structure is determined.

[0150] The abnormal area was inspected and analyzed in detail; chemical analysis revealed that the area contained a higher concentration of impurities; further physical property testing revealed that the conductivity and optical properties of the area were significantly lower than those of the surrounding areas, indicating that the anomaly had a significant impact on the quality and performance of the crystal; based on these analyses, appropriate measures were taken to process and improve the crystal growth process, such as adjusting growth parameters, optimizing crucible design, and strengthening raw material quality control, in order to improve the quality and performance of the crystal.

[0151] Therefore, the abnormal information corresponding to the abnormal part is determined based on the position, corresponding morphology and corresponding internal image of the abnormal part, the growth stage of the crystal and the corresponding time node are determined based on the tracing of the abnormal information corresponding to the abnormal part, and the abnormal growth parameters of the crystal characteristics in the crucible are determined based on the growth stage of the crystal, the corresponding time node and the crucible database, which is compatible with the overall consideration of the crystal growth stage, the corresponding time node and the crucible database, ensuring the accuracy of the abnormal growth parameters of the crystal characteristics in the crucible.

[0152] At this point, the position information, morphological characteristics and internal image data of the abnormal part are combined to analyze in detail and determine the specific abnormal information displayed by the component, including the type of abnormality (such as cracks, inclusions, composition deviations, etc.), severity and impact on the overall performance of the crystal; at this point, the specific position of the abnormal part in the crystal is analyzed to determine whether it is located in a critical area, such as the core, edge or specific crystal face of the crystal; the morphology of the abnormal part is observed, including its size, shape, boundary clarity and comparison with the surrounding normal tissue; high-resolution imaging technology (such as scanning electron microscope, transmission electron microscope, etc.) is used to obtain the internal structure image of the abnormal part and analyze its microscopic characteristics; the type, cause and impact of the abnormality are determined by combining the position, morphology and internal image data.

[0153] After determining the abnormal information of the abnormal part, trace the growth record of the crystal to determine at which growth stage the abnormality occurred and the corresponding specific time node, so as to understand the background and conditions of the abnormality and provide key time clues for subsequent analysis; at this time, consult the crystal growth log, monitoring video, sensor data, etc. to understand the growth history of the crystal; locate the time node of the abnormality based on the characteristics of the abnormal part and key events in the growth record (such as temperature fluctuations, atmosphere changes, raw material addition, etc.); combine the growth cycle and stage characteristics of the crystal to determine at which growth stage the abnormality occurred (such as nucleation period, growth period, cooling period, etc.).

[0154] After determining the growth stage and time node of the abnormal part, the historical data and experimental parameters in the crucible database are used to analyze and determine the crucible growth parameters that caused the anomaly. These parameters include temperature gradient, atmosphere composition, crucible material, shape design, and variables during operation. At this time, the database containing crucible design, material, historical usage records and experimental parameters is accessed. The growth stage and time node when the anomaly occurred are matched with the records in the database to find similar growth conditions and parameter settings. Through comparative analysis, parameters that are significantly different from those under normal growth conditions are identified. The impact of the abnormal parameters is verified by experiments, and the crucible design or growth process is adjusted according to the verification results to eliminate or alleviate the anomaly.

[0155] Specifically, in step S142, the location, morphology, and internal image of an abnormal site in the crystal were determined, and it was preliminarily determined to be an inclusion anomaly. Using high-resolution imaging technology, it was discovered that the inclusion was primarily composed of metallic impurities and located in the core region of the crystal, significantly affecting the crystal's electrical conductivity. A review of growth records revealed that the anomaly occurred in the early stages of crystal growth, within a few minutes after the raw materials were heated to a molten state. During this time, the temperature gradient was large and the oxygen content in the atmosphere was high. Using the crucible database, the parameter settings under normal growth conditions were compared with the records at the time of the anomaly. It was discovered that the anomaly occurred due to improper crucible material selection (poor thermal conductivity), resulting in an excessively large temperature gradient. Furthermore, inaccurate atmosphere control and high oxygen content promoted the oxidation of metallic impurities and the formation of inclusions. Based on these analyses, the crucible material selection was adjusted, the atmosphere control system was optimized, and these improvements were implemented in subsequent growth experiments. The results showed a significant reduction in the number of abnormal sites and improved overall crystal performance.

[0156] In some embodiments of the present application, an abnormal feature matching table is collected, and the abnormal feature matching table is shown in Table 4:

[0157] Table 4 Abnormal feature matching table

[0158] Abnormal characteristics Exception Type Exception details Located at the edge of the crystal, elongated crack Caused by excessive temperature gradient, check heating and cooling rates Internal images show black particles inclusions Caused by metal impurities in the raw materials, the raw material purity and purification process need to be checked

[0159] After determining the abnormal information, use growth logs, sensor data, etc. to trace the growth history of the crystal and determine the growth stage and time node when the abnormality occurred; at this time, check the logs that record key events and parameter changes during the crystal growth process; analyze the real-time data of sensors such as temperature, pressure, and atmosphere to find abnormal values ​​or trends when the abnormality occurs; combine growth logs and sensor data to determine the specific time point when the abnormality occurred and the growth stage it is in.

[0160] After determining the growth stage and time point where the anomaly occurs, the crucible growth parameters that cause the anomaly are determined by weight and score calculation using historical data and experimental parameters in the crucible database. At this time, a database containing crucible design, materials, historical usage records, and experimental parameters is collected. Weights are assigned to each growth parameter that causes the anomaly based on professional knowledge and experience. For example, the temperature gradient has a higher weight because it is known that the temperature gradient has a significant impact on crystal quality. The growth parameters when the anomaly occurs are compared with the normal parameter range in the database, and the score of each parameter is calculated based on the weight. The higher the score, the stronger the correlation between the parameter and the occurrence of the anomaly. Based on the score, the crucible growth parameter that causes the anomaly is determined.

[0161] Collect the score matching table, which is shown in Table 5:

[0162] Table 5 Score matching table

[0163] Growth parameters Weight Abnormal occurrence time Normal range Score (degree of exceeding the range × weight) Temperature gradient 0.5 20°C / cm ≤10°C / cm (20-10) / 10×0.5 = 0.5 Oxygen content in the atmosphere 0.3 5% ≤1% (5-1) / 1×0.3 = 1.2 Raw material purity 0.2 99.8% ≥99.95% (99.95-99.8) / 0.05×0.2 = 0.6

[0164] In this case, the oxygen content in the atmosphere had the highest score (1.2), so it was judged to be the main growth parameter causing the anomaly; based on this judgment, the atmosphere control system was adjusted to reduce the oxygen content and observe whether the anomaly improved.

[0165] refer to Figure 6 In step S15, an abnormal variation is determined based on the traversal of the abnormal portion of the crystal feature, and a local compensation measure of the crucible for the crystal feature is triggered according to the abnormal variation and the spatial position of the abnormal portion of the crystal feature, so as to trigger autonomous optimization of the abnormal portion of the crystal feature;

[0166] In the specific implementation process of the present invention, the specific steps are:

[0167] S151: collecting the abnormal part of the crystal feature, determining the theoretical shape of the part of the crystal feature by matching the abnormal part of the crystal feature with the crucible database, and determining the abnormal change amount by comparing the abnormal part of the crystal feature with the theoretical shape of the part of the crystal feature;

[0168] S152: Determine the abnormality type of the crystal feature in different dimensions based on the abnormal variation and the abnormal location of the crystal feature, determine a first compensation measure for the crucible based on the abnormality type of the crystal feature in different dimensions and the spatial position of the abnormal location, and determine a second compensation measure for the crucible based on the abnormality type of the crystal feature in different dimensions and the current morphology of the crystal;

[0169] S153: In the crucible, a local compensation measure of the crucible for the crystal feature is determined based on the first compensation measure, the second compensation measure and the compensation mapping relationship, the working parameters of the crucible are adjusted based on the execution of the local compensation measure, and the position of the abnormal part of the crystal feature is targetedly adjusted to achieve autonomous optimization of the abnormal part of the crystal feature until the abnormal part of the crystal feature is adjusted to a normal form.

[0170] In an embodiment of the present application, the abnormal part of the crystal feature is collected, and the theoretical morphology of the part of the crystal feature is determined based on the matching of the abnormal part of the crystal feature and the crucible database. The abnormal change amount is determined based on the comparison between the abnormal part of the crystal feature and the theoretical morphology of the part of the crystal feature. This is compatible with the overall consideration of the comparison between the abnormal part of the crystal feature and the theoretical morphology of the part of the crystal feature, ensuring the accuracy of the abnormal change amount.

[0171] At this time, the abnormal parts of the crystal characteristics are collected. After obtaining the data of the abnormal parts, they are matched with the theoretical morphology in the crucible database; the theoretical morphology is established based on crystal growth theory and historical data, and represents the expected morphology of the crystal under ideal conditions; at the same time, a database containing information such as crucible design, crystal growth conditions, and theoretical morphology is accessed; the data of the abnormal parts are compared with the theoretical morphology in the database, and the most similar theoretical morphology is found as a reference; the differences in morphology, size, etc. between the abnormal parts and the theoretical morphology are compared to provide a basis for subsequent analysis.

[0172] After completing the matching of the abnormal part with the theoretical morphology, the amount of abnormal change is determined by comparison, which includes morphological differences, size deviations, and performance changes. At this time, image processing or morphological analysis technology is used to quantify the morphological differences between the abnormal part and the theoretical morphology. The size of the abnormal part is measured and compared with the size of the theoretical morphology to calculate the deviation value. Based on the position, morphology, and size deviation of the abnormal part, its impact on the overall performance of the crystal is evaluated.

[0173] Specifically, suppose that when producing a crystal for semiconductor devices, an obvious concave anomaly is found on its surface; a high-resolution optical microscope is used to image the abnormal part to obtain its detailed morphological and size information; the crucible database is accessed to find a theoretical morphology similar to the current growth conditions as a reference; through comparison, it is found that the morphology of the abnormal part is significantly different from the theoretical morphology, especially in the depth and shape of the concave; the image processing software is used to measure the depth of the concave at the abnormal part and compare it with the corresponding position of the theoretical morphology; the results show that the depth of the concave at the abnormal part is 0.5 mm deeper than the theoretical morphology, and the shape is irregular; in addition, the effect of the concave on the conductive properties of the crystal is also evaluated, and it is believed that it leads to an increase in local resistance; based on the above analysis, the abnormal change is determined to be an increase of 0.5 mm in the concave depth, and corresponding measures are planned to repair or optimize this abnormal part to improve the overall quality and performance of the crystal.

[0174] Furthermore, the abnormal type of the crystal feature in different dimensions is determined based on the abnormal change amount and the abnormal part of the crystal feature, the first compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the spatial position of the abnormal part, and the second compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the current form of the crystal, which is compatible with the overall consideration of the abnormal type of the crystal feature in different dimensions and the current form of the crystal, ensuring the accuracy of the second compensation measure of the crucible.

[0175] At this time, the abnormal type of the crystal feature in different dimensions is determined based on the abnormal change amount and the specific position of the abnormal part in the crystal; the abnormal type includes morphological distortion, size deviation, uneven composition, etc.; at this time, review the abnormal change amount determined in step S151, including differences in morphology, size, etc.; according to the structure and functional requirements of the crystal, the crystal is divided into different dimensions, such as length, width, height, surface, etc.; the abnormal type is determined by combining the abnormal change amount and the part dimension information; for example, if the abnormal change amount is mainly manifested as morphological distortion and is located in the length dimension of the crystal, the abnormal type can be determined as "length dimension morphological distortion".

[0176] After the type of anomaly is determined, the first compensation measure for the crucible is determined based on the spatial position information of the abnormal part; the first compensation measure involves adjusting the crucible design or operating parameters to improve the crystal growth conditions; at this time, the specific position of the abnormal part in the crystal and its impact on the overall structure and performance of the crystal are analyzed; based on the spatial position analysis, consider whether it is necessary to adjust the shape, size or internal structure of the crucible to better adapt to the growth requirements of the crystal; evaluate whether it is necessary to adjust the growth temperature, atmosphere, pressure and other operating parameters to improve the growth environment of the abnormal part.

[0177] In addition to the first compensation measure, a second compensation measure needs to be determined based on the type of anomaly and the current morphology of the crystal. The second compensation measure involves direct intervention in the crystal growth process, such as using a specific post-processing process to repair or optimize the abnormal part. At this time, analyze the relationship between the type of anomaly and the current morphology of the crystal to understand the specific impact of the anomaly on the crystal performance. According to the type of anomaly and the crystal performance requirements, select a suitable post-processing process, such as heat treatment, chemical etching, laser repair, etc.; set the specific parameters of the post-processing process, such as temperature, time, concentration, etc., to ensure that the treatment effect meets the expectations.

[0178] Specifically, assume that when a crystal is produced, an obvious scratch abnormality is found on its surface; based on the abnormal change amount (depth and length of the scratch) and the abnormal location (crystal surface) determined in step S151, the abnormality type is determined to be "surface scratch".

[0179] The first compensatory measure was determined: the spatial location of the scratch was analyzed, and it was found that it was located on a key optical surface of the crystal, with a significant impact on the optical performance. Therefore, it was decided to adjust the design of the crucible and add a protective coating to reduce mechanical damage to the crystal during the growth process. At the same time, the atmosphere control during the growth process was optimized to reduce chemical reactions that aggravate the scratch.

[0180] Determine the second compensatory measure: Considering that the scratch already exists, a post-processing process is needed to repair it; after evaluation, chemical etching is chosen to smooth the scratched surface; the appropriate etching solution concentration and processing time are set to ensure that the scratch is effectively removed while keeping the overall optical performance of the crystal unaffected; by implementing these compensatory measures, it is expected that the scratch abnormality on the crystal surface can be repaired or optimized, and the overall quality and optical performance of the crystal can be improved.

[0181] Therefore, in the crucible, the local compensation measure of the crucible for the crystal feature is determined based on the first compensation measure, the second compensation measure and the compensation mapping relationship, the working parameters of the crucible are adjusted based on the execution of the local compensation measure, and the position of the abnormal part of the crystal feature is targetedly adjusted to achieve autonomous optimization of the abnormal part of the crystal feature until the abnormal part of the crystal feature is adjusted to a normal form, which is compatible with the overall consideration of the first compensation measure, the second compensation measure and the compensation mapping relationship, and ensures the accuracy of the local compensation measure of the crucible for the crystal feature. At the same time, the abnormal part of the crystal feature is introduced to ensure the accuracy of the local compensation measure of the crucible for the crystal feature, and realizes dynamic detection and autonomous optimization of the crystal feature.

[0182] At this time, the first compensation measure and the second compensation measure are integrated, and based on the compensation mapping relationship (i.e., the correspondence between different compensation measures and abnormal parts), a local compensation measure for the crystal feature is determined; the compensation mapping relationship is based on historical data, expert experience or simulation analysis; at this time, the first compensation measure (such as crucible design adjustment, operating parameter optimization) and the second compensation measure (such as post-processing process selection) determined in step S152 are reviewed; the compensation mapping relationship is analyzed to determine which compensation measures are most effective for the specific abnormal part, and the local compensation measure for the crystal feature is determined by considering factors such as the size, shape, position and abnormality type of the abnormal part. The local compensation measure involves adjusting certain parts of the crucible (such as adding a protective layer, changing the shape), optimizing growth parameters (such as temperature, atmosphere, pressure) or adopting a specific post-processing process.

[0183] After determining the local compensation measures, adjust the working parameters of the crucible to implement the local compensation measures. At this time, set the working parameters of the crucible according to the specific requirements of the local compensation measures to ensure the accuracy and feasibility of the parameters; continuously monitor the working status of the crucible during the growth process, and adjust the parameters as needed. Use sensors to monitor key parameters such as temperature, atmosphere, and pressure, and make real-time adjustments through the control system; after implementing local compensation measures and adjusting the working parameters of the crucible, carry out targeted regulation of abnormal parts of the crystal characteristics to promote their autonomous optimization.

[0184] Optionally, lasers, ion beams, or other tools can be used to directly target the abnormal region to promote changes in its morphology, structure, or composition. If direct intervention is not feasible or ineffective, the growth process of the abnormal region can be indirectly influenced by adjusting growth parameters (such as temperature gradient and atmosphere composition). During the regulation process, changes in the abnormal region are continuously monitored to ensure the effectiveness of the regulation measures. Through the implementation and regulation of the above steps, it is expected that the abnormal region of the crystal characteristics will be autonomously optimized and gradually adjusted to a normal morphology. Through multiple iterations and adjustments, the desired optimization effect will be achieved.

[0185] Specifically, assume that when producing a crystal for optoelectronic devices, a tiny protrusion abnormality is found on its surface; based on the first compensation measure (such as optimizing the growth temperature and atmosphere control) and the second compensation measure (such as chemical etching) determined in step S152, and combined with the compensation mapping relationship, a local compensation measure combining growth parameter optimization and chemical etching is adopted; specifically, the temperature gradient during the growth process is adjusted to reduce the driving force for the formation of the protrusion, and chemical etching is used to smooth the protrusion surface after the growth is completed.

[0186] The power distribution of the heating elements was adjusted to form a more uniform temperature gradient during the growth process. At the same time, the atmosphere control system was optimized to ensure that the gas composition and flow in the growth environment were optimal. After the growth was completed, the raised areas were treated using chemical etching technology. By precisely controlling the concentration of the etching solution and the treatment time, the raised surface was successfully smoothed while maintaining the overall structure and performance of the crystal. At the same time, after the implementation and regulation of the above steps, it was observed that the raised anomalies had been significantly reduced and gradually adjusted to a normal form. Through multiple iterations and adjustments, the autonomous optimization of the abnormal parts of the crystal characteristics was finally achieved, thereby improving the overall quality and performance of the crystal.

[0187] In some embodiments of the present application, a compensation measure matching table is collected, and the compensation measure matching table is shown in Table 6:

[0188] Table 6 Compensation Measures Matching Table

[0189] Exception Type Abnormal area First compensation measure Second compensation measure Distorted form crystal surface Adjusting the crucible shape to suit crystal growth Chemical etching for smooth surface Dimensional deviation Crystal length Optimizing growth temperature gradient Laser cutting to adjust size Uneven ingredients Crystal center Adjust raw material ratio and atmosphere control Diffusion treatment homogenization components

[0190] Assume that when producing a silicon crystal used in solar panels, a morphologically distorted anomaly is discovered on its surface. Based on the compensation measure matching table, the following local compensation measures are determined:

[0191] First compensation measure: adjust the crucible shape to suit the crystal growth, reduce stress accumulation during the growth process, and thus reduce the risk of morphological distortion;

[0192] Second compensatory measure: After the growth is completed, chemical etching technology is used to smooth the distorted parts to further improve the morphology of the crystal surface.

[0193] Reference Figure 7 , the dynamic detection system of crystal growth includes:

[0194] An internal image module 21 is used to collect multiple internal images of the crucible in a preset time period when the crystal is in the crucible, and mark the crystal growth area in each internal image;

[0195] A crystal feature module 22 is configured to determine a plurality of crystal features based on crystal identification of each crystal growth region, wherein the plurality of crystal features correspond to different time nodes;

[0196] A crystal growth dynamic graph module 23 is configured to determine a corresponding crystal growth scene based on scene recognition of each crystal growth region, and to determine a crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes;

[0197] An abnormal location module 24 is configured to determine an abnormal location of a crystal feature based on identification of an abnormal region in the crystal growth dynamic graph, and to determine abnormal growth parameters of the crystal feature in the crucible based on tracing back the abnormal location of the crystal feature;

[0198] The local compensation module 25 is used to determine the abnormal change amount based on the traversal of the abnormal part of the crystal feature, and trigger the crucible to take local compensation measures for the crystal feature according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature, so as to trigger the autonomous optimization of the abnormal part of the crystal feature.

[0199] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A dynamic detection method for crystal growth, characterized in that: include: The crystal is in the crucible, multiple internal images of the crucible are collected in a preset time period, and the crystal growth area in each internal image is marked; Determining multiple crystal features based on crystal identification of each crystal growth region, wherein the multiple crystal features correspond to different time nodes; Determine the corresponding crystal growth scene based on scene recognition of each crystal growth area, and determine the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes and corresponding crystal growth scenes; In the crystal growth dynamic graph, the abnormal portion of the crystal feature is determined based on the identification of the abnormal region of the crystal growth dynamic graph, and the abnormal growth parameter of the crystal feature in the crucible is determined based on the tracing back of the abnormal portion of the crystal feature; Based on the traversal of the abnormal part of the crystal feature, the abnormal change amount is determined, and according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature, the crucible triggers the local compensation measure of the crystal feature to trigger the autonomous optimization of the abnormal part of the crystal feature.

2. The dynamic detection method of crystal growth according to claim 1, characterized in that: The crystal is in a crucible, and a plurality of internal images of the crucible in a preset time period are collected, and a crystal growth area in each internal image is marked, including: The crystal is placed on the positioning component of the crucible and dynamically grows in the crucible. At this time, the crucible is monitored in real time and the interior of the crucible is triggered to be photographed within a preset time period to determine multiple internal images of the crucible within the preset time period; In each internal image, the spatial position of the crystal is determined based on detection of the internal image, a first sub-region is determined based on the spatial position of the crystal and the growth stage of the crystal, and a second sub-region is determined based on the spatial position of the crystal and the internal spatial morphology of the crucible; In each internal image, a crystal growth region in each internal image is determined based on a mapping relationship among the first subregion, the second subregion, and the growth region, so as to mark the crystal growth region in each internal image.

3. The dynamic detection method of crystal growth according to claim 1, characterized in that: The multiple crystal features are determined based on the crystal identification of each crystal growth region, and the multiple crystal features correspond to different time nodes, including: In each crystal growth region, the corresponding crystal outer contour is determined based on the traversal of the crystal growth region, and the corresponding crystal recognition mode is determined according to the crystal outer contour and the crystal model, and the crystal recognition mode includes a ring detection mode, a local detection mode and an overall detection mode; triggering crystal recognition of a crystal growth region along the crystal recognition pattern to output a plurality of sub-crystal features; The corresponding crystal feature is determined by synthesizing multiple sub-crystal features and images corresponding to the crystal growth area. At this time, each crystal feature is compared with an internal image, and the shooting time of the internal image is used as the time node corresponding to the crystal feature.

4. The dynamic detection method of crystal growth according to claim 1, characterized in that: The method of determining a corresponding crystal growth scene based on scene recognition of each crystal growth area and determining a crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes includes: In each crystal growth area, the regional position of the crystal growth area is collected, and the corresponding scene area is determined based on the regional position of the crystal growth area, the morphology of the crystal growth area and the position of the crystal characteristics. The corresponding crystal growth scene is determined based on the scene recognition of the scene area.

5. The dynamic detection method of crystal growth according to claim 4, characterized in that: The method further includes determining a corresponding crystal growth scene based on scene recognition of each crystal growth region, and determining a crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes. Determine a dynamic change graph of crystal features based on the morphology of multiple crystal features and corresponding time nodes; determine an environmental change graph of crystal features based on the morphology of multiple crystal features and corresponding crystal growth scenarios; The crystal growth dynamics graph is determined based on the synthesis of the dynamics change graph of the crystal characteristics and the environmental change graph of the crystal characteristics.

6. The dynamic detection method of crystal growth according to claim 1, characterized in that: The method of determining an abnormal portion of a crystal feature based on identification of an abnormal region in the crystal growth dynamic graph, and determining abnormal growth parameters of the crystal feature in the crucible based on tracing back the abnormal portion of the crystal feature, includes: The crystal growth dynamic graph is monitored in real time, a plurality of abnormal sub-regions are determined according to the traversal of the crystal growth dynamic graph, and the abnormal region of the crystal growth dynamic graph is determined according to the synthesis of the regional positions and regional morphologies of the plurality of abnormal sub-regions.

7. The method for dynamic detection of crystal growth according to claim 6, characterized in that: The method further includes: determining an abnormal portion of a crystal feature based on identification of an abnormal region in the crystal growth dynamic graph, and determining abnormal growth parameters of the crystal feature in the crucible based on tracing back the abnormal portion of the crystal feature. Identifying an abnormal region of the crystal growth dynamic graph, determining an abnormal portion of the crystal characteristic based on the identification of the abnormal region of the crystal growth dynamic graph, and determining corresponding abnormal information based on the detection of the abnormal portion; The abnormal information corresponding to the abnormal part is determined based on the position, corresponding morphology and corresponding internal image of the abnormal part. The growth stage and corresponding time node of the crystal are determined based on the tracing of the abnormal information corresponding to the abnormal part. The abnormal growth parameters of the crystal characteristics in the crucible are determined based on the growth stage of the crystal, the corresponding time node and the crucible database.

8. The dynamic detection method of crystal growth according to claim 1, characterized in that: The determining of the abnormal variation based on the traversal of the abnormal portion of the crystal feature, and triggering a local compensation measure of the crucible on the crystal feature according to the abnormal variation and the spatial position of the abnormal portion of the crystal feature to trigger autonomous optimization of the abnormal portion of the crystal feature, includes: Collecting the abnormal part of the crystal feature, determining the theoretical morphology of the part of the crystal feature based on matching the abnormal part of the crystal feature with a crucible database, and determining the abnormal change amount based on comparison between the abnormal part of the crystal feature and the theoretical morphology of the part of the crystal feature; The abnormal type of the crystal feature in different dimensions is determined based on the abnormal change amount and the abnormal location of the crystal feature. The first compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the spatial position of the abnormal location. The second compensation measure of the crucible is determined based on the abnormal type of the crystal feature in different dimensions and the current morphology of the crystal.

9. The method for dynamic detection of crystal growth according to claim 8, characterized in that: The method of determining an abnormal variation based on traversal of the abnormal portion of the crystal feature, and triggering a local compensation measure of the crucible for the crystal feature according to the abnormal variation and the spatial position of the abnormal portion of the crystal feature to trigger autonomous optimization of the abnormal portion of the crystal feature, further includes: In the crucible, a local compensation measure of the crucible for the crystal feature is determined based on the first compensation measure, the second compensation measure and the compensation mapping relationship, the working parameters of the crucible are adjusted based on the execution of the local compensation measure, and the position of the abnormal part of the crystal feature is targetedly adjusted to achieve autonomous optimization of the abnormal part of the crystal feature until the abnormal part of the crystal feature is adjusted to a normal form.

10. A dynamic detection system for crystal growth, characterized in that: The dynamic detection system for crystal growth is applied to the dynamic detection method for crystal growth according to any one of claims 1 to 9, and the dynamic detection system for crystal growth comprises: An internal image module is used to collect multiple internal images of the crucible in a preset time period when the crystal is in the crucible, and mark the crystal growth area in each internal image; A crystal feature module is used to determine multiple crystal features based on crystal identification of each crystal growth region, where the multiple crystal features correspond to different time nodes; A crystal growth dynamic graph module is used to determine the corresponding crystal growth scene based on scene recognition of each crystal growth area, and to determine the crystal growth dynamic graph according to the morphology of multiple crystal features, corresponding time nodes, and corresponding crystal growth scenes; An abnormal location module is used to determine the abnormal location of the crystal feature based on the identification of the abnormal area in the crystal growth dynamic diagram, and to determine the abnormal growth parameters of the crystal feature in the crucible based on the tracing of the abnormal location of the crystal feature; The local compensation module is used to determine the abnormal change amount based on the traversal of the abnormal part of the crystal feature, and trigger the crucible to take local compensation measures for the crystal feature according to the abnormal change amount and the spatial position of the abnormal part of the crystal feature, so as to trigger the autonomous optimization of the abnormal part of the crystal feature.

Images