Silicon wafer mixing and sorting method combining rotating target detection and similarity judgment
Through the combined rotary target detection and similarity determination methods, the problems of misjudgment and misjudgment in silicon wafer mix sorting are solved, efficient silicon wafer sorting is achieved, and processing quality is improved.
Patent Information
- Application Number
- CN202411426700.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-14
- Publication Date
- 2025-02-07
AI Technical Summary
The existing silicon wafer mix sorting methods are prone to misjudgment and misjudgment, which leads to impurity of the silicon wafer and affects subsequent processing.
Using the combined rotation target detection and similarity discrimination method, the rotating target frame with crease angle information is obtained through the silicon wafer crease direction detection model, and combined with the fold angle difference and similarity discrimination technology, abnormal silicon wafers are sorted out .
It realizes efficient mix sorting of all silicon wafers in the same batch, with high degree of automation, reducing the possibility of missed inspection and missed inspection.
Smart Images
Figure CN119399520A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of silicon wafer mixed material sorting, and in particular to a silicon wafer mixed material sorting method combining rotating target detection and similarity discrimination. Background Art
[0002] Silicon wafers are thin, flat, round silicon-based materials and are an important raw material for chip manufacturing. Silicon wafer mixing occurs during the production process of silicon wafers. During the production process of different types of silicon wafers, silicon wafers of a certain series may be mixed with silicon wafers of other series, resulting in impure silicon wafers of that series, which in turn affects the subsequent processing of the silicon wafers of that series.
[0003] At present, the sorting of mixed silicon wafers still mainly adopts the method of manual sampling. Since manual sampling is greatly affected by personal and environmental factors, the sorting of mixed silicon wafers is prone to missed judgments and misjudgments. Summary of the invention
[0004] In order to solve the problem that the existing silicon wafer mixed material sorting is prone to missed judgment and misjudgment, an embodiment of the present invention provides a silicon wafer mixed material sorting method that combines rotating target detection and similarity judgment.
[0005] In a first aspect, an embodiment of the present invention provides a method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination, the method comprising the following steps:
[0006] The images of the same batch of silicon wafers are sequentially input into the silicon wafer crease direction detection model to obtain the rotation target frames corresponding to the same batch of silicon wafers;
[0007] The crease angles of the rotating target frames corresponding to the same batch of silicon wafers are determined to obtain the crease angle determination results corresponding to the same batch of silicon wafers;
[0008] Based on the crease angle determination results corresponding to the same batch of silicon wafers, the images of the same batch of silicon wafers are subjected to crease angle difference or similarity determination to sort out abnormal silicon wafers.
[0009] In the above embodiment, the present invention first obtains a rotating target frame with crease angle information through a silicon wafer crease direction detection model, and then determines the crease angle of the rotating target frame to distinguish between creased silicon wafers and non-crease silicon wafers. Finally, the crease angle difference method is used to perform abnormal judgment on the creased silicon wafers, and silicon wafers of different sources are sorted out from most of the silicon wafers. The non-crease silicon wafers are judged as abnormal by similarity judgment, and silicon wafers of different sources are sorted out from a small part of the silicon wafers, thereby realizing the mixed sorting of all silicon wafers in the same batch.
[0010] As some optional implementations of the present application, the silicon wafer crease direction detection model is obtained by training a rotating target detection model.
[0011] In the above embodiment, the present invention trains the rotating target detection model so that the silicon wafer crease direction detection model can obtain a rotating target frame with crease angle information.
[0012] As some optional implementations of the present application, the process of training the rotating target detection model is as follows:
[0013] Acquire an image corresponding to the silicon wafer with the crease, and annotate the crease on the image corresponding to the silicon wafer with the crease to obtain training data;
[0014] The training data is input into the rotation target detection model for training to obtain a silicon wafer crease direction detection model.
[0015] In the above embodiment, the present invention trains the rotating target detection model so that the silicon wafer crease direction detection model can first extract the feature information about the crease on the silicon wafer surface from shallow to deep through the feature extraction network in the process of obtaining the rotating target frame, and then fuse the feature information of the crease through the feature fusion network layer to form an effective feature map; finally, the feature map is classified, predicted and located through the classification prediction head to obtain the rotating target frame of the crease.
[0016] As some optional implementations of the present application, the rotating target frame carries angle information of the fold.
[0017] As some optional embodiments of the present application, the crease angle determination result includes a silicon wafer without creases, a silicon wafer with creases, and a silicon wafer crease angle.
[0018] In the above embodiment, if the crease angle determination result of the present invention is a crease-free silicon wafer, there is no corresponding silicon wafer crease angle. If the crease angle determination result is a crease-free silicon wafer, there is also an associated silicon wafer crease angle information to facilitate subsequent crease angle difference verification.
[0019] As some optional embodiments of the present application, based on the crease angle determination results corresponding to the same batch of silicon wafers, the crease angles of the images of the same batch of silicon wafers are differentiated or determined by similarity, so as to sort out abnormal silicon wafers as follows:
[0020] According to the crease angle determination result, the silicon wafers with creases are divided into a first array, and the silicon wafers without creases are divided into a second array;
[0021] Obtaining the number of images corresponding to the first array and the second array, and verifying the number of images corresponding to the first array and the second array;
[0022] If the number of images corresponding to the first array is not less than two, the images are detected for anomalies in turn by using the method of difference in crease angles to sort out silicon wafers from different sources; if the number of images corresponding to the second array is not less than two, the images are detected for anomalies in turn by using the method of similarity judgment to sort out silicon wafers from different sources.
[0023] In the above embodiment, the present invention adopts the method of making a difference in fold angles to determine the fold angles of folded silicon wafers in the same batch, and can sort out silicon wafers of different sources from most of the silicon wafers, and adopts the method of similarity judgment to determine the abnormalities of non-folded silicon wafers in the same batch, and can sort out silicon wafers of different sources from a small part of the silicon wafers, thereby realizing the mixed material sorting of all silicon wafers in the same batch, with a high degree of automation, and is not prone to the problems of missed detection and wrong detection.
[0024] As some optional implementations of the present application, if the number of images corresponding to the second array is not less than two, the process of performing abnormality detection on the images in sequence by similarity determination to sort out silicon wafers from different sources is as follows:
[0025] Performing dimension increase processing on two images of silicon wafers without creases, converting the two one-dimensional images into two three-dimensional images;
[0026] Perform feature extraction, average pooling and format conversion on the two three-dimensional images to obtain two feature maps;
[0027] The two feature maps are convolved, squared and dimensionally reduced to obtain the similarity between the two images of the silicon wafer without creases.
[0028] If the similarity corresponding to the images of two crease-free silicon wafers is less than the similarity threshold, it is determined that the two crease-free silicon wafers are of different origins.
[0029] In the above embodiment, the present invention can perform similarity comparison when the crease cannot be obtained. By performing similarity comparison on the detail features, it can be determined whether two silicon wafers without creases are of the same origin.
[0030] In a second aspect, the present invention provides a silicon wafer mixed material sorting system combining rotation target detection and similarity discrimination, the system comprising:
[0031] A rotation angle acquisition unit, wherein the rotation angle acquisition unit is used to sequentially input images of the same batch of silicon wafers into a silicon wafer crease direction detection model to obtain a rotation target frame corresponding to the same batch of silicon wafers;
[0032] A crease angle determination unit, the crease angle determination unit is used to determine the crease angle of the rotating target frame corresponding to the same batch of silicon wafers, so as to obtain the crease angle determination results corresponding to the same batch of silicon wafers;
[0033] The image similarity determination unit performs crease angle difference or similarity determination on images of the same batch of silicon wafers based on the crease angle determination results corresponding to the same batch of silicon wafers, so as to sort out abnormal silicon wafers.
[0034] In a third aspect, the present invention provides a computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the silicon wafer mixed material sorting method combining rotation target detection and similarity judgment.
[0035] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for sorting mixed silicon wafers by combining rotation target detection and similarity determination is implemented.
[0036] The beneficial effects of the present invention are as follows: the present invention first obtains a rotating target frame with crease angle information through a silicon wafer crease direction detection model, then determines the crease angle of the rotating target frame, distinguishes silicon wafers with creases from silicon wafers without creases, and finally uses the crease angle difference method to perform abnormality judgment on silicon wafers with creases, sorts out silicon wafers of different sources from most of the silicon wafers, and uses the similarity judgment method to perform abnormality judgment on silicon wafers without creases, sorts out silicon wafers of different sources from a small part of the silicon wafers, thereby realizing the mixed sorting of all silicon wafers in the same batch. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.
[0038] Figure 1 is a schematic diagram of the computer device structure of the hardware operating environment described in the embodiment of the present invention;
[0039] Figure 2 is a flow chart of the silicon wafer mixed material sorting method according to an embodiment of the present invention;
[0040] Figure 3 is a schematic diagram comparing crease angles of creased silicon wafers according to an embodiment of the present invention;
[0041] Figure 4 is a similar comparison schematic diagram of a crease-free silicon wafer according to an embodiment of the present invention;
[0042] Figure 5 It is a structural block diagram of the silicon wafer mixed material sorting system described in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.
[0044] In order to solve the problem that the existing silicon wafer mixed material sorting is prone to missed judgment and misjudgment, the embodiment of the present invention provides a silicon wafer mixed material sorting method, system, device and storage medium for combined rotation target detection and similarity discrimination. Before introducing the specific technical solution of the present application, the hardware operating environment involved in the embodiment of the present application is first introduced.
[0045] See also Figure 1 , Figure 1 The figure is a schematic diagram of the computer device structure of the hardware operating environment involved in the embodiment of the present invention.
[0046] like Figure 1 As shown, the computer device may include: a processor, such as a central processing unit (CPU), a communication bus, a user interface, a network interface, and a memory. Among them, the communication bus is used to realize the connection and communication between these components. The user interface may include a display screen (Display), an input unit such as a keyboard (Keyboard), and optionally, the user interface may also include a standard wired interface and a wireless interface. The network interface may be a wired interface or a wireless interface (such as a wireless fidelity interface). The memory may be a high-speed random access memory (Random Access Memory, RAM), or a stable non-volatile memory (Non-Volatile Memory, NVM), such as a disk storage. The memory may optionally be a storage device independent of the aforementioned processor.
[0047] Those skilled in the art will understand that Figure 1 The structure shown in the figure does not constitute a limitation on the computer device, and may include more or less components than shown in the figure, or combine some components, or arrange the components differently.
[0048] like Figure 1 As shown, the memory as a storage medium may include an operating system module, a network communication module, a user interface module and a data storage module.
[0049] exist Figure 1In the computer device shown, the network interface is mainly used for data communication with the network server; the user interface is mainly used for data interaction with the user; the processor and memory in the computer device of the present application can be set in the computer device, and the computer device calls the silicon wafer mixing and sorting system stored in the data storage module through the processor, and executes the silicon wafer mixing and sorting method provided in the embodiment of the present application.
[0050] Based on the hardware environment of the above-mentioned embodiment, the embodiment of the present application provides a method for sorting mixed silicon wafers by combining rotation target detection and similarity judgment. Figure 2 , Figure 2 The flowchart of the silicon wafer mixed material sorting method is as follows:
[0051] (1) Collect images of silicon wafers from the same batch, and input the images of silicon wafers from the same batch into a silicon wafer crease direction detection model in sequence to obtain a rotation target frame corresponding to the silicon wafers from the same batch, wherein the rotation target frame carries crease angle information.
[0052] In the embodiment of the present invention, silicon wafers of the same series are cut from the same crystal rod, and 90% of the surfaces have creases in the same direction, and 10% of the surfaces have no creases. At the same time, since they are cut from the same crystal rod, the directions of the creases of silicon wafers of the same series should be consistent if there are creases.
[0053] In an embodiment of the present invention, model training is required before the images of the same batch of silicon wafers are input into the silicon wafer crease direction detection model in sequence, so that the silicon wafer crease direction detection model can obtain a rotating target frame with crease angle information; specifically, the silicon wafer crease direction detection model is obtained by training a rotating target detection model.
[0054] In an embodiment of the present invention, the process of training the rotating target detection model is as follows:
[0055] (1.1) Obtain images corresponding to creased silicon wafers and annotate the creases on the images to obtain training data; wherein the main object of the crease annotation is the creases on the surface of the silicon wafer.
[0056] (1.2) The training data is input into the rotation target detection model for training to obtain a silicon wafer crease direction detection model.
[0057] In an embodiment of the present invention, when the silicon wafer crease direction detection model obtains a rotated target frame, the silicon wafer crease direction detection model first extracts feature information about creases on the silicon wafer surface from shallow to deep through a feature extraction network, and then fuses the feature information of the creases through a feature fusion network layer to form an effective feature map; finally, the feature map is classified, predicted and located through a classification prediction head to obtain a rotated target frame of the crease.
[0058] (2) The crease angle of the rotating target frame corresponding to the same batch of silicon wafers is determined to obtain the crease angle determination results corresponding to the same batch of silicon wafers.
[0059] In an embodiment of the present invention, the crease angle determination result includes a silicon wafer without creases, a silicon wafer with creases, and a silicon wafer crease angle.
[0060] (3) Based on the crease angle determination results corresponding to the same batch of silicon wafers, the images of the same batch of silicon wafers are subjected to crease angle difference or similarity determination to sort out abnormal silicon wafers.
[0061] In an embodiment of the present invention, the process of sorting abnormal silicon wafers is as follows:
[0062] (3.1) According to the crease angle determination result, the information related to the crease-bearing silicon wafer is divided into a first array, and the information related to the non-crease-bearing silicon wafer is divided into a second array.
[0063] (3.2) Obtain the number of images corresponding to the first array and the second array, and verify the number of images corresponding to the first array and the second array.
[0064] (3.3) If the number of images corresponding to the first array is not less than two, the images are detected for anomalies in turn by using the crease angle difference method to sort out silicon wafers from different sources; if the number of images corresponding to the second array is not less than two, the images are detected for anomalies in turn by using the similarity judgment method to sort out silicon wafers from different sources.
[0065] Specifically, when the number of corresponding images in the first array is equal to two, it is checked whether the folding angles of the two images are consistent. If they are consistent, the silicon wafers corresponding to the two images are of the same origin. Otherwise, the silicon wafers corresponding to the two images are of different origins, and it is output that the silicon wafers corresponding to the two images are abnormal.
[0066] Specifically, when the number of corresponding images in the first array is greater than two, the fold angle of the current image is subtracted from the fold angles of the remaining images in turn to obtain several angle difference values about the current image. If the angle difference value of the current image is more than 50% greater than the abnormal value, the silicon wafer corresponding to the current image is determined to be abnormal; otherwise, the silicon wafer corresponding to the current image is determined to be normal, and this process is repeated until all image abnormality detections are completed; wherein, the abnormal value is generally 11 to 170, and the embodiment of the present invention does not impose a limitation on this.
[0067] Specifically, when the number of corresponding images in the second array is greater than or equal to two, the current image is used to perform similarity judgment with the remaining images in turn. If the similarity is greater than the similarity threshold, the silicon wafers corresponding to the two images are of the same source. Otherwise, the silicon wafers corresponding to the two images are of different sources, and the images of the same source and different sources are classified and stored, and the silicon wafers corresponding to the images of a smaller number of categories are judged as abnormal, and the silicon wafers corresponding to the images of a larger number of categories are judged as normal. The similarity threshold is generally 0.95-1, and the embodiment of the present invention does not limit this.
[0068] Specifically, when the number of images corresponding to the first array and the second array is one, the image corresponding to the creased silicon wafer and the image corresponding to the non-creased silicon wafer are judged for similarity. If the similarity is greater than the similarity threshold, the silicon wafers corresponding to the two images are of the same origin, and the silicon wafers corresponding to the two images are normal. Otherwise, the silicon wafers corresponding to the two images are of different origins, and the silicon wafers corresponding to the two images are abnormal.
[0069] Specifically, when the number of images corresponding to the first array or the second array is one, the silicon wafer corresponding to the directly output image is abnormal.
[0070] In the embodiment of the present invention, the process of similarity determination is as follows:
[0071] (3.31) The two images of the uncreased silicon wafer are dimensionally increased to convert the two one-dimensional images into three-dimensional images A and B with height h, width w, and number of channels c.
[0072] (3.32) The residual network is used to first perform feature extraction on the three-dimensional image A and the three-dimensional image B, then perform average pooling, and finally perform format conversion (Float32 bit format under Torch) to obtain feature map A 1 And feature map B 1 .
[0073] (3.33) For the feature map A 1 And feature map B 1 Convolution, square root and dimensionality reduction are performed to obtain the similarity between the two uncreased images.
[0074] (3.34) If the similarity between the two images of silicon wafers without creases is less than the similarity threshold, it is determined that the silicon wafers corresponding to the two images are from different sources.
[0075] Specifically, for feature map A 1 And feature map B 1 The process of convolution, square root and dimensionality reduction is as follows:
[0076] ① For feature map A 1 Perform convolution processing, where the convolution kernel of the convolution processing is the feature map B 1 , that is, using feature map B 1 In feature map A 1 A convolution operation is performed on the , the bias is set to 0, the step size is set to 1, and zero bits are filled to obtain the first similarity Result1.
[0077] ②Use feature map B 1 Multiply by feature map B 1 , get the new feature map B 11 , and use feature map A 1 Multiply by feature map A 1 , get the new feature map A 11 .
[0078] ③ For feature map A 11 Perform convolution processing, where the convolution kernel is a convolution kernel with all elements set to 1 and the same size as feature map B. Set the bias to None, the step size to 1, and fill in zeros to obtain a new feature map A. 12 .
[0079] ④ The feature map B 11 Multiply by feature map A 12 , and perform square root processing to obtain the second similarity Result2.
[0080] ⑤ Divide the first similarity Result1 by the second similarity Result2, and remove the values of height h and width w, and finally obtain the similarity of the two images without creases.
[0081] In summary, the embodiment of the present invention combines the rotating target detection technology and the similarity judgment technology to firstly detect the crease angle of the same batch of silicon wafers to distinguish between creased silicon wafers and non-creased silicon wafers, then judge the crease angle of the creased silicon wafers in the same batch of silicon wafers to sort out silicon wafers of different sources from most of the silicon wafers, and perform similarity judgment on the non-creased silicon wafers in the same batch of silicon wafers to sort out silicon wafers of different sources from a small part of the silicon wafers, thereby realizing the mixed material sorting of all silicon wafers in the same batch. Please refer to Figure 3 , Figure 4 , Figure 3These are two images of creased silicon wafers from different sources. Figure 4 These are comparison images of two crease-free silicon wafers from different sources.
[0082] In addition, in one embodiment, based on the same inventive concept as the above embodiment, the embodiment of the present invention provides a silicon wafer mixed material sorting system combining rotating target detection and similarity discrimination, see Figure 5 , Figure 5 is a structural block diagram of the silicon wafer mixed material sorting system, the system corresponds one-to-one to the method of Example 1, and the system includes:
[0083] A rotation angle acquisition unit is used to sequentially input images of the same batch of silicon wafers into a silicon wafer crease direction detection model to obtain a rotation target frame corresponding to the same batch of silicon wafers.
[0084] A crease angle determination unit is used to perform crease angle determination on the rotating target frame corresponding to the same batch of silicon wafers to obtain crease angle determination results corresponding to the same batch of silicon wafers.
[0085] The image similarity determination unit performs crease angle difference or similarity determination on images of the same batch of silicon wafers based on the crease angle determination results corresponding to the same batch of silicon wafers, so as to sort out abnormal silicon wafers.
[0086] It should be noted that each unit in the silicon wafer mixing and sorting system in this embodiment corresponds one-to-one to each step in the silicon wafer mixing and sorting method in the aforementioned embodiment. Therefore, the specific implementation method and technical effects achieved in this embodiment can refer to the implementation method of the aforementioned silicon wafer mixing and sorting method, and will not be repeated here.
[0087] In addition, in one embodiment, the present application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory, and the computer program implements the method in the aforementioned embodiment when executed by the processor.
[0088] In addition, in one embodiment, the present application further provides a computer storage medium, on which a computer program is stored, and the computer program implements the method in the aforementioned embodiment when executed by a processor.
[0089] In some embodiments, the computer readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or various devices including one or any combination of the above memories. The computer may be various computing devices including intelligent terminals and servers.
[0090] In some embodiments, executable instructions may be in the form of a program, software, software module, script or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a stand-alone program or as a module, component, subroutine or other unit suitable for use in a computing environment.
[0091] As an example, executable instructions may, but need not, correspond to a file in a file system, may be stored as part of a file that stores other programs or data, such as in one or more scripts in a HyperText Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple coordinated files (e.g., files storing one or more modules, subroutines, or code portions).
[0092] By way of example, executable instructions may be deployed to be executed on one computing device, or on multiple computing devices located at one site, or on multiple computing devices distributed across multiple sites and interconnected by a communication network.
[0093] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.
[0094] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.
[0095] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory / random access memory, a magnetic disk, or an optical disk), and includes a number of instructions for enabling a multimedia terminal device (which can be a mobile phone, a computer, a television receiver, or a network device, etc.) to execute the methods described in each embodiment of the present application.
[0096] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.
Claims
1. A method for sorting mixed silicon wafers by combining rotating target detection and similarity discrimination, characterized in that: The method comprises the following steps: The images of the same batch of silicon wafers are sequentially input into the silicon wafer crease direction detection model to obtain the rotation target frames corresponding to the same batch of silicon wafers; The crease angles of the rotating target frames corresponding to the same batch of silicon wafers are determined to obtain the crease angle determination results corresponding to the same batch of silicon wafers; Based on the crease angle determination results corresponding to the same batch of silicon wafers, the images of the same batch of silicon wafers are subjected to crease angle difference or similarity determination to sort out abnormal silicon wafers.
2. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 1, characterized in that: The silicon wafer crease direction detection model is obtained by training a rotation target detection model.
3. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 2, characterized in that: The process of training the rotating target detection model is as follows: Acquire an image corresponding to the silicon wafer with the crease, and annotate the crease on the image corresponding to the silicon wafer with the crease to obtain training data; The training data is input into the rotation target detection model for training to obtain a silicon wafer crease direction detection model.
4. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 1, characterized in that: The rotating target frame carries the angle information of the fold.
5. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 1, characterized in that: The crease angle determination result includes a silicon wafer without creases, a silicon wafer with creases, and a silicon wafer crease angle.
6. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 5, characterized in that: Based on the crease angle determination results corresponding to the same batch of silicon wafers, the crease angles of the images of the same batch of silicon wafers are differentiated or determined in a similarity manner to sort out abnormal silicon wafers as follows: According to the crease angle determination result, the silicon wafers with creases are divided into a first array, and the silicon wafers without creases are divided into a second array; Obtaining the number of images corresponding to the first array and the second array, and verifying the number of images corresponding to the first array and the second array; If the number of images corresponding to the first array is not less than two, the images are detected for anomalies in turn by using the method of difference in crease angles to sort out silicon wafers from different sources; if the number of images corresponding to the second array is not less than two, the images are detected for anomalies in turn by using the method of similarity judgment to sort out silicon wafers from different sources.
7. The method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination according to claim 6, characterized in that: If the number of images corresponding to the second array is not less than two, the images are sequentially detected for abnormality by similarity determination to sort out silicon wafers from different sources. The process is as follows: Performing dimension increase processing on two images of silicon wafers without creases, converting the two one-dimensional images into two three-dimensional images; Perform feature extraction, average pooling and format conversion on the two three-dimensional images to obtain two feature maps; The two feature maps are convolved, squared and dimensionally reduced to obtain the similarity between the two images of the silicon wafer without creases. If the similarity corresponding to the images of two crease-free silicon wafers is less than the similarity threshold, it is determined that the two crease-free silicon wafers are of different origins.
8. A silicon wafer mixed material sorting system combining rotating target detection and similarity discrimination, characterized in that: The system comprises: A rotation angle acquisition unit, wherein the rotation angle acquisition unit is used to sequentially input images of the same batch of silicon wafers into a silicon wafer crease direction detection model to obtain a rotation target frame corresponding to the same batch of silicon wafers; A crease angle determination unit, the crease angle determination unit is used to determine the crease angle of the rotating target frame corresponding to the same batch of silicon wafers, so as to obtain the crease angle determination results corresponding to the same batch of silicon wafers; The image similarity determination unit performs crease angle difference or similarity determination on images of the same batch of silicon wafers based on the crease angle determination results corresponding to the same batch of silicon wafers, so as to sort out abnormal silicon wafers.
9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for sorting mixed silicon wafers by combining rotation target detection and similarity discrimination as described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for sorting mixed silicon wafers by combining rotation target detection and similarity determination as described in any one of claims 1 to 7 is implemented.