Method, device and equipment for detecting melt liquid level position and computer storage medium

Abstract

The invention provides a method, device and equipment for detecting the liquid level position of melt and a computer storage medium, and belongs to the technical field of monocrystalline silicon production. Performing feature point matching on the to-be-measured liquid level image and each reference image in the plurality of reference images, and correcting matching point pairs according to a geometrical relationship among feature points to obtain a plurality of matching point sets; and determining the current position of the liquid level to be measured according to the liquid level position of the target reference image corresponding to the matching point set with the highest matching degree in the plurality of matching point sets. According to the method, the detection efficiency can be improved while the precision of melt liquid level position detection is improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of single-crystal silicon production, and particularly to a method, device, equipment, and computer storage medium for detecting the position of a melt liquid surface. Background Art

[0002] Most single-crystal silicon rods are manufactured by the Czochralski method, also known as the direct pulling method. This method utilizes the principle of melt condensation crystallization drive. At the interface between the solid and the liquid, due to the decrease in the melt temperature, a phase change occurs from liquid to solid. The single-crystal silicon rods grown by the direct pulling method have a relatively high oxygen content and a relatively large diameter, and it is a widely used method at present. However, as the single-crystal silicon rod grows continuously in the solid state, the volume of the molten silicon in the crucible gradually decreases, and the liquid surface of the molten silicon continuously drops, which will affect the growth control of the crystal and the crystal quality. Therefore, during the preparation of the single-crystal silicon rod, it is necessary to monitor the position of the melt liquid surface in the quartz crucible in real time, so that the control of the melt liquid surface position is in a closed-loop control state, and higher control accuracy can be obtained in this way.

[0003] The reflection method is a method for detecting the position of the molten silicon liquid surface based on machine vision. Based on the simple principle of specular reflection, the surface of the molten silicon in a high-temperature molten state has a strong reflection ability. Through the gold-plated protective glass, the reflection of the object above the molten silicon on the surface of the molten silicon can be clearly observed with the naked eye. Fix the reference object above the molten silicon and the observation position, lift and lower the crucible to cause a vertical displacement of the melt surface. The position of the liquid surface and the position of the reflection of the reference object in the field of view are in one-to-one correspondence. According to the corresponding relationship and the position of the reflection of the reference object in the field of view, the position of the liquid surface can be determined.

[0004] However, when using the reflection method to detect the melt liquid surface, how to improve the detection accuracy while reducing the calculation amount has become an urgent problem to be solved. Summary of the Invention

[0005] The present disclosure provides a method, device, equipment, and computer storage medium for detecting the position of a melt liquid surface; it can improve the detection efficiency while improving the detection accuracy of the melt liquid surface position.

[0006] The technical solution of the present disclosure is realized as follows: In a first aspect, the present disclosure provides a method for detecting the position of a melt liquid surface. The method includes: acquiring an image of the liquid surface to be measured; performing feature point matching on the image of the liquid surface to be measured and each reference image in a plurality of reference images, and correcting the matching point pairs according to the geometric relationship between the feature points to obtain a plurality of matching point sets; determining the current position of the liquid surface to be measured according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the plurality of matching point sets.

[0007] In a second aspect, the present disclosure provides a device for detecting the position of a molten liquid surface. The device includes: an acquisition module, a matching and correction module, and a determination module; the acquisition module is configured to acquire an image of the liquid surface to be measured; the matching and correction module is configured to perform feature point matching between the image of the liquid surface to be measured and each reference image among a plurality of reference images, and correct the matching point pairs according to the geometric relationship between the feature points to obtain a plurality of matching point sets; the determination module is configured to determine the current position of the liquid surface to be measured according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the plurality of matching point sets.

[0008] In a third aspect, the present disclosure provides a device for detecting the position of a molten liquid surface. The device includes a reference object, a CCD camera, a memory, and a processor disposed above the liquid surface to be measured inside a single crystal furnace; wherein, the memory is used to store a computer program that can run on the processor; the CCD camera is used to acquire an image of the liquid surface to be measured; the processor is used to execute the steps of the method for detecting the position of the molten liquid surface described in the first aspect when running the computer program.

[0009] In a fourth aspect, the present disclosure provides a computer-readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of the method for detecting the position of the molten liquid surface described in the first aspect are implemented.

[0010] In a fifth aspect, the present disclosure provides a computer program product, wherein the computer program product includes a computer program or instruction. When the computer program product runs on a processor, the processor is caused to execute the computer program or instruction to implement the steps of the method for detecting the position of the molten liquid surface described in the first aspect.

[0011] In a sixth aspect, the present disclosure provides a chip, which includes a processor and a communication interface. The communication interface is coupled to the processor, and the processor is used to run a program or instruction to implement the method for detecting the position of the molten liquid surface described in the first aspect.

[0012] The present disclosure provides a method for detecting the position of a molten liquid surface. Through feature point matching, key information in the image of the liquid surface to be measured can be captured, and geometric relationship correction can eliminate incorrect matching point pairs to ensure the accuracy of the matching point set, thereby improving the accuracy of liquid surface position detection. Moreover, through matching point pair correction and matching degree screening, errors caused by interference in the image (such as radiation, noise, smoke, reflection, etc.) can be reduced, enhancing the robustness of the detection. At the same time, the calculation amount can also be reduced to improve the detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 It is a schematic structural diagram of a single crystal furnace provided by the present disclosure; Figure 2Flow schematic diagram of a method for detecting the position of the molten liquid surface provided by the present disclosure; Figure 3 Schematic diagram of the slope of the matching point pairs provided by the present disclosure; Figure 4 Schematic diagram of the feature points included in the initial image provided by the present disclosure; Figure 5 Schematic diagram of the feature points included in the image of the liquid surface to be measured provided by the present disclosure; Figure 6 Schematic diagram of the visualization interface provided by the present disclosure; Figure 7 Structural block diagram of a device for detecting the position of the molten liquid surface provided by the present disclosure; Figure 8 Hardware structure schematic diagram of a device for detecting the position of the molten liquid surface provided by the present disclosure. Detailed implementation manners

[0014] Next, the technical solutions in the embodiments of the present disclosure will be clearly described in conjunction with the accompanying drawings in the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those of ordinary skill in the art belong to the scope protected by the present disclosure.

[0015] The terms "first", "second", etc. in the specification of the present disclosure are used to distinguish similar objects, rather than to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances so that the present disclosure can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second", etc. are usually of the same category, and the number of objects is not limited. For example, the first object can be one or multiple.

[0016] In industrial fields such as metallurgy, casting, and crystal growth, accurately detecting the position of the molten liquid surface is crucial for ensuring product quality and production safety. Traditional methods for detecting the molten liquid surface mainly include the crucible root ratio method and the weighing method. The crucible root ratio method is a method for determining the liquid surface position by measuring the relative position between the molten liquid surface in the crucible (container) and the bottom of the crucible; the weighing method is a method for calculating the liquid surface position by measuring the total weight of the crucible and the melt inside it, and combining the melt density and the crucible geometry. However, the measurement results of traditional methods are easily affected by factors such as changes in the crucible shape and melt fluctuations, resulting in insufficient measurement accuracy, and the response speed of complex mechanical devices is relatively slow, making it difficult to meet the requirements of real-time monitoring.

[0017] In recent years, non-contact molten liquid surface detection methods based on machine vision have gradually become a research hotspot. Such methods collect images of the molten liquid surface and use image processing algorithms to extract the liquid surface position information, which have the advantages of good real-time performance and no need to contact the melt.

[0018] However, under the interference of high temperature and strong radiation, the reflection of the reference object on the molten liquid surface is not clear, and there is still a problem of low measurement accuracy when using the existing molten liquid surface detection methods.

[0019] Therefore, the present disclosure aims to provide a method for detecting the position of the molten liquid surface that can improve the detection accuracy of the molten liquid surface position. Refer to Figure 1 , which shows a single crystal furnace 1 capable of implementing the technical solution of the embodiment of the present disclosure. The single crystal furnace 1 may include: a furnace body 10, a guiding cylinder 11, and a crucible 12; it should be noted that, Figure 1 The structure of the single crystal furnace 1 shown is not specifically limited. In order to clearly illustrate the technical solution of the embodiment of the present invention, other components required for manufacturing single crystal silicon by the Czochralski method are omitted and not shown, such as a heater for heating the polysilicon raw material contained in the crucible 12, a seed crystal cable for pulling, and a pulling drive component for lifting and rotating the seed crystal cable, etc. The crucible 12 contains a melt, and the molten liquid surface 15 is as shown in the figure. Based on Figure 1 the single crystal furnace 1 shown, an observation window 14 may also be opened in the upper part of the furnace body 10 for an optical observation instrument 13, such as a CCD camera, to observe the inside of the furnace body 10.

[0020] According to the structural example of the single crystal furnace 1 described above, the embodiment of the present disclosure expects to provide a method for detecting the position of the molten liquid surface. As Figure 2 shown, it is a schematic flowchart of the method for detecting the position of the molten liquid surface provided by the present disclosure. The method may include the following steps S201 to step S203.

[0021] In step S201, an image of the liquid surface to be measured is obtained.

[0022] By collecting the real-time image of the molten liquid surface through an optical observation instrument, an image of the liquid surface to be measured is obtained, and the image of the liquid surface to be measured includes the reflection of the reference object. The reference object may be an object located above or around the liquid surface to be measured (such as a camera bracket, a light source, a mechanical device, the guiding cylinder 11, etc.).

[0023] In step S202, the image of the liquid surface to be measured is matched with each reference image in a plurality of reference images, and the matching point pairs are corrected according to the geometric relationship between the feature points to obtain a plurality of matching point sets.

[0024] The reference image is a pre - acquired melt image with a known liquid level position, which is used for comparison with the liquid level image to be measured. Multiple reference images corresponding to different liquid level positions can be pre - acquired according to actual needs, and the specific number of reference images is not limited in this disclosure.

[0025] Feature points can reflect the information of local regions with significant features in the image. Feature point matching is to find similar or identical feature points in the liquid level image to be measured and the reference image. The feature points in the liquid level image to be measured that match the feature points in the reference image form a matching point pair, and all the matching point pairs in the liquid level image to be measured and a reference image form a matching point set. Matching the liquid level image to be measured with multiple reference images can obtain multiple matching point sets.

[0026] Geometric relationship refers to the spatial position relationship of feature points in the image, such as distance, angle, slope, etc. Using the spatial position relationship of feature points, the positions of matching point pairs are adjusted to eliminate errors. For example, if the distance relationship and angle relationship of most matching point pairs in the liquid level image to be measured and the reference image are consistent, while a few matching point pairs do not conform, then determine these few matching point pairs as incorrect matching point pairs and these incorrect matching point pairs can be removed.

[0027] In some embodiments, the geometric relationship includes the slope of the matching point pair in the image coordinate system; the above - mentioned correction of the matching point pair according to the geometric relationship between feature points can specifically be to correct the matching point pair corresponding to the outlier slope value among the slope values of the matching point pairs.

[0028] The slope is the tangent value of the angle between the line connecting the two feature points included in the matching point pair and the horizontal direction, which characterizes the inclination of the line and can intuitively reflect the spatial position relationship between the matching point pairs. Outlier slope values can be identified through statistical methods (such as mean, standard deviation) or clustering methods (such as K - means). Outlier slope values usually correspond to incorrect matching point pairs. For example, if the slope values of most matching point pairs are concentrated within a certain range, while the slope values of a few matching point pairs deviate significantly from this range, the slopes that deviate from this range are outlier slope values, and the matching point pairs corresponding to these outlier slope values are incorrect. By correction, the matching accuracy can be improved.

[0029] Exemplarily, as Figure 3 shown, it is a schematic diagram of the slope of the matching point pair. The reference object reflection 301 in the liquid level image 30 to be measured is matched with the reference object reflection 311 in the reference image 31. The matching point pairs are connected by dotted lines or solid lines in the figure. It can be seen from the figure that each solid line is approximately parallel, but the slope value of the dotted line is quite different from that of the solid line, which belongs to the outlier slope. The matching point pair corresponding to the dotted line is an incorrect matching point pair. Therefore, the matching point pair composed of the two feature points corresponding to the dotted line is removed.

[0030] Since the change in the position of the melt liquid level is a continuous process and has high requirements for real-time performance, while slope correction only needs to calculate and compare the slope values of the matching point pairs, with low computational complexity, being simple and efficient, and slope correction can effectively handle problems such as noise, occlusion, or local deformation in the image, making the current position of the liquid level to be measured determined according to the finally corrected matching point pairs more accurate.

[0031] In some embodiments, the matching point pairs are corrected according to the geometric relationship between the feature points. Specifically, after each feature point is matched, the geometric relationship of the matching point pairs is obtained, and the matching point pairs are corrected according to the geometric relationship of the matching point pairs. That is, correction is performed while matching. For the image of the liquid level to be measured and a reference image, after obtaining the Nth (N is an integer greater than 3) matching point pair, it is determined whether the geometric relationship of the Nth matching point pair is consistent with the geometric relationships of the previous N - 1 matching point pairs. If they are consistent, they are retained; otherwise, the matching point pair is eliminated. This is repeated until the matching between the image of the liquid level to be measured and a reference image is completed, and a set of matching points is obtained.

[0032] In this matching method of correcting while matching, after each new matching point pair is matched, it is immediately corrected according to the geometric relationship, and the incorrect matching point pairs that do not conform to the geometric relationship are eliminated, reducing the number of matching point pairs to be processed in the subsequent matching process, thereby reducing the computational complexity. Since the incorrect matching point pairs are deleted in a timely manner, the memory occupancy is low. In addition, by eliminating the incorrect matching point pairs in real time, it can also avoid the interference of incorrect matching point pairs on the subsequent matching process, thereby improving the overall matching accuracy.

[0033] In some embodiments, the matching point pairs are corrected according to the geometric relationship between the feature points. Specifically, it can also be that after all feature points are matched, the geometric relationship of all matching point pairs is obtained, and the matching point pairs are corrected according to the geometric relationship of all matching point pairs. That is, for the image of the liquid level to be measured and a reference image, after the matching is completed, the geometric relationship is determined through the already matched matching point pairs, and then according to the geometric relationship, the incorrect matching point pairs are screened out from all the matching point pairs.

[0034] In this method of correcting after all matching is completed, after all feature points are matched, correction can be performed based on the global information of the matching point pairs (such as the geometric relationship of all matching point pairs), with a higher tolerance for local noise and outliers, thereby avoiding prematurely eliminating potentially correct matching point pairs.

[0035] In some embodiments, the method for detecting the position of the melt liquid level further includes: after correcting the matching point pairs through the geometric relationship, the incorrect matching point pairs in multiple sets of matching points are eliminated by the random sample consensus algorithm. In this way, the accuracy rate of the matching point pairs can be further improved.

[0036] In step S203, based on the liquid level position of the target reference image corresponding to the matching point set with the highest matching degree among multiple matching point sets, the current position of the liquid level to be measured is determined.

[0037] A matching point set is obtained for the image of the liquid level to be measured and each reference image. Each matching point set corresponds to a matching degree for measuring the similarity between the two images. For example, the matching degree is determined by the number of matching points. The higher the number of matching points in a matching point set, the higher the matching degree, and vice versa.

[0038] The target reference image is the reference image with the highest matching degree with the image to be measured among multiple reference images. Each reference image has the position of the actual molten liquid level corresponding to it. Based on the liquid level position of the target reference image, the current position of the molten liquid level in the image of the liquid level to be measured is deduced.

[0039] Through the method for detecting the position of the molten liquid level provided by the present disclosure, key information in the image of the liquid level to be measured can be captured through feature point matching, and geometric relationship correction can eliminate incorrect matching point pairs to ensure the accuracy of the matching point set, thereby improving the accuracy of liquid level position detection. Moreover, through matching point pair correction and matching degree screening, incorrect matching caused by interference (such as radiation, noise, smoke, reflection, etc.) in the image can be reduced, enhancing the robustness of the detection.

[0040] In some embodiments, the above-mentioned feature point matching of the image of the liquid level to be measured with each reference image among multiple reference images may specifically be to determine multiple key point neighborhoods in the image of the liquid level to be measured; determine the feature points corresponding to the feature vectors for describing the key point neighborhoods according to the brightness changes of each pixel point within the key point neighborhoods; and determine the matching point pairs that match between the feature points in the image of the liquid level to be measured and the feature points of each reference image according to the similarity between them.

[0041] A key point refers to a local area in the image with significant features, such as a corner point, an edge point, or an area with rich texture. A key point neighborhood refers to a local area centered on the key point, usually a window of a fixed size (such as 16x16 pixels or 32x32 pixels), and the pixel points within the key point neighborhood contain the local feature information of the key point.

[0042] In some embodiments, to determine multiple key point neighborhoods in the image of the liquid level to be measured, it may specifically be to perform Gaussian convolution on the image of the liquid level to be measured to obtain a three-dimensional image matrix formed by stacking images of different scales; determine multiple extreme points from the three-dimensional image matrix through difference of Gaussians; and use the preset area centered on each extreme point as the key point neighborhood.

[0043] The liquid level image to be measured is subjected to Gaussian convolution using different convolution kernels (multiple Gaussian blurs are performed) to generate images of different scales. The images of different scales are stacked into a pyramid shape to obtain a three-dimensional image matrix. The images of adjacent scales in the three-dimensional image matrix are differentiated to obtain a three-dimensional Gaussian difference image matrix. Each pixel point is compared with its neighborhood, including m neighborhood points of the same scale and (m + 1)×2 = 18 neighborhood points of adjacent scales, where m is an integer greater than 1, and (m + 1) represents the center point and m neighborhood points. If the pixel point is a local maximum or local minimum, it is marked as a candidate extreme point. Points with low contrast are removed (filtered by a threshold), and points with weak edge responses are removed (the principal curvature is calculated through the Hessian matrix). The finally obtained extreme points are the key points, and a preset region of a certain size centered on the key points is used as the key point neighborhood.

[0044] The brightness change refers to the change in the gray value or color value of the pixel points within the key point neighborhood. The brightness change can reflect the local texture, edge, or shape characteristics of the key points. The brightness change includes the brightness change intensity and the brightness change direction of the pixel points.

[0045] The gradient magnitude can be used to represent the brightness change magnitude of each pixel point, and the gradient direction can be used to represent the brightness change direction of each pixel point. Specifically, the gradient of each pixel point is divided into a horizontal gradient and a vertical gradient. The calculation formula for the horizontal gradient (x direction) of each pixel point is: , and the calculation formula for the vertical gradient (y direction) is: , where represents the brightness value (gray value) of the pixel point ; the calculation formula for the gradient magnitude of each pixel point is: , and the calculation formula for the gradient direction is: .

[0046] The feature vector is a mathematical representation used to describe the features of the key point neighborhood. It is usually a multi-dimensional vector, and each dimension of the feature vector corresponds to a certain feature of the pixel points within the neighborhood (such as the gradient direction, gray value, etc.). The feature point includes the feature vectors of the features of each pixel within the neighborhood.

[0047] Specifically, the key point neighborhood is divided into multiple sub-regions. For each sub-region, the gradient magnitude and direction of each pixel within it are determined. The gradient direction is quantized into p (an integer greater than 1, such as 8) directions, and p interval direction histograms are constructed. The gradient magnitude of each pixel is assigned to the corresponding interval according to its direction and multiplied by a Gaussian weight (the farther away from the key point, the smaller the weight). The p interval direction histograms of each sub-region are spliced into a vector with a dimension equal to the product of the sub-region size and p, and normalization processing is performed, that is, the feature vector corresponding to a matching point is obtained.

[0048] The similarity refers to the degree of similarity of the feature vectors between the feature points in the liquid level image to be measured and the feature points in the reference image. For example, the similarity can be determined by calculating the Euclidean distance between two feature vectors. The smaller the distance, the higher the similarity. The similarity can also be determined by calculating the Hamming distance between two feature vectors. The smaller the distance, the higher the similarity. Or the similarity can also be determined by calculating the cosine value of the angle between two feature vectors. The larger the cosine value, the higher the similarity. Specifically, the method for calculating the similarity is not limited in this disclosure.

[0049] The specific matching method can be, for each feature point in the liquid level image to be measured, finding the feature point in the reference image with the closest feature vector distance to it, which is the feature point matching this feature point. Or, it can also be, for each feature point in the liquid level image to be measured, finding the nearest neighbor and the second nearest neighbor feature points in the reference image, calculating the ratio of the distance between the nearest neighbor feature point and the feature point in the liquid level image to be measured to the distance between the second nearest neighbor feature point and the feature point in the liquid level image to be measured. If the ratio of the distances is less than a preset threshold, it is considered that the nearest neighbor matching point pair is reliable, and the nearest neighbor matching point is determined to match the feature point in the liquid level image to be measured. Specifically, other matching methods can be adopted according to actual needs, and this disclosure does not make limitations.

[0050] In this disclosure, by determining the key points in the liquid level image to be measured that can represent the key information in the image (such as corner points, edge points, or regions with rich texture), and then obtaining the pixel information within the key point neighborhood centered on the key points, the feature points corresponding to the feature vectors are obtained. Therefore, the feature points obtained from the liquid level image to be measured can accurately reflect the key information in the liquid level image to be measured (such as the reflection of the reference object), so that the target reference image matched based on this feature point is also more accurate, and further making the current position of the liquid level to be measured finally determined according to the liquid level position of the target reference image more accurate.

[0051] In some embodiments, the above-mentioned obtaining of the liquid level image to be measured can specifically be collecting an initial image of the liquid level to be measured including the reflection of the reference object; cropping the image area outside the reflection of the reference object from the initial image to obtain the liquid level image to be measured.

[0052] The initial image is the original image directly captured by a CCD camera or other image acquisition device. Due to the high reflectivity of the melt surface, the initial image usually contains the reflection of the reference object. The reference object forms a reflection on the melt surface, and the reflection area appears as a high-brightness or complex texture in the image.

[0053] Cropping refers to removing the unwanted regions from the initial image and retaining the region of interest (i.e., the region of the reference object's reflection). Specifically, it can be manually selecting the liquid surface region and cropping it according to prior knowledge or image features; it can also be automatically identifying the region corresponding to the reference object's reflection using image processing algorithms and cropping. After cropping, the liquid surface image to be measured containing only the region of the reference object's reflection is obtained, removing the interference of the reference object's reflection and the background.

[0054] For automatically identifying the reference object's reflection and cropping, in some embodiments, cropping the image region outside the reference object's reflection from the initial image to obtain the liquid surface image to be measured can be based on the brightness values of the pixels. The image region corresponding to the reference object's reflection is determined from the initial image through threshold segmentation; the image region outside the image region corresponding to the reference object's reflection is cropped to obtain the liquid surface image to be measured.

[0055] The pixel brightness value refers to the gray value or color intensity value of each pixel in the image. In the liquid surface image to be measured, the region of the reference object's reflection usually has a higher brightness value, while the brightness values of the liquid surface region and the background region are lower. Threshold segmentation is based on the pixel brightness value. By setting a brightness threshold, the liquid surface image to be measured is divided into a target region and a background region. For the segmentation of the reference object's reflection region, a relatively high brightness threshold is usually selected to separate the reference object's reflection region from other regions. The threshold segmentation method based on pixel brightness values is simple to calculate, has high calculation efficiency, and is suitable for real-time applications.

[0056] In some embodiments, cropping the image region outside the reference object's reflection from the initial image to obtain the liquid surface image to be measured can also be based on edge features. The image region corresponding to the reference object's reflection is determined from the initial image; the image region outside the image region corresponding to the reference object's reflection is cropped to obtain the liquid surface image to be measured.

[0057] An edge refers to the region where the brightness or color in the image changes significantly, usually corresponding to the boundary of an object or texture changes. In the liquid surface image to be measured, the edge of the reference object's reflection usually appears as an obvious boundary between the high-brightness region and the surrounding region.

[0058] Use edge detection algorithms (such as Canny, Sobel, or Laplacian algorithms) to extract the edges in the initial image. The edge detection result is a binary image, where the edge pixels are white and the non-edge pixels are black. Perform contour extraction on the edge detection result, find the closed contour region, and determine the contour corresponding to the reference object's reflection by analyzing the shape, size, and position of the contour. The method based on edge features can accurately identify the boundary of the reference object's reflection and avoid incorrect cropping.

[0059] If the initial image is not cropped, a large number of invalid feature points will be detected when determining the feature points. This will not only increase the computational complexity but also affect the accuracy of feature point matching, resulting in low efficiency and high error rate. By cropping, only the area of the reflected image of the reference object of interest is retained, which can not only greatly reduce the computational complexity but also make the matching of feature points more accurate, thus making the final detection result more accurate.

[0060] Exemplarily, as Figure 4 shown, in the initial image, there are the reflected image 301 of the reference object and other parts. The feature points detected in the initial image are shown as black solid circles in the figure; as Figure 5 shown, the image of the liquid surface to be measured obtained after cropping the initial image, which only retains the reflected image 301 of the reference object. The feature points detected in the image of the liquid surface to be measured are shown as black solid circles in the figure. Comparing Figure 4 and Figure 5 it can be seen that Figure 5 the number of feature points in

[0061] is greatly reduced.

[0062] Due to various reasons such as liquid surface fluctuations, even the target reference image corresponding to the set of matching points with the highest matching degree and the image of the liquid surface to be measured cannot be exactly the same. That is, there will be a certain degree of error in directly determining the current position of the liquid surface to be measured as the liquid surface position of the target reference image.

[0063] Therefore, in some embodiments, determining the current position of the liquid surface to be measured according to the liquid surface position of the reference image corresponding to the set of matching points with the highest matching degree among multiple sets of matching points includes: determining an intermediate position according to the liquid surface position of the target reference image corresponding to the set of matching points with the highest matching degree among multiple sets of matching points; determining the first centroid of the curve formed by the matching points in the image of the liquid surface to be measured, and the second centroid of the curve formed by the matching points in the target reference image; correcting the intermediate position according to the position difference between the first centroid and the second centroid to obtain the current position.

[0063] The first centroid refers to the geometric centroid of the curve formed by the matching points in the image of the liquid surface to be measured, that is, the average value of the coordinates of all matching points is taken. The second centroid refers to the geometric centroid of the curve formed by the matching points in the target reference image.

[0064] Since it is impossible to ensure that the number of detected matching points is the same each time, it is impossible to directly determine the offset according to the matching points. Therefore, the present disclosure corrects the intermediate position through the relative displacement of the centroids of the reflected images of the liquid surface to be measured and the target reference image. The intermediate position is the position of the liquid surface to be measured preliminarily determined according to the liquid surface position of the target reference image, and it can be the liquid surface position of the target reference image. In this way, through centroid correction, the intermediate position is further corrected to improve the accuracy of liquid surface position detection.

[0065] Since the change in the position of the molten liquid surface is a continuous process, due to the fluctuations of the molten liquid surface (such as temperature changes, mechanical vibrations, or melt flow), the detection results of the liquid surface position may show short-term fluctuations or noise. These fluctuations will cause deformation of the reflected image of the reference object, thereby affecting the accuracy of the liquid surface position calculation. To reduce this impact, in some embodiments, filtering processing is performed on the current position to reduce the impact on the calculation result caused by the deformation of the reflected image of the reference object due to the liquid surface fluctuation.

[0066] Specifically, a filtering window (also called a sliding window) is determined. The size of the window determines the smoothness of the filtering. The larger the window, the more obvious the smoothing effect, but some detailed information may be lost; the smaller the window, the weaker the smoothing effect, but more details can be retained. The mean value of the current position and the k - 1 positions before the current position is determined as the result of filtering the current position, where k is the size of the filtering window and k is an integer greater than 1.

[0067] In addition, the method for detecting the position of the molten liquid surface according to the present disclosure also provides a visual interface, such as Figure 6 shown, area 601 corresponds to displaying the image of the liquid surface to be measured, area 602 corresponds to displaying the matched target reference image, and the user can also manually modify the target reference image. The first centroid, the second centroid, and the offset are also displayed in the interface.

[0068] The method for detecting the position of the molten liquid surface provided by the present disclosure has the following beneficial effects: Improve detection accuracy: Through feature point matching and geometric relationship correction, incorrect matching point pairs can be effectively eliminated, ensuring the accuracy of the matching point set, thereby improving the detection accuracy of the liquid surface position. Especially in a high-temperature and strong-radiation environment, this method can handle problems such as image noise, occlusion, and local deformation, ensuring the reliability of the measurement results.

[0069] Good real-time performance: This method adopts the strategies of slope correction and correction while matching edges, with low computational complexity and fast response speed, and can meet the real-time monitoring requirements of the change in the position of the molten liquid surface.

[0070] Strong robustness: Through the random sample consensus algorithm and centroid correction, incorrect matches caused by interference in the image (such as radiation, noise, smoke, etc.) are further reduced, enhancing the stability and anti-interference ability of the detection.

[0071] Reduce the amount of calculation: By cropping the initial image and only retaining the area of the reflected image of the reference object, the detection of invalid feature points is reduced, the computational complexity is lowered, and the matching efficiency is improved.

[0072] Visual interface: Provide an intuitive visual interface, which is convenient for users to monitor in real time and manually adjust the target reference image, enhancing the user experience and the convenience of operation.

[0073] Figure 7

[0073] is a structural block diagram of a device for detecting the position of the molten liquid surface shown in this disclosure. The device 70 for detecting the position of the molten liquid surface includes: an acquisition module 701, a matching and correction module 702, and a determination module 703. The acquisition module 701 is configured to acquire an image of the liquid surface to be measured. The matching and correction module 702 is configured to perform feature point matching on the image of the liquid surface to be measured with each reference image in a plurality of reference images, and correct the matching point pairs according to the geometric relationship between the feature points to obtain a plurality of matching point sets. The determination module 703 is configured to determine the current position of the liquid surface to be measured according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the plurality of matching point sets.

[0074] In some embodiments, the above-mentioned matching and correction module 702 is specifically configured to, after each feature point is matched, acquire the geometric relationship of the matching point pair, and correct the matching point pair according to the geometric relationship of the matching point pair.

[0075] In some embodiments, the above-mentioned matching and correction module 702 is specifically configured to, after all feature points are matched, acquire the geometric relationship of all matching point pairs, and correct the matching point pairs according to the geometric relationship of all matching point pairs.

[0076] In some embodiments, the geometric relationship includes the slope of the matching point pair in the image coordinate system. The above-mentioned matching and correction module 702 is specifically configured to correct the matching point pair according to the geometric relationship between the feature points, and correct the matching point pair corresponding to the outlier slope value among the slope values of the matching point pairs.

[0077] In some embodiments, the above-mentioned matching and correction module 702 is specifically configured to determine a plurality of key point neighborhoods in the image of the liquid surface to be measured, determine the feature points corresponding to the feature vectors for describing the key point neighborhoods according to the brightness changes of each pixel point in the key point neighborhoods, and determine the matching point pairs matched between the image of the liquid surface to be measured and each reference image according to the similarity between the feature points in the image of the liquid surface to be measured and the feature points of each reference image.

[0078] In some embodiments, the above-mentioned matching and correction module 702 is specifically configured to perform Gaussian convolution on the image of the liquid surface to be measured to obtain a three-dimensional image matrix stacked by images of different scales, determine a plurality of extreme points from the three-dimensional image matrix through Gaussian difference, and use a preset region centered on each extreme point as a key point neighborhood.

[0079] In some embodiments, the above-mentioned acquisition module 701 is specifically configured to collect an initial image of the liquid surface to be measured including the reflection of the reference object, and crop the image region outside the reflection of the reference object from the initial image to obtain an image of the liquid surface to be measured.

[0080] In some embodiments, the obtaining module 701 is specifically configured to determine, based on the luminance value of pixels, an image area corresponding to the reflection of the reference object from the initial image through threshold segmentation; and crop the image area outside the image area corresponding to the reflection of the reference object to obtain an image of the liquid surface to be measured.

[0081] In some embodiments, the obtaining module 701 is specifically configured to determine, based on edge features, an image area corresponding to the reflection of the reference object from the initial image; and crop the image area outside the image area corresponding to the reflection of the reference object to obtain an image of the liquid surface to be measured.

[0082] In some embodiments, the determining module 703 is specifically configured to determine an intermediate position according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among multiple matching point sets; determine a first centroid of the curve formed by the matching points in the image of the liquid surface to be measured, and a second centroid of the curve formed by the matching points in the target reference image; and correct the intermediate position according to the position difference between the first centroid and the second centroid to obtain the current position.

[0083] In some embodiments, the device 70 for detecting the position of the molten liquid surface further includes a rejection module, which is configured to reject mismatched point pairs in multiple matching point sets through the random sample consensus algorithm.

[0084] In the embodiments of the present disclosure, each module can implement the method for detecting the position of the molten liquid surface provided in the above method embodiments, and can achieve the same technical effects. To avoid repetition, details are not described here again.

[0085] The present disclosure also provides a device for detecting the position of the molten liquid surface. Please refer to Figure 8 ., which shows the specific hardware structure of a device 80 for detecting the position of the molten liquid surface that can implement the device 70 for detecting the position of the molten liquid surface provided in the embodiments of the present invention. The device 80 for detecting the position of the molten liquid surface can be applied to Figure 1 the single crystal furnace 1 shown in. The device 80 for detecting the position of the molten liquid surface may include: a CCD camera 801, a memory 802, and a processor 803 disposed in the single crystal furnace 1; and each component can be coupled together through a bus system 804. It can be understood that the bus system 804 is used to realize the connection and communication between these components. In addition to the data bus, the bus system 804 further includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, in Figure 8 all kinds of buses are labeled as the bus system 804. Among them, the CCD camera 801 is configured to obtain an image of the liquid surface to be measured; the memory 802 is configured to store a computer program that can run on the processor 803; The above-mentioned processor 803 is configured to perform the following steps when running the computer program: Perform feature point matching between the liquid level image to be measured and each reference image in a plurality of reference images, correct the matching point pairs according to the geometric relationship between the feature points, and obtain a plurality of matching point sets; determine the current position of the liquid level to be measured according to the liquid level position of the target reference image corresponding to the matching point set with the highest matching degree among the plurality of matching point sets.

[0086] It can be understood that the memory 802 in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM), and direct rambus random access memory (DRRAM). The memory 802 of the systems and methods described herein is intended to include but not be limited to these and any other suitable types of memory.

[0087] The processor 803 may be an integrated circuit chip with the ability to process signals. In the implementation process, the steps of the above method can be completed by the integrated logic circuit in the hardware of the processor 803 or instructions in the form of software. The above-mentioned processor 803 may be a general-purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by a combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor 803 reads the information in the memory 802 and combines its hardware to complete the steps of the above method.

[0088] It can be understood that the embodiments described herein can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more ASICs, DSPs, digital signal processing devices (DSPDs), programmable logic devices (PLDs), FPGAs, general-purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or a combination thereof.

[0089] For software implementation, the techniques described herein can be implemented by modules (such as procedures, functions, etc.) that execute the functions described herein. The software code can be stored in the memory 802 and executed by the processor 803. The memory 802 can be implemented inside or outside the processor 803.

[0090] Specifically, when the processor 803 is further configured to run the computer program, it executes the method steps for detecting the position of the molten liquid surface in the foregoing technical solution, which will not be elaborated herein.

[0091] The present disclosure also provides a computer-readable storage medium storing at least one instruction for being executed by a processor to implement the method for detecting the position of the molten liquid surface as described in the above respective embodiments.

[0092] The present disclosure also provides a computer program product including computer instructions stored in a computer-readable storage medium; a processor of an electronic device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the electronic device executes to implement the method for detecting the position of the molten liquid surface as described in the above respective embodiments.

[0093] Another embodiment of the present disclosure provides a chip including a processor and a communication interface, the communication interface is coupled to the processor, and the processor is configured to run a program or instructions to implement each process of the method embodiment for detecting the position of the molten liquid surface as described above, and can achieve the same technical effect. To avoid repetition, it will not be elaborated here.

[0094] It should be understood that the chip mentioned in the embodiments of the present disclosure may also be referred to as a system-on-chip, system chip, chip system, or system-on-chip, etc.

[0095] In several embodiments provided by the present disclosure, it should be understood that the disclosed systems, devices, servers, and methods may be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other may be through some interfaces, and the indirect couplings or communication connections of devices or units may be in electrical, mechanical, or other forms.

[0096] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0097] In addition, each functional unit in the respective embodiments of the present disclosure may be integrated in one processing unit, or each unit may exist physically alone, or two or more units may be integrated in one unit. The above integrated units may be implemented in the form of hardware or in the form of software functional units.

[0098] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present disclosure. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical discs that can store program codes.

[0099] Those skilled in the art should be able to realize that in the above one or more examples, the functions described in the present disclosure can be implemented by hardware, software, firmware, or any combination thereof. When implemented using software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. The computer-readable medium includes computer storage media and communication media, where the communication media includes any medium that facilitates the transmission of a computer program from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0100] It should be noted that: among the technical solutions recorded in the present disclosure, without conflict, they can be combined arbitrarily.

[0101] As described above, it is only the specific implementation manner of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present disclosure can easily think of changes or substitutions, which should all be covered within the protection scope of the present disclosure.

Claims

1. A method for detecting the position of a melt level, characterized in that: The method for detecting the melt level position comprises: Acquire a liquid surface image to be tested; Matching feature points of the liquid surface image to be tested with each of the multiple reference images, and correcting the matching point pairs according to the geometric relationship between the feature points to obtain multiple matching point sets; The current position of the liquid surface to be measured is determined according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the multiple matching point sets.

2. The method for detecting the melt level position according to claim 1, characterized in that: The correcting of the matching point pairs according to the geometric relationship between the feature points comprises: After completing the matching of a feature point each time, the geometric relationship between the matching point pairs is obtained; The matching point pairs are corrected according to the geometric relationship of the matching point pairs.

3. The method for detecting the melt level position according to claim 1, characterized in that: The correcting of the matching point pairs according to the geometric relationship between the feature points comprises: After completing the matching of all feature points, obtain the geometric relationship of all matching point pairs; According to the geometric relationship of all matching point pairs, the matching point pairs are corrected.

4. The method for detecting the melt level position according to any one of claims 1 to 3, characterized in that: The geometric relationship includes the slope of the matching point pair in the image coordinate system; The correcting of the matching point pairs according to the geometric relationship between the feature points comprises: Among the slope values ​​of the matching point pairs, the matching point pairs corresponding to the outlier slope values ​​are corrected.

5. The method for detecting the melt level position according to claim 1, characterized in that: The step of matching the feature points of the liquid surface image to be measured with each of the multiple reference images comprises: Determining a plurality of key point neighborhoods in the liquid surface image to be measured; Determining a feature point corresponding to a feature vector used to describe the neighborhood of the key point according to a brightness change of each pixel point in the neighborhood of the key point; According to the similarity between the feature points in the liquid surface image to be measured and the feature points in each reference image, the matching point pairs between the liquid surface image to be measured and each reference image are determined.

6. The method for detecting the melt level position according to claim 5, characterized in that: The step of determining a plurality of key point neighborhoods in the liquid surface image to be measured comprises: Performing Gaussian convolution on the liquid surface image to be measured to obtain a three-dimensional image matrix formed by stacking images of different scales; Determine multiple extreme points from the three-dimensional image matrix by Gaussian difference; A preset area centered on each extreme point is used as the key point neighborhood.

7. The method for detecting the melt level position according to claim 1, characterized in that: The step of obtaining the image of the liquid surface to be measured comprises: Acquiring an initial image of the liquid surface to be measured including the reflection of the reference object; The image area outside the reflection of the reference object is cut out from the initial image to obtain the liquid surface image to be measured.

8. The method for detecting the melt level position according to claim 7, characterized in that: The step of cutting out the image area outside the reflection of the reference object from the initial image to obtain the liquid surface image to be measured includes: Based on the brightness values ​​of the pixels, determining the image area corresponding to the reflection of the reference object from the initial image by threshold segmentation; The image area outside the image area corresponding to the reflection of the reference object is cut out to obtain the liquid surface image to be measured.

9. The method for detecting the melt level position according to claim 7, characterized in that: The step of cutting out the image area outside the reflection of the reference object from the initial image to obtain the liquid surface image to be measured includes: Based on edge features, determining an image region corresponding to the reflection of the reference object from the initial image; The image area outside the image area corresponding to the reflection of the reference object is cut out to obtain the liquid surface image to be measured.

10. The method for detecting the melt level position according to claim 1, characterized in that: The step of determining the current position of the liquid surface to be measured according to the liquid surface position of the reference image corresponding to the matching point set with the highest matching degree among the multiple matching point sets comprises: determining the middle position according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the multiple matching point sets; Determine a first centroid of a curve formed by matching points in the liquid surface image to be measured, and a second centroid of a curve formed by matching points in the target reference image; The intermediate position is corrected according to the position difference between the first center of gravity and the second center of gravity to obtain the current position.

11. The method for detecting the melt level position according to claim 1, characterized in that: The method for detecting the melt level position also includes: By using a random sampling consensus algorithm, mismatched point pairs in the plurality of matching point sets are eliminated.

12. A device for detecting the position of a melt surface, characterized in that: The device for detecting the position of the melt level comprises: an acquisition module, a matching and correction module, and a determination module; The acquisition module is configured to acquire an image of the liquid surface to be measured; The matching and correction module is configured to match the liquid surface image to be tested with each of the multiple reference images for feature points, and correct the matching point pairs according to the geometric relationship between the feature points to obtain multiple matching point sets; The determination module is configured to determine the current position of the liquid surface to be measured according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the multiple matching point sets.

13. A device for detecting the position of a melt surface, characterized in that: The device for detecting the position of the melt level is applied to a single crystal furnace, and the device for detecting the position of the melt level comprises: A reference object, a CCD camera, a memory and a processor are arranged above the liquid level to be measured in the single crystal furnace; in, The memory is used to store a computer program that can be run on the processor; The CCD camera is used to obtain the image of the liquid surface to be measured; The processor is configured to perform the following steps when running the computer program: Matching feature points of the liquid surface image to be tested with each of the multiple reference images, and correcting the matching point pairs according to the geometric relationship between the feature points to obtain multiple matching point sets; The current position of the liquid surface to be measured is determined according to the liquid surface position of the target reference image corresponding to the matching point set with the highest matching degree among the multiple matching point sets.

14. A computer-readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by the processor, the steps of the method for detecting the melt level position according to any one of claims 1 to 11 are implemented.

Images