CN112329687A - Automatic recognition method and system for dislocation of silicon carbide substrate - Google Patents
Automatic recognition method and system for dislocation of silicon carbide substrate Download PDFInfo
- Publication number
- CN112329687A CN112329687A CN202011279607.XA CN202011279607A CN112329687A CN 112329687 A CN112329687 A CN 112329687A CN 202011279607 A CN202011279607 A CN 202011279607A CN 112329687 A CN112329687 A CN 112329687A
- Authority
- CN
- China
- Prior art keywords
- dislocation
- silicon carbide
- carbide substrate
- neural network
- network model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/69—Microscopic objects, e.g. biological cells or cellular parts
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The application discloses a method and a system for automatically identifying dislocation of a silicon carbide substrate, which are used for solving the technical problems that the existing dislocation identification method cannot automatically shoot and record and further cannot automatically identify dislocation of the silicon carbide substrate. The method comprises the following steps: automatically scanning the silicon carbide substrate to be detected through a microscope, and automatically photographing and recording a scanning area; receiving a dislocation image which is sent by a microscope and is related to a silicon carbide substrate to be detected through computer equipment, and inputting the dislocation image into a silicon carbide substrate dislocation identification neural network model; identifying the dislocation in the silicon carbide substrate to be detected based on the silicon carbide substrate dislocation identification neural network model; and displaying the dislocation information output by the silicon carbide substrate dislocation identification neural network model on a computer device. According to the method, the automatic photographing record of the dislocation scanning area of the silicon carbide substrate is realized, and the process of automatically identifying the dislocation of the silicon carbide substrate is further realized.
Description
Technical Field
The application relates to the technical field of semiconductor materials, in particular to a method and a system for automatically identifying dislocation of a silicon carbide substrate.
Background
At present, the silicon carbide crystal industrial production mostly adopts a Physical Vapor Transport (PVT) method, and because the growth condition requirement of the silicon carbide crystal is higher, the defects introduced in the growth process limit the improvement and further application and development of the performance of the silicon carbide crystal. Therefore, it is very important to identify and characterize the defects of the silicon carbide substrate during the growth process.
Dislocations, one of the most common defects in the growth of silicon carbide crystals. The existing method for identifying the dislocation defects of the silicon carbide substrate has the problems that automatic recording of a scanning area cannot be realized, and further automatic identification of the dislocation of the silicon carbide substrate cannot be realized.
Disclosure of Invention
The embodiment of the application provides a method and a system for automatically identifying dislocation of a silicon carbide substrate, which are used for solving the technical problems that the existing method for identifying dislocation of the silicon carbide substrate cannot automatically shoot and record and further cannot automatically identify.
In one aspect, an embodiment of the present application provides a method for automatically identifying dislocations of a silicon carbide substrate, including: automatically scanning the silicon carbide substrate to be detected through a microscope, and automatically photographing and recording a scanning area; receiving a dislocation image which is sent by a microscope and is related to a silicon carbide substrate to be detected through computer equipment, and inputting the dislocation image into a silicon carbide substrate dislocation identification neural network model; identifying the dislocation in the silicon carbide substrate to be detected based on the silicon carbide substrate dislocation identification neural network model; and displaying the dislocation information output by the silicon carbide substrate dislocation identification neural network model on a computer device.
According to the automatic dislocation identification method for the silicon carbide substrate, the silicon carbide substrate to be detected is automatically scanned through the microscope, and the automatic photographing and recording function of the scanning area of the silicon carbide substrate to be detected is realized by matching with the fixed-frequency photographing function of the microscope. The dislocation image shot by the microscope is identified through the silicon carbide substrate dislocation identification neural network model on the computer equipment, the participation of detection personnel is avoided, and the automatic proceeding of the silicon carbide substrate dislocation identification process is realized.
In one implementation of the present application, before the microscope automatically scans the silicon carbide substrate to be tested, the method further includes: determining a silicon carbide substrate to be detected; corroding the silicon carbide substrate to be detected by using molten hot alkali liquor to obtain the silicon carbide substrate to be detected; the silicon carbide substrate to be tested comprises a plurality of silicon carbide substrate dislocation corrosion pits.
In the embodiment of the application, the silicon carbide substrate to be detected is corroded by molten hot alkali liquor, so that dislocation defects contained in the silicon carbide substrate are displayed in a corrosion pit mode. Different dislocation types correspond to different etch pit shapes, so that the dislocation is detected in the form of the etch pits, the dislocation type can be identified conveniently in the later period, and the convenience and the feasibility of the identification method provided by the embodiment of the application are ensured.
In an implementation of this application, carry out the automatic scanning to the carborundum substrate that awaits measuring through the microscope to carry out the record of shooing voluntarily to the scanning area, specifically include: the microscope objective table drives the silicon carbide substrate to be detected to perform fixed-frequency movement at a first preset frequency through the microscope objective table movement auxiliary device; the microscope objective table moving auxiliary device mainly comprises a first connecting piece, a second connecting piece, a corner speed reducer, a driving motor and a touch display screen; taking a fixed-frequency photograph of the silicon carbide substrate to be measured at a second preset frequency through a microscope; the first preset frequency is equal to the second preset frequency.
In one implementation of the present application, one end of the first connecting member includes a limiting portion for mating with a knob of a microscope stage; the other end of the first connecting piece comprises a plug part which is used for being matched with the second connecting piece; the limiting part comprises a limiting hole and a limiting groove; the limiting groove is arranged on the inner surface of the limiting part and used for limiting a knob of the microscope objective table in the inserting direction; the limiting hole is arranged on the outer surface of the limiting part and used for limiting a knob of the microscope objective table in the direction perpendicular to the inserting direction; the insertion part comprises a limiting column; the limiting column is arranged on the outer surface of the insertion part and is used for being matched with the second connecting piece.
In one implementation of the present application, one end of the second connector has a mating insertion part for mating with the first connector; the other end of the second connecting piece is provided with a through hole which is used for being matched with an output shaft of the corner speed reducer; the matching insertion part comprises a limiting groove and a limiting hole; the limiting groove is arranged on the inner surface of the matching insertion part and used for limiting the first connecting piece in the insertion direction; the limiting hole is arranged on the outer surface of the matching insertion part and used for limiting the first connecting piece in the direction perpendicular to the insertion direction; the end, provided with the through hole, of the second connecting piece comprises a limiting hole used for limiting an output shaft of the corner speed reducer in the vertical direction of the inserting direction.
According to the automatic identification method for dislocation of the silicon carbide substrate, the object stage is driven to move through the microscope object stage movement auxiliary device, so that the silicon carbide substrate to be detected is driven to move in a fixed-film mode, the fixed-film photographing function of a microscope is matched, and automatic photographing recording of a scanning area of the silicon carbide substrate to be detected can be achieved.
In one implementation of the present application, prior to inputting the dislocation image into the silicon carbide substrate dislocation identification neural network model, the method further comprises: acquiring a plurality of image data related to the dislocation corrosion pits of the silicon carbide substrate; marking the dislocation corrosion pits of the silicon carbide substrate in the image data in the form of a marking frame; constructing a training data set of the neural network model based on the labeled image data; and inputting the training data set into the neural network model, and training until the output converges to obtain the silicon carbide substrate dislocation recognition neural network model.
In an implementation manner of the present application, identifying dislocations in a silicon carbide substrate to be detected based on a silicon carbide substrate dislocation identification neural network model specifically includes: inputting the dislocation image into the silicon carbide substrate dislocation recognition neural network model through an input layer of the silicon carbide substrate dislocation recognition neural network model; identifying the convolution layer of the neural network model through the dislocation of the silicon carbide substrate to obtain a characteristic image corresponding to the dislocation image; inputting the characteristic image into a pooling layer for pooling, and inputting the pooled characteristic image into a classification layer through a full connection layer; and fitting the characteristic image output by the full connection layer with a pre-stored defect image through the classification layer to determine a fitting score value corresponding to the characteristic image.
In one implementation manner of the present application, after determining the fitting score value corresponding to the feature image, the method further includes: determining the size relation between the fitting score value and a preset threshold value; and under the condition that the fitting score value is larger than a preset threshold value, determining that the dislocation type corresponding to the dislocation image is consistent with the dislocation type corresponding to the defect image.
In one implementation of the present application, the dislocation information includes at least any one or more of: dislocation type, dislocation position, number of dislocations, dislocation density, and size of dislocation etch pits.
According to the automatic identification method for the dislocation of the silicon carbide substrate, the type of each dislocation in the silicon carbide substrate to be detected is determined through the neural network model, and the accuracy of automatic identification of the dislocation of the silicon carbide substrate is guaranteed. In addition, the output of the neural network model can be adjusted to realize the classified statistics of the dislocation of the silicon carbide substrate to be detected, so that detection personnel can conveniently determine the performance of the silicon carbide substrate to be detected directly according to the output data, and the practicability of the automatic identification method for the dislocation of the silicon carbide substrate provided by the embodiment of the application is ensured.
On the other hand, the embodiment of the present application further provides an automatic dislocation identification system for a silicon carbide substrate, including: microscopes, computer equipment; the microscope is used for automatically scanning the silicon carbide substrate to be detected and automatically photographing and recording a scanning area; the computer equipment is used for receiving dislocation images which are sent by the microscope and are related to the silicon carbide substrate to be detected, and is used for inputting the dislocation images into the silicon carbide substrate dislocation identification neural network model; the computer equipment is also used for identifying the dislocation in the silicon carbide substrate to be detected based on the silicon carbide substrate dislocation identification neural network model, and displaying dislocation information output by the silicon carbide substrate dislocation identification neural network model.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
fig. 1 is a flowchart of a method for automatically identifying dislocations in a silicon carbide substrate according to an embodiment of the present disclosure;
fig. 2 is a schematic overall structure diagram of a microscope stage movement assisting device according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of a first connecting element according to an embodiment of the present disclosure;
FIG. 4 is a schematic diagram of a second connecting member according to an embodiment of the present disclosure;
FIG. 5 is a graph illustrating the results of annotation of image data relating to dislocation etch pits in a silicon carbide substrate according to an embodiment of the present application;
FIG. 6 is a schematic diagram of an output result of a silicon carbide substrate dislocation recognition neural network model provided in an embodiment of the present application;
fig. 7 is a schematic structural diagram of an automatic dislocation identification system for a silicon carbide substrate according to an embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the technical solutions of the present application will be described in detail and completely with reference to the following specific embodiments of the present application and the accompanying drawings. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Silicon carbide crystals are widely concerned due to their excellent semi-insulating properties, and particularly for high-power semiconductor devices with special requirements, silicon carbide crystals are potential materials of choice for these devices due to their high temperature, high frequency, high power, and other characteristics.
At present, the silicon carbide crystal is produced by adopting a Physical Vapor Transport (PVT) method in the industrial production, but the improvement of the performance and the further application and development of the silicon carbide crystal are limited by defects introduced in the growth process due to the higher requirement of the growth condition of the silicon carbide crystal. Therefore, the characterization and statistics of the defects of the silicon carbide substrate in the growth process become the first prerequisite for improving the defects. Dislocations, which are a type of line defect, can be classified into Threading Edge Dislocations (TED), Threading Screw Dislocations (TSD), and Basal Plane Dislocations (BPD) according to their formation mechanism and the resulting difference in the half atomic planes. The magnitude of the different dislocations and their density also have different effects on subsequent epitaxial growth. Therefore, accurately distinguishing the various types of dislocations is critical to determining the quality of the silicon carbide substrate. In order to characterize the growth quality of the silicon carbide substrate, the surface thereof is often carefully surface-scanned using an optical microscope. However, automatic photographing and recording cannot be performed in the surface scanning process, and further automatic identification of dislocation of the silicon carbide substrate cannot be realized, so that a tester needs to focus on an eyepiece of a microscope or an external display thereof, which brings great inconvenience to the tester.
The embodiment of the application provides a method and a system for automatically identifying dislocation of a silicon carbide substrate, wherein the silicon carbide substrate is automatically scanned through a microscope, and automatic photographing record is realized by matching with a fixed-frequency photographing function of the microscope; and detecting the image shot by the microscope through the silicon carbide substrate dislocation identification neural network model, determining the dislocation therein, and realizing the automatic identification process of the silicon carbide substrate dislocation. The method solves the technical problems that automatic photographing recording cannot be realized in the existing silicon carbide substrate detection method, and further automatic detection of dislocation of the silicon carbide substrate cannot be realized.
The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.
Fig. 1 is a flowchart of a method for automatically identifying dislocations in a silicon carbide substrate according to an embodiment of the present application. As shown in fig. 1, the method for automatically identifying dislocations on a silicon carbide substrate according to the embodiment of the present application mainly includes the following steps:
and 101, automatically photographing and recording the silicon carbide substrate to be detected through a microscope to obtain a dislocation image.
According to the automatic dislocation identification method for the silicon carbide substrate, the silicon carbide substrate to be detected is obtained firstly, then the microscope designed and provided in the embodiment of the application drives the silicon carbide substrate to be detected to move on the microscope objective table at a fixed frequency, and the automatic photographing record of the silicon carbide substrate to be detected is realized by matching with the fixed frequency photographing function of the microscope, so that a dislocation image corresponding to the silicon carbide substrate to be detected is obtained.
As will be apparent to those skilled in the art, the silicon carbide substrate is of great importance in the growth of silicon carbide crystals, and dislocation defects on the silicon carbide substrate directly affect the properties of the grown silicon carbide crystal.
Specifically, firstly, determining a silicon carbide substrate to be detected, and then corroding the silicon carbide substrate to be detected through molten hot alkali liquor so that different types of dislocations on the silicon carbide substrate are displayed in different forms of corrosion pits to obtain the detected silicon carbide substrate, so as to facilitate subsequent detection and identification. It should be noted that the etch pits appear as follows, because the different types of dislocations have different mechanisms of formation and different resulting half atomic planes, and also have different bernoulli vectors: threading Edge Dislocation (TED) is a small rounded etch pit, Threading Screw Dislocation (TSD) is a large hexagonal etch pit and Basal Plane Dislocation (BPD) is a small drop etch pit.
Further, after the silicon carbide substrate to be detected is obtained, the silicon carbide substrate to be detected is placed on a microscope stage for detection. It should be noted that the placing of the silicon carbide substrate to be tested on the microscope stage may be implemented by the existing operation manual or operation method, which is not described herein in detail in the embodiments of the present application. Then, the microscope movement auxiliary device provided in the embodiment of the application is installed on a microscope, so that the microscope objective table is driven to move at a first preset frequency by the microscope objective table movement auxiliary device, and further the silicon carbide substrate to be detected on the objective table is driven to move at a fixed frequency at the first preset frequency; and matching with the fixed-frequency photographing function of the microscope, namely performing fixed-frequency photographing on the scanning area (the current detection area corresponding to the silicon carbide substrate to be detected) of the silicon carbide substrate to be detected at a second preset frequency through the microscope so as to realize automatic photographing record of the scanning area of the silicon carbide substrate to be detected. It should be noted that, in order to enable the microscope to take a picture of the entire region of the silicon carbide substrate to be measured, in the embodiment of the present application, the first preset frequency and the second preset frequency are set to be the same value, so that each time the microscope takes a picture, different scanning regions corresponding to the silicon carbide substrate to be measured can be taken.
In one embodiment of the application, the microscope stage movement assisting device mainly comprises a first connecting piece, a second connecting piece, a corner speed reducer, a driving motor and a touch display screen. The overall structure is shown in fig. 2. Fig. 2 is a schematic view of an overall structure of a microscope stage movement assisting device according to an embodiment of the present disclosure. As shown in fig. 2, the first connecting member 2 of the microscope stage movement assisting device provided in the embodiment of the present application is used for connecting a knob and a second connecting member 3 of the microscope stage, the second connecting member 3 is used for connecting a corner speed reducer 4, the corner speed reducer 4 is connected to a driving motor 5, and the driving motor 5 is connected to a touch display screen 7 through a wire. Further, can be through the rotational speed of 7 control driving motor 5 of touch display screen in this application embodiment, then rotate the level of driving motor 5 through corner speed reducer 4 and convert vertical rotation into, the adjustment rotational speed simultaneously, through second connecting piece 3, the knob that first connecting piece 2 drove microscope objective table rotates, and then realizes the fixed frequency of microscope objective table and remove.
The structure of the first connecting member 2 in the embodiment of the present application is shown in fig. 3, and the structure of the second connecting member 3 is shown in fig. 4.
Fig. 3 is a schematic structural diagram of a first connecting element according to an embodiment of the present disclosure. As shown in fig. 3, one end of the first connecting member 2 includes a limiting portion for cooperating with a knob of the microscope stage; the other end of the first connector 2 comprises a plug-in part for cooperation with the second connector 3. Furthermore, the limiting part comprises a limiting hole 2-2 and a limiting groove 2-1; the limiting groove 2-1 is arranged on the inner surface of the limiting part and used for limiting a knob of the microscope objective table in the inserting direction; the limiting hole 2-2 is arranged on the outer surface of the limiting part and used for limiting a knob of the microscope objective table in the direction perpendicular to the inserting direction. The inserting part comprises a limiting column 2-3 which is arranged on the outer surface of the inserting part and is used for being matched with the second connecting piece 3. As can be seen from fig. 3, the semi-cylindrical recess of the position-limiting groove 2-1 of the first connecting member 2 can cooperate with the semi-cylindrical protrusion of the knob of the microscope stage to achieve position limitation. Meanwhile, the two limiting holes 2-2 are circular through holes and are symmetrically arranged on the first connecting piece 2, and the knob of the microscope objective table can be further limited in the opposite direction through the jackscrews, so that the firmness of connection between the first connecting piece and the knob of the microscope objective table is ensured.
It should be noted that, in the embodiment of the present application, the shape of the limiting groove on the first connecting member is set to match with the knob of the microscope stage, and may be appropriately adjusted according to the protruding columnar shape of the outer surface of the knob, as long as the limiting groove can be matched with the knob of the microscope stage and can limit the knob of the microscope stage, which is not limited in the embodiment of the present application.
Fig. 4 is a schematic structural diagram of a second connecting member according to an embodiment of the present disclosure. As shown in fig. 4, one end of the second connector 3 has a mating insertion part for mating with the first connector; the other end of the second connecting piece 3 is provided with a through hole which is used for being matched with an output shaft of the corner speed reducer 4. Further, the matching insertion part comprises a limiting groove 3-3 and a limiting hole 3-2; the limiting groove 3-3 is arranged on the inner surface of the matching insertion part and used for limiting the first connecting piece 2 in the insertion direction; the limiting hole 3-2 is arranged on the outer surface of the matching insertion part and used for limiting the first connecting piece in the direction perpendicular to the insertion direction. One end of the second connecting piece 3, which is provided with a through hole, comprises a limiting hole 3-1 used for limiting the output shaft of the corner speed reducer 4 in the direction perpendicular to the plugging direction. Specifically, the limiting groove 3-3 on the second connecting piece 3 is used for matching with the limiting post 2-3 on the first connecting piece 2, so that the first connecting piece 2 is connected with the second connecting piece 3, and the jackscrew is inserted through the limiting hole 3-2 on the second connecting piece 3 to limit the first connecting piece 2, and the firmness of the connection relationship between the first connecting piece 2 and the second connecting piece 3 is ensured. And the through-hole on the second connecting piece 3 can cooperate with the output shaft of the corner speed reducer 4 to the corner speed reducer 4 drives the second connecting piece 3 and the first connecting piece 2 to rotate, and then drives the knob of the microscope objective table to rotate, so that the fixed-frequency movement of the microscope objective table is realized. In addition, the limiting hole 3-1 on the second connecting piece 3 can also limit the output shaft of the corner speed reducer 4 in the direction perpendicular to the plugging direction, so as to ensure the connection firmness and avoid the situation that the output shaft of the corner speed reducer 4 cannot drive the second connecting piece 3 to rotate.
This application embodiment removes auxiliary device through above-mentioned microscope objective table, drives the carborundum substrate that awaits measuring on the microscope objective table and removes, and the function of shooing is frequently decided to the cooperation microscope realizes the record of shooing the automation of the carborundum substrate scanning area that awaits measuring, obtains the dislocation image that the carborundum substrate that awaits measuring corresponds.
And 102, inputting the dislocation image into a silicon carbide substrate dislocation identification neural network model, and identifying the dislocation in the silicon carbide substrate to be detected to obtain dislocation information.
And after the dislocation image corresponding to the silicon carbide substrate to be detected is obtained, outputting the dislocation image to a silicon carbide substrate dislocation identification neural network model so as to identify the dislocation on the silicon carbide substrate to be detected and obtain dislocation information.
In one embodiment of the present application, the neural network model first needs to be trained before inputting the dislocation images into the silicon carbide substrate dislocation recognition neural network model. Specifically, a plurality of image data related to the silicon carbide substrate dislocation corrosion pits are obtained and marked, the corrosion pits in the image data are marked by using a marking frame, and the dislocation types corresponding to the corrosion pits are marked. It should be noted that, in the embodiment of the present application, the silicon carbide substrate to be detected is corroded by the molten hot alkali solution, and the dislocations on the silicon carbide substrate are already displayed in the form of corrosion pits, so when the silicon carbide substrate to be detected is detected, the obtained dislocation image is also displayed in the form of corrosion pits. Therefore, in the embodiment of the application, the image related to the dislocation etching pits after the silicon carbide substrate is etched is not a pure silicon carbide substrate image, so that the effectiveness of training a neural network model is ensured. It should be further noted that, in the embodiment of the present application, labeling of image data may be implemented by labellmg software.
The labeling result of the acquired image data is shown in fig. 5. Fig. 5 is a schematic diagram illustrating the labeling result of image data related to dislocation etching pits of a silicon carbide substrate according to an embodiment of the present application. As shown in fig. 5, each etch pit is marked in the form of a mark frame, and different types of dislocations corresponding to etch pits of different shapes are explained.
Further, after the image data is labeled, a training data set is constructed based on the labeled image, and the training data set is input into the neural network model for training; training until the output converges to obtain the silicon carbide substrate dislocation recognition neural network model.
Further, after obtaining the silicon carbide substrate dislocation identification neural network model, inputting a dislocation image obtained through a microscope into the model to identify the dislocation on the silicon carbide substrate to be measured. Specifically, an input layer of a neural network model is identified through dislocation of a silicon carbide substrate, and dislocation images are input into the model; and then, obtaining a characteristic image corresponding to the dislocation image through the convolution layer, performing pooling treatment through the pooling layer, and entering the classification layer through the full-connection layer. In the classification layer, the feature images corresponding to the dislocation images are fitted with feature images corresponding to pre-stored defect images (images in a training data set), and a fitting score value is determined. And under the condition that the fitting score value is larger than a preset threshold value, determining that the dislocation type in the dislocation image is consistent with the dislocation type in the defect image, and further determining the dislocation type corresponding to each etch pit in the dislocation image. It should be noted that, since the silicon carbide substrate to be measured may have different corrosion degrees corresponding to different dislocations during corrosion by a molten hot alkali solution, such as a molten hot potassium hydroxide solution, when the fitting score value (1 is completely similar, and 0 is completely different) is greater than 0.9, it is considered to be consistent with the dislocation type in the defect image.
The output results of the silicon carbide substrate dislocation recognition neural network model in the embodiment of the present application are shown in fig. 6. Fig. 6 is a schematic diagram of an output result of a silicon carbide substrate dislocation identification neural network model according to an embodiment of the present application. As shown in fig. 6, the output result of the silicon carbide substrate dislocation identification neural network model is in the form of an image, and each etch pit on the output image contains a mark frame, and each mark frame contains a dislocation type corresponding to the etch pit. Meanwhile, the statistical value of each type of dislocation is also included in the output image.
And step 103, displaying the dislocation information on the computer equipment.
And after the dislocation recognition in the dislocation image is completed by the silicon carbide substrate dislocation neural network model, displaying the output image on a computer device. Meanwhile, dislocation information can be output and displayed on computer equipment.
In one embodiment of the present application, the dislocation information includes at least any one or more of: dislocation type, number of dislocations, dislocation position, dislocation density, and size of dislocation etch pits. It should be noted that the output result of the silicon carbide substrate dislocation identification neural network model in the embodiment of the present application may be adjusted according to actual requirements, for example, dislocation density information may also be output through the model, and the statistical result is output in Excel, which greatly facilitates statistics of dislocations of a silicon carbide substrate to be detected, and also facilitates detection personnel to quickly determine the defect degree of the silicon carbide substrate to be detected.
According to the automatic dislocation identification method for the silicon carbide substrate, provided by the embodiment of the application, the dislocation of the silicon carbide substrate can be exposed in the form of a dislocation pit by carrying out molten hot alkali corrosion on the silicon carbide substrate, so that the detection and identification are facilitated; through the microscope objective table movement auxiliary device of design, can realize driving the carborundum substrate under optical microscope and carry out the frequency movement of deciding, the function of shooing is frequently decided to the cooperation microscope to realize shooing and record in a large scale fast. And then, the machine learning neural network model is utilized to automatically identify the dislocations of different types of the silicon carbide substrate, so that the rapid identification and classification statistics of the dislocations of the silicon carbide substrate can be realized. In summary, the identification method in the embodiment can greatly save the labor cost and the dislocation identification and classification time, and realize dislocation identification, classification and statistics on the whole surface of the silicon carbide substrate.
The above is an embodiment of the method in the present application, and based on the same inventive concept, the embodiment of the present application further provides an automatic dislocation identification system for a silicon carbide substrate, and the structure of the automatic dislocation identification system is shown in fig. 7.
Fig. 7 is a schematic structural diagram of an automatic dislocation identification system for a silicon carbide substrate according to an embodiment of the present application. As shown in fig. 7, the system includes a microscope 702, a computer device 701; the microscope 702 is used for automatically scanning the silicon carbide substrate to be detected and automatically photographing and recording a scanning area; the computer device 701 is used for receiving a dislocation image which is sent by the microscope 702 and related to the silicon carbide substrate to be detected, and inputting the dislocation image into the silicon carbide substrate dislocation identification neural network model; the computer device 701 is further configured to identify dislocations in the silicon carbide substrate to be detected based on the silicon carbide substrate dislocation identification neural network model, and to display dislocation information output by the silicon carbide substrate dislocation identification neural network model.
The embodiments in the present application are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, as for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The above description is only an example of the present application and is not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A method for automatically identifying dislocations in a silicon carbide substrate, the method comprising:
automatically scanning the silicon carbide substrate to be detected through a microscope, and automatically photographing and recording a scanning area;
receiving a dislocation image which is sent by a microscope and is related to a silicon carbide substrate to be detected through computer equipment, and inputting the dislocation image into a silicon carbide substrate dislocation identification neural network model;
identifying the dislocation in the silicon carbide substrate to be detected based on a silicon carbide substrate dislocation identification neural network model;
and displaying the dislocation information output by the silicon carbide substrate dislocation identification neural network model on computer equipment.
2. The method for automatically identifying dislocation of a silicon carbide substrate as claimed in claim 1, wherein before the microscope automatically scans the silicon carbide substrate to be tested, the method further comprises:
determining a silicon carbide substrate to be detected;
corroding the silicon carbide substrate to be detected by using molten hot alkali liquor to obtain the silicon carbide substrate to be detected; the silicon carbide substrate to be tested comprises a plurality of silicon carbide substrate dislocation corrosion pits.
3. The method for automatically identifying dislocation of a silicon carbide substrate according to claim 1, wherein the silicon carbide substrate to be tested is automatically scanned through a microscope, and the scanning area is automatically photographed and recorded, and the method specifically comprises the following steps:
the microscope objective table drives the silicon carbide substrate to be detected to perform fixed-frequency movement at a first preset frequency through the microscope objective table movement auxiliary device; the microscope objective table moving auxiliary device mainly comprises a first connecting piece, a second connecting piece, a corner speed reducer, a driving motor and a touch display screen;
taking a fixed-frequency photograph of the silicon carbide substrate to be tested at a second preset frequency through the microscope; wherein the first preset frequency is equal to the second preset frequency.
4. The automatic identification method for dislocation of silicon carbide substrate as claimed in claim 3, wherein one end of said first connecting member includes a stopper for engaging with a knob of said microscope stage; the other end of the first connecting piece comprises a plug-in part which is used for being matched with the second connecting piece;
the limiting part comprises a limiting hole and a limiting groove; the limiting groove is arranged on the inner surface of the limiting part and used for limiting a knob of the microscope objective table in the inserting direction; the limiting hole is arranged on the outer surface of the limiting part and used for limiting a knob of the microscope objective table in the direction perpendicular to the inserting direction;
the insertion part comprises a limiting column; the limiting column is arranged on the outer surface of the inserting part and is used for being matched with the second connecting piece.
5. The automatic recognition method of dislocation of silicon carbide substrate as claimed in claim 3, wherein one end of said second connection member has a fitting insertion part for fitting with said first connection member; the other end of the second connecting piece is provided with a through hole which is used for being matched with an output shaft of the corner speed reducer;
the matching insertion part comprises a limiting groove and a limiting hole; the limiting groove is arranged on the inner surface of the matching insertion part and used for limiting the first connecting piece in the insertion direction; the limiting hole is arranged on the outer surface of the matching insertion part and used for limiting the first connecting piece in the direction perpendicular to the insertion direction;
and one end of the second connecting piece, which is provided with the through hole, comprises a limiting hole, and the limiting hole is used for limiting the output shaft of the corner speed reducer in the vertical direction of the plugging direction.
6. The automatic identification method of dislocation of silicon carbide substrate as claimed in claim 1, wherein before inputting the dislocation image into a silicon carbide substrate dislocation identification neural network model, the method further comprises:
acquiring a plurality of image data related to the dislocation corrosion pits of the silicon carbide substrate;
marking the dislocation corrosion pits of the silicon carbide substrate in the image data in a form of a marking frame;
constructing a training data set of the neural network model based on the labeled image data;
and inputting the training data set into a neural network model, and training until the output converges to obtain the silicon carbide substrate dislocation recognition neural network model.
7. The method for automatically identifying dislocations in a silicon carbide substrate according to claim 1, wherein identifying dislocations in the silicon carbide substrate to be detected based on a silicon carbide substrate dislocation identification neural network model specifically comprises:
inputting the dislocation image into a silicon carbide substrate dislocation recognition neural network model through an input layer of the silicon carbide substrate dislocation recognition neural network model;
identifying a convolution layer of a neural network model through the dislocation of the silicon carbide substrate to obtain a characteristic image corresponding to the dislocation image;
inputting the characteristic image into a pooling layer for pooling, and inputting the pooled characteristic image into a classification layer through a full connection layer;
and fitting the characteristic image output by the full connection layer with a pre-stored defect image through a classification layer to determine a fitting score value corresponding to the characteristic image.
8. The method for automatically identifying dislocations in a silicon carbide substrate as claimed in claim 7, wherein after determining the corresponding fit score value of the characteristic image, the method further comprises:
determining the size relation between the fitting score value and a preset threshold value;
and determining that the dislocation type corresponding to the dislocation image is consistent with the dislocation type corresponding to the defect image under the condition that the fitting score value is larger than a preset threshold value.
9. The method for automatically identifying dislocations of a silicon carbide substrate as claimed in claim 1, wherein the dislocation information includes at least one or more of: dislocation type, dislocation position, number of dislocations, dislocation density, and size of dislocation etch pits.
10. An automatic dislocation identification system for a silicon carbide substrate, the system comprising: microscopes, computer equipment;
the microscope is used for automatically scanning the silicon carbide substrate to be detected and automatically photographing and recording a scanning area;
the computer equipment is used for receiving dislocation images which are sent by a microscope and are related to the silicon carbide substrate to be detected, and inputting the dislocation images into a silicon carbide substrate dislocation identification neural network model;
the computer equipment is also used for identifying the dislocation in the silicon carbide substrate to be detected based on the silicon carbide substrate dislocation identification neural network model, and displaying the dislocation information output by the silicon carbide substrate dislocation identification neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011279607.XA CN112329687B (en) | 2020-11-16 | 2020-11-16 | Automatic recognition method and system for dislocation of silicon carbide substrate |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011279607.XA CN112329687B (en) | 2020-11-16 | 2020-11-16 | Automatic recognition method and system for dislocation of silicon carbide substrate |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112329687A true CN112329687A (en) | 2021-02-05 |
| CN112329687B CN112329687B (en) | 2022-10-18 |
Family
ID=74318716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011279607.XA Active CN112329687B (en) | 2020-11-16 | 2020-11-16 | Automatic recognition method and system for dislocation of silicon carbide substrate |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112329687B (en) |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2359708Y (en) * | 1999-02-12 | 2000-01-19 | 罗特军 | Automatic driving device for optical microscope stage |
| CN2679703Y (en) * | 2003-12-31 | 2005-02-16 | 石家庄钢铁股份有限公司 | Auto scanner of microscope |
| CN203904356U (en) * | 2014-01-09 | 2014-10-29 | 程涛 | Intelligent food bacteria detection device based on microscopic vision |
| CN105005140A (en) * | 2015-07-28 | 2015-10-28 | 苏州南光电子科技有限公司 | Microscope object stage specimen fixator automatic control system |
| CN107356606A (en) * | 2017-05-27 | 2017-11-17 | 中国电子科技集团公司第十三研究所 | The detection method of semiconductor wafer dislocation density |
| CN107451997A (en) * | 2017-07-31 | 2017-12-08 | 南昌航空大学 | A kind of automatic identifying method of the welding line ultrasonic TOFD D scanning defect types based on deep learning |
| CN108169228A (en) * | 2017-11-28 | 2018-06-15 | 中国工程物理研究院电子工程研究所 | A kind of method of accurate discrimination single-crystal silicon carbide Types of Dislocations |
| CN110222681A (en) * | 2019-05-31 | 2019-09-10 | 华中科技大学 | A kind of casting defect recognition methods based on convolutional neural networks |
| CN110796631A (en) * | 2018-07-25 | 2020-02-14 | Fei 公司 | Training artificial neural networks using simulated sample images |
| CN110889823A (en) * | 2019-10-08 | 2020-03-17 | 山东天岳先进材料科技有限公司 | SiC defect detection method and system |
| CN111238910A (en) * | 2020-01-15 | 2020-06-05 | 山东天岳先进材料科技有限公司 | Dislocation identification method of silicon carbide crystal |
| CN111340787A (en) * | 2020-02-28 | 2020-06-26 | 广东电网有限责任公司 | Method and device for detecting and identifying defects of conducting wire of power transmission line and computer equipment |
| CN111652846A (en) * | 2020-04-30 | 2020-09-11 | 成都数之联科技有限公司 | Semiconductor defect identification method based on characteristic pyramid convolution neural network |
-
2020
- 2020-11-16 CN CN202011279607.XA patent/CN112329687B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN2359708Y (en) * | 1999-02-12 | 2000-01-19 | 罗特军 | Automatic driving device for optical microscope stage |
| CN2679703Y (en) * | 2003-12-31 | 2005-02-16 | 石家庄钢铁股份有限公司 | Auto scanner of microscope |
| CN203904356U (en) * | 2014-01-09 | 2014-10-29 | 程涛 | Intelligent food bacteria detection device based on microscopic vision |
| CN105005140A (en) * | 2015-07-28 | 2015-10-28 | 苏州南光电子科技有限公司 | Microscope object stage specimen fixator automatic control system |
| CN107356606A (en) * | 2017-05-27 | 2017-11-17 | 中国电子科技集团公司第十三研究所 | The detection method of semiconductor wafer dislocation density |
| CN107451997A (en) * | 2017-07-31 | 2017-12-08 | 南昌航空大学 | A kind of automatic identifying method of the welding line ultrasonic TOFD D scanning defect types based on deep learning |
| CN108169228A (en) * | 2017-11-28 | 2018-06-15 | 中国工程物理研究院电子工程研究所 | A kind of method of accurate discrimination single-crystal silicon carbide Types of Dislocations |
| CN110796631A (en) * | 2018-07-25 | 2020-02-14 | Fei 公司 | Training artificial neural networks using simulated sample images |
| CN110222681A (en) * | 2019-05-31 | 2019-09-10 | 华中科技大学 | A kind of casting defect recognition methods based on convolutional neural networks |
| CN110889823A (en) * | 2019-10-08 | 2020-03-17 | 山东天岳先进材料科技有限公司 | SiC defect detection method and system |
| CN111238910A (en) * | 2020-01-15 | 2020-06-05 | 山东天岳先进材料科技有限公司 | Dislocation identification method of silicon carbide crystal |
| CN111340787A (en) * | 2020-02-28 | 2020-06-26 | 广东电网有限责任公司 | Method and device for detecting and identifying defects of conducting wire of power transmission line and computer equipment |
| CN111652846A (en) * | 2020-04-30 | 2020-09-11 | 成都数之联科技有限公司 | Semiconductor defect identification method based on characteristic pyramid convolution neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112329687B (en) | 2022-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2017215241A1 (en) | Circuit board element missing detection method and system | |
| Liggieri et al. | An automatic technique for measuring the surface tension of liquid metals | |
| CN110991360B (en) | Robot inspection point position intelligent configuration method based on visual algorithm | |
| Kwon et al. | Practical guide to machine vision software: an introduction with LabVIEW | |
| CN112085089A (en) | Intelligent temperature measurement method for transformer substation equipment based on deep learning algorithm | |
| CN113083804A (en) | Laser intelligent derusting method and system and readable medium | |
| CN114894642B (en) | Fatigue crack propagation rate testing method and device based on deep learning | |
| CN114565552A (en) | Method and device for detecting surface defects of optical element and electronic equipment | |
| CN112329687B (en) | Automatic recognition method and system for dislocation of silicon carbide substrate | |
| CN115423755A (en) | Wafer micromachining structure defect detection method and device, equipment and medium thereof | |
| JPS63219085A (en) | Separation of jamming information from effective information of microscope image | |
| CN113569679A (en) | Method, device and system for measuring elongation at break | |
| CN108346592B (en) | Method and device for simulating defects on back of wafer | |
| CN114913086B (en) | Face image quality enhancement method based on generation countermeasure network | |
| JPWO2008111290A1 (en) | Template matching device, camera equipped with template matching device, computer matching program for computer | |
| CN116931247A (en) | Multi-region slice scanning-oriented rapid focusing method | |
| US20030185432A1 (en) | Method and system for image registration based on hierarchical object modeling | |
| Kim et al. | An efficient method of estimating edge locations with subpixel accuracy in noisy images | |
| JP5076567B2 (en) | Template matching device, camera equipped with template matching device, computer matching program for computer | |
| CN111223077A (en) | Test method and test system of circuit board | |
| Adiga et al. | Metrology of 3D NAND in electron micrographs by scale space snakes | |
| US20220373468A1 (en) | Method of automatically setting optical parameters and automated optical inspection system using the same | |
| WO2022196292A1 (en) | Method for evaluating crystal defects of silicon carbide single crystal wafer | |
| CN115876803A (en) | Milling cutter coating defect detection method and system based on zoom microscopy | |
| CN115424099B (en) | Model training method, recognition method and device for recognizing silicon carbide dislocation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |