WO2024066413A1 - 加料时机检测方法、装置、电子设备及存储介质 - Google Patents

加料时机检测方法、装置、电子设备及存储介质 Download PDF

Info

Publication number
WO2024066413A1
WO2024066413A1 PCT/CN2023/096088 CN2023096088W WO2024066413A1 WO 2024066413 A1 WO2024066413 A1 WO 2024066413A1 CN 2023096088 W CN2023096088 W CN 2023096088W WO 2024066413 A1 WO2024066413 A1 WO 2024066413A1
Authority
WO
WIPO (PCT)
Prior art keywords
feeding
target image
image
timing detection
detection model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2023/096088
Other languages
English (en)
French (fr)
Inventor
郭力
张伟建
王正远
李广砥
杨正华
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Longi Green Energy Technology Co Ltd
Original Assignee
Longi Green Energy Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Longi Green Energy Technology Co Ltd filed Critical Longi Green Energy Technology Co Ltd
Publication of WO2024066413A1 publication Critical patent/WO2024066413A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • C—CHEMISTRY; METALLURGY
    • C30—CRYSTAL GROWTH
    • C30B—SINGLE-CRYSTAL GROWTH; UNIDIRECTIONAL SOLIDIFICATION OF EUTECTIC MATERIAL OR UNIDIRECTIONAL DEMIXING OF EUTECTOID MATERIAL; REFINING BY ZONE-MELTING OF MATERIAL; PRODUCTION OF A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; SINGLE CRYSTALS OR HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; AFTER-TREATMENT OF SINGLE CRYSTALS OR A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; APPARATUS THEREFOR
    • C30B15/00—Single-crystal growth by pulling from a melt, e.g. Czochralski method
    • C30B15/02—Single-crystal growth by pulling from a melt, e.g. Czochralski method adding crystallising materials or reactants forming it in situ to the melt
    • C—CHEMISTRY; METALLURGY
    • C30—CRYSTAL GROWTH
    • C30B—SINGLE-CRYSTAL GROWTH; UNIDIRECTIONAL SOLIDIFICATION OF EUTECTIC MATERIAL OR UNIDIRECTIONAL DEMIXING OF EUTECTOID MATERIAL; REFINING BY ZONE-MELTING OF MATERIAL; PRODUCTION OF A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; SINGLE CRYSTALS OR HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; AFTER-TREATMENT OF SINGLE CRYSTALS OR A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; APPARATUS THEREFOR
    • C30B15/00—Single-crystal growth by pulling from a melt, e.g. Czochralski method
    • C—CHEMISTRY; METALLURGY
    • C30—CRYSTAL GROWTH
    • C30B—SINGLE-CRYSTAL GROWTH; UNIDIRECTIONAL SOLIDIFICATION OF EUTECTIC MATERIAL OR UNIDIRECTIONAL DEMIXING OF EUTECTOID MATERIAL; REFINING BY ZONE-MELTING OF MATERIAL; PRODUCTION OF A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; SINGLE CRYSTALS OR HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; AFTER-TREATMENT OF SINGLE CRYSTALS OR A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; APPARATUS THEREFOR
    • C30B15/00—Single-crystal growth by pulling from a melt, e.g. Czochralski method
    • C30B15/20—Controlling or regulating
    • C—CHEMISTRY; METALLURGY
    • C30—CRYSTAL GROWTH
    • C30B—SINGLE-CRYSTAL GROWTH; UNIDIRECTIONAL SOLIDIFICATION OF EUTECTIC MATERIAL OR UNIDIRECTIONAL DEMIXING OF EUTECTOID MATERIAL; REFINING BY ZONE-MELTING OF MATERIAL; PRODUCTION OF A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; SINGLE CRYSTALS OR HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; AFTER-TREATMENT OF SINGLE CRYSTALS OR A HOMOGENEOUS POLYCRYSTALLINE MATERIAL WITH DEFINED STRUCTURE; APPARATUS THEREFOR
    • C30B29/00—Single crystals or homogeneous polycrystalline material with defined structure characterised by the material or by their shape
    • C30B29/02—Elements
    • C30B29/06—Silicon

Definitions

  • the present application relates to the technical field of crystal preparation, and in particular to a method for detecting a feeding timing, a device for detecting a feeding timing, an electronic device, and a storage medium.
  • the preparation process of single crystal silicon material is mainly based on the Czochralski process (CZ), which uses the Czochralski process to refine polysilicon raw materials into single crystal silicon.
  • CZ Czochralski process
  • the process of generating rod-shaped single crystal silicon crystals in the Czochralski process is divided into the steps of loading, heating melt, temperature adjustment, seeding, shoulder release, shoulder rotation, equal diameter, and finishing.
  • seeding is to contact the seed crystal (that is, a single crystal processed into a certain shape) installed at the end of the wire rope in advance with the liquid surface.
  • the silicon molecules will grow along the lattice direction of the seed crystal to form a single crystal.
  • Shouldering is to gradually grow the crystal diameter to the required diameter for generation. In the process of shouldering, a section of crystal will be pulled out as the length gradually increases and the diameter gradually increases to about the required diameter, so as to eliminate crystal dislocation.
  • the crystal When the crystal grows to the diameter required for production during the shouldering process, it enters the shoulder turning process. Shoulder turning is to control the crystal diameter to the diameter required for production. After the shoulder turning is completed, it enters the equal diameter control step. In this step, the crystal will grow equal to the set diameter through automatic control of the pulling speed and temperature.
  • the secondary feeding is divided into the feeding of the first stage and the feeding of the circulation stage.
  • the secondary feeding of the first stage refers to the feeding process after the first loading and melting
  • the feeding of the circulation stage refers to the feeding process after the crystal rod is taken out. Due to the limited capacity of the furnace body, the amount of material fed at one time is limited. After the crystal is successfully pulled this time, it is necessary to feed again, and production will continue after the feeding is completed.
  • the direct pulling single crystal equipment repeatedly feeds and pulls crystals for production.
  • the material needs to be fed several times.
  • the on-site operator observes the solid-liquid ratio in the field of view of the heat shield in real time to determine whether the material can be fed. If the added raw materials are not completely melted, the furnace space is not available for adding materials. If the proportion of liquid is large, adding materials will cause silicon splashing and lead to wire breakage, so adding materials cannot be performed.
  • the feeding stage requires the operator to intervene throughout the process, which takes up a lot of operator time and has low work efficiency.
  • an embodiment of the present application is proposed to provide a method for detecting the timing of feeding that overcomes the above problems or at least partially solves the above problems, so as to solve the problem that the operator needs to be fully involved in the feeding stage, which takes up a lot of working hours and has low work efficiency.
  • the embodiment of the present application also provides a feeding timing detection device, an electronic device and a storage medium to ensure the implementation and application of the above method.
  • the present application discloses a method for detecting the timing of adding material, comprising:
  • the feeding timing detection model outputs a feeding timing detection result in the feeding stage.
  • obtaining a target image of the lower opening of the heat shield of the CZ single crystal device includes:
  • the target image is obtained by removing the area image of the heat screen from the captured image.
  • removing the area image of the heat screen in the captured image to obtain the target image includes:
  • the captured image is cut according to a preset cutting method to obtain the target image.
  • the step of cutting the captured image in a preset cutting manner according to the melt region in the binary image to obtain the target image includes:
  • the longest horizontal line segment that crosses the melt area is determined. and the longest longitudinal line segment longitudinally passing through the melt region;
  • the captured image is cut according to the target rectangle to obtain the target image.
  • the method before inputting the target image into the feeding timing detection model, the method further includes:
  • the target image is converted into a target image with a preset size and a preset number of channels.
  • the feeding timing detection model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer, the convolution kernel sizes of the first convolutional layer and the second convolutional layer are the same, the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer, the parameters of the first pooling layer and the second pooling layer are the same, and the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • the method further comprises:
  • the feeding timing detection model is trained by using the feeding label marked for the target image sample, until the loss value of the feeding timing detection model is less than the set loss value, so as to obtain the trained feeding timing detection model.
  • the embodiment of the present application also discloses a feeding timing detection device, comprising:
  • An image acquisition module used for acquiring a target image of the lower opening of the heat shield of the CZ single crystal device after the first feeding of the CZ single crystal in the feeding stage;
  • An input module used for inputting the target image into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and correspondingly marked feeding labels;
  • the result determination module is used to output the feeding timing detection result in the feeding stage according to the target image by the feeding timing detection model.
  • the image acquisition module includes:
  • the image acquisition submodule is used to acquire the thermal shield of the CZ single crystal device after the first feeding.
  • the image removal submodule is used to remove the area image of the thermal screen in the captured image to obtain the target image.
  • the image removal submodule includes:
  • a binarization unit used for performing binarization processing on the captured image to obtain a binarized image
  • the cutting unit is used to cut the captured image according to the molten material area in the binary image in a preset cutting manner to obtain the target image.
  • the cutting unit comprises:
  • a line segment determination subunit used to determine, according to the binary image, the longest horizontal line segment that passes through the molten material region horizontally, and the longest vertical line segment that passes through the molten material region vertically;
  • a rectangle determination subunit used to determine a target rectangle within the melt region by taking the intersection of the longest horizontal line segment and the longest vertical line segment as the center of the rectangle;
  • the cutting subunit is used to cut the captured image according to the target rectangle to obtain the target image.
  • the device further comprises:
  • a conversion module is used to convert the target image into a target image with a preset size and a preset number of channels before inputting the target image into the feeding timing detection model.
  • the feeding timing detection model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer, the convolution kernel sizes of the first convolutional layer and the second convolutional layer are the same, the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer, the parameters of the first pooling layer and the second pooling layer are the same, and the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • the device further comprises:
  • a sample acquisition module used to acquire a target image sample and a corresponding label indicating whether or not to add material after the first feeding in the feeding stage of the CZ single crystal and before acquiring a target image of the lower opening of the heat shield of the CZ single crystal device;
  • a sample input module used for inputting the target image sample and the corresponding label of whether to add material into the feeding time detection model
  • a training module is used to use the label of whether to add material to the target image sample to train
  • the feeding timing detection model is trained until the loss value of the feeding timing detection model is less than the set loss value, thereby obtaining the trained feeding timing detection model.
  • the embodiment of the present application also discloses an electronic device, which includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;
  • Memory used to store computer programs
  • the processor is used to implement the above-mentioned method steps when executing the program stored in the memory.
  • the embodiments of the present application also disclose a readable storage medium.
  • the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute one or more of the feeding timing detection methods described in the embodiments of the present application.
  • a target image of the lower opening of the heat shield of the Czochralski single crystal equipment is obtained, and the target image is input into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and corresponding marked feeding permission labels, and the feeding timing detection model outputs a feeding timing detection result in the feeding stage according to the target image, so that the feeding timing is automatically detected in the feeding stage without the need for the operator to intervene throughout the process, thereby avoiding the problem of occupying more working hours of the operator and improving work efficiency, and real-time automatic detection can prevent the operator from missing the feeding opportunity, thereby improving the accuracy of determining the feeding timing.
  • FIG1 is a flow chart of steps of an embodiment of a method for detecting a feeding timing of the present application
  • FIG2 is a schematic diagram of a feeding timing detection model
  • FIG3 is a flow chart of steps of an embodiment of a method for detecting a feeding timing according to the present application
  • FIG4 is a schematic diagram of an image before and after binarization
  • FIG5 is a schematic diagram of image cutting
  • FIG6 is a structural block diagram of an embodiment of a feeding timing detection device of the present application.
  • FIG. 7 is a diagram showing a computing device for detecting a timing for adding material according to an exemplary embodiment.
  • Step 101 after the first feeding of the feeding stage of the CZ single crystal, a target image of the lower opening of the heat shield of the CZ single crystal equipment is obtained.
  • the Czochralski process is a process of refining raw materials into single crystals using the Czochralski method, for example, a process of Czochralski single crystal silicon.
  • the Czochralski process can be divided into a feeding stage, a melting stage, a temperature adjustment stage, a seeding stage, and the like.
  • the feeding stage of the CZ single crystal process needs to be carried out in multiple times. For example, in one feeding stage, a total of 5 barrels of material need to be added. After the first barrel of material is added, the first barrel of material begins to melt. During the melting process of the first barrel of material, when the ratio of solid and liquid is appropriate, the next barrel of material can be added. The material is added in this way until all 5 barrels of material are added.
  • the CZ single crystal device is a device for CZ single crystal, for example, a single crystal furnace for CZ single crystal silicon.
  • the heat shield is a device for shielding heat diffusion on the crucible in the CZ single crystal device. The inside of the crucible can be observed through the opening at the lower end of the heat shield, and when the material is added to the crucible, the process of the crystal melting from solid to molten can be observed.
  • the target image of the lower opening of the heat shield may be an image taken from somewhere above the heat shield to the lower opening of the heat shield, and the taken image may include the interior of the crucible and a portion of the heat shield, or the taken image may only include the interior of the crucible and not include the heat shield.
  • the target image of the lower opening of the heat shield may also be an image taken from somewhere above the heat shield to the lower opening of the heat shield, and then the regional image of the heat shield in the taken image is removed to obtain a target image, which only includes images of solids and liquids formed by melting raw materials.
  • a suitable target image may be selected according to actual needs, and the embodiment of the present application does not limit this.
  • the added raw materials after the first addition of the raw materials in the feeding stage, the added raw materials begin to melt, and the target images are acquired multiple times, for example, once every 5 seconds, or multiple target images are acquired at multiple set time points, so as to continuously detect the raw materials during the melting process. Measure in order to determine the time to add material in time.
  • Step 102 input the target image into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and correspondingly marked feeding or not labels.
  • the timing of feeding refers to the timing of feeding determined based on the target image when the target image is acquired.
  • the timing of feeding in the feeding stage can be determined by machine learning, and a feeding timing detection model that can detect the timing of feeding is obtained based on the correlation between the target image in the feeding stage and whether feeding can be performed.
  • the charging label is a manually labeled label of whether charging can be added or not.
  • the target image samples can be obtained through multiple experiments, and the charging labels can be manually marked according to the target image samples.
  • the target image samples and the labeled charging labels can also be selected from historical data.
  • the feeding timing detection model can adopt a binary classification model based on a convolutional neural network. Specifically, any applicable model can be adopted, and the embodiment of the present application does not impose any limitation on this.
  • the input of the feeding timing detection model is the acquired target image, and each time the target image is acquired, the feeding timing detection model is input.
  • the model outputs whether feeding is currently possible, for example, outputting 0 means that feeding is currently not possible, and outputting 1 means that feeding is currently possible.
  • the first feeding in the feeding stage of the Czochralski single crystal before obtaining the target image of the lower opening of the heat shield of the Czochralski single crystal equipment, it can also include: obtaining a target image sample and a corresponding marked feeding label; inputting the target image sample and the corresponding marked feeding label into a feeding timing detection model; using the feeding label marked with the target image sample, training the feeding timing detection model until the loss value of the feeding timing detection model is less than the set loss value, thereby obtaining the trained feeding timing detection model.
  • the input is the target image sample and the corresponding labeled charging label.
  • the labeled charging label is used to train the neural network of the charging timing detection model.
  • each time the model outputs the result value of whether to charge it is compared with the actual value (that is, the labeled charging label), and the comparison result is input into the loss function to calculate the loss value.
  • the convergence conditions of the model can be: the loss value is less than the set loss value, or the maximum number of iterations is reached. For example, a relatively small set loss value is set, and each time training is performed, At the same time, calculate the size of the loss value.
  • the model When the loss value is less than the set loss value, the model can be considered to have converged, and the training can be terminated.
  • a relatively large maximum number of iterations is pre-set, such as 100 iterations, or 10,000 iterations, or 1,000,000 iterations, etc., which needs to be selected according to the actual situation. The embodiment of this application does not limit this. After the model completes the specified number of trainings, the model training can be considered complete.
  • the feeding timing detection model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer
  • the convolution kernel size of the first convolutional layer and the second convolutional layer is the same
  • the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer
  • the parameters of the first pooling layer and the second pooling layer are the same
  • the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • the feeding timing detection model can be an image binary classification model based on a convolutional neural network. Through experiments, the model is designed and optimized.
  • the model consists of a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer.
  • the setting of various parameters in the model will affect the operation speed and accuracy of the model. Due to different tasks and hardware conditions, the parameters set are also different to achieve the optimization between operation speed and accuracy.
  • the operation speed and accuracy are optimized.
  • the convolution kernel size of the first convolutional layer and the second convolutional layer is the same
  • the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer
  • the parameters of the first pooling layer and the second pooling layer are the same
  • the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • FIG. 2 a schematic diagram of the feeding timing detection model.
  • the size of the target image is 224 ⁇ 224 ⁇ 3, where 224 is the length and width of the image, and 3 is the number of channels.
  • the number of channels of the output of the first convolutional layer is 32, the size of the convolution kernel is 3 ⁇ 3, and the activation function is the relu (rectified linear unit) function.
  • the kernel size of the first pooling layer is 2 ⁇ 2, and the maximum value pooling is selected.
  • the number of channels of the output of the second convolutional layer is 64, the size of the convolution kernel is 3 ⁇ 3, and the activation function is the relu function.
  • the kernel size of the second pooling layer is 2 ⁇ 2, and the maximum value pooling is selected. Then the two-dimensional output is converted to one dimension, and the number of channels of the output of the fully connected layer is 128, and then the softmax (normalized exponential function) function is used for binary classification, which is classified as feedable and non-feedable.
  • Step 103 outputting the feeding timing detection model according to the target image The results of the feeding timing test during the stage.
  • the feeding timing detection model detects the input target image and generates a detection result of whether feeding is allowed or not, which is recorded as the feeding timing detection result.
  • the output of the feeding timing detection model is obtained to determine whether it is the current feeding timing.
  • the visual system of the CZ equipment obtains the feeding signal, the visual system starts the feeding timing detection model, automatically sets the target image area, and saves the target image of this area locally in real time.
  • the feeding timing detection model obtains the saved target image in real time, and the model makes a judgment and gives the feeding timing detection result of whether feeding can be done.
  • the visual system sends the feeding signal to the control system, and the control system automatically controls the feeding.
  • the visual system automatically shuts down the feeding timing detection model.
  • a target image of the lower opening of the heat shield of the Czochralski single crystal equipment is obtained, and the target image is input into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and corresponding marked feeding permission labels, and the feeding timing detection model outputs a feeding timing detection result in the feeding stage according to the target image, so that the feeding timing is automatically detected in the feeding stage without the need for the operator to intervene throughout the process, thereby avoiding the problem of occupying more working hours of the operator and improving work efficiency, and real-time automatic detection can prevent the operator from missing the feeding opportunity, thereby improving the accuracy of determining the feeding timing.
  • Step 201 after the first feeding, obtain a photographic image of the lower opening of the heat shield of the CZ single crystal device.
  • the lower opening of the heat shield of the CZ single crystal device is photographed to obtain a photographed image. Since the photographed image is usually a regular rectangle, and the shape of the solid and liquid regions formed by the melting of the raw materials in the photographed image is irregular, the photographed image includes parts outside the solid and liquid regions formed by the melting of the raw materials, such as a part of the heat shield.
  • Step 202 removing the area image of the heat screen in the captured image to obtain the target image.
  • the captured image includes the solid and liquid areas formed by the melting of the raw materials.
  • the parts outside the domain such as the regional image of the heat screen, are invalid information that is not helpful for the model to detect the timing of adding fuel, and have a negative impact on the accuracy of the model for detecting the timing of adding fuel.
  • the area image of the heat screen in the captured image is removed to obtain the target image.
  • the target image is an image with the area image of the heat screen removed.
  • the solid and liquid areas formed by the melting of the raw materials in the captured image can be identified, and a square or rectangular image can be cut out from the area.
  • any applicable cutting method can be used, and the embodiment of the present application does not limit this.
  • a specific implementation method of removing the regional image of the heat screen in the captured image to obtain the target image includes: binarizing the captured image to obtain a binary image; and cutting the captured image according to a preset cutting method based on the molten material area in the binary image to obtain the target image.
  • the captured image is binarized.
  • One part of the binarized image is the melt area, and the other part is the non-melt area.
  • the melt area includes the solid and liquid areas formed by the melting of the raw materials.
  • the non-melt area includes the heat shield.
  • the schematic diagram before and after the image is binarized the left side is the captured image, and the right side is the binarized image.
  • the white part in the binarized image is the melt area, and the black part is the non-melt area.
  • the melt area is irregular, and the melt area in different captured images is also different.
  • the images processed are required to be of uniform specifications, so the images need to be cut.
  • the melt area can be obtained.
  • the final image area can be determined according to the preset cutting method. Then the determined image area is cut out on the captured image to obtain the target image.
  • the preset cutting method can be a method of cutting a larger rectangular area within the molten material area, or a method of cutting a larger area of other shapes within the molten material area, or any other applicable method of cutting an area within the molten material area, and the embodiments of the present application are not limited to this.
  • the captured image is cut according to a preset cutting method to obtain a specific implementation method of the target image, including: according to the binary image, determining the longest horizontal line segment that passes through the molten material area horizontally, and the longest vertical line segment that passes through the molten material area vertically; taking the intersection of the longest horizontal line segment and the longest vertical line segment as the center of the rectangle, determining the target rectangle in the molten material area; according to the A target rectangle is used to cut the captured image to obtain the target image.
  • the rectangle can be a square.
  • the captured image is cut to obtain the image of the area within the target rectangle, and the target image is obtained.
  • the target image is obtained.
  • an identical square area is cut out in the captured image as the target image.
  • Step 203 convert the target image into a target image with a preset size and a preset number of channels.
  • the specifications of the target image obtained by the above cutting may not be uniform.
  • the target image is converted to a uniform preset size and preset number of channels. In both the training and use stages of the model, the processed target image and the target image sample need to be processed in the same way.
  • Step 204 input the target image into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and correspondingly marked feeding or not labels.
  • Step 205 Based on the target image, the feeding timing detection model outputs a feeding timing detection result in the feeding stage.
  • a photographic image of the lower end opening of the heat shield of the CZ single crystal device is obtained, and the regional image of the heat shield in the photographic image is removed to obtain the target image, and the target image is converted into a target image of a preset size and a preset number of channels, and the target image is input into the feeding timing detection model, wherein the feeding timing
  • the machine detection model is trained through target image samples and corresponding marked feeding labels.
  • the feeding timing detection model outputs the feeding timing detection result in the feeding stage, so that the feeding timing is automatically detected during the feeding stage without the need for operator intervention throughout the process, thus avoiding the problem of occupying more operator working hours and improving work efficiency.
  • real-time automatic detection can prevent the operator from missing the feeding opportunity, thereby improving the accuracy of determining the feeding timing.
  • FIG. 6 a structural block diagram of an embodiment of a feeding timing detection device of the present application is shown, which may specifically include the following modules:
  • An image acquisition module 301 is used to acquire a target image of the lower opening of the heat shield of the CZ single crystal device after the first feeding of the CZ single crystal in the feeding stage;
  • An input module 302 is used to input the target image into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and correspondingly marked feeding labels;
  • the result determination module 303 is used to output the feeding timing detection result in the feeding stage according to the target image by the feeding timing detection model.
  • the image acquisition module includes:
  • An image acquisition submodule used for acquiring a photographic image of the lower opening of the heat shield of the CZ single crystal device after the first feeding;
  • the image removal submodule is used to remove the area image of the thermal screen in the captured image to obtain the target image.
  • the image removal submodule includes:
  • a binarization unit used for performing binarization processing on the captured image to obtain a binarized image
  • the cutting unit is used to cut the captured image according to the molten material area in the binary image in a preset cutting manner to obtain the target image.
  • the cutting unit comprises:
  • a line segment determination subunit used to determine, according to the binary image, the longest horizontal line segment that passes through the molten material region horizontally, and the longest vertical line segment that passes through the molten material region vertically;
  • a rectangle determination subunit used to determine a target rectangle within the melt region by taking the intersection of the longest horizontal line segment and the longest vertical line segment as the center of the rectangle;
  • the cutting subunit is used to cut the captured image according to the target rectangle to obtain the target image.
  • the device further comprises:
  • a conversion module is used to convert the target image into a target image with a preset size and a preset number of channels before inputting the target image into the feeding timing detection model.
  • the feeding timing detection model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer, the convolution kernel sizes of the first convolutional layer and the second convolutional layer are the same, the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer, the parameters of the first pooling layer and the second pooling layer are the same, and the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • the device further comprises:
  • a sample acquisition module used to acquire a target image sample and a corresponding label indicating whether or not to add material after the first feeding in the feeding stage of the CZ single crystal and before acquiring a target image of the lower opening of the heat shield of the CZ single crystal device;
  • a sample input module used for inputting the target image sample and the corresponding label of whether to add material into the feeding time detection model
  • the training module is used to train the feeding timing detection model by using the feeding label marked on the target image sample, until the loss value of the feeding timing detection model is less than the set loss value, so as to obtain the trained feeding timing detection model.
  • a target image of the lower opening of the heat shield of the Czochralski single crystal device is obtained, and the target image is input into a feeding timing detection model, wherein the feeding timing detection model is trained by target image samples and corresponding marked feeding labels, and the feeding timing detection model outputs the feeding timing detection result in the feeding stage according to the target image, so that the feeding timing is automatically detected in the feeding stage without the need for the operator to intervene throughout the process, thereby avoiding the problem of occupying more working hours of the operator and improving the work efficiency.
  • the real-time automatic detection can prevent the operator from missing the feeding time, thereby improving the accuracy of the feeding time determination.
  • the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
  • Fig. 7 is a block diagram of an electronic device 400 for detecting a charging time according to an exemplary embodiment.
  • the electronic device 400 may be a mobile phone, a computer, a digital broadcast terminal, a messaging device, a game console, a tablet device, a medical device, a fitness device, a personal digital assistant, etc.
  • the electronic device 400 may include one or more of the following components: a processing component 402 , a memory 404 , a power component 406 , a multimedia component 408 , an audio component 410 , an input/output (I/O) interface 412 , a sensor component 414 , and a communication component 416 .
  • the processing component 402 generally controls the overall operation of the electronic device 400, such as operations associated with display, phone calls, data communications, camera operations, and recording operations.
  • the processing component 402 may include one or more processors 420 to execute instructions to complete all or part of the steps of the above-mentioned feeding timing detection method.
  • the processing component 402 may include one or more modules to facilitate the interaction between the processing component 402 and other components.
  • the processing component 402 may include a multimedia module to facilitate the interaction between the multimedia component 408 and the processing component 402.
  • the memory 404 is configured to store various types of data to support operations on the device 400. Examples of such data include instructions for any application or method operating on the electronic device 400, contact data, phone book data, messages, pictures, videos, etc.
  • the memory 404 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.
  • SRAM static random access memory
  • EEPROM electrically erasable programmable read-only memory
  • EPROM erasable programmable read-only memory
  • PROM programmable read-only memory
  • ROM read-only memory
  • magnetic memory flash memory
  • flash memory magnetic disk or optical disk.
  • the power component 404 provides power to the various components of the electronic device 400.
  • the power component 404 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the electronic device 400.
  • the multimedia component 408 includes a screen that provides an output interface between the electronic device 400 and the user.
  • the screen may include a liquid crystal display (LCD) and a touch panel. (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from the user.
  • the touch panel includes one or more touch sensors to sense touch, slide, and gestures on the touch panel. The touch sensor can not only sense the boundaries of the touch or slide action, but also detect the duration and pressure associated with the touch or slide operation.
  • the multimedia component 408 includes a front camera and/or a rear camera. When the electronic device 400 is in an operating mode, such as a shooting mode or a video mode, the front camera and/or the rear camera can receive external multimedia data. Each front camera and rear camera can be a fixed optical lens system or have focal length and optical zoom capabilities.
  • the audio component 410 is configured to output and/or input audio signals.
  • the audio component 410 includes a microphone (MIC), and when the electronic device 400 is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode, the microphone is configured to receive an external audio signal.
  • the received audio signal can be further stored in the memory 404 or sent via the communication component 416.
  • the audio component 410 also includes a speaker for outputting audio signals.
  • I/O interface 412 provides an interface between processing component 402 and peripheral interface modules, such as keyboards, click wheels, buttons, etc. These buttons may include but are not limited to: a home button, a volume button, a start button, and a lock button.
  • the sensor assembly 414 includes one or more sensors for providing various aspects of status assessment for the electronic device 400.
  • the sensor assembly 414 can detect the open/closed state of the device 400, the relative positioning of the components, such as the display and keypad of the electronic device 400, and the sensor assembly 414 can also detect the position change of the electronic device 400 or a component of the electronic device 400, the presence or absence of contact between the user and the electronic device 400, the orientation or acceleration/deceleration of the electronic device 400, and the temperature change of the electronic device 400.
  • the sensor assembly 414 may include a proximity sensor configured to detect the presence of nearby objects without any physical contact.
  • the sensor assembly 414 may also include an optical sensor, such as a CMOS or CCD image sensor, for use in imaging applications.
  • the sensor assembly 414 may also include an accelerometer, a gyroscope sensor, a magnetic sensor, a pressure sensor, or a temperature sensor.
  • the communication component 416 is configured to facilitate wired or wireless communication between the electronic device 400 and other devices.
  • the electronic device 400 can access a wireless network based on a communication standard, such as WiFi, 2G or 3G, or a combination thereof.
  • the communication component 414 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel.
  • the communication component 414 further includes a near field communication (NFC) module to facilitate short-range communication.
  • the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.
  • RFID radio frequency identification
  • IrDA infrared data association
  • UWB ultra-wideband
  • Bluetooth Bluetooth
  • the electronic device 400 can be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components to perform the above-mentioned method for detecting the timing of adding materials.
  • ASICs application-specific integrated circuits
  • DSPs digital signal processors
  • DSPDs digital signal processing devices
  • PLDs programmable logic devices
  • FPGAs field programmable gate arrays
  • controllers microcontrollers, microprocessors or other electronic components to perform the above-mentioned method for detecting the timing of adding materials.
  • a non-transitory computer-readable storage medium including instructions is also provided, such as a memory 404 including instructions, and the instructions can be executed by a processor 420 of an electronic device 400 to complete the above-mentioned feeding timing detection method.
  • the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.
  • a non-transitory computer-readable storage medium when the instructions in the storage medium are executed by a processor of a terminal, enables the terminal to perform a method for detecting a charging time, the method comprising:
  • the feeding timing detection model outputs a feeding timing detection result in the feeding stage.
  • obtaining a target image of the lower opening of the heat shield of the CZ single crystal device includes:
  • the target image is obtained by removing the area image of the heat screen from the captured image.
  • removing the area image of the heat screen in the captured image to obtain the target image includes:
  • the captured image is cut according to a preset cutting method to obtain the target image.
  • the molten material area in the binary image is cut according to a preset cutting method.
  • the method further comprises: cutting the captured image to obtain the target image, comprising:
  • the captured image is cut according to the target rectangle to obtain the target image.
  • the method before inputting the target image into the feeding timing detection model, the method further includes:
  • the target image is converted into a target image with a preset size and a preset number of channels.
  • the feeding timing detection model includes a first convolutional layer, a first pooling layer, a second convolutional layer, a second pooling layer and a fully connected layer, the convolution kernel sizes of the first convolutional layer and the second convolutional layer are the same, the number of channels of the output of the second convolutional layer is twice the number of channels of the output of the first convolutional layer, the parameters of the first pooling layer and the second pooling layer are the same, and the number of channels of the output of the fully connected layer is twice the number of channels of the output of the second convolutional layer.
  • the method further comprises:
  • the feeding timing detection model is trained by using the feeding label marked for the target image sample, until the loss value of the feeding timing detection model is less than the set loss value, so as to obtain the trained feeding timing detection model.
  • the embodiments of the present application can be provided as methods, devices, or computer program products. Therefore, the embodiments of the present application can be in the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the embodiments of the present application can be implemented in a computer program implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The form of the sequence product.
  • each process and/or box in the flowchart and/or block diagram, and the combination of the process and/or box in the flowchart and/or block diagram can be realized by computer program instructions.
  • These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device produce a device for realizing the function specified in one process or multiple processes in the flowchart and/or one box or multiple boxes in the block diagram.
  • These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing terminal device to operate in a predictable manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.
  • These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce computer-implemented processing, so that the instructions executed on the computer or other programmable terminal device provide steps for implementing the functions specified in one or more processes in the flowchart and/or one or more boxes in the block diagram.

Landscapes

  • Chemical & Material Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Crystallography & Structural Chemistry (AREA)
  • Materials Engineering (AREA)
  • Metallurgy (AREA)
  • Organic Chemistry (AREA)
  • Investigating Or Analyzing Materials Using Thermal Means (AREA)
  • Image Analysis (AREA)

Abstract

一种加料时机检测方法、装置、设备及介质。该方法包括：在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果，使得加料阶段时自动检测加料时机，无需操作人员全程介入，避免了占用操作人员的工时较多的问题，提高了工作效率，而且实时自动检测能够避免操作人员错过加料时机，继而提高了加料时机确定的准确性。

Description

加料时机检测方法、装置、电子设备及存储介质
本申请要求在2022年9月30日提交中国专利局、申请号为202211208292.9、名称为“加料时机检测方法、装置、电子设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及晶体制备技术领域，特别是涉及一种加料时机检测方法、一种加料时机检测装置、一种电子设备以及一种存储介质。
背景技术
单晶硅材料的制备工艺以直拉法(Czochralski process/CZ)为主，利用直拉法将多晶硅原料提炼成单晶硅。在直拉单晶过程中生成棒状单晶硅晶体的过程分为装料、加热熔料、调温、引晶、放肩、转肩、等径、收尾等步骤。
其中，当多晶硅原料融化完成后，还不能马上开始引晶，因为这时的温度要高于引晶温度，还必须经过降温，将温度调整到引晶的温度。引晶是将事先装到钢丝绳末端的籽晶(也就是加工成一定形状的单晶)与液面接触，在引晶温度下，硅分子将沿着籽晶的晶格方向生长，从而形成单晶。放肩是将晶体直径逐步生长到生成所要求的直径，在放肩的过程中将拉出随着长度逐渐变长，直径逐渐变大到要求的直径左右的一段晶体，以便消除晶体位错。当晶体在放肩过程中生长到生产要求的直径后，进入转肩过程。转肩是将晶体直径控制在生产所要求的直径。当转肩完成后进入等径控制步骤，在该步骤中，通过对拉速和温度的自动控制，让晶体将按照设定的直径等径生长。
在单晶硅棒生产中，二次加料分为首段和循环段的加料。首段的二次加料是指首次装料并熔料完成后进行的加料过程，循环段的加料是指晶棒提出后进行的加料过程。由于炉体的容量有限，因此一次投料量有限。待此次晶体拉制成功后，需要再次进行加料，加料完成后继续生产。直拉单晶设备反复的加料拉晶来进行生产。
现有技术中，一个加料阶段中需要分多次进行加料，现场操作人员实时观察热屏视野内的固体与液体比例，判断是否可以加料。如果固体的占比较 大，加入的原料未熔完，炉台空间有效不能进行加料。如果液体得占比较大，加料会引起溅硅，导致断线发生，故也不能进行加料。目前的加料阶段需要操作人员全程介入，存在占用操作人员的工时较多，工作效率低下的问题。
发明内容
鉴于上述问题，提出了本申请实施例以便提供一种克服上述问题或者至少部分地解决上述问题的一种加料时机检测方法，以解决加料阶段需要操作人员全程接入，占用工时较多，工作效率低的问题。
相应的，本申请实施例还提供了一种加料时机检测装置、一种电子设备以及一种存储介质，用以保证上述方法的实现及应用。
为了解决上述问题，本申请实施例公开了一种加料时机检测方法，包括：
在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
可选地，所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，包括：
在所述首次加料之后，获取直拉单晶设备的热屏的下端开口的拍摄图像；
去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
可选地，所述去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像，包括：
对所述拍摄图像进行二值化处理，得到二值化图像；
根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
可选地，所述根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像，包括：
根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以 及纵向穿过所述熔料区域的纵向最长线段；
以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；
根据所述目标矩形，切割所述拍摄图像，得到所述目标图像。
可选地，在所述将所述目标图像输入加料时机检测模型之前，所述方法还包括：
将所述目标图像转换为预设尺寸和预设通道数的目标图像。
可选地，所述加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
可选地，在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，所述方法还包括：
获取目标图像样本，以及对应标记的可否加料标签；
将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；
采用针对所述目标图像样本标记的可否加料标签，训练所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
本申请实施例还公开了一种加料时机检测装置，包括：
图像获取模块，用于在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
输入模块，用于将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
结果确定模块，用于根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
可选地，所述图像获取模块，包括：
图像获取子模块，用于在所述首次加料之后，获取直拉单晶设备的热屏 的下端开口的拍摄图像；
图像去除子模块，用于去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
可选地，所述图像去除子模块，包括：
二值化单元，用于对所述拍摄图像进行二值化处理，得到二值化图像；
切割单元，用于根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
可选地，所述切割单元，包括：
线段确定子单元，用于根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以及纵向穿过所述熔料区域的纵向最长线段；
矩形确定子单元，用于以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；
切割子单元，用于根据所述目标矩形，切割所述拍摄图像，得到所述目标图像。
可选地，所述装置还包括：
转换模块，用于在所述将所述目标图像输入加料时机检测模型之前，将所述目标图像转换为预设尺寸和预设通道数的目标图像。
可选地，所述加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
可选地，所述装置还包括：
样本获取模块，用于在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，获取目标图像样本，以及对应标记的可否加料标签；
样本输入模块，用于将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；
训练模块，用于采用针对所述目标图像样本标记的可否加料标签，训练 所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
本申请实施例还公开了一种电子设备，其中，包括处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；
存储器，用于存放计算机程序；
处理器，用于执行存储器上所存放的程序时，实现如上所述的方法步骤。
本申请实施例还公开了一种可读存储介质，当所述存储介质中的指令由电子设备的处理器执行时，使得电子设备能够执行本申请实施例中一个或多个所述的加料时机检测方法。
本申请实施例包括以下优点：
依据本申请实施例，通过在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果，使得加料阶段时自动检测加料时机，无需操作人员全程介入，避免了占用操作人员的工时较多的问题，提高了工作效率，而且实时自动检测能够避免操作人员错过加料时机，继而提高了加料时机确定的准确性。
附图说明
为了更清楚地说明本申请实施例的技术方案，下面将对本申请实施例的描述中所需要使用的附图作简单地介绍。
图1是本申请的一种加料时机检测方法实施例的步骤流程图；
图2是加料时机检测模型的示意图；
图3是本申请的一种加料时机检测方法实施例的步骤流程图；
图4是图像二值化前后的示意图；
图5是图像切割的示意图；
图6是本申请的一种加料时机检测装置实施例的结构框图；
图7是根据一示例性实施例示出的一种用于加料时机检测的计算设备 的结构框图。
具体实施例
为使本申请的上述目的、特征和优点能够更加明显易懂，下面结合附图和具体实施方式对本申请作进一步详细的说明。
参照图1，示出了本申请的一种加料时机检测方法实施例的步骤流程图，具体可以包括如下步骤：
步骤101，在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像。
在本申请实施例中，直拉单晶过程是利用直拉法将原料提炼成单晶的过程，例如，直拉单晶硅的过程。直拉单晶过程可以划分为加料阶段、熔料阶段、调温阶段、引晶阶段等。
在本申请实施例中，直拉单晶过程的加料阶段需要分多次进行加料，例如，一个加料阶段中，共需加5桶料，加第一桶料后，第一桶料开始熔化，在第一桶料的熔化过程中，待固体和液体的比例合适，就可以再加下一桶料，按照这种方式加料，直至5桶料全部加完。
在本申请实施例中，直拉单晶设备是用于直拉单晶的设备，例如，直拉单晶硅的单晶炉。热屏就是直拉单晶设备中坩埚上屏蔽热量扩散的装置。通过热屏的下端开口可以观察到坩埚内部，当加料到坩埚内后，可以观察到晶体从固态熔化成熔液的过程。
在本申请实施例中，热屏的下端开口的目标图像可以是从热屏上方某处向热屏的下端开口处拍摄的图像，拍摄的图像中可以包括坩埚内部和热屏的一部分，或者拍摄的图像中仅包括坩埚内部，不包括热屏。热屏的下端开口的目标图像也可以是从热屏上方某处向热屏的下端开口处拍摄的图像，再去除所述拍摄图像中热屏的区域图像，得到目标图像，仅包括原料熔化形成的固体和液体的图像。具体可以根据实际需要选择合适的目标图像，本申请实施例对此不做限制。
在本申请实施例中，在加料阶段的首次加料之后，加入的原料开始熔化，多次获取目标图像，例如，每隔5秒获取一次目标图像，或者在设定的多个时间点获取多个目标图像等多种方式，以便在原料熔化的过程中不断进行检 测，以便及时确定加料时机。
步骤102，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到。
在本申请实施例中，加料时机是指获取目标图像时，根据目标图像判断的可以加料的时机。加料阶段中的加料时机可以采用机器学习的方式，根据加料阶段中的目标图像和可否加料之间的相关关系，得到一个可以检测加料时机的加料时机检测模型。
为了训练加料时机检测模型，需要准确样本数据以及对应的标签数据，即在加料阶段中获取的目标图像样本，以及对应标记的可否加料标签。可否加料标签为人工标记的可加料或不可加料的标签。具体可以通过多次实验，获得目标图像样本，人工根据目标图像样本标记可否加料标签，也可以从历史数据中选取目标图像样本以及标记的可否加料标签。
在本申请实施例中，加料时机检测模型可以采用基于卷积神经网络的二分类模型，具体可以采用任意适用的模型，本申请实施例对此不做限制。
在本申请实施例中，在加料阶段的首次加料后，加料时机检测模型的输入为获取的目标图像，每次获取到目标图像后，输入加料时机检测模型。该模型输出当前是否可加料，例如，输出0代表当前不可加料，输出1代表当前可加料。
在本申请的一种可选实施例中，在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，还可以包括：获取目标图像样本，以及对应标记的可否加料标签；将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；采用针对所述目标图像样本标记的可否加料标签，训练所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
对加料时机检测模型进行训练时，输入为目标图像样本，以及对应标记的可否加料标签。采用标记的可否加料标签，对加料时机检测模型的神经网络进行训练。也就是说，模型每次输出可否加料的结果值后，与实际值(即标记的可否加料标签)进行比较，并将比较的结果输入到损失函数中，计算得到损失值。模型的收敛条件可以有：损失值小于设定损失值，或者达到最大迭代次数。例如，设定一个比较小的设定损失值，每一次训练的时候，都 同时计算一下损失值的大小，当损失值小于设定损失值，就可以认为模型收敛了，那么就可以结束训练。预先设定一个比较大的最大迭代次数，比如迭代100次，或者10000次，或者1000000次等，需要根据实际情况来选择，本申请实施例对此不做限制。模型完成规定次数的训练之后，就可以认为模型训练完毕。
在本申请的一种可选实施例中，加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
加料时机检测模型可以是一个基于卷积神经网络的图像二分类模型。通过试验，对模型进行设计和优化。该模型由第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层组成。模型中各种参数的设置将影响模型的运算速度和准确性。由于任务以及硬件条件的不同，设置的参数也不同，以达到运算速度和准确性之间的最优化。对于加料时机检测任务，通过大量试验，设置参数时，满足下述关系，对加料时机检测任务来说，运算速度和准确性之间达到最优化。具体为：第一卷积层和第二卷积层的卷积核大小相同，第二卷积层的输出的通道数为第一卷积层的输出的通道数的两倍，第一池化层和第二池化层的参数相同，全连接层的输出的通道数为第二卷积层的输出的通道数的两倍。具体的数值还受到硬件条件的影响，本申请实施例对此不做限制。
例如，如图2所示的加料时机检测模型的示意图。目标图像的尺寸为224×224×3，其中224是图像的长和宽，3为通道数。第一卷积层的输出的通道数为32，卷积核的大小是3×3，激活函数为relu(Rectified Linear Unit，纠正线性单元)函数。第一池化层的核大小是2×2，选择最大值池化。第二卷积层的输出的通道数为64，卷积核大小为3×3，激活函数为relu函数。第二池化层的核大小是2×2，选择最大值池化。然后将二维的输出转化为一维，利用全连接层的输出的通道数为128，再利用softmax(归一化指数函数)函数进行二分类，归为可加料和不可加料。
步骤103，根据所述目标图像，由所述加料时机检测模型输出所述加料 阶段中的加料时机检测结果。
在本申请实施例中，加料时机检测模型根据输入的目标图像进行检测，生成可加料或不可加料的检测结果，记为加料时机检测结果。获得加料时机检测模型的输出，确定当前是否加料时机。
例如，直拉单晶设备的视觉系统获取加料信号，视觉系统启动加料时机检测模型，自动设置目标图像的区域，将此区域的目标图像实时保存在本地。加料时机检测模型实时获取保存的目标图像，模型进行判断，给出是否可以加料的加料时机检测结果，当判断到可以加料时，视觉系统将加料的信号发送至控制系统，控制系统自动控制进行加料。当加料阶段完成后，视觉系统自动关闭加料时机检测模型。
依据本申请实施例，通过在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果，使得加料阶段时自动检测加料时机，无需操作人员全程介入，避免了占用操作人员的工时较多的问题，提高了工作效率，而且实时自动检测能够避免操作人员错过加料时机，继而提高了加料时机确定的准确性。
参照图3，示出了本申请的一种加料时机检测方法实施例的步骤流程图，具体可以包括如下步骤：
步骤201，在所述首次加料之后，获取直拉单晶设备的热屏的下端开口的拍摄图像。
在本申请实施例中，对直拉单晶设备的热屏的下端开口进行拍摄，得到拍摄图像。由于拍摄图像通常为规则的矩形，而拍摄图像中原料熔化形成的固体和液体的区域的形状是不规则的，拍摄图像中包括了原料熔化形成的固体和液体的区域之外的部分，例如，热屏的一部分等。
步骤202，去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
在本申请实施例中，拍摄图像中包括有原料熔化形成的固体和液体的区 域之外的部分，如热屏的区域图像，这一部分对模型检测加料时机没有帮助的无效信息，对于加料时机检测模型的准确性还有负面影响。
在本申请实施例中，去除所述拍摄图像中热屏的区域图像，从而得到目标图像。目标图像是一个去除了热屏的区域图像的图像。具体可以识别出拍摄图像中原料熔化形成的固体和液体的区域，从该区域中切割出一个正方形或长方形的图像，具体可以采用任意适用的切割方式，本申请实施例对此不做限制。
在本申请的一种可选实施例中，去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像的一种具体实现方式中，包括：对所述拍摄图像进行二值化处理，得到二值化图像；根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
为了将拍摄图像分为熔料区域和非熔料区域，对拍摄图像进行二值化处理。二值化图像中中一部分为熔料区域，另一部分为非熔料区域。其中，熔料区域包括原料熔化形成的固体和液体的区域。非熔料区域包括热屏的部分。例如，如图4所示的图像二值化前后的示意图，左侧为拍摄图像，右侧为二值化图像，二值化图像中白色部分为熔料区域，黑色部分为非熔料区域。
熔料区域是不规则的，而且不同的拍摄图像中熔料区域也不相同。模型在训练和检测两个阶段中，要求处理的图像都是统一规格的，因此，还需要对图像进行切割。根据二值化图像，可以得到熔料区域。以二值化图像中的熔料区域为依据，按照预设切割方式，可以确定出最终采用的图像区域。再在拍摄图像上将该确定的图像区域切割出来，得到目标图像。
其中，预设切割方式可以是在熔料区域内，切割一个较大的矩形区域的方式，也可以是在熔料区域内，切割一个较大的其他形状区域的方式，或者其他任意适用的在熔料区域内切割一个区域的方式，本申请实施例对此不做限制。
在本申请的一种可选实施例中，根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像的一种具体实现方式中，包括：根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以及纵向穿过所述熔料区域的纵向最长线段；以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；根据所述 目标矩形，切割所述拍摄图像，得到所述目标图像。
根据二值化图像，找出横向穿过熔料区域的最长线段，记为横向最长线段，找出纵向穿过熔料区域的最长线段，记为纵向最长线段。以横向最长线段和纵向最长线段的交点，作为矩形中心。以该矩形中心，确定熔料区域内的目标矩形。例如，先以该矩形中心，确定一个较大的矩形，如果该矩形中包括非熔料区域，则缩小该矩形，缩小后，再确定缩小后的矩形中是否包括非熔料区域，如果缩小后的矩形中包括非熔料区域，则继续缩小该矩形，直至缩小后的矩形不包括非熔料区域，从而得到一个熔料区域内的目标矩形。矩形可以采用正方形。
根据二值化图像确定的目标矩形，对拍摄图像进行切割，切割出目标矩形内的区域的图像，得到目标图像。例如，如图5所示的图像切割的示意图，按照二值化图像中确定的正方形，在拍摄图像中切割出一个相同的正方形区域，作为目标图像。
步骤203，将所述目标图像转换为预设尺寸和预设通道数的目标图像。
在本申请实施例中，上述切割得到的目标图像的规格可能是不统一的，为了满足模型的处理要求，将目标图像转换为统一的预设尺寸和预设通道数。在模型的训练和使用两个阶段，所处理的目标图像和目标图像样本都需要采用相同的处理方式。
步骤204，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到。
在本申请实施例中，具体实现方式可以参见前述实施例中的描述，此处不另赘述。
步骤205，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
在本申请实施例中，具体实现方式可以参见前述实施例中的描述，此处不另赘述。
依据本申请实施例，通过在所述首次加料之后，每隔设定时长，获取直拉单晶设备的热屏的下端开口的拍摄图像，去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像，将所述目标图像转换为预设尺寸和预设通道数的目标图像，将所述目标图像输入加料时机检测模型，其中，所述加料时 机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果，使得加料阶段时自动检测加料时机，无需操作人员全程介入，避免了占用操作人员的工时较多的问题，提高了工作效率，而且实时自动检测能够避免操作人员错过加料时机，继而提高了加料时机确定的准确性。
需要说明的是，对于方法实施例，为了简单描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本申请实施例并不受所描述的动作顺序的限制，因为依据本申请实施例，某些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于优选实施例，所涉及的动作并不一定是本申请实施例所必须的。
参照图6，示出了本申请的一种加料时机检测装置实施例的结构框图，具体可以包括如下模块：
图像获取模块301，用于在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
输入模块302，用于将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
结果确定模块303，用于根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
可选地，所述图像获取模块，包括：
图像获取子模块，用于在所述首次加料之后，获取直拉单晶设备的热屏的下端开口的拍摄图像；
图像去除子模块，用于去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
可选地，所述图像去除子模块，包括：
二值化单元，用于对所述拍摄图像进行二值化处理，得到二值化图像；
切割单元，用于根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
可选地，所述切割单元，包括：
线段确定子单元，用于根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以及纵向穿过所述熔料区域的纵向最长线段；
矩形确定子单元，用于以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；
切割子单元，用于根据所述目标矩形，切割所述拍摄图像，得到所述目标图像。
可选地，所述装置还包括：
转换模块，用于在所述将所述目标图像输入加料时机检测模型之前，将所述目标图像转换为预设尺寸和预设通道数的目标图像。
可选地，所述加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
可选地，所述装置还包括：
样本获取模块，用于在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，获取目标图像样本，以及对应标记的可否加料标签；
样本输入模块，用于将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；
训练模块，用于采用针对所述目标图像样本标记的可否加料标签，训练所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
依据本申请实施例，通过在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到，根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果，使得加料阶段时自动检测加料时机，无需操作人员全程介入，避免了占用操作人员的工时较多的问题，提高了工 作效率，而且实时自动检测能够避免操作人员错过加料时机，继而提高了加料时机确定的准确性。
对于装置实施例而言，由于其与方法实施例基本相似，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。
图7是根据一示例性实施例示出的一种用于加料时机检测的电子设备400的结构框图。例如，电子设备400可以是移动电话，计算机，数字广播终端，消息收发设备，游戏控制台，平板设备，医疗设备，健身设备，个人数字助理等。
参照图7，电子设备400可以包括以下一个或多个组件：处理组件402，存储器404，电源组件406，多媒体组件408，音频组件410，输入/输出(I/O)的接口412，传感器组件414，以及通信组件416。
处理组件402通常控制电子设备400的整体操作，诸如与显示，电话呼叫，数据通信，相机操作和记录操作相关联的操作。处理组件402可以包括一个或多个处理器420来执行指令，以完成上述的加料时机检测方法的全部或部分步骤。此外，处理组件402可以包括一个或多个模块，便于处理组件402和其他组件之间的交互。例如，处理部件402可以包括多媒体模块，以方便多媒体组件408和处理组件402之间的交互。
存储器404被配置为存储各种类型的数据以支持在设备400的操作。这些数据的示例包括用于在电子设备400上操作的任何应用程序或方法的指令，联系人数据，电话簿数据，消息，图片，视频等。存储器404可以由任何类型的易失性或非易失性存储设备或者它们的组合实现，如静态随机存取存储器(SRAM)，电可擦除可编程只读存储器(EEPROM)，可擦除可编程只读存储器(EPROM)，可编程只读存储器(PROM)，只读存储器(ROM)，磁存储器，快闪存储器，磁盘或光盘。
电力组件404为电子设备400的各种组件提供电力。电力组件404可以包括电源管理系统，一个或多个电源，及其他与为电子设备400生成、管理和分配电力相关联的组件。
多媒体组件408包括在所述电子设备400和用户之间的提供一个输出接口的屏幕。在一些实施例中，屏幕可以包括液晶显示器(LCD)和触摸面板 (TP)。如果屏幕包括触摸面板，屏幕可以被实现为触摸屏，以接收来自用户的输入信号。触摸面板包括一个或多个触摸传感器以感测触摸、滑动和触摸面板上的手势。所述触摸传感器可以不仅感测触摸或滑动动作的边界，而且还检测与所述触摸或滑动操作相关的持续时间和压力。在一些实施例中，多媒体组件408包括一个前置摄像头和/或后置摄像头。当电子设备400处于操作模式，如拍摄模式或视频模式时，前置摄像头和/或后置摄像头可以接收外部的多媒体数据。每个前置摄像头和后置摄像头可以是一个固定的光学透镜系统或具有焦距和光学变焦能力。
音频组件410被配置为输出和/或输入音频信号。例如，音频组件410包括一个麦克风(MIC)，当电子设备400处于操作模式，如呼叫模式、记录模式和语音识别模式时，麦克风被配置为接收外部音频信号。所接收的音频信号可以被进一步存储在存储器404或经由通信组件416发送。在一些实施例中，音频组件410还包括一个扬声器，用于输出音频信号。
I/O接口412为处理组件402和外围接口模块之间提供接口，上述外围接口模块可以是键盘，点击轮，按钮等。这些按钮可包括但不限于：主页按钮、音量按钮、启动按钮和锁定按钮。
传感器组件414包括一个或多个传感器，用于为电子设备400提供各个方面的状态评估。例如，传感器组件414可以检测到设备400的打开/关闭状态，组件的相对定位，例如所述组件为电子设备400的显示器和小键盘，传感器组件414还可以检测电子设备400或电子设备400一个组件的位置改变，用户与电子设备400接触的存在或不存在，电子设备400方位或加速/减速和电子设备400的温度变化。传感器组件414可以包括接近传感器，被配置用来在没有任何的物理接触时检测附近物体的存在。传感器组件414还可以包括光传感器，如CMOS或CCD图像传感器，用于在成像应用中使用。在一些实施例中，该传感器组件414还可以包括加速度传感器，陀螺仪传感器，磁传感器，压力传感器或温度传感器。
通信组件416被配置为便于电子设备400和其他设备之间有线或无线方式的通信。电子设备400可以接入基于通信标准的无线网络，如WiFi，2G或3G，或它们的组合。在一个示例性实施例中，通信部件414经由广播信道接收来自外部广播管理系统的广播信号或广播相关信息。在一个示例性实 施例中，所述通信部件414还包括近场通信(NFC)模块，以促进短程通信。例如，在NFC模块可基于射频识别(RFID)技术，红外数据协会(IrDA)技术，超宽带(UWB)技术，蓝牙(BT)技术和其他技术来实现。
在示例性实施例中，电子设备400可以被一个或多个应用专用集成电路(ASIC)、数字信号处理器(DSP)、数字信号处理设备(DSPD)、可编程逻辑器件(PLD)、现场可编程门阵列(FPGA)、控制器、微控制器、微处理器或其他电子元件实现，用于执行上述加料时机检测方法。
在示例性实施例中，还提供了一种包括指令的非临时性计算机可读存储介质，例如包括指令的存储器404，上述指令可由电子设备400的处理器420执行以完成上述加料时机检测方法。例如，所述非临时性计算机可读存储介质可以是ROM、随机存取存储器(RAM)、CD-ROM、磁带、软盘和光数据存储设备等。
一种非临时性计算机可读存储介质，当所述存储介质中的指令由终端的处理器执行时，使得终端能够执行一种加料时机检测方法，所述方法包括：
在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
可选地，所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，包括：
在所述首次加料之后，获取直拉单晶设备的热屏的下端开口的拍摄图像；
去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
可选地，所述去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像，包括：
对所述拍摄图像进行二值化处理，得到二值化图像；
根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
可选地，所述根据所述二值化图像中的熔料区域，按照预设切割方式切 割所述拍摄图像，得到所述目标图像，包括：
根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以及纵向穿过所述熔料区域的纵向最长线段；
以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；
根据所述目标矩形，切割所述拍摄图像，得到所述目标图像。
可选地，在所述将所述目标图像输入加料时机检测模型之前，所述方法还包括：
将所述目标图像转换为预设尺寸和预设通道数的目标图像。
可选地，所述加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
可选地，在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，所述方法还包括：
获取目标图像样本，以及对应标记的可否加料标签；
将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；
采用针对所述目标图像样本标记的可否加料标签，训练所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
本说明书中的各个实施例均采用递进的方式描述，每个实施例重点说明的都是与其他实施例的不同之处，各个实施例之间相同相似的部分互相参见即可。
本领域内的技术人员应明白，本申请实施例的实施例可提供为方法、装置、或计算机程序产品。因此，本申请实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且，本申请实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程 序产品的形式。
本申请实施例是参照根据本申请实施例的方法、终端设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理终端设备的处理器以产生一个机器，使得通过计算机或其他可编程数据处理终端设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理终端设备以预测方式工作的计算机可读存储器中，使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品，该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理终端设备上，使得在计算机或其他可编程终端设备上执行一系列操作步骤以产生计算机实现的处理，从而在计算机或其他可编程终端设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
尽管已描述了本申请实施例的优选实施例，但本领域内的技术人员一旦得知了基本创造性概念，则可对这些实施例做出另外的变更和修改。所以，所附权利要求意欲解释为包括优选实施例以及落入本申请实施例范围的所有变更和修改。
最后，还需要说明的是，在本文中，诸如第一和第二等之类的关系术语仅仅用来将一个实体或者操作与另一个实体或操作区分开来，而不一定要求或者暗示这些实体或操作之间存在任何这种实际的关系或者顺序。而且，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、物品或者终端设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、物品或者终端设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、物品或者终 端设备中还存在另外的相同要素。
以上对本申请所提供的一种加料时机检测方法和装置、一种电子设备以及一种可读储存介质，进行了详细介绍，本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本申请的限制。

Claims (10)

  1. 一种加料时机检测方法，其中，包括：
    在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
    将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
    根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
  2. 根据权利要求1所述的方法，其中，所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像，包括：
    在所述首次加料之后，获取直拉单晶设备的热屏的下端开口的拍摄图像；
    去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像。
  3. 根据权利要求2所述的方法，其中，所述去除所述拍摄图像中所述热屏的区域图像，得到所述目标图像，包括：
    对所述拍摄图像进行二值化处理，得到二值化图像；
    根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像。
  4. 根据权利要求3所述的方法，其中，所述根据所述二值化图像中的熔料区域，按照预设切割方式切割所述拍摄图像，得到所述目标图像，包括：
    根据所述二值化图像，确定横向穿过所述熔料区域的横向最长线段，以及纵向穿过所述熔料区域的纵向最长线段；
    以所述横向最长线段和纵向最长线段的交点为矩形中心，确定所述熔料区域内的目标矩形；
    根据所述目标矩形，切割所述拍摄图像，得到所述目标图像。
  5. 根据权利要求2所述的方法，其中，在所述将所述目标图像输入加料时机检测模型之前，所述方法还包括：
    将所述目标图像转换为预设尺寸和预设通道数的目标图像。
  6. 根据权利要求1所述的方法，其中，所述加料时机检测模型包括第一卷积层，第一池化层，第二卷积层，第二池化层和全连接层，所述第一卷积层和所述第二卷积层的卷积核大小相同，所述第二卷积层的输出的通道数 为所述第一卷积层的输出的通道数的两倍，所述第一池化层和第二池化层的参数相同，所述全连接层的输出的通道数为所述第二卷积层的输出的通道数的两倍。
  7. 根据权利要求1所述的方法，其中，在所述在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像之前，所述方法还包括：
    获取目标图像样本，以及对应标记的可否加料标签；
    将所述目标图像样本和对应标记的可否加料标签输入加料时机检测模型；
    采用针对所述目标图像样本标记的可否加料标签，训练所述加料时机检测模型，直至所述加料时机检测模型的损失值小于设定损失值，得到训练好的所述加料时机检测模型。
  8. 一种加料时机检测装置，其中，包括：
    图像获取模块，用于在直拉单晶的加料阶段的首次加料之后，获取直拉单晶设备的热屏的下端开口的目标图像；
    输入模块，用于将所述目标图像输入加料时机检测模型，其中，所述加料时机检测模型通过目标图像样本，以及对应标记的可否加料标签训练得到；
    结果确定模块，用于根据所述目标图像，由所述加料时机检测模型输出所述加料阶段中的加料时机检测结果。
  9. 一种电子设备，其中，包括处理器、通信接口、存储器和通信总线，其中，处理器，通信接口，存储器通过通信总线完成相互间的通信；
    存储器，用于存放计算机程序；
    处理器，用于执行存储器上所存放的程序时，实现权利要求1～7任一所述的方法步骤。
  10. 一种可读存储介质，其中，当所述存储介质中的指令由电子设备的处理器执行时，使得电子设备能够执行如方法权利要求1～7中一个或多个所述的加料时机检测方法。
PCT/CN2023/096088 2022-09-30 2023-05-24 加料时机检测方法、装置、电子设备及存储介质 Ceased WO2024066413A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202211208292.9 2022-09-30
CN202211208292.9A CN117845320A (zh) 2022-09-30 2022-09-30 加料时机检测方法、装置、电子设备及存储介质

Publications (1)

Publication Number Publication Date
WO2024066413A1 true WO2024066413A1 (zh) 2024-04-04

Family

ID=90475839

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/096088 Ceased WO2024066413A1 (zh) 2022-09-30 2023-05-24 加料时机检测方法、装置、电子设备及存储介质

Country Status (2)

Country Link
CN (1) CN117845320A (zh)
WO (1) WO2024066413A1 (zh)

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111222360A (zh) * 2018-11-23 2020-06-02 隆基绿能科技股份有限公司 硅料熔化状态的检测方法、设备及存储介质
CN113122919A (zh) * 2021-04-16 2021-07-16 上海新昇半导体科技有限公司 二次加料方法及装置、单晶硅生长方法及装置
CN113673529A (zh) * 2021-08-16 2021-11-19 连城凯克斯科技有限公司 语义分割模型训练方法、硅熔融状态检测方法及电子设备
CN113781474A (zh) * 2021-09-30 2021-12-10 无锡唯因特数据技术有限公司 塌料状态检测方法、装置和存储介质
CN114387248A (zh) * 2022-01-12 2022-04-22 苏州天准科技股份有限公司 一种硅料熔化度监测方法、存储介质、终端和拉晶设备
TWI766600B (zh) * 2021-03-02 2022-06-01 環球晶圓股份有限公司 熔湯的檢測方法
CN115992381A (zh) * 2022-12-28 2023-04-21 浙江晶盛机电股份有限公司 单晶炉复投方法、装置、计算机设备和存储介质

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111222360A (zh) * 2018-11-23 2020-06-02 隆基绿能科技股份有限公司 硅料熔化状态的检测方法、设备及存储介质
TWI766600B (zh) * 2021-03-02 2022-06-01 環球晶圓股份有限公司 熔湯的檢測方法
CN113122919A (zh) * 2021-04-16 2021-07-16 上海新昇半导体科技有限公司 二次加料方法及装置、单晶硅生长方法及装置
CN113673529A (zh) * 2021-08-16 2021-11-19 连城凯克斯科技有限公司 语义分割模型训练方法、硅熔融状态检测方法及电子设备
CN113781474A (zh) * 2021-09-30 2021-12-10 无锡唯因特数据技术有限公司 塌料状态检测方法、装置和存储介质
CN114387248A (zh) * 2022-01-12 2022-04-22 苏州天准科技股份有限公司 一种硅料熔化度监测方法、存储介质、终端和拉晶设备
CN115992381A (zh) * 2022-12-28 2023-04-21 浙江晶盛机电股份有限公司 单晶炉复投方法、装置、计算机设备和存储介质

Also Published As

Publication number Publication date
CN117845320A (zh) 2024-04-09

Similar Documents

Publication Publication Date Title
CN104065878B (zh) 拍摄控制方法、装置及终端
RU2651240C1 (ru) Способ и устройство для обработки фотографий
RU2634907C2 (ru) Метод и устройство для получения фотографии
CN105611413B (zh) 一种添加视频段类别标记的方法和装置
US9891814B2 (en) Display control apparatus and method for displaying information related to icons
CN112312016B (zh) 拍摄处理方法、装置、电子设备和可读存储介质
CN106599070B (zh) 获取第一应用程序中信息的方法、装置及终端设备
CN104967788A (zh) 拍摄方法和装置
CN113215651A (zh) 一种拉晶控制方法和设备、单晶炉以及计算机存储介质
CN105631467A (zh) 显示图片的方法及装置
CN112954214A (zh) 拍摄的方法、装置、电子设备及存储介质
WO2024066413A1 (zh) 加料时机检测方法、装置、电子设备及存储介质
EP4304190A1 (en) Focus chasing method, electronic device, and storage medium
CN117702265A (zh) 加料提示信息确定方法、装置、电子设备及存储介质
CN116145237B (zh) 拉速调整方法、装置、电子设备及存储介质
US10440260B2 (en) Display control apparatus to enable a user to check a captured image after image processing
CN104811619A (zh) 成像校正方法及装置、电子设备
CN116361637B (zh) 氧含量预测方法、装置、电子设备及存储介质
CN116189814B (zh) 功率设定方法、装置、电子设备及存储介质
CN117702266A (zh) 加料时机确定方法、装置、电子设备及存储介质
CN109934264A (zh) 样本采样方法、装置及存储介质
CN118516748A (zh) 放肩控制方法、装置、电子设备及存储介质
CN117853262A (zh) 停电处理方法、装置、电子设备及存储介质
CN116356417A (zh) 转肩启动方法、装置、电子设备及存储介质
WO2023221576A1 (zh) 氧含量控制方法、装置、电子设备及存储介质

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23869666

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 23869666

Country of ref document: EP

Kind code of ref document: A1