Control method for image generation in crystal growth process

Abstract

The invention provides a control method for image generation in a crystal growth process, and relates to the technical field of crystal growth image.The control method comprises the following steps that axial acceleration, temperature gradient and interface morphology data in the crystal growth process are collected in real time, the current growth stage of a crystal is judged, and if the crystal enters a transition stage, the crystal is subjected to image generation; if yes, historical rotating speed data and historical temperature field distribution characteristics of the crystal are called to carry out time sequence prediction, a multi-parameter data stream of the crystal in a future preset time period is generated, and shooting parameters of corresponding imaging units are obtained and comprise a shooting sequence and adjacent imaging intervals; the shooting sequence comprises a shooting sequence and a shooting angle of each imaging unit; and controlling each imaging unit to image the crystal growth process under the shooting parameters in a future preset time period so as to realize multi-view image splicing. According to the method, image missing in a key deformation stage is avoided, the crystal image splicing success rate and splicing efficiency are improved, and then the real-time monitoring effect on the crystal morphology is guaranteed.

Description

Technical Field

[0001] This application relates to the technical field of crystal growth image, and particularly to a control method for image generation during crystal growth. Background Art

[0002] During the crystal growth (such as the preparation of single crystal silicon by the Czochralski method), real-time monitoring of the crystal interface morphology is the key to controlling the growth quality. Traditional methods usually rely on imaging control at a fixed frequency. However, crystal growth has a steady state section (stable growth) and a transition section (dynamic processes such as sudden diameter change and crystallization rate jump). Traditional methods are prone to missing images in critical deformation stages (for example, when the crystal diameter suddenly changes, fixed-frequency imaging may lose interface defect data due to failure to adjust the shooting angle in time). In addition, traditional methods often result in mismatched shooting angles due to sudden changes in the growth trajectory in the transition section (for example, when the crystal rotation speed suddenly increases, there are splicing gaps between adjacent imaging units due to motion blur). In summary, traditional control imaging methods often lead to splicing failure due to image loss or the existence of splicing gaps, affecting the real-time monitoring effect of the crystal interface morphology. Summary of the Invention

[0003] In view of the above-mentioned defects or deficiencies in the prior art, this application aims to provide a control method for image generation during crystal growth to improve the crystal image splicing effect. The control method includes the following steps: Real-time collect the axial acceleration, temperature gradient and interface morphology data during crystal growth; According to the axial acceleration and the temperature gradient, judge the growth stage that the current crystal is in, and the growth stage includes a steady state section and a transition section; If the crystal is in the steady state section, control a plurality of imaging units arranged in an array circumferentially around the crystal to perform multi-angle synchronous imaging at a fixed time interval; After judging the growth stage that the current crystal is in, the following steps are further included: If the crystal enters the transition section, retrieve the historical rotation speed data and the historical temperature field distribution characteristics of the crystal to perform time series prediction and generate a multi-parameter data stream of the crystal within a future preset time period; the multi-parameter data stream includes a plurality of growth trajectory positioning identifiers of the crystal within the future preset time period, and the predicted acceleration change rate and the predicted interface morphology curvature corresponding to each growth trajectory positioning identifier; According to the multi-parameter data stream, obtain the shooting parameters of the imaging units within the future preset time period, and the shooting parameters include a shooting sequence and an adjacent imaging interval; the shooting sequence includes the shooting order and the shooting angle of each imaging unit; Control each of the imaging units to image the crystal growth process under the shooting parameters within the preset future time period, so as to achieve multi-view image stitching.

[0004] According to the technical solution provided by the present application, the judging the growth stage in which the current crystal is located includes the following steps: If the fluctuation amplitude of the axial acceleration is within the preset steady-state threshold range and the change rate of the temperature gradient is less than the first threshold, it is determined as the steady-state section.

[0005] According to the technical solution provided by the present application, the obtaining the shooting parameters of the imaging unit according to the multi-parameter data stream includes the following steps: Calibrate the target imaging unit corresponding to each of the growth trajectory positioning identifiers; Sort the target imaging units according to the sequence of each of the growth trajectory positioning identifiers to obtain the shooting sequence; Calculate the target focus distance of each of the target imaging units according to the predicted interface morphology curvature corresponding to each of the growth trajectory positioning identifiers, so as to obtain the corresponding shooting angle, and further obtain the shooting sequence.

[0006] According to the technical solution provided by the present application, the obtaining the shooting parameters of the imaging unit within the preset future time period according to the multi-parameter data stream includes the following steps: Divide the target imaging units into high-priority and low-priority according to the change rate of the predicted interface morphology curvature of the growth trajectory positioning identifier; Obtain the first interval of the low-priority target imaging units according to the predicted acceleration change rate; insert supplementary shooting frames in the first interval to obtain the second interval of the high-priority target imaging units, and the first interval and the second interval form the adjacent imaging intervals.

[0007] According to the technical solution provided by the present application, the dividing the target imaging units into high-priority and low-priority according to the change rate of the predicted interface morphology curvature of the growth trajectory positioning identifier includes the following steps: Calibrate the target imaging units corresponding to the growth trajectory positioning identifiers with the change rate of the predicted interface morphology curvature greater than the second threshold as high-priority.

[0008] According to the technical solution provided by the present application, after generating the multi-parameter data stream of the crystal within the preset future time period for time series prediction, the following steps are further included: Judge whether there is an unstable region in the growth process of the crystal within the preset future time period according to the multi-parameter data stream; Controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period includes the following steps: If not, control each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period.

[0009] According to the technical solution provided by the present application, determining whether there is an instability region in the crystal during the growth process within the future preset time period according to the multi-parameter data stream includes the following steps: When the change rate of the predicted interface morphology curvature exceeds the third threshold, it is determined that an interface instability phenomenon occurs, and the region where the interface instability phenomenon occurs is used as the instability region.

[0010] According to the technical solution provided by the present application, after determining whether there is an instability region in the crystal during the growth process within the future preset time period according to the multi-parameter data stream, the following steps are further included: If not, according to the spatial distribution characteristics of the instability region, shorten the adjacent imaging interval of the corresponding target imaging unit to 1 / 3 - 1 / 2 of the original interval to obtain the updated adjacent imaging interval; Increase the exposure times of the imaging units corresponding to the angles of the instability region in the shooting sequence, and adjust the shooting angles of the target imaging units after the instability region according to the predicted curvature change after each exposure to obtain the updated shooting angles; Based on the updated adjacent imaging interval and the updated shooting angles, obtain the updated shooting parameters; Control each of the target imaging units to image the crystal growth process under the updated shooting parameters within the future preset time period.

[0011] According to the technical solution provided by the present application, retrieving the historical rotation speed data and historical temperature field distribution characteristics of the crystal to perform time series prediction and generate a multi-parameter data stream of the crystal within the future preset time period includes the following steps: According to the historical rotation speed data and historical temperature field distribution characteristics of the crystal, predict the predicted rotation speed data and predicted temperature field distribution characteristics of the crystal within the future preset time period, and then obtain the multi-parameter data stream; Controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period includes the following steps: Real-time monitor the real-time temperature field distribution characteristics, and obtain the deviation degree between the real-time temperature field distribution characteristics and the predicted temperature field distribution characteristics at the corresponding moment; the deviation degree includes the axial temperature gradient deviation and the radial temperature gradient deviation; When the axial temperature gradient deviation exceeds the fourth threshold or the radial temperature gradient deviation exceeds the fifth threshold, a temperature field perturbation compensation mode is triggered to eliminate or reduce the interference of the temperature field deviation on the crystal growth image capture.

[0012] According to the technical solution provided by the present application, the temperature field perturbation compensation mode includes: Pausing the current capture sequence and starting the fast scanning imaging unit to perform a fan-shaped scan on the abnormal temperature region to obtain a scan result; Re-calibrating the growth trajectory positioning identifier according to the scan result to obtain an updated growth trajectory positioning identifier; Using the updated growth trajectory positioning identifier to obtain an updated multi-parameter data stream and then resuming the capture.

[0013] Compared with the prior art, the beneficial effects of the present application are as follows: The present application accurately identifies the steady state section and the transition section through the combined criterion of axial acceleration and temperature gradient. When entering the transition section, the future growth trajectory is predicted based on historical data, and a parameter data stream including the acceleration change rate and the interface curvature is generated in advance, and then the corresponding capture parameters are obtained, so that when entering the transition section, the images are acquired according to the capture order, capture angle, and adjacent capture interval of each imaging unit, adapting to the rapid changes in crystal growth in the transition section, and avoiding the problems of missing images in the critical deformation stage and mismatch of capture angles caused by fixed-frequency imaging and fixed capture angles in the traditional method. The accurate judgment of the growth stage and the dynamic adjustment of the capture parameters in the present application can obtain more complete and accurate image data at different stages of crystal growth. In the transition section, the missing images in the critical deformation stage are avoided, ensuring the continuity and integrity of the image information, providing rich and effective materials for multi-view image stitching, improving the stitching success rate and stitching efficiency, and thus ensuring the real-time monitoring effect of the crystal interface morphology. Description of the Drawings

[0014] Figure 1 It is a flowchart of the steps of the control method for image generation during crystal growth provided by the present application. Detailed Embodiments

[0015] The present application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related invention and are not a limitation to the invention. In addition, it should be noted that for the sake of description, only the parts related to the invention are shown in the drawings.

[0016] It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The present application will be described in detail below with reference to the drawings and embodiments.

[0017] Embodiment 1 As mentioned in the background art, in view of the problems in the prior art, the present application proposes a control method for image generation during crystal growth, which is mainly applied to the monitoring of the Czochralski single crystal growth process. There are 6 high-resolution CCD imaging units (numbered U1-U6 respectively) surrounding the crystal in the crystal growth furnace, and each imaging unit is equipped with an axial acceleration sensor, an infrared temperature measurement array, and a laser ranging module. Each sensor collects data in real time at a sampling frequency of 50Hz and transmits it to the control host through the industrial Ethernet.

[0018] As Figure 1 shown, the control method includes the following steps: S100. Collect the axial acceleration, temperature gradient, and interface morphology data during the crystal growth process in real time; Specifically, the axial acceleration is measured by a MEMS acceleration sensor (such as ADXL355) installed on the base of the lifting rod. Six groups of infrared temperature measurement probes (such as Heitronics KT15.99D) arranged in a ring are used to set a temperature measurement point every 2 cm in the axial direction of the crystal to obtain the temperature gradient. The interface morphology data is obtained by scanning the solid-liquid interface with a laser triangulation rangefinder (such as Keyence LK-G5000).

[0019] S200. Judge the growth stage in which the current crystal is located according to the axial acceleration and the temperature gradient, and the growth stage includes a steady state section and a transition section; Furthermore, the step of judging the growth stage in which the current crystal is located includes the following steps: If the fluctuation amplitude of the axial acceleration is within the preset steady state threshold range and the change rate of the temperature gradient is less than the first threshold, it is determined to be the steady state section.

[0020] Specifically, the preset steady state threshold is that the axial acceleration fluctuation ≤ 0.05g (g = 9.8m / s²), and the temperature gradient change rate threshold is set to 1℃ / cm·min. When the fluctuation amplitude of the axial acceleration is within the preset steady state threshold range and the change rate of the temperature gradient is less than the first threshold for three consecutive sampling periods (600ms), the growth stage in which the crystal is located is defined as the steady state section. In the steady state section of the growth stage, the atoms or molecules inside the crystal are arranged in a relatively stable and regular manner, enabling the crystal to grow continuously under relatively stable conditions, and the growth rate, interface morphology and other parameters of the crystal are also relatively stable.

[0021] S310. If the crystal is in the steady state section, control multiple imaging units arranged in an array circumferentially around the crystal to perform multi-angle synchronous imaging at a fixed time interval; Specifically, since crystal growth is stable during the growth stage of the steady state section, by controlling multiple imaging units to perform multi-angle synchronous imaging at fixed time intervals, a complete and seamless spliced crystal growth image can be obtained.

[0022] Specifically, when the amplitude of axial acceleration fluctuation exceeds the preset steady state threshold range, or the change rate of temperature gradient is greater than or equal to the first threshold, the stage in which the crystal is located is the transition section. During the growth stage of the transition section, the state of the crystal growth system has changed significantly. The axial acceleration fluctuation exceeding the range may be due to changes in the external environment, such as minor vibrations of the equipment, air flow changes, and other factors having a greater impact on crystal growth. The increase in the temperature gradient change rate indicates that the heat distribution in the crystal growth environment has changed significantly, which may cause obvious changes in parameters such as crystal growth rate and interface morphology. Crystal growth may change from one stable state to another, or be in an unstable transition process. Therefore, during the transition section, the shape and position of the crystal may change significantly in a short period of time. This will cause misalignment and mismatch when splicing the images taken according to the preset shooting parameters and fixed time intervals, making it difficult to achieve seamless splicing of multi-view images and affecting the complete presentation of the overall crystal growth process. Based on this, this embodiment provides an imaging control method for the crystal entering the transition section.

[0023] After determining the growth stage in which the current crystal is located, the following steps are further included: S320. If the crystal enters the transition section, retrieve the historical rotation speed data and historical temperature field distribution characteristics of the crystal to perform time series prediction and generate a multi-parameter data stream of the crystal within a preset future time period; the multi-parameter data stream includes multiple growth trajectory positioning identifiers of the crystal within the preset future time period, and the predicted acceleration change rate and predicted interface morphology curvature corresponding to each growth trajectory positioning identifier. Specifically, first, data preprocessing is carried out: extract time-series data such as crystal rotation speed, temperature field distribution (such as axial / radial temperature gradient), interface morphology (such as curvature, roughness), etc. from the historical database, and label the corresponding growth stage labels (steady-state segment / transition segment). Introduce crystal material properties (such as crystal type, thermal expansion coefficient) and environmental parameters (such as chamber pressure, solution saturation) as static features to enhance the physical relevance of the model; perform sliding window processing on the historical rotation speed and temperature gradient data to extract statistical features such as mean, variance, and trend terms. Use Fourier transform or wavelet analysis to extract the periodic features of the temperature field distribution, and combine the geometric features of the interface morphology (such as the curvature change rate) as input variables; then construct a prediction model: based on the crystal growth kinetics equation (such as the Ginzburg-Landau theory describes the evolution of the phase change interface), transform the physical relationship between the temperature field distribution and the interface morphology into a regularization term of the model to constrain the physical rationality of the prediction results. Use the long short-term memory network (LSTM) to capture the non-linear time-series dependence relationship between the rotation speed and temperature gradient in the transition segment, predict the acceleration change rate and curvature change trend in the future period, and use the mean square error (MSE) and physical consistency index (such as the matching degree between the temperature field distribution and the crystal morphology) as the objective function. Adjust the hyperparameters such as the hidden layer structure and learning rate of the LSTM through Bayesian optimization to ensure the generalization ability of the model in the transition segment; then generate a multi-parameter data stream: based on the acceleration change rate and interface curvature predicted by the prediction model, combine the crystal growth rate equation (such as the Burton-Cabrera-Frank model), dynamically divide the key points of the growth trajectory in the future period (positioning marks, divide the points with significant growth state changes and key interface morphology change points into key points of the growth trajectory), and calculate the temperature field gradient thresholds corresponding to each mark. Use Monte Carlo Dropout or Bayesian neural network to output the confidence interval of the predicted value, providing a basis for dynamically adjusting the shooting parameters (such as the interval time) of the imaging unit to avoid image stitching failure caused by prediction deviation. Among them, the prediction model involves model verification and online update. Use the historical transition segment data to cross-validate the model, compare the error between the predicted value and the true value (such as MAE < 5%), and verify the physical consistency through finite element simulation (such as the Cz method temperature field distribution simulation). Deploy an online learning module to collect new data in the transition segment in real time and update the model parameters. Use transfer learning technology to transfer the steady-state segment model parameters to the transition segment to improve the adaptability of the model to sudden operating conditions.

[0024] S321. Obtain the shooting parameters of the imaging unit in the future preset period according to the multi-parameter data stream, where the shooting parameters include a shooting sequence and an adjacent imaging interval; the shooting sequence includes the shooting order and shooting angle of each imaging unit; Specifically, the goal of the shooting sequence planning is to determine the shooting order, angles, and coverage ranges of multiple imaging units (such as CCD cameras distributed in a circular array). Locate the markers based on the predicted growth trajectory, and calculate the optimal observation angle corresponding to each marker through a geometric projection model (such as the mapping relationship between the crystal growth axis and the imaging plane). Exemplarily, if it is predicted that the interface will shift to a certain side (the curvature increases), then the imaging unit corresponding to the corresponding angle is preferentially scheduled to perform high-density shooting on this area. Dynamically adjust the priorities of the imaging units according to the predicted acceleration change rate and curvature change rate, allocate more imaging units to high-change-rate areas (such as acceleration mutation points), and shorten the adjacent shooting intervals. For low-change-rate areas, reduce the shooting frequency to save resources. Use a greedy algorithm or reinforcement learning (Q-Learning) to optimize the shooting order and maximize the coverage probability of key areas. The adjacent imaging interval is the time interval between two adjacent shootings, ensuring that the image sequence can capture the continuous changes in the interface morphology. According to the predicted curvature change rate of the interface morphology (such as the curvature derivative), through the formula (wherein, is the maximum allowable curvature change amount (calibrated by experiments), is a small constant to prevent the denominator from being zero) dynamically adjust the adjacent imaging intervals. Through a hardware-level time synchronization protocol (such as the PTP precision clock protocol), ensure that multiple imaging units are triggered in a unified time sequence to avoid image misalignment caused by time drift.

[0025] S322. Control each of the imaging units to image the crystal growth process under the shooting parameters within the preset future time period to achieve multi-view image stitching.

[0026] Specifically, this control relies on a master controller: receiving the predicted data stream and sending shooting instructions (angles, time points) to each imaging unit. Slave control unit: Each imaging unit is built with an FPGA or a microcontroller, which parses the instructions in real time and adjusts the mechanical pan-tilt (rotation angle) and camera parameters (exposure time, focal length). According to the real-time feedback of the temperature field distribution (such as infrared sensor data) and crystal vibration (accelerometer data), correct the shooting angle and exposure time: Vibration compensation: If the axial acceleration exceeds the threshold, trigger the high-speed imaging mode (shorten the exposure time to prevent motion blur).

[0027] Specifically, the implementation method for stitching images at different angles and time points into a panoramic view of the crystal growth interface in continuous space-time includes: using timestamps and predicted trajectories, grouping multi-angle images at the same growth stage (such as near the same positioning identifier) for time alignment; extracting features such as the crystal interface edge and curvature extreme points in the images, and using SIFT or deep learning features (such as SuperPoint) for matching for spatial alignment; calculating the homography matrix through the RANSAC algorithm to align images from different perspectives; for the overlapping areas of multi-perspective images, using weighted averaging or Poisson fusion to eliminate the stitching seams; if an image from a certain perspective fails due to occlusion or vibration, generating virtual patches using the predicted interface morphology data (such as curvature changes), thus forming a panoramic view of the crystal growth interface in continuous space-time.

[0028] Furthermore, obtaining the shooting parameters of the imaging unit according to the multi-parameter data stream includes the following steps: Calibrating the target imaging unit corresponding to each growth trajectory positioning identifier; Specifically, mapping the growth trajectory positioning identifier to the field of view of the imaging unit, and taking the imaging unit with the growth trajectory positioning identifier at the center of the field of view as the target imaging unit. Exemplarily, when the identifier P5 is at the center of the field of view of the imaging unit U3, the imaging unit U3 is defined as the target imaging unit of P5.

[0029] Sorting the target imaging units according to the sequence of each growth trajectory positioning identifier to obtain the shooting order; Calculating the target focusing distance of each target imaging unit according to the predicted interface morphology curvature corresponding to each growth trajectory positioning identifier to obtain the corresponding shooting angle, and further obtaining the shooting sequence.

[0030] Specifically, through a pre-trained model, inputting the predicted interface morphology curvature into the model to obtain the target focusing distance of the target imaging unit, and then according to the relational formula between the focusing distance d, the curvature radius R and the shooting angle θ , calculating to obtain the shooting angle θ.

[0031] In a preferred embodiment, obtaining the shooting parameters of the imaging unit in the future preset time period according to the multi-parameter data stream includes the following steps: Dividing the target imaging units into high-priority and low-priority according to the change rate of the predicted interface morphology curvature of the growth trajectory positioning identifier; Furthermore, dividing the target imaging units into high-priority and low-priority according to the change rate of the predicted interface morphology curvature of the growth trajectory positioning identifier includes the following steps: The target imaging unit corresponding to the growth trajectory positioning mark whose change rate of the predicted interface morphology curvature is greater than a second threshold is marked as a high priority. The second threshold can be selected as the curvature change rate ΔR / Δt=0.2mm / s.

[0032] According to the predicted acceleration change rate, a first interval of the target imaging unit with a low priority is obtained; a supplementary shooting frame is inserted into the first interval to obtain a second interval of the target imaging unit with a high priority, and the first interval and the second interval constitute the adjacent imaging interval.

[0033] Specifically, the number and positions of inserted supplementary shooting frames can be determined according to the importance of high-priority target imaging units and the size of the predicted interface morphology curvature change rate. This embodiment can reasonably allocate imaging resources according to different states of crystal growth, increase the shooting frequency in areas with drastic changes (insert supplementary shooting frames, shorten the interval between adjacent imaging), improve the resolution and information content of the image, and help to more accurately monitor the crystal growth process.

[0034] In a preferred embodiment, after performing timing prediction to generate a multi-parameter data stream of a crystal within a future preset time period, the method further includes the following steps: According to the multi-parameter data stream, determining whether there is an unstable region in the growth process of the crystal within a future preset time period; Further, judging whether there is an unstable region in the growth process of the crystal in a future preset time period according to the multi-parameter data stream includes the following steps: When the change rate of the predicted interface morphology curvature exceeds a third threshold, it is determined that interface instability occurs, and the area where the interface instability occurs is regarded as an unstable area; wherein the third threshold may be selected as 0.5 mm / s.

[0035] Furthermore, the controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period comprises the following steps: If not, each of the imaging units is controlled to image the crystal growth process under the shooting parameters within the future preset time period.

[0036] Specifically, during the crystal growth process, abnormal changes in the crystal growth state often occur, resulting in local instability. If no local instability is predicted within a preset future period of time (that is, there is no unstable area during the crystal growth process within the preset future period of time), then the above scheme can maximize the seamless stitching of multi-view images.

[0037] Further, after determining whether there is an instability region during the growth of the crystal in a future preset period according to the multi-parameter data stream, the following steps are further included: If not, according to the spatial distribution characteristics of the instability region, shorten the adjacent imaging interval of the corresponding target imaging unit to 1 / 3 - 1 / 2 of the original interval to obtain an updated adjacent imaging interval; Specifically, since each imaging unit has its specific field of view range, by comparing the spatial coordinates of the instability region with the coordinate ranges of the fields of view of each imaging unit, the target imaging unit corresponding to the instability region (the target imaging unit whose field of view covers the instability region) is found. For example, using a coordinate matching algorithm, it is determined whether the boundary coordinates of the instability region fall within the boundary coordinate range of the field of view of a certain imaging unit. Before determining the instability region, the original adjacent imaging interval of the target imaging unit has been determined according to the multi-parameter data stream. This interval is determined based on the prediction information under the normal crystal growth state and can be determined according to the severity and change trend of the instability region. If the instability region changes violently, a shorter interval can be selected as the updated adjacent imaging interval.

[0038] Increase the exposure times of the imaging unit at the angle corresponding to the instability region in the shooting sequence, and adjust the shooting angle of the target imaging unit after the instability region according to the predicted curvature change each time after exposure to obtain an updated shooting angle; Specifically, in the shooting sequence, for the imaging unit at the angle corresponding to the instability region, increase its exposure times. This helps to obtain more image information about the instability region and improve the success rate of seamless stitching. Each time after exposure, by retrieving a pre-established mapping relationship between curvature and shooting angle (for example, the optimal shooting angle at different curvatures obtained through experiments or theoretical models). When the predicted curvature changes, calculate the corresponding shooting angle according to this mapping relationship to obtain an updated shooting angle.

[0039] Based on the updated adjacent imaging interval and the updated shooting angle, obtain updated shooting parameters; Control each target imaging unit to image the crystal growth process under the updated shooting parameters in the future preset period.

[0040] In this embodiment, by shortening the adjacent imaging interval and increasing the exposure times, image information of the instability region can be obtained more frequently, finer changes can be captured, feature changes caused by time differences can be reduced, and the accuracy of matching multiple captured images in the process of stitching images can be improved. In addition, by adjusting the shooting angle in real time, the imaging unit can obtain the clearest images that can best reflect the surface characteristics of the crystal, provide more accurate feature information for image matching, and further improve the accuracy of multi-angle imaging seamless stitching.

[0041] In a preferred embodiment, retrieving the historical rotation speed data and historical temperature field distribution characteristics of the crystal to perform time series prediction and generate a multi-parameter data stream of the crystal within a preset future period includes the following steps: Predict the predicted rotation speed data and predicted temperature field distribution characteristics of the crystal within a preset future period based on the historical rotation speed data and historical temperature field distribution characteristics of the crystal, and then obtain a multi-parameter data stream; Controlling each of the imaging units to image the crystal growth process under the shooting parameters within the preset future period includes the following steps: Real-time monitor the real-time temperature field distribution characteristics, and obtain the deviation degree between the real-time temperature field distribution characteristics and the predicted temperature field distribution characteristics at the corresponding moment; the deviation degree includes the axial temperature gradient deviation and the radial temperature gradient deviation; Specifically, compare the temperature field distribution characteristics obtained by real-time monitoring with the predicted temperature field distribution characteristics at the corresponding moment. The axial temperature gradient deviation is obtained by subtracting the predicted axial temperature gradient from the real-time axial temperature gradient, and the radial temperature gradient deviation is obtained in the same way.

[0042] When the axial temperature gradient deviation exceeds the fourth threshold or the radial temperature gradient deviation exceeds the fifth threshold, trigger the temperature field disturbance compensation mode to eliminate or reduce the interference of the temperature field deviation on the crystal growth image shooting. Among them, the fourth threshold can be selected as 5 °C / cm, and the fifth threshold can be selected as 6 °C / cm.

[0043] Further, the temperature field disturbance compensation mode includes: Pause the current shooting sequence, and start the fast-scanning imaging unit to perform fan-shaped scanning on the abnormal temperature area to obtain a scanning result; Re-calibrate the growth trajectory positioning identifier according to the scanning result to obtain an updated growth trajectory positioning identifier; Resume shooting after obtaining an updated multi-parameter data stream with the updated growth trajectory positioning identifier.

[0044] Specifically, once the temperature field perturbation compensation mode is triggered, the currently ongoing shooting sequence is immediately paused to avoid a decline in image quality caused by continued shooting under abnormal temperature field conditions. The fast-scanning imaging unit is activated to perform a fan-shaped scan of the abnormal temperature region. The fast-scanning imaging unit has a high scanning speed and resolution, capable of obtaining detailed image information of the abnormal temperature region within a short period and yielding a scan result. Based on the result obtained from the fan-shaped scan, the key positions and states of the crystal within a preset future time period are redetermined using the above method (through the prediction model), that is, the growth trajectory positioning identifier is recalibrated to obtain an updated growth trajectory positioning identifier. Based on the updated growth trajectory positioning identifier, in combination with, for example, the updated predicted rotation speed data, predicted temperature field distribution characteristics, etc., an updated multi-parameter data stream is generated. The shooting parameters are adjusted according to the updated multi-parameter data stream, and then the shooting of the crystal growth process is resumed.

[0045] Specifically, abnormal changes in the temperature field may cause changes in the optical properties of the crystal, thereby affecting the clarity and accuracy of the image. Through the temperature field perturbation compensation mode, it is possible to promptly detect and eliminate or reduce the interference of temperature field deviation on the shooting of the crystal growth image, ensuring that the multi-angle images captured can truly and accurately reflect the growth state of the crystal and providing high-quality image materials for subsequent seamless stitching. When the temperature field is abnormal, the growth morphology and surface characteristics of the crystal may change. The fast-scanning imaging unit performs a fan-shaped scan of the abnormal temperature region, which can more accurately capture these changes, enabling the updated multi-parameter data stream to more precisely reflect the actual future growth of the crystal. Adjusting the shooting parameters based on the accurate multi-parameter data stream enables the imaging unit to capture the key features of the crystal at the appropriate time and angle, improving the matching degree between multi-angle images and facilitating seamless stitching.

[0046] During the growth process of a 300-mm diameter silicon single crystal in this application, the missed detection rate of interface defects is reduced from 15% of the traditional method to below 3%, while the image stitching efficiency is increased by 40% (from processing 20 frames per minute to 28 frames per minute).

[0047] In this article, specific examples are used to elaborate on the principle and implementation manner of this application. The description of the above embodiments is only for helping to understand the method and its core idea of this application. The above is only the preferred implementation manner of this application. It should be noted that due to the limitation of literal expression and objectively existing infinite specific structures, for those of ordinary skill in the art, without departing from the principle of this invention, several improvements, refinements or changes can be made, or the above technical features can be combined in an appropriate manner; these improvements, refinements, changes or combinations, or directly applying the concept and technical solution of the invention to other occasions without improvement, should all be regarded as the protection scope of this application.

Claims

1. A method for controlling image generation during crystal growth, characterized in that: The following steps are involved: Real-time collection of axial acceleration, temperature gradient and interface morphology data during crystal growth; According to the axial acceleration and the temperature gradient, determining the current growth stage of the crystal, wherein the growth stage includes a steady-state stage and a transition stage; If the crystal is in the steady-state section, a plurality of imaging units distributed in an array around the crystal are controlled to perform multi-angle synchronous imaging at a fixed time interval; After determining the current growth stage of the crystal, the following steps are also included: If the crystal enters the transition section, the historical rotation speed data and the historical temperature field distribution characteristics of the crystal are retrieved to perform timing prediction and generate a multi-parameter data stream of the crystal in a future preset time period; the multi-parameter data stream includes a plurality of growth trajectory positioning marks of the crystal in the future preset time period, and a predicted acceleration change rate and a predicted interface morphology curvature corresponding to each growth trajectory positioning mark; According to the multi-parameter data stream, obtaining shooting parameters of the imaging unit within the future preset time period, the shooting parameters including a shooting sequence and adjacent imaging intervals; the shooting sequence including a shooting order and a shooting angle of each of the imaging units; Controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period to achieve multi-view image stitching.

2. The method for controlling image generation during crystal growth according to claim 1, characterized in that: The step of determining the current growth stage of the crystal comprises the following steps: If the fluctuation amplitude of the axial acceleration is within the preset steady-state threshold range and the rate of change of the temperature gradient is less than a first threshold, it is determined to be a steady-state segment.

3. The method for controlling image generation during crystal growth according to claim 1, characterized in that: The step of obtaining the shooting parameters of the imaging unit according to the multi-parameter data stream comprises the following steps: Calibrate the target imaging unit corresponding to each growth trajectory positioning mark; According to the sequence of each growth trajectory positioning mark, the target imaging units are sorted to obtain the shooting sequence; According to the predicted interface morphology curvature corresponding to each growth trajectory positioning mark, the target focus distance of each target imaging unit is calculated to obtain the corresponding shooting angle, and then the shooting sequence is obtained.

4. The method for controlling image generation during crystal growth according to claim 1, characterized in that: The step of obtaining the shooting parameters of the imaging unit within the future preset time period according to the multi-parameter data stream comprises the following steps: Classifying target imaging units into high priority and low priority according to the change rate of the predicted interface morphology curvature of the growth trajectory positioning marker; According to the predicted acceleration change rate, a first interval of the target imaging unit with a low priority is obtained; a supplementary shooting frame is inserted into the first interval to obtain a second interval of the target imaging unit with a high priority, and the first interval and the second interval constitute the adjacent imaging interval.

5. The method for controlling image generation during crystal growth according to claim 4, characterized in that: The method of dividing the target imaging units into high priority and low priority according to the change rate of the predicted interface morphology curvature marked by the growth trajectory positioning marker comprises the following steps: The target imaging unit corresponding to the growth trajectory positioning identifier whose change rate of the predicted interface morphology curvature is greater than a second threshold is marked as a high priority.

6. The method for controlling image generation during crystal growth according to claim 3, characterized in that: After performing timing prediction and generating a multi-parameter data stream of a crystal within a future preset time period, the following steps are also included: According to the multi-parameter data stream, determining whether there is an unstable region in the growth process of the crystal within a future preset time period; The controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period comprises the following steps: If not, each of the imaging units is controlled to image the crystal growth process under the shooting parameters within the future preset time period.

7. The method for controlling image generation during crystal growth according to claim 6, characterized in that: The method of judging whether there is an unstable region in the growth process of the crystal in a future preset time period according to the multi-parameter data stream comprises the following steps: When the change rate of the predicted interface morphology curvature exceeds a third threshold, it is determined that interface instability occurs, and the region where the interface instability occurs is regarded as an instability region.

8. The method for controlling image generation during crystal growth according to claim 6, characterized in that: After determining whether there is an unstable region in the growth process of the crystal within a future preset time period according to the multi-parameter data stream, the following steps are also included: If not, shorten the adjacent imaging interval of the corresponding target imaging unit to 1 / 3-1 / 2 of the original interval according to the spatial distribution characteristics of the unstable area to obtain an updated adjacent imaging interval; In the shooting sequence, the number of exposures of the imaging unit corresponding to the angle of the unstable area is increased, and after each exposure, the shooting angle of the target imaging unit behind the unstable area is adjusted according to the predicted curvature change to obtain an updated shooting angle; Based on the updated adjacent imaging interval and the updated shooting angle, obtaining updated shooting parameters; Control each of the target imaging units to image the crystal growth process under the updated shooting parameters within the future preset time period.

9. The method for controlling image generation during crystal growth according to claim 1, characterized in that: The method of retrieving the historical rotation speed data and the historical temperature field distribution characteristics of the crystal to perform timing prediction and generate a multi-parameter data stream of the crystal within a future preset time period includes the following steps: According to the historical rotation speed data and the historical temperature field distribution characteristics of the crystal, the predicted rotation speed data and the predicted temperature field distribution characteristics of the crystal in the future preset time period are predicted, thereby obtaining a multi-parameter data stream; The controlling each of the imaging units to image the crystal growth process under the shooting parameters within the future preset time period comprises the following steps: Real-time monitoring of the real-time temperature field distribution characteristics, and obtaining the deviation between the real-time temperature field distribution characteristics and the predicted temperature field distribution characteristics at the corresponding moment; the deviation includes the axial temperature gradient deviation and the radial temperature gradient deviation; When the axial temperature gradient deviation exceeds a fourth threshold or the radial temperature gradient deviation exceeds a fifth threshold, the temperature field disturbance compensation mode is triggered to eliminate or reduce the interference of the temperature field deviation on the crystal growth image shooting.

10. The method for controlling image generation during crystal growth according to claim 9, characterized in that: The temperature field disturbance compensation mode includes: Pause the current shooting sequence, and start the fast scanning imaging unit to perform sector scanning on the abnormal temperature area to obtain the scanning result; Recalibrate the growth trajectory positioning mark according to the scanning result to obtain an updated growth trajectory positioning mark; The updated growth trajectory is used to locate the marker, and the shooting is resumed after obtaining the updated multi-parameter data stream.

Images