WO2020103826A1 - 一种熔料状态检测方法、装置及设备 - Google Patents

一种熔料状态检测方法、装置及设备

Info

Publication number
WO2020103826A1
WO2020103826A1 PCT/CN2019/119464 CN2019119464W WO2020103826A1 WO 2020103826 A1 WO2020103826 A1 WO 2020103826A1 CN 2019119464 W CN2019119464 W CN 2019119464W WO 2020103826 A1 WO2020103826 A1 WO 2020103826A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
melt
sub
state
current image
Prior art date
Application number
PCT/CN2019/119464
Other languages
English (en)
French (fr)
Inventor
郭力
周锐
李侨
徐战军
Original Assignee
隆基绿能科技股份有限公司
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
Priority claimed from CN201811408346.XA external-priority patent/CN111222360B/zh
Priority claimed from CN201910770397.5A external-priority patent/CN112488980B/zh
Application filed by 隆基绿能科技股份有限公司 filed Critical 隆基绿能科技股份有限公司
Publication of WO2020103826A1 publication Critical patent/WO2020103826A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis

Definitions

  • the invention generally relates to the technical field of solar photovoltaic power generation, in particular to a method, device and equipment for detecting the state of molten material.
  • silicon material melting control is an important part of crystal growth.
  • the detection of the melting state of the silicon material helps to analyze the progress of the material process.
  • multiple feedings are required.
  • the heating power needs to be adjusted according to the melting state of the silicon material, and the appropriate feeding timing needs to be selected. Therefore, the detection of the melting process of silicon material has important practical significance in the single crystal silicon industry.
  • the single crystal silicon is prepared by the Czochralski method in a single crystal furnace, heating and melting the silicon material in the crucible, and then infiltrating the seed crystal into the solution, turning the seed crystal and the crucible while pulling the seed crystal, in order to proceed at the lower end of the seed crystal
  • Single crystal silicon rods are prepared by the steps of seeding, shoulder setting, shoulder turning, equal diameter and finishing.
  • the melting process of silicon material is one of the important processes of crystal generation. In this process, the silicon material is put into the crucible. The heater around the crucible is used to heat and melt the silicon material. The heating power is adjusted according to the melting state of the silicon material, and the appropriate feeding time is selected.
  • Correlation calculation is performed based on the base image and the image collected from the current melting of the silicon material, and whether the melting state of the silicon material has been changed is determined according to the magnitude of the correlation. Specifically, a three-dimensional tensor is first constructed, and high-order singular value decomposition of the tensor The base image used to characterize the previous melting state; then, a correlation calculation is performed on the currently collected silicon material melting state image and the base image, and whether the silicon material melting state is changed is determined according to the magnitude of the correlation.
  • the correlation calculation is performed based on the base image and the image collected by the current melting of the silicon material, and the method of determining whether the melting state of the silicon material changes according to the magnitude of the correlation has the problem of low detection accuracy.
  • the invention provides a method, device and equipment for detecting a molten material state, aiming at performing correlation calculation based on the base image and the image collected by the current melting of the silicon material in the prior art, and determining whether the melting state of the silicon material has changed according to the magnitude of the correlation
  • the method has the problem of low detection accuracy.
  • the invention discloses a method for detecting a molten material state.
  • the method includes:
  • the melt state of the melt corresponding to the current image is determined according to the image processing result.
  • the image processing of the current image includes:
  • the determining the melt state of the melt corresponding to the current image according to the image processing result includes: determining the melt state of each sub-region according to the minimum gray value and the variance of the gray value, the The molten state of the sub-region includes liquid or solid;
  • the melt state of the melt corresponding to the current image includes one of solid, liquid or solid-liquid mixture .
  • performing image processing on the current image includes:
  • the determining the molten material state of the molten material corresponding to the current image according to the image processing result includes:
  • the melt state corresponding to the current image is determined according to the degree of similarity.
  • the dividing the current image into multiple sub-regions includes:
  • the current image is a grayscale image
  • the dividing the current image into multiple sub-regions further includes:
  • the dividing the current image into multiple sub-regions further includes:
  • the current image is a color image
  • the current color image divided into a plurality of sub-regions is grayed.
  • the dividing the current image into multiple sub-regions includes:
  • the detection area is set according to the position of the melt in the current image, the detection area is rectangular, and the multiple The sub-regions have the same size.
  • the determining the melt state of each sub-region according to the minimum gray value and the gray value variance includes:
  • the preset condition is that the minimum gray value of the pixels in the sub-region is greater than the first threshold, and the variance of the gray value of the pixels in the sub-region is smaller than the second threshold.
  • the determining the melt state of the melt corresponding to the current image according to the melt state of each sub-region includes:
  • N the number of sub-regions with solid state in all sub-regions
  • the melt state of the melt corresponding to the current image is determined.
  • the solid material of the molten material is added to the molten material.
  • the method further includes;
  • the method before acquiring the current image, the method further includes:
  • performing image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image includes:
  • the method before the normalizing the first image to obtain the second image, the method further includes:
  • the normalizing the first image to obtain the second image includes: normalizing the filtered image to obtain the second image.
  • the performing gray-scale processing on the current image to obtain the first image includes:
  • the pixel value of each pixel in the image measurement area is gray-scaled using a first formula to obtain a first image.
  • the first formula includes :
  • (i, j) represents the coordinates of a pixel in the image measurement area
  • R (i, j) represents the pixel value of the red channel pixel (i, j)
  • G (i, j) represents the green channel pixel
  • B (i, j) represents the pixel value of the blue channel pixel point (i, j)
  • f (i, j) represents the pixel point (i, j) after grayscale processing ) Gray value.
  • the filtering the first image to obtain a filtered image includes:
  • a second formula is used to average the pixel values of each pixel in the first image to obtain a filtered image.
  • the second formula includes:
  • K (i, j) represents the gray value of the pixel point (i, j) after averaging in the neighborhood
  • m * n represents the size of the statistical window
  • (s, t) represents the pixel point in the statistical window
  • the coordinates, f (s, t) represents the gray value of the pixel point (s, t) in the statistical window after gray processing.
  • the normalizing the filtered image to obtain the second image includes:
  • a third formula is used to normalize the pixel value of each pixel in the filtered image to obtain a second image.
  • the third formula includes:
  • K '(i, j) K (i, j)-[K max -K min ] * (f max -f min ) / 255
  • K '(i, j) represents the gray value of the pixel (i, j) after normalization
  • K (i, j) represents the gray of the pixel (i, j) after the neighborhood average Value
  • K max represents the maximum value of pixel gray in the filtered image
  • K min represents the minimum value of pixel gray in the filtered image
  • f max represents the maximum value of pixel gray in the first image
  • f min represents the maximum value of pixel gray in the first image Pixel gray minimum.
  • the similarity degree is a similarity coefficient between the first feature vector and each feature vector in the feature vector list
  • Said determining the state of the melt corresponding to the current image according to the degree of similarity includes:
  • the molten material state corresponding to the second feature vector is determined to be the molten material state corresponding to the current image, and the second feature vector is the most similar to the first feature vector.
  • the invention also discloses a molten material state detection device, which comprises:
  • Acquisition module for acquiring the current image of the melt
  • An image processing module configured to perform image processing on the current image
  • the melt state determination module is configured to determine the melt state of the melt corresponding to the current image according to the image processing result.
  • the image processing module includes:
  • An image processing sub-module which is used to divide the current image into a plurality of sub-regions; and also used to separately obtain the minimum gray value of the pixels in each sub-region and the variance of the gray value of the pixels;
  • the melt state determination module includes: a melt state determination sub-module for determining the melt state of each sub-region according to the minimum gray value and the variance of the gray value, the melt of each sub-region
  • the state of the material includes liquid or solid; it is also used to determine the state of the material corresponding to the current image according to the state of the material in each subregion; the state of the material corresponding to the current image includes the state of solid , Liquid or solid-liquid mixture.
  • the invention also discloses a molten material state detection device.
  • the molten material state detection device includes: an interface, a bus, a first memory and a first processor, and the interface, the first memory and the first processor pass the A bus is connected, the first memory is used to store an executable program, and the first processor is configured to run the executable program to implement the steps of the melt state detection method.
  • the invention also discloses a molten material state detection device.
  • the molten material state detection device includes: an interface, a bus, a second memory and a second processor, and the interface, the second memory and the second processor pass the A bus is connected, the second memory is used to store an executable program, and the second processor is configured to run the executable program to implement the steps of the melt state detection method.
  • the invention also discloses a computer-readable storage medium.
  • the computer-readable storage medium stores a first executable program, and the first executable program is executed by the first processor to implement the melt state detection method A step of.
  • the invention also discloses a computer-readable storage medium.
  • the computer-readable storage medium stores a second executable program, and the second executable program is executed by a second processor to implement the molten material state detection method. A step of.
  • the present application discloses a method for detecting a molten material state.
  • the method includes acquiring a current image of the molten material; performing image processing on the current image; and determining the molten material state of the molten material corresponding to the current image according to the image processing result.
  • only one current image at a time can be used to determine the melt state of the melt, and there is no need to perform correlation calculation with the base image before this time.
  • the brightness of the furnace body is different, so there is no problem of low detection accuracy.
  • the accuracy of the melt state detection of the present application is not affected by the brightness of the single crystal furnace, and is higher than the accuracy.
  • FIG. 1 shows a flowchart of steps of a method for detecting a melt state in Embodiment 1 of the present invention
  • FIG. 2 shows a schematic structural diagram of a molten material state detection device in Embodiment 1 of the present invention
  • FIG. 3 shows a schematic diagram of dividing a current image into sub-regions in Embodiment 1 of the present invention
  • FIG. 4 shows a flowchart of steps of a method for detecting a melt state in Embodiment 2 of the present invention
  • FIG. 5 shows a schematic structural diagram of a molten material state detection device in Embodiment 2 of the present invention
  • FIG. 6 shows a schematic diagram of the current image of the molten material obtained by the camera in the second embodiment of the present invention
  • FIG. 8 shows a schematic diagram of the detection area corresponding to FIG. 7 in Embodiment 2 of the present invention.
  • Embodiment 9 shows a schematic diagram of the variance distribution corresponding to each sub-region in another detection region in Embodiment 2 of the present invention.
  • FIG. 10 is a schematic diagram of the detection area corresponding to FIG. 9 in Embodiment 2 of the present invention.
  • FIG. 11 shows a schematic structural diagram of a molten material state detection device in Embodiment 3 of the present invention.
  • FIG. 12 shows a schematic diagram of a logical structure of a molten material state detection device in the third embodiment of the present invention.
  • FIG. 13 is a flowchart of a method for detecting a molten material state provided by Embodiment 4 of the present disclosure
  • FIG. 14 is a schematic diagram of filtering processing of pixel grayscale provided by Embodiment 4 of the present disclosure.
  • FIG. 15 is a schematic diagram of an image frame acquired when silicon material is melted, provided by Embodiment 4 of the present disclosure.
  • FIG. 16 is a schematic diagram of a method for detecting a melt state provided by Embodiment 4 of the present disclosure.
  • FIG. 17 is a schematic diagram of a method for detecting a molten material state provided by Embodiment 4 of the present disclosure.
  • FIG. 18 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 19 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 20 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 21 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 22 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 23 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 24 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 25 is a structural diagram of a molten material detection device provided in Embodiment 4 of the present disclosure.
  • FIG. 1 shows a flowchart of steps of a method for detecting a melt state according to Embodiment 1 of the present invention.
  • the method includes:
  • Step 101 Obtain the current image of the melt.
  • FIG. 2 shows a schematic structural diagram of a molten material state detection device in Embodiment 1 of the present invention.
  • the state detection device is provided in the single crystal furnace 11 and includes an image collector, specifically a camera 12, a crucible 14 is placed in the single crystal furnace, a melt 13 is placed in the crucible 14, and the camera 12 faces the The orientation of the crucible 14 is used to obtain a current image of the melt 13 in the crucible 14.
  • the molten material 13 may be formed by melting silicon material in the crucible 14, and the molten material 13 includes melted liquid silicon material and unmelted solid silicon material.
  • the placement direction of the camera 12 is subject to being able to obtain the current image of the melt, and the camera 12 is used to obtain the current image of the melt 13, specifically, the camera 12 may be set according to a preset period Acquire the current image of the melt, optionally, the preset period may be 1s.
  • Step 102 Divide the current image into multiple sub-regions.
  • the current image in order to accurately detect the melt state corresponding to the current image, the current image is divided into a plurality of sub-regions, and one sub-region is set to correspond to a melt state.
  • the melt state includes liquid and solid. It can be understood that the smaller the sub-region is divided, the higher the accuracy of the molten material state in the sub-region obtained by calculation, but the calculation complexity is also increased accordingly. Conversely, the larger the sub-region is divided, the accuracy of the melt state in the calculated sub-region will be reduced accordingly, but the calculation complexity will be reduced accordingly.
  • the factors of calculation complexity and accuracy are comprehensively considered, and the sub-regions of the current image are divided. Referring to FIG. 3, a schematic diagram of sub-region division in the first embodiment is shown, and the current image with pixels of 1000pix ⁇ 300pix is divided into 10 ⁇ 10 sub-regions. The pixels of each sub-region are 100pix ⁇ 30pix.
  • Step 103 Obtain the minimum gray value of each pixel and the variance of the gray value of the pixel in each of the sub-regions.
  • obtaining the minimum gray value of the sub-region may include: traversing the gray value of each pixel of the sub-region line by line starting from the first pixel of the sub-region, when the current pixel gray is obtained When the degree value is less than the gray value of the previous pixel, the gray value of the current pixel is saved.
  • the gray value of the pixel obtained is the minimum gray value of the pixel in the sub-region. It can be understood that this embodiment does not limit how to obtain the minimum gray value of the pixels in the sub-region, as long as the minimum gray value of the pixels in the sub-region can be obtained.
  • Obtaining the gray value variance of the pixels in the sub-region may include: first calculating the average value M of the gray values of all pixels in the entire sub-region, obtaining the gray value of each pixel in the sub-region, and recording it as x i , According to the formula The gray value variance S 2 of the pixels in the sub-region is calculated.
  • Step 104 Determine the melt state of each sub-region according to the minimum gray value and the variance of the gray value, and the melt state of each sub-region includes a liquid state and a solid state.
  • the minimum gray value and the gray value variance of each sub-region are obtained through calculation, and determining the melt state of each sub-region according to the minimum gray value and the gray value variance may include: first set A predetermined condition is determined. When both the minimum gray value and the variance of the gray value satisfy the predetermined condition, it is determined that the molten state of the sub-region is liquid, otherwise, the molten state of the sub-region is determined to be solid. Of course, other preset conditions may also be set under actual conditions. When the minimum gray value and the variance of the gray value satisfy the preset conditions, it is determined that the melt state of the sub-region is solid, otherwise, it is determined The melt state in the sub-region is liquid. In this regard, the present invention is not limited.
  • the state of the molten material in the sub-region is determined according to the minimum gray value of the sub-region and the variance of the gray value because when the molten material is in a solid state or a liquid state, there is a large difference in the gray value.
  • the gray value corresponding to the pixels in the sub-region is smaller, and the image region is darker when it appears on the image.
  • the sub-region is liquid, the gray value corresponding to the pixels in the sub-region is larger, and the image region is brighter when it appears on the image.
  • this embodiment obtains the molten material state corresponding to the sub-region by setting corresponding preset conditions.
  • the prior art it is necessary to determine the base image before acquiring the current image, and then calculate the state change of the melt according to the correlation between the base image and the current image. Since the time when the base image is acquired is different from the time when the current image is acquired, the two times The light conditions will be different, which will result in different brightness in the single crystal furnace at two moments, so the detection results obtained in the prior art will be affected by the light, but this application only uses the current image at one moment to detect the current image Corresponding to the state of the molten material, the current image is an image acquired at a time, and there is no need to calculate the correlation with the base image before this time. There are no different fluctuations and different reflections in the single crystal furnace at different times. Interference with different brightness in the furnace. That is, this application only uses the current image to detect the state of the molten material, which can improve the accuracy of the detection.
  • Step 105 Determine the melt state of the melt corresponding to the current image according to the melt state of each sub-region; the melt state of the melt corresponding to the current image includes solid, liquid or solid-liquid mixed Kind of.
  • the number of liquid-sub-regions and the number of solid sub-regions in the image are calculated, according to the number of solid sub-regions in the image
  • the sum of the number of sub-regions in the liquid state determines the melt state of the melt corresponding to the current image. For example, if the number of the sub-regions is 100, the number of solid-state sub-regions in the current image is 30, and the number of liquid sub-regions is 70, then the molten material corresponding to the current image
  • the state of the melt is solid-liquid mixing, and the proportion of the liquid melt in the melt is 70%. For example, if the number of the sub-regions is 100, the number of solid-state sub-regions in the current image is 0, and the number of liquid sub-regions is 100, then the current The melt state is liquid.
  • the method is applied to a molten material state detection device.
  • the molten material state detection device includes a single crystal furnace, including an image collector, and may specifically be a camera.
  • a molten material is placed in the single crystal furnace; the method includes: acquiring the current image of the molten material through the camera; dividing the current image into a plurality of sub-regions; and obtaining the smallest pixel in each of the sub-regions.
  • the gray value and the gray value variance of the pixel the melt state of each sub-region is determined according to the minimum gray value and the gray value variance, the melt state of each sub-region includes liquid and solid; according to the The melt state of the sub-region determines the melt state of the melt corresponding to the current image.
  • the melt state of the melt corresponding to the current image includes one of solid, liquid or solid-liquid mixture.
  • the current image of the melt in the single crystal furnace is collected, and the current image is divided into a plurality of sub-regions, and multiple corresponding sub-regions are determined according to the minimum gray value and the variance of the gray value of the multiple sub-regions.
  • the molten state of the region and then determine the molten state of the molten material corresponding to the current image according to the molten state of the multiple sub-regions, because the application determines the molten state of different sub-regions separately, according to the molten state of different sub-regions Determine the melt state of the melt corresponding to the current image and improve the detection accuracy.
  • FIG. 4 shows a flowchart of steps of a method for detecting a molten material state according to Embodiment 2 of the present invention.
  • the method includes:
  • Step 201 Obtain the current image of the melt.
  • FIG. 5 shows a schematic structural diagram of a molten material state detection device according to Embodiment 2 of the present invention.
  • the device is provided in the single crystal furnace 11 and includes an image collector, specifically a camera 12, a crucible 14 is placed in the single crystal furnace, a melt 13 is placed in the crucible 14, and the camera 12 faces the crucible Placed in the direction of 14, the single crystal furnace is further provided with an observation window 15, and the camera 12 is disposed near the observation window 15.
  • the observation window 15 is provided on the furnace wall of the single crystal furnace, and can be used to pass a part of the camera 12 to enable the camera 12 to obtain the current image of the melt 13 placed in the crucible 14.
  • FIG. 6 shows the current image of the molten material acquired by the camera 12. It can be seen that the gray value of the edge area of the acquired current image of the molten material is 0.
  • a detection area is first set in the acquired current image, and the detection area is an area corresponding to the molten material in the crucible in the current image.
  • the region corresponding to the molten material in the crucible in the current image can be selected to detect the state of the molten material, without detecting the entire current image , Improve the detection efficiency.
  • only the region corresponding to the melt in the crucible in the current image is selected to detect the state of the melt, which can avoid the image detection without the presence of the melt and interfere with the subsequent accuracy of the state of the melt.
  • the detection area can be set as the white border in the current image Out of the rectangular area. It can be seen that the gray values of different pixels in the detection area have obvious changes and have the characteristics of the melt in the molten state. It can be understood that, in this embodiment, the detection area may also be provided in other shapes, as long as the detection area can be guaranteed to have the characteristics of the molten material in the molten state. This embodiment does not limit the shape of the detection area.
  • the detection area is divided into multiple sub-areas, wherein the detection area is set according to the position of the melt in the current image, and the detection area It is rectangular, and the plurality of sub-regions have the same size.
  • Step 2021 when the current image is a grayscale image, divide the grayscale image into a plurality of sub-regions.
  • the current image of the material obtained by the camera is a grayscale image.
  • a detection area is first set on the acquired grayscale image, and then the detection area is divided into a plurality of sub-areas, and the sub-areas have the same size. It can be understood that, according to different situations, The sub-regions can be set to different sizes. In this embodiment, in order to reduce the calculation complexity, the sub-regions are set to the same size.
  • Step 2022 when the current image is a color image, grayscale the color image to obtain a grayscale image
  • Step 2023 When the current image is a color image, divide the current color image into a plurality of sub-regions; divide the current color image into a plurality of sub-regions into grayscales.
  • the color image may be grayscaled to obtain a grayscale image, and then the grayscale image may be divided into multiple sub-regions. You can also choose to divide the current color image into multiple sub-regions, and then grayscale the current color image divided into multiple sub-regions.
  • the present invention is not limited.
  • Step 203 Obtain the minimum gray value of the pixels in each sub-region and the variance of the gray value of the pixels, respectively.
  • step 103 is the same as step 103 in the previous embodiment, which will not be repeated in this embodiment.
  • Step 204 When the minimum gray value and the variance of the gray value satisfy the preset condition, determine that the melt state of the sub-region is liquid; when the minimum gray value and the variance of the gray value do not satisfy the preset condition At this time, it is determined that the melt state of the sub-region is solid.
  • the preset condition is that the minimum gray value of the pixels in the sub-region is greater than the first threshold, and the variance of the gray value of the pixels in the sub-region is smaller than the second threshold.
  • the first threshold is 20, and the second threshold is 10.
  • the molten material is a silicon material melted in a crucible.
  • the silicon material in the crucible exists in a solid-liquid two-phase state
  • the liquid silicon material and the solid silicon material have a difference in gray value in the grayscale image
  • the liquid silicon material is in
  • the gray value in the image is larger than the solid silicon material in the image.
  • the darker area corresponds to the solid silicon material
  • the lighter area corresponds to the liquid silicon material.
  • a suitable threshold is set to define the solid silicon material and the liquid silicon material, and the molten material state of the sub-region can be accurately obtained.
  • the variance of the gray value of the sub-region indicates the uniformity of the gray distribution among the pixels of the sub-region.
  • the variance value of the sub-region When the variance value of the sub-region is large, it indicates that the gray distribution among the pixels of the sub-region is not uniform, which further explains The melt state corresponding to the sub-region is a two-phase state. When the variance value of the sub-region is small, it means that the gray distribution among the pixels of the sub-region is uniform, which further indicates that the state of the melt corresponding to the sub-region is a one-phase state.
  • An appropriate threshold is set to distinguish whether the melt state is one-phase or two-phase state, which has a positive effect on accurately acquiring the melt state of the sub-region.
  • the state of the silicon material corresponding to the sub-region in the crucible is obtained by setting the first threshold and the second threshold.
  • the minimum gray value of the pixel in the sub-region when the minimum gray value of the pixel in the sub-region is greater than 20, it means that the gray value of the pixel in the sub-region is large at this time, and the molten material in the sub-region is liquid. Further, the sub-region The variance of the gray value of the pixel is less than 10, which means that the pixels in the sub-region are uniform at this time, and the difference between the pixels is very small. Through the above two conditions, it can be determined that the melt state of the sub-region is liquid. As another example, when the minimum gray value of the pixel in the sub-region is less than 20, it means that the gray value of the pixel in the sub-region is small at this time, and most of the melt in the sub-region is solid. At this time, It is directly judged that the melt state of the sub-region is solid.
  • step 205 the number of sub-regions in which the state of the molten material is liquid in all the sub-regions is obtained, which is recorded as M.
  • the melt state of each sub-region is obtained.
  • set M M + 1 to traverse the entire detection area to get the state of the melt in all sub-regions.
  • Number of sub-regions M It can be understood that, in this embodiment, other methods may also be used to obtain the number of the sub-regions in which the state of the melt is liquid in all the sub-regions, which is not limited in this embodiment.
  • step 206 the number of sub-regions in which the state of the molten material is solid in all the sub-regions is obtained, which is recorded as N.
  • step 204 obtains the melt state of each sub-region.
  • Step 207 Determine the melt state corresponding to the current image according to N and M.
  • the melt state of the melt corresponding to the current image includes one of solid, liquid, or solid-liquid mixture.
  • the state of the melt corresponding to the current image may be determined according to the ratio between the N and M and the total number of the sub-regions.
  • the detection area is divided into 100 sub-areas of the same size.
  • FIG. 7 shows the variance distribution corresponding to each sub-area in the detection area, where the abscissa Each sub-region is represented, and the ordinate is the variance of the gray value corresponding to each sub-region;
  • FIG. 8 shows the detection region. It can be seen that the gray value variance of each sub-region in FIG. 7 is distributed between 1-70, and it can also be seen from the detection region in FIG. 8 that the state of the melt at this time is solid-liquid mixing. By calculating the detection area in FIG. 8 through the above steps, it is obtained that the proportion of liquid silicon material in the melt corresponding to the detection area at this time is 15%.
  • the detection area is also divided into 100 sub-areas of the same size.
  • FIG. 9 shows the variance distribution corresponding to each sub-area in the detection area, where, The abscissa represents each sub-region, and the ordinate represents the variance of the gray value corresponding to each sub-region;
  • FIG. 10 shows the detection region. It can be seen that the gray value variance of each sub-region in FIG. 9 is distributed between 1-2.5, and it can also be seen from the detection region in FIG. 10 that the melt state at this time is liquid. By calculating the detection area in FIG. 10 through the above steps, it is obtained that the proportion of liquid silicon material in the melt corresponding to the detection area at this time is 100%.
  • Step 208 when the molten material state of the molten material corresponding to the current image satisfies the first condition, the solid raw material of the molten material is added.
  • the silicon material in the crucible of the single crystal furnace when the silicon material in the crucible of the single crystal furnace is completely melted, it is easy to cause silicon splashing, and the splashed solution may splash the heater, insulation barrel, and crucible in the single crystal furnace, so that the heater and the insulation Cracks occur in the barrel and crucible, so it is necessary to adjust the heating power according to the melting state of the silicon material and select the appropriate feeding timing.
  • the solid raw material of the molten material is added to the crucible of the single crystal furnace; specifically, the suitable molten material reaches The state of the molten material may be that the proportion of the liquid silicon material in the molten material is 50-70%, and when the proportion of the liquid silicon material in the molten material is greater than 70%, the molten metal is added to the crucible.
  • the liquid silicon material in the crucible is too much, causing silicon splashing; when the proportion of liquid silicon material in the melt is less than 50%, it means that most of the solid silicon material has not begun to melt at this time.
  • the above first condition is that the proportion of liquid silicon material in the melt is 50-70%, and when the proportion of liquid silicon material in the melt is 50-70%, add the content to the crucible When the solid material of the melt is described, it will not cause silicon splash, nor will it affect the melting speed of the silicon material.
  • Step 209 When the melt state of the melt corresponding to the current image satisfies the second condition, control the melt to enter a temperature stabilization phase; after the temperature stabilization phase ends, perform seeding.
  • temperature stabilization needs to be performed to ensure that the silicon liquid is stabilized to an appropriate seeding temperature, so that the seed crystal and the molten silicon liquid are fused to perform crystal growth in the seeding process.
  • all the molten material in the crucible is liquid, that is to say, the state of the molten material is liquid at this time.
  • the above second condition may be that the state of the melt is liquid.
  • a molten material is placed in the single crystal furnace; the method includes: acquiring a current image of the molten material; dividing the current image into a plurality of sub-regions; and separately acquiring the smallest pixel in each of the sub-regions
  • the gray value and the gray value variance of the pixel the melt state of each sub-region is determined according to the minimum gray value and the gray value variance, the melt state of each sub-region includes liquid and solid; according to the The melt state of the sub-region determines the melt state of the melt corresponding to the current image, and the melt state of the melt corresponding to the current image includes one of solid, liquid, or solid-liquid mixture.
  • the current image of the melt in the single crystal furnace is collected, and the current image is divided into a plurality of sub-regions, and multiple corresponding sub-regions are determined according to the minimum gray value and the variance of the gray value of the multiple sub-regions
  • the molten state of the region and then determine the molten state of the molten material corresponding to the current image according to the molten state of the multiple sub-regions, because the application determines the molten state of different sub-regions separately, according to the molten state of different sub-regions Determine the solid-liquid distribution of the melt corresponding to the current image, which improves the detection accuracy, and in calculating the melt state of different sub-regions in this application, only the minimum gray value and the gray value variance of multiple sub-regions are considered to reduce The calculation complexity is improved, and the detection efficiency is improved.
  • FIG. 11 shows a molten material state detection device in Embodiment 3 of the present invention.
  • the device includes: an image detection unit.
  • the image detection unit may be a camera 12, and the device is provided
  • the single crystal furnace includes a crucible 14, the melt 13 is placed in the crucible 14, and the camera 12 is placed toward the crucible 14 for collecting a current image of the melt;
  • the device further includes a detection unit 16, which is connected to the camera 12 and is used to obtain a current image of the melt material collected by the camera.
  • connection method of the camera 12 and the detection unit 16 includes a wired connection and a wireless connection, and this embodiment of the present invention is not limited.
  • the camera 12 is used to collect the current image of the melt 13;
  • the detection unit 16 is used to divide the current image into a plurality of sub-regions, and then separately obtain the minimum gray value of the pixels in each sub-region and the variance of the gray value of the pixels;
  • the detection unit 16 is further configured to determine the melt state of each sub-region according to the minimum gray value and the variance of the gray value, and further determine the melt corresponding to the current image according to the melt state of each sub-region
  • the molten state of each sub-region includes liquid and solid.
  • the molten state of the molten material corresponding to the current image includes one of solid, liquid or solid-liquid mixture.
  • the detection unit 16 is used to divide the current image into a plurality of sub-regions when the current image is a grayscale image
  • the detection unit 16 is further configured to divide the current image into a plurality of sub-regions after the current image is grayscaled when the current image is a color image;
  • the detection unit 16 is further configured to, when the current image is a color image, divide the current image into multiple sub-regions, and then grayscale the current image divided into multiple sub-regions.
  • an observation window 15 is provided on the furnace wall of the single crystal furnace, and the camera 12 is disposed near the observation window 15 to obtain a current image of the melt 13 in the crucible 14;
  • the detection unit 16 is further configured to set a detection area of the current image of the melt and divide the detection area into a plurality of sub-areas; wherein the detection area is set according to the position of the melt in the current image It is determined that the detection area is rectangular, and the plurality of sub-areas have the same size.
  • the detection unit 16 is further configured to determine that the melt state of the sub-region is liquid when the minimum gray value and the variance of the gray value satisfy a preset condition; when the minimum gray When the variance of the value and the gray value does not satisfy the preset condition, it is determined that the melt state of the sub-region is solid; wherein, the preset condition is that the minimum gray value of the pixel in the sub-region is greater than the first threshold, so The variance of the gray value of the pixels in the sub-region is smaller than the second threshold.
  • the detection unit 16 is further used to obtain the number of sub-regions in which the state of the melt is liquid in all sub-regions, which is denoted as M; It is denoted as N; further, according to N and M, the melt state corresponding to the current image is determined.
  • the device further includes a control unit 17 and a feeder 18.
  • the control unit 17 is used to issue a first prompt when the melt state of the melt corresponding to the detection area satisfies the first condition signal;
  • the first control signal is used to prompt that the state of the molten material satisfies the first timing, and the first timing is the timing to add the solid material of the molten material into the single crystal furnace.
  • the solid material of the molten material is added into the single crystal furnace through the feeder 18 to prevent the human body from being injured when the solid material is added manually.
  • the device further includes a crystal pulling rope 19, and the control unit 17 is further configured to control the entry of the molten material when the molten material state corresponding to the detection area satisfies the second condition A temperature stabilization phase; and when the temperature stabilization phase ends, a second prompt signal is issued;
  • the second prompt signal is used to prompt that the state of the molten material satisfies the second timing, and the second timing is the timing for crystallizing the molten material in the single crystal furnace.
  • the seeding rope 19 is used for seeding.
  • the functions of the various parts in the melt state detection device can be referred to the related records in the foregoing embodiments, and the same beneficial effects can be achieved. In order to avoid repetition, they will not be repeated here.
  • the molten material state detection device includes: a single crystal furnace and a camera, the single crystal furnace includes a crucible, the molten material is placed in the crucible, and the camera faces the crucible Placed in the direction for collecting the current image of the melt; the device further includes a detection unit connected to the camera for acquiring the current image of the melt collected by the camera.
  • the detection unit is used to divide the current image into a plurality of sub-regions, and then separately obtain the minimum gray value of the pixels in each sub-region and the gray value variance of the pixels; the detection unit is also used to The minimum gray value and the variance of the gray value determine the melt state of each sub-region, and then determine the melt state of the melt corresponding to the current image according to the melt state of each sub-region;
  • the melt state includes liquid and solid.
  • the current image of the melt in the single crystal furnace is collected, and the current image is divided into a plurality of sub-regions, and multiple corresponding sub-regions are determined according to the minimum gray value and the variance of the gray value of the multiple sub-regions.
  • the molten state of the region and then determine the molten state of the molten material corresponding to the current image according to the molten state of the multiple sub-regions, because the application determines the molten state of different sub-regions separately, according to the molten state of different sub-regions Determine the melt state of the melt corresponding to the current image and improve the detection accuracy.
  • FIG. 12 shows a schematic diagram of a logical structure of a molten material state detection device according to an embodiment of the present invention.
  • the melt state detection device provided by the embodiment of the present invention may include: an interface 41, a first processor 42, a second memory 43, and a bus 44; wherein, the bus 44 is used to implement the interface 41.
  • the executable program stored in 43 to realize the steps of the method for detecting the melt state in the first embodiment or the second embodiment shown in FIG. 1 or FIG. 4, and can achieve the same or similar effects. Repeat again.
  • the present invention also provides a computer-readable storage medium that stores one or more first executable programs, and the one or more first executable programs can be used by one or more first
  • the processor executes to realize the steps of the method for detecting the molten material state in the first embodiment or the second embodiment as shown in FIG. 1 or FIG. 4, and can achieve the same or similar effects.
  • An embodiment of the present disclosure provides a method for detecting a molten material state. As shown in FIG. 13, the method for detecting a molten material state includes the following steps:
  • acquiring the current image includes receiving the collected current image sent by the CCD camera.
  • the CCD camera can collect image frames in which the silicon material is in different melting states.
  • the embodiment of the present disclosure uses the current image as an example to describe the detection method of the melt state corresponding to the current image.
  • performing image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image includes:
  • step S4 is further included before step S2: performing filtering processing on the first image to obtain a filtered image. Then, step S2 performing normalization processing on the first image to obtain the second image includes: performing normalization processing on the filtered image to obtain the second image.
  • performing gray-scale processing on the current image to obtain the first image includes: determining the image measurement area from the current image; according to the pixel value of each pixel in the image measurement area, using the first formula to the image measurement area The pixel value of each pixel is gray-scaled to obtain the first image.
  • the first formula includes:
  • (i, j) represents the coordinates of a pixel in the image measurement area
  • R (i, j) represents the pixel value of the red channel pixel (i, j)
  • G (i, j) represents the green channel pixel
  • B (i, j) represents the pixel value of the blue channel pixel point (i, j)
  • f (i, j) represents the pixel point (i, j) after grayscale processing ) Gray value.
  • the CCD camera Since the CCD camera is set above the side of the single crystal furnace, the CCD camera may have an area that is blocked by the thermal screen of the single crystal furnace or some other components in the acquired image frame. Therefore, the image measurement needs to be determined from the current image In the area, in view of the silicon material melting, the silicon material near the edge of the heat shield of the single crystal furnace melts first, so the image measurement area can be set near the edge of the heat shield, so that the melting state of the silicon material can be accurately determined. By averaging the pixel values of each pixel in the image measurement area, the gray value of each pixel in the measurement area is obtained.
  • the purpose of the gray scale processing of the image measurement area is to convert the color image into a gray image To facilitate subsequent image processing.
  • filtering the first image to obtain the filtered image includes: setting the size of the statistical window; according to the size of the statistical window, using the second formula to average the pixel values of each pixel in the first image to obtain To filter the image, the second formula includes:
  • the second formula includes:
  • K (i, j) represents the gray value of the pixel point (i, j) after averaging in the neighborhood
  • m * n represents the size of the statistical window
  • (s, t) represents the pixel point in the statistical window
  • the coordinates, f (s, t) represents the gray value of the pixel point (s, t) in the statistical window after gray processing.
  • the size of the statistical window can be determined according to the size of the selected image measurement area or can be selected based on experience. Generally, the size of the statistical window is 3 ⁇ 3, 5 ⁇ 5, 7 ⁇ 7, etc.
  • the gray value of each pixel in the first image is set to the average value of the gray values of all pixels in the pixel statistical window, so as to smooth the pixels with a sudden change in gray Filtering. As shown in FIG.
  • the size of the image measurement area is 9 ⁇ 9
  • each pixel in the image measurement area is represented by “ ⁇ ”
  • the size of the statistical window is 3 ⁇ 3
  • the portion shown by the dotted frame is the size of the statistical window Taking the neighborhood averaging of black bold pixels ⁇ as an example, the average value of the gray values of all pixels in the statistical window is calculated, and the average value is determined as the gray value of the pixels of black bold pixels ⁇ .
  • step S2 normalizing the filtered image to obtain the second image includes:
  • the third formula is used to normalize the pixel value of each pixel in the filtered image to obtain a second image.
  • the third formula includes:
  • K '(i, j) K (i, j)-[K max -K min ] * (f max -f min ) / 255
  • K '(i, j) represents the gray value of the pixel (i, j) after normalization
  • K (i, j) represents the gray of the pixel (i, j) after the neighborhood average Value
  • K max represents the maximum value of pixel gray in the filtered image
  • K min represents the minimum value of pixel gray in the filtered image
  • f max represents the maximum value of pixel gray in the first image
  • f min represents the maximum value of pixel gray in the first image Pixel gray minimum.
  • the gray histogram of the first image can be obtained, and the gray value of the pixel in the first image can be obtained according to the gray histogram
  • the maximum value f max and the pixel gray minimum value f min it is also possible to sort the gray values of all pixels in the first image in order from large to small or small to large, so as to obtain the pixels in the first image
  • the maximum value of gray value f max and the minimum value of pixel gray value f min .
  • performing feature extraction on the second image to obtain the first feature vector of the current image includes: performing feature extraction on the gray value in the second image to obtain the first feature vector of the current image.
  • the feature vector list is preset. Specifically, before step 101, the method further includes: obtaining at least P image frames of different melt states in P single crystal furnaces, P ⁇ 1, Q ⁇ 1; according to a preset algorithm, P * Q image frames Perform image processing on each image frame in to obtain the feature vector of each image frame; classify the feature vectors of P * Q image frames to obtain a list of feature vectors.
  • the preset algorithm described here is the same as the preset algorithm described in step 102.
  • each of the P * Q image frames is processed to obtain each image frame Corresponding feature vectors; then, all feature vectors of the P * Q image frames are classified according to preset rules, and different categories correspond to different melt states, so that the corresponding relationship between each feature vector and the melt state is obtained Feature vector list.
  • step 1304 specifically includes: determining the melt state corresponding to the second feature vector as the current image Corresponding melt state.
  • the second feature vector is the most similar to the first feature vector, which means that the first feature vector is closest to the second feature vector, and the feature vector list contains the correspondence between each feature vector and the state of the melt, Therefore, the melt state corresponding to the second feature vector is determined as the melt state corresponding to the current image.
  • the similarity coefficient includes the similarity parameter and the dissimilarity parameter. Both parameters can measure the similarity.
  • the difference is that the value of the similarity parameter directly reflects the degree of similarity between the two feature vectors. Larger means more similar, and the value of the dissimilarity parameter reflects the degree of difference between the two feature vectors. The smaller the value, the more similar.
  • the similarity coefficient between the first feature vector and each feature vector in the feature vector list may be the distance coefficient between the two feature vectors or the angle cosine between the two feature vectors, which is selected according to the actual situation. The embodiments of the present disclosure do not impose any restrictions on this.
  • the method for detecting the molten material state includes acquiring the current image; performing image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image; acquiring each of the first feature vector and the feature vector list
  • the similarity degree of the feature vectors, the feature vector list contains the correspondence between each feature vector and the melt state; the melt state corresponding to the current image is determined according to the similarity degree.
  • This method uses two similar feature vectors to determine the current molten state of the image, the calculation is simple, and it can adapt to different single crystal furnaces, determine the melting state of the silicon material in real time, and prompt the operator to perform steps such as feeding or power control. Improve the accuracy of test results.
  • another embodiment of the present disclosure provides a method for detecting the molten material state.
  • the method for detecting the molten material state provided in this embodiment includes the following steps: an offline learning program Storage, online image acquisition and determination of the degree of completion of the melt, the melt described here is silicon material, and the judgment of the degree of completion of the melt is to determine the current state of the melt in the single crystal furnace.
  • the first step is to store the offline learning program.
  • the specific steps mainly include the following:
  • a CCD camera is used to collect images of different furnace bodies in different melting states.
  • the specific collected images are shown in FIG. 16, and some silicon materials 41 in the furnace body are not completely melted. Due to the blockage of the heat shield and other components in the furnace body, a part of the silicon material melting image could not be displayed.
  • the side arc region in FIG. 16 shows the captured edge image 42 of the heat shield, the upper arc region and the lower arc region are in the furnace body Other structural images.
  • the CCD camera collects the image, it is input into the industrial control computer through the circuit, and the single crystal growth image is processed by the image processing program of the industrial control computer. Second, set the image measurement area. As shown in FIG.
  • the image measurement area 43 is disposed near the edge 42 of the thermal screen, and the measurement area is automatically identified by the feature recognition module included in the image processing program.
  • the silicon material near the edge of the heat shield is melted first, so this embodiment selects the measurement area near the edge image of the heat shield to determine the state of the melt in time.
  • image processing includes performing gray-scale processing, smooth filtering processing and normalization processing on the images located in the measurement area among the collected images of different melt states of different furnaces. In this embodiment, the following methods are used to process all the collected images of different furnaces in different melt states:
  • Gray-scale processing the images collected in different furnaces with different melt conditions are processed by three-component brightness averaging method to gray-scale the image in the measurement area.
  • the calculation principle is as follows:
  • (i, j) represents the coordinates of a pixel in the image measurement area
  • R (i, j) represents the pixel value of the red channel pixel (i, j)
  • G (i, j) represents the green channel pixel
  • B (i, j) represents the pixel value of the blue channel pixel point (i, j)
  • f (i, j) represents the pixel point (i, j) after grayscale processing ) Gray value.
  • the other pixels in the measurement area adopt the same processing method.
  • K (i, j) represents the gray value of the pixel point (i, j) after averaging in the neighborhood
  • m * n represents the size of the statistical window
  • (s, t) represents the pixel point in the statistical window
  • the coordinates, f (s, t) represents the gray value of the pixel point (s, t) in the statistical window after gray processing.
  • c. Normalization Obtain the grayscale histogram of the image after grayscale and smooth filtering in the measurement area of all collected images, and perform normalization.
  • each image that has undergone grayscale and smooth filter processing is normalized.
  • the normalized calculation principle is as follows:
  • K '(i, j) K (i, j)-[K max -K min ] * (f max -f min ) / 255
  • K '(i, j) represents the gray value of the pixel (i, j) after normalization
  • K (i, j) represents the gray of the pixel (i, j) after the neighborhood average Value
  • K max represents the maximum value of pixel gray in the filtered image
  • K min represents the minimum value of pixel gray in the filtered image
  • f max represents the maximum value of pixel gray in the first image
  • f min represents the maximum value of pixel gray in the first image Pixel gray minimum.
  • the feature vectors are extracted and classified, and stored in the offline learning program.
  • feature vector extraction is performed on the gray values of all collected images in the measurement area. Specifically, after all the collected images of different furnaces with different melt conditions are processed above, feature vectors are extracted separately, and the extracted feature vectors are divided into multiple categories and stored in the offline learning program. Different categories correspond to different melts Material status.
  • the offline learning program is an offline classifier, and of course it may be other stored programs.
  • the offline classifier includes two parameters: one is the sample array X, the number of rows is equal to the number of collected samples, and the number of columns is equal to the length of the feature vector, and each row is a feature vector.
  • the actual structure of X is equivalent to arranging each feature vector line by line to form an array.
  • Another parameter is the category vector y, whose elements can only take certain types.
  • all samples are divided into 5 categories, which can be represented by 1, 2, 3, 4, and 5, indicating that the proportion of liquid is 20%, 40%, 60%, 80%, and 100% , Corresponding to different melt progress.
  • all samples can also be divided into more than 5 categories or less than 5 categories, and the corresponding feature vectors can be divided into different categories.
  • the second step is online image acquisition and processing.
  • the specific steps mainly include the following:
  • a CCD camera is used to collect real-time images of the melt state in the furnace.
  • the measurement area of the online real-time image is the same as the image measurement area set in step 5 of the offline learning program storage.
  • the three-component brightness averaging method is used to grayscale the image in the measurement area; the neighborhood smoothing filter method is used to smooth the image in the measurement area, and the processing method is the same as that in the offline learning program storage step.
  • the real-time image processing in the measurement area also includes acquiring the grayscale histogram of the real-time image in the measurement area and performing normalization processing.
  • the processing method is the same as in the offline learning program storage step.
  • the third step is to judge the degree of completion of the melt.
  • the vector angle method is used to compare feature vectors to determine the degree of completion of the melt.
  • the main contents are as follows:
  • the extracted feature vector of the current frame and the multiple class feature vectors in the offline classifier are respectively calculated by the vector angle method.
  • a value close to 1 represents that the current feature vector is the most similar to a certain feature vector in the class feature vector.
  • the degree of completion of the melt corresponding to the feature vector of this category is the current degree of completion of the silicon melt.
  • the calculation principle of the vector angle method is: the distance between two vectors will be compared to the cosine of the angle between the two vectors. Because the cosine of the angle between the vectors is between 0 and 1, it is better to normalize the distance than the distance between the vectors, so it is easy to determine the classification threshold.
  • the method for detecting the molten material state obtained by the embodiment of the present disclosure obtains the current image; performs image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image; and obtains the first feature vector and each feature in the feature vector list
  • the feature vector list contains the correspondence between each feature vector and the melt state; the melt state corresponding to the current image is determined according to the similarity.
  • This method uses two similar feature vectors to determine the current molten state of the image, the calculation is simple, and it can adapt to different single crystal furnaces, determine the melting state of the silicon material in real time, and prompt the operator to perform steps such as feeding or power control. Improve the accuracy of test results.
  • the molten material state detection device 60 includes: a first acquisition module 601, an image processing module 602, a second acquisition module 603, and a determination module 604
  • the first acquisition module 601 is used to acquire the current image
  • the image processing module 602 is used to perform image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image
  • the second acquisition module 603 is used to acquire the first A similarity degree between a feature vector and each feature vector in the feature vector list
  • a determination module 604 is used to determine the melt state corresponding to the current image according to the similarity degree.
  • the molten material state detection device 60 further includes: a classification module 605;
  • the first obtaining module 601 is used to obtain at least P image frames of different melt states in P single crystal furnaces before obtaining the current image, P ⁇ 1, Q ⁇ 1;
  • the image processing module 602 is configured to perform image processing on each of the P * Q image frames according to a preset algorithm to obtain the feature vector of each image frame;
  • the classification module 605 is used to classify all feature vectors of P * Q image frames to obtain a feature vector list.
  • the image processing module 602 includes: a grayscale processing submodule 6021, a normalization submodule 6022, and a feature extraction submodule 6023;
  • the grayscale processing submodule 6021 is used to grayscale the current image to obtain the first image
  • the normalization submodule 6022 is used to normalize the first image to obtain the second image
  • the feature extraction submodule 6023 is used to perform feature extraction on the second image to obtain the first feature vector of the current image.
  • the image processing module 2102 further includes: a filtering submodule 21024;
  • the filtering sub-module 21024 is used to filter the first image to obtain a filtered image
  • the normalization submodule 21022 is used for normalizing the filtered image to obtain a second image.
  • the grayscale processing submodule 6021 includes: a determination unit 71 and a grayscale processing unit 72;
  • the determining unit 71 is configured to determine an image measurement area from the current image
  • the gradation processing unit 72 is used to perform gradation processing on the pixel value of each pixel in the image measurement area according to the pixel value of each pixel in the image measurement area to obtain the first image, the first
  • the formula includes:
  • (i, j) represents the coordinates of a pixel in the image measurement area
  • R (i, j) represents the pixel value of the red channel pixel (i, j)
  • G (i, j) represents the green channel pixel
  • B (i, j) represents the pixel value of the blue channel pixel point (i, j)
  • f (i, j) represents the pixel point (i, j) after grayscale processing ) Gray value.
  • the filtering sub-module 6024 includes: a setting unit 81 and a neighbor averaging unit 82; a setting unit 81 for setting the size of the statistical window; a neighbor averaging unit 82 for setting the statistics window
  • the size of the pixel average value of each pixel in the first image using the second formula to obtain a filtered image, the second formula includes:
  • K (i, j) represents the gray value of the pixel point (i, j) after averaging in the neighborhood
  • m * n represents the size of the statistical window
  • (s, t) represents the pixel point in the statistical window
  • the coordinates, f (s, t) represents the gray value of the pixel point (s, t) in the statistical window after gray processing.
  • the normalization submodule 6022 includes: an acquisition unit 91 and a normalization unit 92;
  • the obtaining unit 91 is configured to obtain the maximum value of the pixel gray scale in the first image and the minimum value of the pixel gray scale in the first image and the maximum value of the pixel gray scale in the filtered image and the minimum value of the pixel gray scale in the filtered image.
  • the transforming unit 92 is used to normalize the pixel value of each pixel in the filtered image using the third formula to obtain the second image.
  • the third formula includes:
  • K '(i, j) K (i, j)-[K max -K min ] * (f max -f min ) / 255
  • K '(i, j) represents the gray value of the pixel (i, j) after normalization
  • K (i, j) represents the gray of the pixel (i, j) after the neighborhood average Value
  • K max represents the maximum value of pixel gray in the filtered image
  • K min represents the minimum value of pixel gray in the filtered image
  • f max represents the maximum value of pixel gray in the first image
  • f min represents the maximum value of pixel gray in the first image Pixel gray minimum.
  • the device for detecting the melt state provided in the above embodiment is only exemplified by the division of the above functional modules. In practical applications, the above The function allocation is completed by different function modules, that is, the internal structure of the master node is divided into different function modules to complete all or part of the functions described above.
  • the device for detecting the molten material state provided in the above embodiment and the embodiment for the method for detecting the molten material state belong to the same concept. For the specific implementation process, see the method embodiment for details.
  • the detection device obtains the current image; performs image processing on the current image according to a preset algorithm to obtain the first feature vector of the current image; obtains the similarity between the first feature vector and each feature vector in the feature vector list, and the feature vector list contains each The corresponding relationship between the feature vectors and the melt state; determining the melt state corresponding to the current image according to the similarity degree.
  • This method uses two similar feature vectors to determine the current molten state of the image, the calculation is simple, and it can adapt to different single crystal furnaces, determine the melting state of the silicon material in real time, and prompt the operator to perform steps such as feeding or power control. Improve the accuracy of test results.
  • an embodiment of the present disclosure also provides a molten material state detection device.
  • the molten material state detection device includes a receiver 1201, a transmitter 1202, a second memory 1203, and a second processor 1204.
  • the transmitter 1202 and the second memory 1203 are respectively connected to the second processor 1204.
  • the second memory 1203 stores at least one second computer instruction.
  • the second processor 1204 is used to load and execute at least one second computer instruction to implement The method for detecting the state of the melt described in the embodiments corresponding to FIGS. 13, 15 and 16 above.
  • an embodiment of the present disclosure also provides a computer-readable storage medium, for example, a non-transitory computer-readable storage medium may It is read-only memory (English: Read Only Memory, ROM), random access memory (English: Random Access Memory, RAM), CD-ROM, magnetic tape, floppy disk, and optical data storage device. At least one second computer instruction is stored on the storage medium for executing the method for detecting the melt state described in the embodiments corresponding to FIG. 13, FIG. 15 and FIG. 16 above.
  • the program may be stored in a computer-readable storage medium.
  • the mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
  • any reference signs between parentheses should not be constructed as limitations on the claims.
  • the word “comprising” does not exclude the presence of elements or steps not listed in a claim.
  • the word “a” or “one” before an element does not exclude the presence of multiple such elements.
  • the invention can be realized by means of hardware including several different elements and by means of a suitably programmed computer. In the unit claims enumerating several devices, several of these devices may be embodied by the same hardware item.
  • the use of the words first, second, and third does not indicate any order. These words can be interpreted as names.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Crystals, And After-Treatments Of Crystals (AREA)

Abstract

本申请公开了一种熔料状态检测方法、装置及设备，所述方法包括获取熔料当前图像；对所述当前图像进行图像处理；根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。本申请只采用一个时刻的一张当前图像即可确定所述熔料的熔料状态，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰，所以不存在检测精度低的问题，进而，本申请的熔料状态检测准确性不受单晶炉内亮度的影响，相比准确性更高。

Description

一种熔料状态检测方法、装置及设备
本申请要求在2019年8月20日提交中国专利局、申请号为201910770397.5、发明名称为“一种熔料状态检测方法、装置及设备”；在2018年11月23日提交中国专利局、申请号为201811408346.X、发明名称为“硅料熔化状态的检测方法、设备及存储媒介”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
本发明一般涉及太阳能光伏发电技术领域，特别是涉及一种熔料状态检测方法、装置及设备。
背景技术
在制备单晶硅的过程中，硅料熔化控制是晶体生长的重要环节。硅料熔化状态的检测有助于分析熔料过程的进度。在熔料过程还需要进行多次的加料，在多次加料的过程中，需要依据硅料的熔化状态调整加热功率、选择合适的加料时机。因此，硅料熔化进程的检测在单晶硅产业中具有重要的现实意义。
直拉法制备单晶硅是在单晶炉中，加热融化坩埚中的硅料，然后将籽晶侵入到溶体中，转动籽晶及坩埚的同时提拉籽晶，以在籽晶下端依次进行引晶、放肩、转肩、等径及收尾等步骤，制备单晶硅棒。硅料融化环节是晶体生成的重要环节之一，该过程将硅料至于坩埚中，利用坩埚外围的加热器加热融化硅料，并依据硅料的融化状态调整加热功率、选择合适的加料时间。在现有技术中，采用基于高阶奇异值分解的方法进行硅料熔化进程的检测，但是该方法不仅计算复杂，而且在实际的硅料熔化过程中，熔液的波动及熔液对光线的反射使得不同炉体内的亮度不同，导致图像信息复杂，影响检测结果的准确性。
依据基图像与当前硅料熔化所采集的图像进行相关性计算，依据相关性的大小判定硅料熔化状态是否改变，具体为：先构成三维张量，对该张量进行高阶奇异值分解得到用于表征之前熔化状态的基图像；然后，对当前所采集的硅料熔化状态图像与所述基图像进行相关性计算，依据相关性的大小判定硅料熔化状态是否改变。
但上述现有技术中依据基图像与当前硅料熔化所采集的图像进行相关性计算，依据相关性的大小判定硅料熔化状态是否改变的方法，存在检测精度低的问题。
发明内容
本发明提供一种熔料状态检测方法、装置及设备，旨在现有技术中依据基图像与当前硅料熔化所采集的图像进行相关性计算，依据相关性的大小判 定硅料熔化状态是否改变的方法，存在检测精度低的问题。
本发明公开了一种熔料状态检测方法，所述方法包括：
获取熔料当前图像；
对所述当前图像进行图像处理；
根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。
可选地，所述对所述当前图像进行图像处理包括：
将所述当前图像划分为多个子区域；
分别获取各子区域中像素的最小灰度值和像素的灰度值方差；
所述根据所述图像处理结果确定当前图像所对应的熔料的熔料状态包括：根据所述最小灰度值和所述灰度值方差确定所述各子区域的熔料状态，所述各子区域的熔料状态包括液态或固态；
根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
可选地，对所述当前图像进行图像处理包括：
按照预设算法对所述当前图像进行图像处理，得到所述当前图像的第一特征向量；
获取所述第一特征向量与特征向量列表中每个特征向量的相似程度，所述特征向量列表包含每个特征向量与熔料状态的对应关系；
所述根据所述图像处理结果确定当前图像所对应的熔料的熔料状态包括：
根据所述相似程度确定所述当前图像对应的熔料状态。
可选地，所述将所述当前图像划分为多个子区域包括：
当所述当前图像为灰度图像时，将所述灰度图像划分为多个子区域；
所述将所述当前图像划分为多个子区域还包括：
当所述当前图像为彩色图像时，将所述彩色图像灰度化，得到灰度图像；
将所述灰度图像划分为多个子区域；
所述将所述当前图像划分为多个子区域还包括：
当所述当前图像为彩色图像时，将所述当前彩色图像划分为多个子区域；
将所述划分为多个子区域的当前彩色图像灰度化。
可选地，所述将所述当前图像划分为多个子区域包括：
设定所述当前图像的检测区域，将所述检测区域划分为多个子区域；其中，所述检测区域根据所述当前图像中熔料的位置设定，所述检测区域为矩形，所述多个子区域具有相同的大小。
可选地，所述根据所述最小灰度值和灰度值方差确定各子区域的熔料状态包括：
当所述最小灰度值和所述灰度值方差满足预设条件时，确定所述子区域的熔料状态为液态；
当所述最小灰度值和所述灰度值方差不满足预设条件时，确定所述子区域的熔料状态为固态；
其中，所述预设条件为所述子区域中像素的最小灰度值大于第一阈值，所述子区域中像素的灰度值方差小于第二阈值。
可选地，所述根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态包括：
获取所有子区域中熔料状态为液态的子区域个数，记为M；
获取所有子区域中熔料状态为固态的子区域个数，记为N；
根据N与M，确定所述当前图像所对应熔料的熔料状态。
可选地，当所述当前图像所对应的熔料的熔料状态满足第一条件时，向所述熔料中加入所述熔料的固体原料。
可选地，所述方法还包括；
当所述当前图像所对应的熔料的熔料状态满足第二条件时，控制所述熔料进入稳温阶段；
在所述稳温阶段结束后，进行引晶。
可选地，所述获取当前图像之前，所述方法还包括：
获取至少P个单晶炉中Q个不同熔料状态的图像，P≥1，Q≥1；
按照所述预设算法对每个所述图像进行图像处理，得到所述P*Q个图像中的每个图像的特征向量；
对所述P*Q个图像的所有特征向量进行分类，得到所述特征向量列表。
可选地，所述按照预设算法对所述当前图像进行图像处理，得到所述当前图像的第一特征向量包括：
对所述当前图像进行灰度化处理，得到第一图像；
对所述第一图像进行归一化处理，得到第二图像；
对所述第二图像进行特征提取，得到所述当前图像的第一特征向量。
可选地，所述对所述第一图像进行归一化处理得到第二图像之前，所述方法还包括：
对所述第一图像进行滤波处理，得到滤波图像；
所述对所述第一图像进行归一化处理，得到第二图像包括：对所述滤波图像进行归一化处理，得到所述第二图像。
可选地，所述对所述当前图像进行灰度化处理，得到第一图像包括：
从所述当前图像中确定图像测量区域；
根据所述图像测量区域中每个像素点的像素值，利用第一公式对所述图像测量区域中每个像素点的像素值进行灰度化处理，得到第一图像，所述第一公式包括：
f(i,j)＝(R(i,j)+G(i,j)+B(i,j))/3
其中，(i,j)表示所述图像测量区域中一像素点的坐标，R(i,j)表示红色通道像素点(i,j)的像素值，G(i,j)表示绿色通道像素点(i,j)的像素值，B(i,j)表示蓝色通道像素点(i,j)的像素值，f(i,j)表示经过灰度化处理后像素点(i,j)的灰度值。
可选地，所述对所述第一图像进行滤波处理，得到滤波图像包括：
设置统计窗口大小；
根据所述统计窗口的大小，利用第二公式对所述第一图像中每个像素点的像素值进行邻域平均，得到滤波图像，所述第二公式包括：
Figure PCTCN2019119464-appb-000001
其中，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，m*n表示所述统计窗口的大小，(s,t)表示所述统计窗口内像素点的坐标，f(s,t)表示所述统计窗口内经过灰度化处理的像素点(s,t)的灰度值。
可选地，所述对所述滤波图像进行归一化处理，得到第二图像包括：
获取所述第一图像中的像素灰度最大值和所述第一图像中的像素灰度最小值以及所述滤波图像中的像素灰度最大值和所述滤波图像中的像素灰度最小值；
利用第三公式对所述滤波图像中的每个像素点的像素值进行归一化处理，得到第二图像，所述第三公式包括：
K'(i,j)＝K(i,j)-[K max-K min]*(f max-f min)/255
其中，K'(i,j)表示经过归一化处理后像素点(i,j)的灰度值，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，K max表示滤波图像中的像素灰度最大值，K min表示滤波图像中的像素灰度最小值，f max表示第一图像中的像素灰度最大值，f min表示第一图像中的像素灰度最小值。
可选地，所述相似程度为所述第一特征向量与所述特征向量列表中每个特征向量的相似系数；
所述根据所述相似程度确定所述当前图像对应的熔料状态包括：
将第二特征向量对应的熔料状态确定为所述当前图像对应的熔料状态，所述第二特征向量与所述第一特征向量的相似程度最大。
本发明还公开了一种熔料状态检测装置，所述装置包括：
获取模块，用于获取熔料当前图像；
图像处理模块，用于对所述当前图像进行图像处理；
熔料状态确定模块，用于根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。
可选地，所述图像处理模块包括：
图像处理子模块，用于将所述当前图像划分为多个子区域；还用于分别获取各子区域中像素的最小灰度值和像素的灰度值方差；
所述熔料状态确定模块包括：熔料状态确定子模块，用于根据所述最小灰度值和所述灰度值方差确定所述各子区域的熔料状态，所述各子区域的熔料状态包括液态或固态；还用于根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
本发明还公开了一种熔料状态检测设备，所述熔料状态检测设备包括：接口，总线，第一存储器与第一处理器，所述接口、第一存储器与第一处理器通过所述总线相连接，所述第一存储器用于存储可执行程序，所述第一处理器被配置为运行所述可执行程序实现所述的熔料状态检测方法的步骤。
本发明还公开了一种熔料状态检测设备，所述熔料状态检测设备包括：接口，总线，第二存储器与第二处理器，所述接口、第二存储器与第二处理器通过所述总线相连接，所述第二存储器用于存储可执行程序，所述第二处理器被配置为运行所述可执行程序实现所述的熔料状态检测方法的步骤。
本发明还公开了一种计算机可读存储介质，所述计算机可读存储介质上存储第一可执行程序，所述第一可执行程序被第一处理器运行实现所述的熔料状态检测方法的步骤。
本发明还公开了一种计算机可读存储介质，所述计算机可读存储介质上存储第二可执行程序，所述第二可执行程序被第二处理器运行实现所述的熔料状态检测方法的步骤。
本申请公开了一种熔料状态检测方法，所述方法包括获取熔料当前图像；对所述当前图像进行图像处理；根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。本申请只采用一个时刻的一张当前图像即可确定所述熔料的熔料状态，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰，所以不存在检测精度低的问题，进而，本申请的熔料状态检测准确性不受单晶炉内亮度的影响，相比准确性更高。
上述说明仅是本发明技术方案的概述，为了能够更清楚了解本发明的技 术手段，而可依照说明书的内容予以实施，并且为了让本发明的上述和其它目的、特征和优点能够更明显易懂，以下特举本发明的具体实施方式。
附图说明
为了更清楚地说明本发明实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作一简单地介绍，显而易见地，下面描述中的附图是本发明的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1示出了本发明实施例一中的一种熔料状态检测方法的步骤流程图；
图2示出了本发明实施例一中的一种熔料状态检测装置的结构示意图；
图3示出了本发明实施例一中对当前图像进行子区域划分的示意图；
图4示出了本发明实施例二中的一种熔料状态检测方法的步骤流程图；
图5示出了本发明实施例二中的一种熔料状态检测装置的结构示意图；
图6示出了本发明实施例二中通过摄像头获取的熔料当前图像示意图；
图7示出了本发明实施例二中的检测区域中各子区域所对应的方差分布示意图；
图8示出了本发明实施例二中图7所对应的检测区域示意图；
图9示出了本发明实施例二中的另一个检测区域中各子区域所对应的方差分布示意图；
图10示出了本发明实施例二中图9所对应的检测区域示意图；
图11示出了本发明实施例三中的一种熔料状态检测装置的结构示意图；
图12示出了本发明实施例的三中的一种熔料状态检测设备的逻辑结构示意图；
图13是本公开实施例四提供的一种熔料状态的检测方法的流程图；
图14是本公开实施例四提供的一种对像素点灰度进行滤波处理的示意图；
图15是本公开实施例四提供的一种在硅料熔化时采集的图像帧的示意图；
图16是本公开实施例四提供的一种熔料状态的检测方法的示意图；
图17是本公开实施例四提供的一种熔料状态的检测方法的示意图；
图18是本公开实施例四提供的一种熔料状态的检测装置的结构图；
图19是本公开实施例四提供的一种熔料状态的检测装置的结构图；
图20是本公开实施例四提供的一种熔料状态的检测装置的结构图；
图21是本公开实施例四提供的一种熔料状态的检测装置的结构图；
图22是本公开实施例四提供的一种熔料状态的检测装置的结构图；
图23是本公开实施例四提供的一种熔料状态的检测设备的结构图；
图24是本公开实施例四提供的一种熔料状态的检测设备的结构图；
图25是本公开实施例四提供的一种熔料状态的检测设备的结构图。
具体实施例
为使本发明实施例的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例是本发明一部分实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。
需要说明的是，在不冲突的情况下，本申请中的实施例及实施例中的特征可以相互组合。下面将参考附图并结合实施例来详细说明本申请。
实施例一
参照图1，图1示出了本发明实施例一的一种熔料状态检测方法的步骤流程图，所述方法包括：
步骤101，获取所述熔料当前图像。
本发明实施例中，所述方法可以应用于熔料状态检测装置中，参照图2，图2示出了本发明实施例一中的一种熔料状态检测装置的结构示意图，所述熔料状态检测装置设置在单晶炉11中，包括图像采集器，具体可以是摄像头12，所述单晶炉内放置有坩埚14，所述坩埚14内放置有熔料13，所述摄像头12朝向所述坩埚14的方向放置，用于获取所述坩埚14内熔料13的当前图像。
具体的，所述熔料13可以为硅料在坩埚14内熔化形成的，所述熔料13中包括已熔化的液体硅料和未熔化的固体硅料。将所述摄像头12的放置方向以能够获取到所述熔料的当前图像为准，所述摄像头12用于获取所述熔料13当前图像，具体的，可以设置所述摄像头12按照预设周期获取所述熔料当前图像，可选地，所述预设周期可以为1s。
步骤102，将所述当前图像划分为多个子区域。
本实施例中，为了准确检测当前图像所对应的熔料状态，将所述当前图像划分为多个子区域，并设定一个子区域对应一个熔料状态。所述熔料状态包括液态和固态。可以理解，将所述子区域划分的越小，计算获得的所述子区域中熔料状态的准确度也就越高，但是，也相应的增加了计算复杂度。相反，将所述子区域划分的越大，计算获得的所述子区域中熔料状态的准确度会相应降低，但是，计算复杂度会相应减小。本实施例中，综合考虑了计算复杂度和准确度的因素，对所述当前图像的子区域进行划 分。参照图3，示出了本实施例一的子区域划分示意图，将像素为1000pix×300pix的当前图像划分为10×10个子区域。每个子区域的像素为100pix×30pix。
步骤103，分别获取各所述子区域中像素的最小灰度值和像素的灰度值方差。
本实施例中，在获取所述子区域中像素的最小灰度值和各子区域中像素的灰度值方差之前，需要保证所述子区域所对应的图像为灰度图像，如果不是灰度图像，需要采用相应算法对所述子区域所对应的图像进行灰度化。可选地，获取所述子区域的最小灰度值可以包括：从所述子区域的第一个像素点开始逐行遍历所述子区域各个像素的灰度值，当获取到的当前像素灰度值小于上一个像素的灰度值时，保存当前像素灰度值，当遍历完整个子区域后，获取到的像素的灰度值就是所述子区域中像素的最小灰度值。可以理解，本实施例对如何得到所述子区域中像素的最小灰度值的方法不做限定，只要是能获取到所述子区域中像素的最小灰度值就行。获取所述子区域中像素的灰度值方差可以包括：首先计算整个子区域中所有像素灰度值的平均值M，获取所述子区域中每个像素的灰度值，记为x i，根据公式
Figure PCTCN2019119464-appb-000002
计算出所述子区域像素的灰度值方差S 2。
步骤104，根据所述最小灰度值和灰度值方差确定各子区域的熔料状态，所述各子区域的熔料状态包括液态和固态。
本发明实施例中，前述通过计算得到了各子区域的最小灰度值和灰度值方差，根据所述最小灰度值和灰度值方差确定各子区域的熔料状态可以包括：首先设定预设条件，当所述最小灰度值和灰度值方差都满足预设条件时，确定所述子区域的熔料状态为液态，否则，确定所述子区域的熔料状态为固态。当然，也可以实际情况，设定其它预设条件，当所述最小灰度值和灰度值方差都满足预设条件时，确定所述子区域的熔料状态为固态，否则，确定所述子区域的熔料状态为液态。对此，本发明不作限制。
具体的，本实施例设定根据子区域最小灰度值和灰度值方差确定子区域的熔料状态是因为所述熔料在固态和液态时，灰度值存在较大差别，当所述子区域为固态时，所述子区域中像素所对应的灰度值较小，表现在图像上就是图像区域较暗。当所述子区域为液态时，所述子区域中像素所对应的灰度值较大，表现在图像上就是图像区域较亮。本实施例针对此特 点，通过设定相应的预设条件，来获取所述子区域所对应的熔料状态。
现有技术中需要在采集当前图像前确定基图像，再根据基图像与当前图像的相关性来计算熔料的状态变化，由于获取基图像的时刻和获取当前图像的时刻不同，两个时刻的光线条件会有所不同，会导致两个时刻的单晶炉内的亮度不同，故现有技术中获得的检测结果会受光线的影响，但本申请只利用一个时刻的当前图像来检测当前图像所对应的熔料状态，所述当前图像为一个时刻获取的一张图像，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰。即本申请只采用当前图像进行熔料状态的检测能够提高检测的准确性。
步骤105，根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
本发明实施例中，在获取到各子区域的熔料状态之后，计算所述图像中为液态的子区域个数和为固态的子区域个数，根据所述图像中为固态的子区域个数和为液态的子区域个数确定所述当前图像所对应的熔料的熔料状态。例如，所述子区域的个数为100，所述当前图像中为固态的子区域个数为30个，为液态的子区域个数为70个，则所述当前图像所对应的熔料的熔料状态为固液混合，且所述熔料中液态熔料所占的比例为70％。例如：所述子区域的个数为100，所述当前图像中为固态的子区域个数为0个，为液态的子区域个数为100个，则所述当前图像所对应的熔料的熔料状态为液态。
在本发明实施例中，该方法应用于熔料状态检测装置中，所述熔料状态检测装置包括设置在单晶炉中，包括图像采集器，具体可以是摄像头。所述单晶炉内放置有熔料；所述方法包括：通过所述摄像头获取所述熔料当前图像；将所述当前图像划分为多个子区域；分别获取各所述子区域中像素的最小灰度值和像素的灰度值方差；根据所述最小灰度值和灰度值方差确定各子区域的熔料状态，所述各子区域的熔料状态包括液态和固态；根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态。所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。本申请中，采集所述单晶炉中熔料的当前图像，并将所述当前图像划分为多个子区域，根据所述多个子区域的最小灰度值和灰度值方差确定多对应的子区域的熔料状态，再根据多个子区域的熔料状态确定当前图像所对应的熔料的熔料状态，由于本申请分别确定不同子区域的熔料状态，在根据不同子区域的熔料状态确定当前图像所对应的熔料的熔料状 态，提高了检测精度，又本申请在计算不同子区域的熔料状态时，只用考虑多个子区域的最小灰度值和灰度值方差，减少了计算复杂度，提高了检测效率。同时，现有技术中，由于至少需要获取两个不同时刻熔料的图像，然而，由于两个不同时刻所对应的单晶炉内的熔液存在波动及熔液对光线的反射使得两个时刻的炉体内的亮度不同，使得图像信息复杂，进而导致相关性计算的阈值难以设定，影响检测结果的准确性。而本申请只采用一个时刻的一张当前图像即可确定所述熔料的熔料状态，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰，所以不存在检测精度低的问题，进而，本申请的熔料状态检测准确性不受单晶炉内亮度的影响，相比准确性更高。
实施例二
参考图4，图4示出了本发明实施例二中的一种熔料状态检测方法的步骤流程图，所述方法包括：
步骤201，获取所述熔料当前图像。
本发明实施例中所述方法应用于熔料状态检测装置中，参照图5，图5示出了本发明实施例二中的一种熔料状态检测装置的结构示意图，所述熔料状态检测装置设置在单晶炉11内包括图像采集器，具体的可以是摄像头12，所述单晶炉内放置有坩埚14，所述坩埚14内放置有熔料13，所述摄像头12朝向所述坩埚14的方向放置，所述单晶炉还设置有观察窗口15，所述摄像头12靠近所述观察窗口15设置。
具体的，所述观察窗口15设置在所述单晶炉的炉壁，可以用于通过所述摄像头12的一部分，使摄像头12能够获取放置在坩埚14内熔料13的当前图像。参照图6，图6示出了通过摄像头12获取的熔料当前图像，可以看出，获取到的熔料当前图像的边缘区域的灰度值为0，可选地，本实施例在对熔料状态进行检测前，先在获取的当前图像中设置检测区域，所述检测区域为所述当前图像中坩埚内熔料所对应的区域。由于坩埚本体或者坩埚外的区域并不存在熔料，故本实施例中可以只选取所述当前图像中坩埚内熔料所对应的区域进行熔料状态的检测，无需对整张当前图像进行检测，提升了检测效率。同时，只选取当前图像中坩埚内熔料所对应的区域，进行熔料状态的检测，能够避免采用不存在熔料的图像检测，干扰后续的熔料状态准确性的问题。
例如，参考图6，由于图6中，当前图像中坩埚内熔料所对应的区域熔料主要位于当前图像中白色边框框出的矩形区域，因此，检测区域可以设置为当前图像中白色边框框出的矩形区域。可以看出，所述检测区域中不 同像素的灰度值具有明显的变化，具备熔料在熔融状态时的特征。可以理解，本实施例也可以将所述检测区域设置为其它形状，只要能够保证所述检测区域具备熔料在熔融状态时的特征就行，本实施例对检测区域的形状不做限制。
作为一种可选地示例，对所述检测区域进行划分，将所述检测区域划分为多个子区域，其中，所述检测区域根据所述当前图像中熔料的位置设定，所述检测区域为矩形，所述多个子区域具有相同的大小。
步骤2021，当所述当前图像为灰度图像时，将所述灰度图像划分为多个子区域。
具体的，当所述摄像头为黑白摄像头时，所述摄像头获取的熔料当前图像为灰度图像。可选地，本实施例先在获取到的灰度图像上设置检测区域，再将所述检测区域划分为多个子区域，所述子区域具有相同的大小，可以理解，根据不同的情况，也可以将所述子区域设置为不同的大小。本实施为了减小计算复杂度，将所述子区域设置相同的大小。
步骤2022，当所述当前图像为彩色图像时，将所述彩色图像灰度化，得到灰度图像；
将所述灰度图像划分为多个子区域。
步骤2023，当所述当前图像为彩色图像时，将所述当前彩色图像划分为多个子区域；将所述划分为多个子区域的当前彩色图像灰度化。
具体的，当获取到的熔料当前图像为彩色图像，本实施例可以选择先对所述彩色图像进行灰度化后，得到灰度图像，再将所述灰度图像划分为多个子区域。也可以选择将当前彩色图像划分为多个子区域，再对划分为多个子区域的当前彩色图像灰度化处理。对此，本发明不做限制。
步骤203，分别获取所述各子区域中像素的最小灰度值和像素的灰度值方差。
本步骤与上个实施例的步骤103相同，本实施例对此不再赘述。
步骤204，当所述最小灰度值和灰度值方差满足预设条件时，确定所述子区域的熔料状态为液态；当所述最小灰度值和灰度值方差不满足预设条件时，确定所述子区域的熔料状态为固态。
具体的，其中，所述预设条件为所述子区域中像素的最小灰度值大于第一阈值，所述子区域中像素的灰度值方差小于第二阈值。作为一种具体的示例，所述第一阈值为20，所述第二阈值为10。
可选地，所述熔料为在坩埚中进行熔化的硅料。所述硅料在熔化过程中，坩埚中的硅料存在固液两相状态，液态的硅料和固态的硅料在灰度图像中存在灰度值上的大小差异，且液态的硅料在图像中的灰度值较固态的 硅料在图像中的灰度值大，表现在灰度图上就是颜色较深的区域对应于固态的硅料，颜色较浅的区域对应于液态的硅料。设置一个合适的阈值，将固态硅料和液态硅料进行界定，能够准确的获取所述子区域的熔料状态。所述子区域的灰度值方差说明了子区域像素间灰度分布的均匀程度，当所述子区域的方差值较大时，说明所述子区域像素间灰度分布不均匀，进而说明所述子区域所对应的熔料状态为两相状态。当所述子区域的方差值较小时，说明所述子区域像素间灰度分布均匀，进而说明所述子区域所对应的熔料状态为一相状态。设置一个合适的阈值，来区分所述熔料状态为一相还是两相状态，对准确获取所述子区域的熔料状态具有促进作用。本实施例通过设置第一阈值和第二阈值来获得所述子区域所对应的硅料在坩埚中的状态。作为一种示例，当所述子区域的像素最小灰度值大于20，说明此时所述子区域的像素灰度值较大，所述子区域的熔料为液态，进一步，所述子区域的像素灰度值方差小于10，说明此时所述子区域的像素均匀，像素间的差别很小，通过上述两个条件，可以判定所述子区域的熔料状态为液态。作为另一种示例，当所述子区域的像素最小灰度值小于20，说明此时所述子区域的像素灰度值较小，所述子区域的大部分熔料为固态，此时，直接判断所述子区域的熔料状态为固态。
步骤205，获取所有子区域中熔料状态为液态的子区域个数，记为M。
具体的，步骤204获取了各子区域的熔料状态，可选地，本实施例设定所述熔料状态为液态的子区域个数为M，并初始设定M＝0，从第一个子区域开始，逐行遍历每个子区域，当所述子区域的熔料状态为液态时，设定M＝M+1，遍历完整个检测区域，得到所有子区域中熔料状态为液态的子区域个数M。可以理解，本实施也可以采用其它方法获取所述所有子区域中熔料状态为液态的子区域个数，对此，本实施例不作限制。
步骤206，获取所有子区域中熔料状态为固态的子区域个数，记为N。
具体的，步骤204获取了各子区域的熔料状态，可选地，本实施例设定所述熔料状态为固态的子区域个数为N，并初始设定N＝0，从第一个子区域开始，逐行遍历每个子区域，当所述子区域的熔料状态为固态时，设定N＝N+1，遍历完整个检测区域，得到所有子区域中熔料固态为液态的子区域个数N。可以理解，本实施也可以采用其它方法获取所述所有子区域中熔料状态为固态的子区域个数，对此，本实施例不作限制。
步骤207，根据N与M，确定所述当前图像所对应的熔料状态，所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
具体的，可以根据所述N与M与所述子区域的总个数之间的比例确定所述当前图像所对应的熔料状态。
作为一种具体的示例，将所述检测区域划分为100个相同大小的子区域，参照图7和图8，图7示出了检测区域中各子区域所对应的方差分布，其中，横坐标表示各子区域，纵坐标为各子区域所对应的灰度值方差；图8示出了所述检测区域。可以看出，图7中的各子区域的灰度值方差分布在1-70之间，从图8中的检测区域也能够看出此时的熔料状态为固液混合。通过上述步骤对图8中检测区域进行计算，得到此时检测区域所对应的熔料中液态硅料所占的比例为15％。
作为另一种具体的示例，同样将所述检测区域划分为100个相同大小的子区域，参照图9和图10，图9示出了检测区域中各子区域所对应的方差分布，其中，横坐标表示各子区域，纵中坐标为各子区域所对应的灰度值方差；图10示出了所述检测区域。可以看出，图9中的各子区域的灰度值方差分布在1-2.5之间，从图10中的检测区域也能够看出此时的熔料状态为液态。通过上述步骤对图10中检测区域进行计算，得到此时检测区域所对应的熔料中液态硅料所占的比例为100％。
步骤208，当所述当前图像所对应的熔料的熔料状态满足第一条件时，加入所述熔料的固体原料。
具体的，当所述单晶炉的坩埚内的硅料完全熔化时进行加料容易引起硅溅，飞溅的溶液可能溅到单晶炉中的加热器、保温桶、坩埚帮，使得加热器、保温桶、坩埚帮产生裂痕，因此需要依据硅料的熔化状态调整加热功率、选择合适的加料时机。可选地，当单晶炉的坩埚中的熔料达到合适的熔料状态时，向所述单晶炉的坩埚中加入所述熔料的固体原料；具体的，所述合适的熔料达到熔料状态可以为所述熔料中液态硅料所占的比例为50-70％，当所述熔料中液态硅料所占的比例大于70％时，向所述坩埚中加入所述熔料的固体原料时，会由于坩埚中液态硅料过多，引起硅溅；当所述熔料中液态硅料所占的比例小于50％，说明此时大部分固态硅料还没开始熔化，如果再次加入固态硅料，会增加坩埚中的负担，硅料熔化的速度也会减慢。上述第一条件为所述熔料中液态硅料所占的比例为50-70％，在所述熔料中液态硅料所占的比例为50-70％时，向所述坩埚中加入所述熔料的固体原料时，不会引起硅溅，也不会影响硅料的熔化速度。
步骤209，当所述当前图像所对应的熔料的熔料状态满足第二条件时，控制所述熔料进入稳温阶段；在所述稳温阶段结束后，进行引晶。
具体的，熔料完成后需要进行稳温保证硅液稳定到合适的引晶温度，使籽晶与熔融硅液熔接以进行引晶工艺的晶体生长。所述熔料完成为所述坩埚中的熔料全部为液态，也就是说此时熔料状态为液态。上述第二条件可以为所述熔料状态为液态。
在本发明实施例中所述单晶炉内放置有熔料；所述方法包括：获取熔料当前图像；将所述当前图像划分为多个子区域；分别获取各所述子区域中像素的最小灰度值和像素的灰度值方差；根据所述最小灰度值和灰度值方差确定各子区域的熔料状态，所述各子区域的熔料状态包括液态和固态；根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态，所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。本申请中，采集所述单晶炉中熔料的当前图像，并将所述当前图像划分为多个子区域，根据所述多个子区域的最小灰度值和灰度值方差确定多对应的子区域的熔料状态，再根据多个子区域的熔料状态确定当前图像所对应的熔料的熔料状态，由于本申请分别确定不同子区域的熔料状态，在根据不同子区域的熔料状态确定当前图像所对应的熔料的固液分布，提高了检测精度，又本申请在计算不同子区域的熔料状态时，只用考虑多个子区域的最小灰度值和灰度值方差，减少了计算复杂度，提高了检测效率。同时，现有技术中，由于至少需要获取两个不同时刻熔料的图像，然而，由于两个不同时刻所对应的单晶炉内的熔液存在波动及熔液对光线的反射使得两个时刻的炉体内的亮度不同，使得图像信息复杂，进而导致相关性计算的阈值难以设定，影响检测结果的准确性。而本申请只采用一个时刻的一张当前图像即可确定所述熔料的熔料状态，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰，所以不存在检测精度低的问题，进而，本申请的熔料状态检测准确性不受单晶炉内亮度的影响，相比准确性更高。
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都是本申请实施例所必须的。
实施例三
参照图11，图11示出了本发明实施例三中的一种熔料状态检测装置，所述装置包括：图像检测单元，具体的，所述图像检测单元可以是摄像头12，所述装置设置在所述单晶炉内包括坩埚14，所述熔料13放置在所述坩埚14内，所述摄像头12朝向所述坩埚14的方向放置，用于采集所述熔料的当前图像；所述装置还包括检测单元16，所述检测单元与所述摄像头12相连接，用于获取所述摄像头采集的所述熔料的当前图像。
本发明实施例中，所述摄像头12与所述检测单元16的连接方式包括有线连接和无线连接，对此，本发明实施例不作限制。
所述摄像头12用于采集所述熔料13的当前图像；
所述检测单元16用于将所述当前图像划分为多个子区域，进而分别获取各所述子区域中像素的最小灰度值和像素的灰度值方差；
所述检测单元16还用于根据所述最小灰度值和灰度值方差确定各子区域的熔料状态，进而根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述各子区域的熔料状态包括液态和固态，所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
本发明实施例中，所述检测单元16用于当所述当前图像为灰度图像时，将所述当前图像划分为多个子区域；
所述检测单元16还用于，当所述当前图像为彩色图像时，将所述当前图像灰度化后，划分为多个子区域；
所述检测单元16还用于，当所述当前图像为彩色图像时，将所述当前图像划分为多个子区域后，再对所述划分为多个子区域当前图像进行灰度化。本发明实施例中，所述单晶炉炉壁上设置有观察窗口15，所述摄像头12靠近所述观察窗口15设置用于获取所述坩埚14内熔料13的当前图像；
所述检测单元16还用于设定所述熔料的当前图像的检测区域，并将所述检测区域划分为多个子区域；其中，所述检测区域根据所述当前图像中熔料的位置设定，所述检测区域为矩形，所述多个子区域具有相同的大小。
本发明实施例中，所述检测单元16还用于当所述最小灰度值和灰度值方差满足预设条件时，确定所述子区域的熔料状态为液态；当所述最小灰度值和灰度值方差不满足预设条件时，确定所述子区域的熔料状态为固态；其中，所述预设条件为所述子区域中像素的最小灰度值大于第一阈值，所述子区域中像素的灰度值方差小于第二阈值。
本发明实施例中，所述检测单元16还用于获取所有子区域中熔料状态为液态的子区域个数，记为M；获取所有子区域中熔料状态为固态的子区域个数，记为N；进而根据N与M，确定所述当前图像所对应的熔料状态。
本发明实施例中，所述装置还包括控制单元17和加料器18，所述控制单元17用于当所述检测区域所对应的熔料的熔料状态满足第一条件时，发出第一提示信号；
所述第一控制信号用于提示所述熔料状态满足第一时机，所述第一时机为向所述单晶炉内加入熔料的固体原料的时机。
本发明实施例中，通过加料器18向所述单晶炉内加入熔料的固体原料，用于防止人工加入固体原料时，对人体造成伤害。
本发明实施例中，所述装置还包括引晶绳19，所述控制单元17还用于当所述检测区域所对应的熔料的熔料状态满足第二条件时，控制所述熔料进入稳温阶段；并当所述稳温阶段结束后，发出第二提示信号；
所述第二提示信号用于提示所述熔料状态满足第二时机，所述第二时机为对所述单晶炉中的熔料进行引晶的时机。
本发明实施例中，利用引晶绳19进行引晶。
在本发明实施例中，该熔料状态检测装置中各个部分的功能，具体可以参照前述实施例中的相关记载，且能达到相同的有益效果，为了避免重复，此处不再赘述。
在本发明实施例中，所述熔料状态检测装置包括：单晶炉、摄像头，所述单晶炉内包括坩埚，所述熔料放置在所述坩埚内，所述摄像头朝向所述坩埚的方向放置，用于采集所述熔料的当前图像；所述装置还包括检测单元，所述检测单元与所述摄像头相连接，用于获取所述摄像头采集的所述熔料的当前图像。所述检测单元用于将所述当前图像划分为多个子区域，进而分别获取各所述子区域中像素的最小灰度值和像素的灰度值方差；所述检测单元还用于根据所述最小灰度值和灰度值方差确定各子区域的熔料状态，进而根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述各子区域的熔料状态包括液态和固态。本申请中，采集所述单晶炉中熔料的当前图像，并将所述当前图像划分为多个子区域，根据所述多个子区域的最小灰度值和灰度值方差确定多对应的子区域的熔料状态，再根据多个子区域的熔料状态确定当前图像所对应的熔料的熔料状态，由于本申请分别确定不同子区域的熔料状态，在根据不同子区域的熔料状态确定当前图像所对应的熔料的熔料状态，提高了检测精度，又本申请在计算不同子区域的熔料状态时，只用考虑多个子区域的最小灰度值和灰度值方差，减少了计算复杂度，提高了检测效率。同时，现有技术中，由于至少需要获取两个不同时刻熔料的图像，然而，由于两个不同时刻所对应的单晶炉内的熔液存在波动及熔液对光线的反射使得两个时刻的炉体内的亮度不同，使得图像信息复杂，进而导致相关性计算的阈值难以设定，影响检测结果的准确性。而本申请只采用一个时刻的一张当前图像即可确定所述熔料的熔料状态，不用与此时刻之前的基图像进行相关性计算，不存在不同时刻单晶炉内不同波动及不同反射使得两个时刻炉体内的亮度不同的干扰，进而，本申请的熔料状态检测准确性不受单晶炉内亮度的影响，准确性更高。
图12示出了本发明实施例的一种熔料状态检测设备的逻辑结构示意图。如图12所示，本发明实施例提供的熔料状态检测设备可以包括：接口41、第一处理器42、第二存储器43及总线44；其中，所述总线44，用于实现所述接口41、所述第一处理器42和所述第一存储器43之间的连接通信；所述第一存储器43存储有可执行程序，所述第一处理器42，用于执行所述第一存储器43中存储的可执行程序，以实现如图1或图4，实施例一或实施例二中的熔料状态检测方法的步骤，并能达到相同或相似的效果，为了避免重复，此处不再赘述。
本发明还提供一种计算机可读存储介质，所述计算机可读存储介质存储有一个或者多个第一可执行程序，所述一个或者多个第一可执行程序可被一个或者多个第一处理器执行，以实现如图1或图4，实施例一或实施例二中的熔料状态检测方法的步骤，并能达到相同或相似的效果，为了避免重复，此处不再赘述。
实施例四
本公开实施例提供一种熔料状态的检测方法，如图13所示，该熔料状态的检测方法包括以下步骤：
1301、获取当前图像。
在本公开实施例中，获取当前图像包括：接收CCD相机发送的采集的当前图像。在硅料熔化过程中，CCD相机可以采集到硅料处于不同熔化状态的图像帧，本公开实施例以当前图像为例，对当前图像对应的熔料状态的检测方法进行说明。
1302、按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量。
在本公开实施例中，按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量包括：
S1、对当前图像进行灰度化处理，得到第一图像；
S2、对第一图像进行归一化处理，得到第二图像；
S3、对第二图像进行特征提取，得到当前图像的第一特征向量。
为了更好对图像进行处理，在步骤S2之前还包括步骤S4：对第一图像进行滤波处理，得到滤波图像。那么，步骤S2对第一图像进行归一化处理，得到第二图像包括：对滤波图像进行归一化处理，得到第二图像。
下面对步骤S1-步骤S4进行具体描述。对于步骤S1，对当前图像进行灰度化处理，得到第一图像包括：从当前图像中确定图像测量区域；根据图像测量区域中每个像素点的像素值，利用第一公式对图像测量区域中每个像素点的像素值进行灰度化处理，得到第一图像，第一公式包括：
f(i,j)＝(R(i,j)+G(i,j)+B(i,j))/3
其中，(i,j)表示所述图像测量区域中一像素点的坐标，R(i,j)表示红色通道像素点(i,j)的像素值，G(i,j)表示绿色通道像素点(i,j)的像素值，B(i,j)表示蓝色通道像素点(i,j)的像素值，f(i,j)表示经过灰度化处理后像素点(i,j)的灰度值。
由于CCD相机设置在单晶炉的侧上方，因此，CCD相机在采集得到的图像帧中可能有被单晶炉的热屏或其他一些构件遮挡的区域，因此，需要从当前图像中确定图像测量区域，鉴于硅料在熔化过程中，靠近单晶炉热屏边缘的硅料最先熔解，因此，可以将图像测量区域设置在热屏边缘附近，这样，能够准确判断硅料的熔化状态。通过对图像测量区域中每个像素点的像素值进行平均，求得测量区域内每个像素点的灰度值，对图像测量区域进行灰度化处理的目的在于将彩色图像转化成灰度图像，便于后续的图像处理。
对于步骤S4，对第一图像进行滤波处理，得到滤波图像包括：设置统计窗口大小；根据统计窗口的大小，利用第二公式对第一图像中每个像素点的像素值进行邻域平均，得到滤波图像，第二公式包括：
所述第二公式包括：
Figure PCTCN2019119464-appb-000003
其中，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，m*n表示所述统计窗口的大小，(s,t)表示所述统计窗口内像素点的坐标，f(s,t)表示所述统计窗口内经过灰度化处理的像素点(s,t)的灰度值。
对于统计窗口的大小，可以根据选取的图像测量区域的大小确定，也可以根据经验选取，一般统计窗口的大小为3×3、5×5、7×7等。在统计窗口的大小确定之后，将第一图像中每个像素点的灰度值设置为该像素点统计窗口内的所有像素点灰度值的平均值，从而对灰度突变的像素点进行平滑滤波。如图14所示，图像测量区域的大小为9×9，图像测量区域中的每个像素点用“×”表示，统计窗口的大小为3×3，虚线框所示部分为统计窗口的大小，以对黑色加粗像素点×进行邻域平均为例，计算统计窗口内所有像素点的灰度值的平均值，将该平均值确定为黑色加粗像素点×的像素灰度值。
对于步骤S2，对滤波图像进行归一化处理，得到第二图像包括：
获取第一图像中的像素灰度最大值和第一图像中的像素灰度最小值以及滤波图像中的像素灰度最大值和滤波图像中的像素灰度最小值；
利用第三公式对滤波图像中的每个像素点的像素值进行归一化处理， 得到第二图像，第三公式包括：
K'(i,j)＝K(i,j)-[K max-K min]*(f max-f min)/255
其中，K'(i,j)表示经过归一化处理后像素点(i,j)的灰度值，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，K max表示滤波图像中的像素灰度最大值，K min表示滤波图像中的像素灰度最小值，f max表示第一图像中的像素灰度最大值，f min表示第一图像中的像素灰度最小值。
对于获取第一图像中的像素灰度值最大值f max和像素灰度最小值f min，可以获取第一图像的灰度直方图，根据灰度直方图得到第一图像中的像素灰度值最大值f max和像素灰度最小值f min；也可以是对第一图像中所有像素点的灰度值按照从大到小或者从小到大的顺序进行排序，从而得到第一图像中的像素灰度值最大值f max和像素灰度最小值f min。对于获取滤波图像中的像素灰度最大值K max和像素灰度最小值K min，可以参考对第一图像中的像素灰度值最大值f max和像素灰度最小值f min的获取方法，此处不再赘述。对于步骤S3，对第二图像进行特征提取，得到当前图像的第一特征向量包括：对第二图像中的灰度值进行特征提取，得到当前图像的第一特征向量。
1303、获取第一特征向量与特征向量列表中每个特征向量的相似程度。
在本公开实施例中，特征向量列表是预先设置的。具体的，在步骤101之前，该方法还包括：获取至少P个单晶炉中Q个不同熔料状态的图像帧，P≥1，Q≥1；按照预设算法对P*Q个图像帧中的每个图像帧进行图像处理，得到每个图像帧的特征向量；对P*Q个图像帧的特征向量进行分类，得到特征向量列表。此处所描述的预设算法与步骤102所描述的预设算法相同，参考步骤102中对当前图像的处理方式，对P*Q个图像帧中的每个图像帧进行处理，得到每个图像帧对应的特征向量；然后，对P*Q个图像帧的所有特征向量按照预设规则进行分类，不同类别对应不同的熔料状态，这样，得到包含每个特征向量与熔料状态的对应关系的特征向量列表。
1304、根据相似程度确定当前图像对应的熔料状态。
在本公开的一个实施例中，当相似程度为第一特征向量与特征向量列表中每个特征向量的相似系数时，步骤1304具体包括：将第二特征向量对应的熔料状态确定为当前图像对应的熔料状态。其中，第二特征向量与第一特征向量的相似程度最大，这就意味着第一特征向量最接近于第二特征向量，而特征向量列表中包含每个特征向量与熔料状态的对应关系，因此，将第二特征向量对应的熔料状态确定为当前图像对应的熔料状态。在实际应用中，相似系数包括相似性参数和相异性参数，两种参数都可以衡 量相似性，其区别在于，相似性参数的数值大小直接反映两个特征向量之间的相似程度，其数值越大表示越相似，而相异性参数的数值大小则反映两个特征向量之间的差异程度，其数值越小表示越相似。对于第一特征向量与特征向量列表中每个特征向量的相似系数可以是两个特征向量之间的距离系数，也可以是两个特征向量之间的夹角余弦，具体根据实际情况进行选择，本公开实施例对此不加任何限定。
本公开实施例提供的熔料状态的检测方法，包括获取当前图像；按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量；获取第一特征向量与特征向量列表中每个特征向量的相似程度，特征向量列表包含每个特征向量与熔料状态的对应关系；根据所述相似程度确定所述当前图像对应的熔料状态。该方法采用两个特征向量相似的方法确定当前图像的熔料状态，计算简单，而且能够适应不同的单晶炉，实时判断硅料的熔化状态，及时提醒操作人员进行加料或功率控制等步骤，提高检测结果的准确性。基于上述图1对应的实施例提供的熔料状态的检测方法，本公开另一实施例提供一种熔料状态的检测方法，本实施例提供的熔料状态的检测方法包括以下步骤离线学习程序存储、在线图像采集和熔料完成度判断，此处所描述的熔料即硅料，熔料完成度判断即判断当前单晶炉中的熔料状态。
第一步，离线学习程序存储。如图15所示，具体步骤主要包括以下内容：
首先，收集不同熔料状态的图像。
本实施例，采用CCD相机采集不同炉体处于不同熔化状态的图像，具体采集的图像如图16所示，炉体内有部分硅料41未完全熔化。由于炉体内热屏等其他构件的遮挡，一部分硅料熔化图像未能显示，图16中侧面弧形区域所示为拍摄的热屏边缘图像42，上部弧形区域和下部弧形区域为炉体内其它结构图像。本实施例，CCD相机采集完图像后通过电路输入至工控机中，由工控机的图像处理程序对单晶生长图像进行处理。其次，设置图像测量区域。如图16所示，本实施例，图像测量区域43设置在靠近热屏边缘图像42附近，利用图像处理程序包含的特征识别模块自动识别测量区域。在熔料过程中，靠近热屏边缘的硅料最先熔解，因此本实施例将测量区域选择热屏边缘图像附近，能够及时判断熔料状态。再次，进行图像处理。图像处理包括对收集的不同炉台不同熔料状态的图像中位于测量区域内的图像进行灰度化处理、平滑过滤处理和归一化处理。本实施例，对收集到的不同炉台不同熔料状态的所有图像采用以下方法进行处理：
a、灰度化处理：将收集到的不同炉台不同熔料状态的图像采用三分量 亮度求平均方法对测量区域内的图像进行灰度化处理，计算原理如下所示：
f(i,j)＝(R(i,j)+G(i,j)+B(i,j))/3
其中，(i,j)表示所述图像测量区域中一像素点的坐标，R(i,j)表示红色通道像素点(i,j)的像素值，G(i,j)表示绿色通道像素点(i,j)的像素值，B(i,j)表示蓝色通道像素点(i,j)的像素值，f(i,j)表示经过灰度化处理后像素点(i,j)的灰度值。测量区域内其他像素点采用相同的处理方式。
b、平滑过滤处理：将收集到的不同炉台不同熔料状态的所有图像中位于测量区域内的经过灰度化处理的图像采用邻域平滑滤波方法进行平滑过滤处理，本实施例采用以下原理进行处理：
设定图像处理程序中滤波器窗口的大小为m×n，计算窗口区域的像素均值，然后将均值赋值给窗口中心点处的像素，计算原理如下所示：
Figure PCTCN2019119464-appb-000004
其中，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，m*n表示所述统计窗口的大小，(s,t)表示所述统计窗口内像素点的坐标，f(s,t)表示所述统计窗口内经过灰度化处理的像素点(s,t)的灰度值。
c、归一化处理：获取收集的所有图像的测量区域内进行灰度化和平滑过滤处理后的图像的灰度直方图，并进行归一化处理。
在本实施例中，对每一个经过灰度化和平滑过滤处理后的图像进行归一化处理。以对其中的一个图像为例进行说明，归一化计算原理如下所示：
K'(i,j)＝K(i,j)-[K max-K min]*(f max-f min)/255
其中，K'(i,j)表示经过归一化处理后像素点(i,j)的灰度值，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，K max表示滤波图像中的像素灰度最大值，K min表示滤波图像中的像素灰度最小值，f max表示第一图像中的像素灰度最大值，f min表示第一图像中的像素灰度最小值。
再次、提取特征向量并对特征向量分类，并存储在离线学习程序中。
在归一化处理后，对收集的所有图像位于测量区域内图像的灰度值进行特征向量提取。具体的，将收集的不同炉台不同熔料状态的所有图像进行以上处理后，分别提取特性向量，并将提取的特征向量分为多个类别，并存储在离线学习程序中，不同类别对应不同熔料状态。在本实施例中，离线学习程序为离线分类器，当然也可以为其他存储程序。
在本实施例中，离线分类器中包括有2个参数：一个为样本数组X，其行数等于收集的样本数，其列数等于特征向量的长度，每一行即一个特征向量。X的实际结构，相当于把每一个特征向量一行一行地排下去，形成一个数组。另一个参数是类别向量y，其元素只能取特定的类型。在本实施 例中，将所有的样本分为5类，这5类可以用1、2，3、4、5表示，表示液体所占比20％，40％，60％，80％和100％，对应不同的熔料进度。当然，也可以将所有的样本分为大于5类或小于5类，对应的特征向量分为不同的类别。
第二步，在线图像采集处理。如图17所示，具体步骤主要包括以下内容：
(1)采用CCD相机在线实时采集炉体内熔料状态的图像。
(2)设置测量区域，在线实时图像的测量区域与离线学习程序存储步骤5中设置的图像测量区域相同。
(3)对测量区域内的实时图像进行灰度化和平滑过滤处理。采用三分量亮度求平均方法对测量区域内的图像进行灰度化处理；采用邻域平滑滤波方法对测量区域内的图像进行平滑过滤处理，处理方法与离线学习程序存储步骤中相同。
(4)对测量区域内的实时图像处理还包括获取测量区域内实时图像的灰度直方图，并进行归一化处理，处理方法与离线学习程序存储步骤中相同。
(5)将归一化处理后的在线实时图像测量区域内图像的灰度值进行特征向量提取。
第三步，熔料完成度判断。在本实施例中，利用向量夹角法进行特征向量对比，判断熔料完成度，主要内容如下所述：
首先，在当前工控机的视觉系统中加载离线分类器。
其次，将提取的当前帧的特征向量与离线分类器中的多个类别特征向量分别进行向量夹角法计算，值靠近1代表当前特征向量与该类别特征向量中的某个特征向量相似程度最高，该类别特征向量对应的熔料完成度即为当前的硅料熔料完成度。
向量夹角法计算原理为：将比较两个向量之间的距离，变为比较两个向量之间的夹角的余弦。因向量夹角的余弦值在0和1之间，比向量之间的距离归一化程度更好，因而容易确定分类阈值。
本公开实施例提供的熔料状态的检测方法，获取当前图像；按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量；获取第一特征向量与特征向量列表中每个特征向量的相似程度，特征向量列表包含每个特征向量与熔料状态的对应关系；根据所述相似程度确定所述当前图像对应的熔料状态。该方法采用两个特征向量相似的方法确定当前图像的熔料状态，计算简单，而且能够适应不同的单晶炉，实时判断硅料的熔化状态，及时提醒操作人员进行加料或功率控制等步骤，提高检测结果的准确性。
基于上述图13、图15、图16对应的实施例中所描述的熔料状态的检 测方法，下述为本公开装置实施例，可以用于执行本公开方法实施例。
本公开实施例提供一种熔料状态的检测装置，如图18所示，该熔料状态的检测装置60包括：第一获取模块601、图像处理模块602、第二获取模块603和确定模块604；第一获取模块601，用于获取当前图像；图像处理模块602，用于按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量；第二获取模块603，用于获取第一特征向量与特征向量列表中每个特征向量的相似程度；确定模块604，用于根据相似程度确定当前图像对应的熔料状态。
在一个实施例中，如图19所示，熔料状态的检测装置60还包括：分类模块605；
第一获取模块601，用于在获取当前图像之前，获取至少P个单晶炉中Q个不同熔料状态的图像帧，P≥1，Q≥1；
图像处理模块602，用于按照预设算法对P*Q个图像帧中的每个图像帧进行图像处理，得到每个图像帧的特征向量；
分类模块605，用于对P*Q个图像帧的所有特征向量进行分类，得到特征向量列表。
在一个实施例中，如图20所示，图像处理模块602包括：灰度处理子模块6021、归一化子模块6022和特征提取子模块6023；
灰度处理子模块6021，用于对当前图像进行灰度化处理，得到第一图像；
归一化子模块6022，用于对第一图像进行归一化处理，得到第二图像；
特征提取子模块6023，用于对第二图像进行特征提取，得到当前图像的第一特征向量。
在一个实施例中，如图21所示，图像处理模块2102还包括：滤波子模块21024；
滤波子模块21024，用于对第一图像进行滤波处理，得到滤波图像；
归一化子模块21022，用于对滤波图像进行归一化处理，得到第二图像。
在一个实施例中，如图22所示，灰度处理子模块6021包括：确定单元71和灰度处理单元72；
确定单元71，用于从当前图像中确定图像测量区域；
灰度处理单元72，用于根据图像测量区域中每个像素点的像素值，利用第一公式对图像测量区域中每个像素点的像素值进行灰度化处理，得到第一图像，第一公式包括：
f(i,j)＝(R(i,j)+G(i,j)+B(i,j))/3
其中，(i,j)表示所述图像测量区域中一像素点的坐标，R(i,j)表示红色 通道像素点(i,j)的像素值，G(i,j)表示绿色通道像素点(i,j)的像素值，B(i,j)表示蓝色通道像素点(i,j)的像素值，f(i,j)表示经过灰度化处理后像素点(i,j)的灰度值。
在一个实施例中，如图23所示，滤波子模块6024包括：设置单元81和邻域平均单元82；设置单元81，用于设置统计窗口大小；邻域平均单元82，用于根据统计窗口的大小，利用第二公式对第一图像中每个像素点的像素值进行邻域平均，得到滤波图像，第二公式包括：
Figure PCTCN2019119464-appb-000005
其中，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，m*n表示所述统计窗口的大小，(s,t)表示所述统计窗口内像素点的坐标，f(s,t)表示所述统计窗口内经过灰度化处理的像素点(s,t)的灰度值。
在一个实施例中，如图24所示，归一化子模块6022包括：获取单元91和归一化单元92；
获取单元91，用于获取第一图像中的像素灰度最大值和第一图像中的像素灰度最小值以及滤波图像中的像素灰度最大值和滤波图像中的像素灰度最小值归一化单元92，用于利用第三公式对滤波图像中的每个像素点的像素值进行归一化处理，得到第二图像，第三公式包括：
K'(i,j)＝K(i,j)-[K max-K min]*(f max-f min)/255
其中，K'(i,j)表示经过归一化处理后像素点(i,j)的灰度值，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，K max表示滤波图像中的像素灰度最大值，K min表示滤波图像中的像素灰度最小值，f max表示第一图像中的像素灰度最大值，f min表示第一图像中的像素灰度最小值。
需要说明的是，上述实施例提供的熔料状态的检测装置在确定当前时刻图像对应的熔料状态时，仅以上述各功能模块的划分进行举例说明，实际应用中，可以根据需要而将上述功能分配由不同的功能模块完成，即将主节点的内部结构划分成不同的功能模块，以完成以上描述的全部或者部分功能。另外，上述实施例提供的熔料状态的检测装置与熔料状态的检测方法实施例属于同一构思，其具体实现过程详见方法实施例，这里不再赘述本公开实施例提供的熔料状态的检测装置，获取当前图像；按照预设算法对当前图像进行图像处理，得到当前图像的第一特征向量；获取第一特征向量与特征向量列表中每个特征向量的相似程度，特征向量列表包含每个特征向量与熔料状态的对应关系；根据所述相似程度确定所述当前图像对应的熔料状态。该方法采用两个特征向量相似的方法确定当前图像的熔料状态，计算简单，而且能够适应不同的单晶炉，实时判断硅料的熔化状态，及时提醒操作人员 进行加料或功率控制等步骤，提高检测结果的准确性。
参考图25所示，本公开实施例还提供了一种熔料状态的检测设备，该熔料状态的检测设备包括接收器1201、发射器1202、第二存储器1203和第二处理器1204，该发射器1202和第二存储器1203分别与第二处理器1204连接，第二存储器1203中存储有至少一条第二计算机指令，第二处理器1204用于加载并执行至少一条第二计算机指令，以实现上述图13、图15和图16对应的实施例中所描述的熔料状态的检测方法。
基于上述图13、图15和图16对应的实施例中所描述的熔料状态的检测方法，本公开实施例还提供一种计算机可读存储介质，例如，非临时性计算机可读存储介质可以是只读存储器(英文：Read Only Memory，ROM)、随机存取存储器(英文：Random Access Memory，RAM)、CD-ROM、磁带、软盘和光数据存储装置等。该存储介质上存储有至少一条第二计算机指令，用于执行上述图13、图15和图16对应的实施例中所描述的熔料状态的检测方法。
本领域普通技术人员可以理解实现上述实施例的全部或部分步骤可以通过硬件来完成，也可以通过程序来指令相关的硬件完成，所述的程序可以存储于一种计算机可读存储介质中，上述提到的存储介质可以是只读存储器，磁盘或光盘等。
以上描述仅为本申请的较佳实施例以及对所运用技术原理的说明。本领域技术人员应当理解，本申请中所涉及的发明范围，并不限于上述技术特征的特定组合而成的技术方案，同时也应涵盖在不脱离发明构思的情况下，由上述技术特征或其等同特征进行任意组合而形成的其它技术方案。例如上述特征与本申请中公开的(但不限于)具有类似功能的技术特征进行互相替换而形成的技术方案。
本文中所称的“一个实施例”、“实施例”或者“一个或者多个实施例”意味着，结合实施例描述的特定特征、结构或者特性包括在本发明的至少一个实施例中。此外，请注意，这里“在一个实施例中”的词语例子不一定全指同一个实施例。
在此处所提供的说明书中，说明了大量具体细节。然而，能够理解，本发明的实施例可以在没有这些具体细节的情况下被实践。在一些实例中，并未详细示出公知的方法、结构和技术，以便不模糊对本说明书的理解。
在权利要求中，不应将位于括号之间的任何参考符号构造成对权利要求的限制。单词“包含”不排除存在未列在权利要求中的元件或步骤。位于元件之前的单词“一”或“一个”不排除存在多个这样的元件。本发明可以借助于包括有若干不同元件的硬件以及借助于适当编程的计算机来实现。在列举了若 干装置的单元权利要求中，这些装置中的若干个可以是通过同一个硬件项来具体体现。单词第一、第二、以及第三等的使用不表示任何顺序。可将这些单词解释为名称。
最后应说明的是：以上实施例仅用以说明本发明的技术方案，而非对其限制；尽管参照前述实施例对本发明进行了详细的说明，本领域的普通技术人员应当理解：其依然可以对前述各实施例所记载的技术方案进行修改，或者对其中部分技术特征进行等同替换；而这些修改或者替换，并不使相应技术方案的本质脱离本发明各实施例技术方案的精神和范围。

Claims (22)

  1. 一种熔料状态检测方法，其特征在于，所述方法包括：
    获取熔料当前图像；
    对所述当前图像进行图像处理；
    根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。
  2. 根据权利要求1所述的检测方法，其特征在于，所述对所述当前图像进行图像处理包括：
    将所述当前图像划分为多个子区域；
    分别获取各子区域中像素的最小灰度值和像素的灰度值方差；
    所述根据所述图像处理结果确定当前图像所对应的熔料的熔料状态包括：根据所述最小灰度值和所述灰度值方差确定所述各子区域的熔料状态，所述各子区域的熔料状态包括液态或固态；
    根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
  3. 根据权利要求1所述的检测方法，其特征在于，对所述当前图像进行图像处理包括：
    按照预设算法对所述当前图像进行图像处理，得到所述当前图像的第一特征向量；
    获取所述第一特征向量与特征向量列表中每个特征向量的相似程度，所述特征向量列表包含每个特征向量与熔料状态的对应关系；
    所述根据所述图像处理结果确定当前图像所对应的熔料的熔料状态包括：
    根据所述相似程度确定所述当前图像对应的熔料状态。
  4. 根据权利要求2所述的方法，其特征在于，所述将所述当前图像划分为多个子区域包括：
    当所述当前图像为灰度图像时，将所述灰度图像划分为多个子区域；
    所述将所述当前图像划分为多个子区域还包括：
    当所述当前图像为彩色图像时，将所述彩色图像灰度化，得到灰度图像；
    将所述灰度图像划分为多个子区域；
    所述将所述当前图像划分为多个子区域还包括：
    当所述当前图像为彩色图像时，将所述当前彩色图像划分为多个子区域；
    将所述划分为多个子区域的当前彩色图像灰度化。
  5. 根据权利要求2所述的方法，其特征在于，
    所述将所述当前图像划分为多个子区域包括：
    设定所述当前图像的检测区域，将所述检测区域划分为多个子区域；其中，所述检测区域根据所述当前图像中熔料的位置设定，所述检测区域为矩形，所述多个子区域具有相同的大小。
  6. 根据权利要求5所述的方法，其特征在于，所述根据所述最小灰度值和灰度值方差确定各子区域的熔料状态包括：
    当所述最小灰度值和所述灰度值方差满足预设条件时，确定所述子区域的熔料状态为液态；
    当所述最小灰度值和所述灰度值方差不满足预设条件时，确定所述子区域的熔料状态为固态；
    其中，所述预设条件为所述子区域中像素的最小灰度值大于第一阈值，所述子区域中像素的灰度值方差小于第二阈值。
  7. 根据权利要求6所述的方法，其特征在于，所述根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态包括：
    获取所有子区域中熔料状态为液态的子区域个数，记为M；
    获取所有子区域中熔料状态为固态的子区域个数，记为N；
    根据N与M，确定所述当前图像所对应熔料的熔料状态。
  8. 根据权利要求7所述的方法，其特征在于，所述方法还包括：
    当所述当前图像所对应的熔料的熔料状态满足第一条件时，向所述熔料中加入所述熔料的固体原料。
  9. 根据权利要求7所述的方法，其特征在于，所述方法还包括；
    当所述当前图像所对应的熔料的熔料状态满足第二条件时，控制所述熔料进入稳温阶段；
    在所述稳温阶段结束后，进行引晶。
  10. 根据权利要求3所述的方法，其特征在于，所述获取当前图像之前，所述方法还包括：
    获取至少P个单晶炉中Q个不同熔料状态的图像，P≥1，Q≥1；
    按照所述预设算法对每个所述图像进行图像处理，得到所述P*Q个图像中的每个图像的特征向量；
    对所述P*Q个图像的所有特征向量进行分类，得到所述特征向量列表。
  11. 根据权利要求3或10所述的方法，其特征在于，所述按照预设算法对所述当前图像进行图像处理，得到所述当前图像的第一特征向量包括：
    对所述当前图像进行灰度化处理，得到第一图像；
    对所述第一图像进行归一化处理，得到第二图像；
    对所述第二图像进行特征提取，得到所述当前图像的第一特征向量。
  12. 根据权利要求11所述的方法，其特征在于，所述对所述第一图像进行归一化处理得到第二图像之前，所述方法还包括：
    对所述第一图像进行滤波处理，得到滤波图像；
    所述对所述第一图像进行归一化处理，得到第二图像包括：对所述滤波图像进行归一化处理，得到所述第二图像。
  13. 根据权利要求12所述的方法，其特征在于，所述对所述当前图像进行灰度化处理，得到第一图像包括：
    从所述当前图像中确定图像测量区域；
    根据所述图像测量区域中每个像素点的像素值，利用第一公式对所述图像测量区域中每个像素点的像素值进行灰度化处理，得到第一图像，所述第一公式包括：
    f(i,j)＝(R(i,j)+G(i,j)+B(i,j))/3
    其中，(i,j)表示所述图像测量区域中一像素点的坐标，R(i,j)表示红色通道像素点(i,j)的像素值，G(i,j)表示绿色通道像素点(i,j)的像素值，B(i,j)表示蓝色通道像素点(i,j)的像素值，f(i,j)表示经过灰度化处理后像素点(i,j)的灰度值。
  14. 根据权利要求13所述的方法，其特征在于，所述对所述第一图像进行滤波处理，得到滤波图像包括：
    设置统计窗口大小；
    根据所述统计窗口的大小，利用第二公式对所述第一图像中每个像素点的像素值进行邻域平均，得到滤波图像，所述第二公式包括：
    Figure PCTCN2019119464-appb-100001
    其中，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，m*n表示所述统计窗口的大小，(s,t)表示所述统计窗口内像素点的坐标，f(s,t)表示所述统计窗口内经过灰度化处理的像素点(s,t)的灰度值。
  15. 根据权利要求14所述的方法，其特征在于，所述对所述滤波图像进行归一化处理，得到第二图像包括：
    获取所述第一图像中的像素灰度最大值和所述第一图像中的像素灰度最小值以及所述滤波图像中的像素灰度最大值和所述滤波图像中的像素灰度最小值；
    利用第三公式对所述滤波图像中的每个像素点的像素值进行归一化处理，得到第二图像，所述第三公式包括：
    K'(i,j)＝K(i,j)-[K max-K min]*(f max-f min)/255
    其中，K'(i,j)表示经过归一化处理后像素点(i,j)的灰度值，K(i,j)表示经过邻域平均后像素点(i,j)的灰度值，K max表示滤波图像中的像素灰度最大值，K min表示滤波图像中的像素灰度最小值，f max表示第一图像中的像素灰度最大值，f min表示第一图像中的像素灰度最小值。
  16. 根据权利要求3所述的方法，其特征在于，所述相似程度为所述第一特征向量与所述特征向量列表中每个特征向量的相似系数；
    所述根据所述相似程度确定所述当前图像对应的熔料状态包括：
    将第二特征向量对应的熔料状态确定为所述当前图像对应的熔料状态，所述第二特征向量与所述第一特征向量的相似程度最大。
  17. 一种熔料状态检测装置，其特征在于，所述装置包括：
    获取模块，用于获取熔料当前图像；
    图像处理模块，用于对所述当前图像进行图像处理；
    熔料状态确定模块，用于根据所述图像处理结果确定当前图像所对应的熔料的熔料状态。
  18. 根据权利要求17所述的检测装置，其特征在于，所述图像处理模块包括：
    图像处理子模块，用于将所述当前图像划分为多个子区域；还用于分别获取各子区域中像素的最小灰度值和像素的灰度值方差
    所述熔料状态确定模块包括：熔料状态确定子模块，用于根据所述最小灰度值和所述灰度值方差确定所述各子区域的熔料状态，所述各子区域的熔料状态包括液态或固态；还用于根据所述各子区域的熔料状态确定所述当前图像所对应的熔料的熔料状态；所述当前图像所对应的熔料的熔料状态包括固态、液态或固液混合中的一种。
  19. 一种熔料状态检测设备，其特征在于，所述熔料状态检测设备包括：接口，总线，第一存储器与第一处理器，所述接口、第一存储器与第一处理器通过所述总线相连接，所述第一存储器用于存储可执行程序，所述第一处理器被配置为运行所述可执行程序实现如权利要求1-2、4-9中任一项所述的熔料状态检测方法的步骤。
  20. 一种熔料状态检测设备，其特征在于，所述熔料状态检测设备包括：接口，总线，第二存储器与第二处理器，所述接口、第二存储器与第二处理器通过所述总线相连接，所述第二存储器用于存储可执行程序，所述第二处理器被配置为运行所述可执行程序实现如权利要求1、3、10-16中任一项所述的熔料状态检测方法的步骤。
  21. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质 上存储第一可执行程序，所述第一可执行程序被第一处理器运行实现如权利要求1-2、4-9中任一项所述的熔料状态检测方法的步骤。
  22. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质上存储第二可执行程序，所述第二可执行程序被第二处理器运行实现如权利要求1、3、10-16中任一项所述的熔料状态检测方法的步骤。
PCT/CN2019/119464 2018-11-23 2019-11-19 一种熔料状态检测方法、装置及设备 WO2020103826A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN201811408346.X 2018-11-23
CN201811408346.XA CN111222360B (zh) 2018-11-23 2018-11-23 硅料熔化状态的检测方法、设备及存储介质
CN201910770397.5 2019-08-20
CN201910770397.5A CN112488980B (zh) 2019-08-20 2019-08-20 一种熔料状态检测方法、装置及设备

Publications (1)

Publication Number Publication Date
WO2020103826A1 true WO2020103826A1 (zh) 2020-05-28

Family

ID=70773252

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/119464 WO2020103826A1 (zh) 2018-11-23 2019-11-19 一种熔料状态检测方法、装置及设备

Country Status (1)

Country Link
WO (1) WO2020103826A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118330194A (zh) * 2024-06-13 2024-07-12 山东鲁玻玻璃科技有限公司 一种硼硅玻璃制造用监测分析系统

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020067857A1 (en) * 2000-12-04 2002-06-06 Hartmann Alexander J. System and method for classification of images and videos
CN108230321A (zh) * 2018-01-19 2018-06-29 深圳市亿图视觉自动化技术有限公司 缺陷检测方法及装置
CN108734704A (zh) * 2018-05-07 2018-11-02 西安工程大学 基于灰度方差归一化的输电导线断股检测技术

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20020067857A1 (en) * 2000-12-04 2002-06-06 Hartmann Alexander J. System and method for classification of images and videos
CN108230321A (zh) * 2018-01-19 2018-06-29 深圳市亿图视觉自动化技术有限公司 缺陷检测方法及装置
CN108734704A (zh) * 2018-05-07 2018-11-02 西安工程大学 基于灰度方差归一化的输电导线断股检测技术

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118330194A (zh) * 2024-06-13 2024-07-12 山东鲁玻玻璃科技有限公司 一种硼硅玻璃制造用监测分析系统

Similar Documents

Publication Publication Date Title
CN106101697B (zh) 图像清晰度检测方法、装置及测试设备
DE69803932T2 (de) Verfahren und anlage zur steuerung der züchtung eines siliziumkristalls
TWI447271B (zh) 控制單晶矽碇之成長程序的方法與裝置
WO2020253827A1 (zh) 评估图像采集精度的方法及装置、电子设备和存储介质
WO2020108287A1 (zh) 一种硅棒的晶线生长状态检测方法、装置及设备
CN112488980B (zh) 一种熔料状态检测方法、装置及设备
CN105654898B (zh) 图像采集设备成像摩尔纹抑制方法及其应用
CN111222360B (zh) 硅料熔化状态的检测方法、设备及存储介质
TW201807403A (zh) 結晶缺陷之評估方法、矽晶圓之製造方法以及結晶缺陷之評估裝置
CN113215651B (zh) 一种拉晶控制方法和设备、单晶炉以及计算机存储介质
CN104990510B (zh) 一种直拉单晶硅直径测量方法
JP2013025650A5 (zh)
CN117187943B (zh) 一种拉晶过程中熔料检测方法、装置及电子设备
WO2020103826A1 (zh) 一种熔料状态检测方法、装置及设备
CN107818559B (zh) 晶体接种状态检测方法和晶体接种状态图像的采集装置
CN111218714B (zh) 一种断线测量方法、设备及存储介质
CN114369868B (zh) 晶棒断线监测方法、存储介质、终端和拉晶设备
CN115125609A (zh) 一种基于图像处理算法的蓝宝石引晶控制工艺和系统
CN117788448A (zh) 一种基于拉普拉斯算法的锂电池增强方法
CN105260997A (zh) 一种自动获取目标图像的方法
JP2017003491A (ja) 処理プログラム、処理方法、及び処理装置
CN114252018B (zh) 晶体直径检测方法、系统及计算机程序产品
CN114387251B (zh) 饱满点的监测方法、存储介质、终端和拉晶设备
CN114244965B (zh) 一种应用于高拍仪的高精准曝光度调控方法、系统及介质
CN113029014B (zh) 一种单晶硅棒直径的检测方法及检测装置

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: 19886845

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: 19886845

Country of ref document: EP

Kind code of ref document: A1