Wafer defect detection method and related equipment
The method and system enhance wafer defect detection by using a deep learning network that adapts to new defects through incremental training and traditional algorithms, ensuring efficient and accurate detection of relevant defects on semiconductor wafers.
Patent Information
- Application Number
- CN202411064736.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-05
- Publication Date
- 2025-01-03
AI Technical Summary
The challenge lies in improving the deep learning algorithm's ability to detect wafer defects efficiently and accurately, especially with the increasing complexity and size of semiconductor wafers, and the need for continuous training to adapt to new defects in real-time.
A method and system that utilizes a preset deep learning network for defect detection, which can identify defects without annotations and incrementally trains the network when annotations are present, combining with traditional algorithms for enhanced detection and parameter control, leveraging GPU and CPU processing for efficient defect determination.
Enhances the deep learning algorithm's ability to detect wafer defects continuously, improving online detection efficiency and accuracy by selectively identifying relevant defects and reducing unnecessary processing, thus enhancing the quality and production efficiency of semiconductor wafers.
Smart Images
Figure CN119251136A_ABST
Abstract
Description
Background Art
[0002] Wafer defect detection is a key link in the semiconductor manufacturing process. It involves detecting various defects on the wafer surface, such as scratches, contamination, cracks, oxidation, etc., to ensure the quality and performance of the wafer. With the advancement of semiconductor technology, the manufacturing size of wafers is getting smaller and smaller, and the difficulty of defect detection is also increasing accordingly.
[0003] In the prior art, wafer defect images are acquired by image acquisition of wafer defects. The field of view of each high-precision image acquired is generally less than 1mm x 1mm, and the wafer area to be detected ranges from 6 inches to 8 inches, and currently 12-inch wafers are common. Therefore, the algorithm detection platform needs to process a large number of images in real time. Therefore, deep learning algorithms are often used to batch identify wafer defect images, which can detect all defects in each wafer defect image. The application of deep learning algorithms in wafer defect detection greatly improves the automation level and efficiency of detection, while also reducing the complexity and error rate of manual operations. Deep learning algorithms have been widely used in the wafer defect detection process, and deep learning algorithms need to continuously improve defect detection capabilities. The deep learning algorithm modules in the equipment need to be continuously trained, upgraded, and iterated according to the on-site process conditions. Only in this way can the good online detection effect of the equipment be guaranteed.
[0004] Therefore, continuously improving the deep learning algorithm's ability to detect wafer defects and improving the online detection effect of wafer defects are areas that we urgently need to research. Summary of the invention
[0005] In order to continuously improve the deep learning algorithm's ability to detect wafer defects and achieve good online detection results for wafer defects, the present application provides a wafer defect detection method and related equipment.
[0006] In the first aspect, in order to solve the above technical problems, the present application provides a wafer defect detection method, comprising:
[0007] Acquire a captured image containing defects of the wafer to be tested;
[0008] When the captured image does not contain annotations, the type and parameter information of the defect is obtained based on the preset deep learning network;
[0009] When the captured image contains annotations, incremental training is performed on the preset deep learning network using the captured image to obtain a new preset deep learning network, and the type and parameter information of the defect are obtained;
[0010] Determine the defect detection status based on the detection range and parameter information corresponding to the type;
[0011] Based on the detection situation, the detected defects of the wafer to be tested are determined.
[0012] In a second aspect, the present application also provides a wafer defect detection device, which applies a wafer defect detection method.
[0013] In a third aspect, the present application also provides a wafer defect detection system, comprising:
[0014] An acquisition module, used for acquiring a photographed image containing defects of the wafer to be tested;
[0015] A recognition module, used to obtain the type and parameter information of the defect based on a preset deep learning network when no annotation is included in the captured image;
[0016] An incremental training module is used to perform incremental training on a preset deep learning network using captured images when the captured images contain annotations, obtain a new preset deep learning network, and obtain defect type and parameter information;
[0017] A determination module is used to determine the detection status of defects based on the detection range and parameter information corresponding to the type;
[0018] The detection module is used to determine the detection defects of the wafer to be tested based on the detection situation.
[0019] In a fourth aspect, the present application also provides a computing device, including a memory, a processor, and a program stored in the memory and running on the processor, and when the processor executes the program, the steps of a wafer defect detection method as described above are implemented.
[0020] In a fifth aspect, the present application also provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes the steps of a wafer defect detection method.
[0021] The beneficial effects of the present application are: the annotations are characterized as the marking annotations corresponding to the new wafer process defects that are discovered for the first time, and based on the annotations, it can be understood whether the defects in the captured image can be identified by the existing preset deep learning network. That is, the method of the present application directly uses the trained preset deep learning network to identify and check the defects in the captured image when the captured image does not contain annotations, and obtains the type and parameter information of the defects. When it is detected that the captured image contains annotations, the preset deep learning network is incrementally trained in the captured image to learn the defect characteristics of the new wafer process defects in the captured image. In this way, as long as new wafer process defects appear, the preset deep learning network is trained and learned, so that the deep learning algorithm can continuously improve its ability to detect wafer defects. It is convenient to determine the detection situation of the defect based on the detection range and parameter information corresponding to the type in the subsequent detection, and determine the detection defects of the wafer to be tested based on the detection situation, so as to improve the good online detection effect of wafer defects. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Figure 1 A schematic diagram of a process flow of a wafer defect detection method for this application;
[0023] Figure 2 A system design diagram of a wafer defect detection method for this application;
[0024] Figure 3 This is a structural schematic diagram of a wafer defect detection system for the present application. DETAILED DESCRIPTION
[0025] The following examples are provided to further explain and supplement the present application and do not constitute any limitation to the present application.
[0026] A wafer defect detection method and related equipment according to an embodiment of the present application are described below in conjunction with the accompanying drawings.
[0027] A wafer defect detection method in an embodiment of the present application is applied to a terminal device. In the present application scheme, the terminal device is used as the execution subject to illustrate the present application scheme. The terminal device is used to execute the steps of a wafer defect detection method.
[0028] like Figure 1 As shown, the present application provides a wafer defect detection method, comprising:
[0029] Step S11, acquiring a captured image containing defects of the wafer to be tested;
[0030] Step S12, when the captured image does not contain any annotation, obtaining the type and parameter information of the defect based on a preset deep learning network;
[0031] Step S13, when the captured image contains annotations, incrementally training the preset deep learning network using the captured image to obtain a new preset deep learning network, and obtaining the type and parameter information of the defect;
[0032] Step S14, determining the defect detection status based on the detection range and parameter information corresponding to the type;
[0033] Step S15, determining the detected defects of the wafer to be tested based on the detected situation.
[0034] A wafer defect detection method of the present embodiment, the annotation is characterized by a mark annotation corresponding to the wafer new process defect discovered for the first time, and according to the annotation, it can be understood whether the defect in the captured image can be recognized by the existing preset deep learning network. That is, the method of the present application directly uses the trained preset deep learning network to identify and check the defects in the captured image when the captured image does not contain the annotation, and obtains the type and parameter information of the defect. When it is detected that the captured image contains the annotation, the preset deep learning network is incrementally trained in the captured image to learn the defect characteristics of the wafer new process defect in the captured image. In this way, as long as a wafer new process defect occurs, the preset deep learning network is trained and learned, so that the deep learning algorithm can continuously improve the detection ability of wafer defects. It is convenient to determine the detection situation of the defect based on the detection range and parameter information corresponding to the type in the subsequent detection, and determine the detection defect of the wafer to be tested based on the detection situation, so as to improve the good online detection effect of wafer defects. Among them, the types of defects include linear scratches, pits, cracks, and oxide layer defects. Pits include depressions and holes on the wafer surface, cracks include fracture lines on the wafer surface, and oxide layer defects include holes, cracks, and uneven oxide layer defects in the oxide layer on the wafer surface.
[0035] Optionally, obtaining a captured image containing defects of the wafer to be tested includes:
[0036] Determine an area to be photographed that includes defects of the wafer to be tested;
[0037] The area to be photographed is photographed to obtain a photographed image.
[0038] In this embodiment, targeted photography is performed on the area to be photographed that contains defects in the wafer to be tested, so that the captured image only contains an image range with recognition significance, thereby reducing the interference of useless image information, facilitating subsequent recognition detection or training in a preset deep learning network, thereby improving the efficiency and accuracy of recognition detection, or improving the efficiency of training and the recognition accuracy of the new model obtained through training.
[0039] Optionally, when no annotation is included in the captured image, the type and parameter information of the defect is obtained based on a preset deep learning network, including:
[0040] When the captured image does not contain annotations, the captured image is input into a preset deep learning network to extract features of defects in the captured image to obtain defect information;
[0041] Based on the defect information, the type and parameter information of the defect are obtained.
[0042] In this embodiment, when the captured image does not contain annotations, it means that the defect in the captured image is an existing defect that has been learned by the preset deep learning network. Then, the preset deep learning network is used to extract features from the captured image that does not contain annotations, and defect information that meets the accuracy requirements in the captured image can be extracted. Based on the defect information, all information about the defects in the captured image can be understood. Based on the defect information, the defect type and parameter information obtained can meet the accuracy requirements, thereby improving the accuracy of subsequent defect detection.
[0043] Optionally, when the captured image contains annotations, incremental training is performed on a preset deep learning network using the captured image to obtain a new preset deep learning network, and the type and parameter information of the defect are obtained, including:
[0044] Obtain a target training set based on the captured images containing annotations;
[0045] Input the target training set into the preset deep learning network for training, adjust the network parameters in the preset deep learning network, and obtain a new preset deep learning network;
[0046] The annotation includes a defect type annotation and a parameter information annotation, the type annotation is used as the defect type, and the parameter information annotation is used as the parameter information of the defect.
[0047] In this embodiment, the target training set can be a training set consisting of an image training set used in the training of the preset deep learning network and annotated captured images, or a training set consisting of annotated captured images. When the captured images contain annotations, it means that the defects in the captured images are new wafer process defects that have not been learned by the preset deep learning network. At this time, the preset deep learning network is incrementally trained with the target training set of the annotated captured images, and the network parameters in the preset deep learning network are adjusted to obtain a new preset deep learning network, so that the new preset deep learning network can identify the new wafer process defects, thereby continuously improving the deep learning algorithm's ability to detect wafer defects, thereby improving the good online detection effect of wafer defects.
[0048] Optionally, the detection range includes a shape range and an area range, and the parameter information includes a defect shape and a defect area; based on the detection range and parameter information corresponding to the type, determining the detection status of the defect includes:
[0049] When the defect shape is within the shape range and the defect area is within the area range, the detection status of the defect is determined to be pending detection;
[0050] When the defect shape is not within the shape range and / or the defect area is not within the area range, the defect detection status is determined to be not detected.
[0051] In this embodiment, the detection range is the detection limit of the corresponding defect, that is, the defects within the detection range need to be detected and processed, and the defects that are not completely within the detection range can be considered as qualified defects that do not need to be detected and processed. Then, when the defect shape is within the shape range and the defect area is within the area range, the detection situation of the defect is determined as to be detected, and when the defect shape is not within the shape range, and / or the defect area is not within the area range, the detection situation of the defect is determined as not detected, so as to facilitate the subsequent determination of the detected defects of the wafer to be tested according to the detection situation, and improve the good online detection effect of wafer defects. At the same time, selectively only partially detecting the defects that affect the quality of the wafer, and ignoring the defects that do not affect the quality of the wafer as qualified defects, can not only reduce the secondary processing time of the wafer defects, but also improve the product qualification rate of the wafer, so as to save the production cost of the enterprise under the premise of ensuring the quality of the wafer.
[0052] Optionally, based on the detection situation, determining the detected defects of the wafer to be tested includes:
[0053] When the detection condition is pending detection, the defect is regarded as a detected defect of the wafer to be tested;
[0054] When the detection condition is no detection, it is determined that the detection defect of the wafer to be tested is empty.
[0055] In this embodiment, when the detection condition is to be detected, the defect is detected as a detected defect of the wafer to be tested. When the detection condition is not detected, it is determined that the detected defect of the wafer to be tested is empty, which means that the wafer to be tested has no defects that need to be detected, and the wafer to be tested is a qualified product. In this way, by automatically identifying whether there are detected defects that need to be detected on the wafer to be tested, compared with the human eye determining whether there are detected defects that need to be detected on the wafer to be tested, the method in this embodiment can improve the detection efficiency of the wafer.
[0056] A wafer defect detection device according to an embodiment of the present application applies the above-mentioned wafer defect detection method.
[0057] Figure 2 A system design diagram of a wafer defect detection method for this application is shown in FIG. Figure 2As shown, the system design diagram includes an industrial camera, a deep learning detection module, an image database, a deep learning training module, a traditional algorithm detection module A, a traditional algorithm detection module B, and a software module. The industrial camera is connected to the deep learning detection module, and the deep learning detection module is respectively connected to the image database, the deep learning training module, the traditional algorithm detection module A and the traditional algorithm detection module B, and the image database is connected to the deep learning training module; the software module is respectively connected to the traditional algorithm detection module A and the traditional algorithm detection module B.
[0058] The steps for implementing a wafer defect detection method of the present application based on the above system design are as follows:
[0059] The industrial camera is responsible for taking images containing defects of the wafer to be tested, and sending the images to the deep learning detection module; the images are either annotated images or unannotated images. When the production line operator discovers a new process defect of the wafer on the production line, the production line operator annotates the defect information of the new process defect to form an annotated image.
[0060] The deep learning detection module includes a preset deep learning network trained in the deep learning training module. The deep learning detection module first identifies whether the captured image contains annotations. When the captured image does not contain annotations, the preset deep learning network in the deep learning detection module performs defect detection on the captured image to obtain the type and parameter information of the defect, and transmits the type and parameter information of the defect to the traditional algorithm detection module A, and transmits the captured image without annotations to the image database for storage. When the captured image contains annotations, the type and parameter information of the defect is obtained according to the type annotation and parameter information annotation of the defect contained in the annotation, and the type and parameter information of the defect is transmitted to the traditional algorithm detection module B; at the same time, the captured image containing annotations is transmitted to the image database for storage, and the preset deep learning network is incrementally trained in the deep learning training module using the target training set obtained based on the captured image containing annotations in the image database, and the characteristics of the defects in the captured image containing annotations are learned to adjust the network parameters in the preset deep learning network to obtain a new preset deep learning network. Among them, the deep learning detection module is implemented using the tensorfow platform, which is an open source software library mainly used for data flow programming and different types of calculations. It allows researchers and developers to easily develop, train and deploy deep learning models.
[0061] The deep learning training module is implemented using the tensorflow platform and is based on the multi-GPU operation mode of CUDA (Compute Unified Device Architecture). Using the incremental learning method, the deep learning training module conducts learning and training of new wafer new process defects on the preset deep learning network, completes the defect detection upgrade of the preset deep learning network, obtains a new preset deep learning network, and improves the detection capability of wafer defects. After the upgrade, the preset deep learning network in the deep learning detection module is updated to the new preset deep learning network, and the deep learning detection module is updated.
[0062] The traditional algorithm detection module A is implemented by morphological analysis and C++ programming based on CUDA, and runs on GPU (Graphics Processing Unit). The traditional algorithm detection module A receives the type and parameter information of the defects detected by the deep learning module, and performs parameter card control through SIFT (Scale-Invariant Feature Transform) and SURF (Speeded Up Robust Features), realizes the detection range and parameter information corresponding to the type, determines the detection of the defect, and finally confirms whether the defect is a detected defect or a non-detected defect, and passes the result to the software module. At this time, the traditional algorithm detection module B is in an idle state. Among them, GPU is an electronic component specially used for processing and rendering graphics, which is widely used in computers, workstations, game consoles and some mobile devices. GPU has a high degree of parallel processing capability. GPU has a large number of computing units and can handle multiple tasks at the same time, which makes it very efficient in processing image data. The analysis method of morphological analysis is to detect the length, width, area and perimeter of the defect area, whether it is linear or dot-shaped, whether the defect area is circular, spindle-shaped, elliptical, polygonal, etc., and the direction of grayscale change in the defect area. The significance of morphological analysis is that some card control conditions can be set manually, such as an area greater than 20 pixels, a length and width greater than 5 pixels, etc. These conditions can be used to improve the defect detection capability on the production line based on human experience, because deep learning algorithms have problems such as underfitting and overfitting, and cannot identify new defects. Parameter card control refers to monitoring and adjusting various parameters (such as width, length, area, shape, etc.) of the defect features in the captured images to ensure that defects on the wafer that meet the detection standards are detected.
[0063] The traditional algorithm detection module B is implemented by C++ programming and runs on a CPU (Central Processing Unit / Processor). When the traditional algorithm detection module A is in an idle state, the traditional algorithm detection module B is in a working state and completes the above work of the traditional algorithm detection module A.
[0064] The software module implements the switching from the traditional algorithm detection module A to the traditional algorithm detection module B. The software module obtains the detected defects, and saves, displays and uploads the statistical information of the detected defects: (1) the defect image is presented through the software interface corresponding to the software module, and the type of the detected defect or no defect is detected; (2) the detected defects are saved in the local database; (3) the detected defects are reported to the factory that produces the wafer to be tested through the CIM (Computer Integrated Manufacturing) system.
[0065] This application makes full use of the parallel computing performance of CUDA-based GPUs to achieve rapid processing of industrial camera-collected images and online upgrades of deep learning defect detection algorithms. At the same time, this application uses a relatively common GPU graphics card to achieve parallel processing of image detection and training upgrades, and uses traditional algorithm modules to achieve defect control while upgrading and training deep learning algorithms, thereby improving the online detection time of the equipment and enabling on-site online training. At a relatively low hardware cost, the online detection equipment's ability to detect new process defects is solved, improving the work efficiency of both equipment A and B.
[0066] like Figure 3 As shown, the present application provides a wafer defect detection system, comprising:
[0067] An acquisition module, used for acquiring a photographed image containing defects of the wafer to be tested;
[0068] A recognition module, used to obtain the type and parameter information of the defect based on a preset deep learning network when no annotation is included in the captured image;
[0069] An incremental training module is used to perform incremental training on a preset deep learning network using captured images when the captured images contain annotations, obtain a new preset deep learning network, and obtain defect type and parameter information;
[0070] A determination module is used to determine the detection status of defects based on the detection range and parameter information corresponding to the type;
[0071] The detection module is used to determine the detection defects of the wafer to be tested based on the detection situation.
[0072] Optionally, obtain a module, specifically for:
[0073] Determine an area to be photographed that includes defects of the wafer to be tested;
[0074] The area to be photographed is photographed to obtain a photographed image.
[0075] Optionally, the identification module is specifically used to:
[0076] When the captured image does not contain annotations, the captured image is input into a preset deep learning network to extract features of defects in the captured image to obtain defect information;
[0077] Based on the defect information, the type and parameter information of the defect are obtained.
[0078] Optionally, the incremental training module is specifically used to:
[0079] Obtain a target training set based on the captured images containing annotations;
[0080] Input the target training set into the preset deep learning network for training, adjust the network parameters in the preset deep learning network, and obtain a new preset deep learning network;
[0081] The annotation includes a defect type annotation and a parameter information annotation, the type annotation is used as the defect type, and the parameter information annotation is used as the parameter information of the defect.
[0082] Optionally, the detection range includes a shape range and an area range, and the parameter information includes a defect shape and a defect area; the determination module is specifically used for:
[0083] When the defect shape is within the shape range and the defect area is within the area range, the detection status of the defect is determined to be pending detection;
[0084] When the defect shape is not within the shape range and / or the defect area is not within the area range, the defect detection status is determined to be not detected.
[0085] Optionally, the detection module is specifically used to:
[0086] When the detection condition is pending detection, the defect is regarded as a detected defect of the wafer to be tested;
[0087] When the detection condition is no detection, it is determined that the detection defect of the wafer to be tested is empty.
[0088] A computing device according to an embodiment of the present application includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, some or all steps of the above-mentioned wafer defect detection method are implemented.
[0089] Among them, the computing device can be a computer, and correspondingly, its program is computer software. The above-mentioned parameters and steps in a computing device of the present application can refer to the parameters and steps in an embodiment of a wafer defect detection method mentioned above, and will not be repeated here.
[0090] In an embodiment of the present application, a computer-readable storage medium is provided, in which instructions are stored. When the instructions are executed, the steps of the above-mentioned wafer defect detection method are executed.
[0091] The computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.
[0092] The technical solution of the embodiment of the present disclosure can be embodied in the form of a software product, which is stored in a storage medium and includes one or more instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method of the embodiment of the present disclosure. The aforementioned computer-readable storage medium may be a non-transitory computer-readable storage medium, including: a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a disk or an optical disk, and other media that can store program codes, or a transient computer-readable storage medium.
[0093] The flowchart and block diagram in the accompanying drawings illustrate the possible architecture, functions and operations of the system, method and computer program product according to various embodiments of the present application. Wherein, each box in the flowchart or block diagram can represent a module, a program segment, or a part of the code, and the above-mentioned module, program segment, or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0094] Those skilled in the art know that the present application can be implemented as a system, method or computer program product. Therefore, the present disclosure can be specifically implemented in the following forms, namely: it can be complete hardware, it can be complete software (including firmware, resident software, microcode, etc.), or it can be a combination of hardware and software, generally referred to as "module" or "system" herein. In addition, in some embodiments, the present application can also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code. Computer-readable storage media can be, for example, but not limited to - electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above.
[0095] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.
[0096] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.
Claims
1. A wafer defect detection method, characterized in that: include: Acquire a captured image containing defects of the wafer to be tested; When the captured image does not contain any annotation, obtaining the type and parameter information of the defect based on a preset deep learning network; When the captured image contains a label, incrementally training the preset deep learning network using the captured image to obtain a new preset deep learning network, and obtaining the type and parameter information of the defect; Determining the detection status of the defect based on the detection range corresponding to the type and the parameter information; Based on the detection situation, the detected defects of the wafer to be tested are determined.
2. The method according to claim 1, characterized in that The step of acquiring a captured image containing defects of the wafer to be tested comprises: Determine an area to be photographed that includes defects of the wafer to be tested; The area to be photographed is photographed to obtain a photographed image.
3. The method according to claim 1, characterized in that When the captured image does not contain a label, obtaining the type and parameter information of the defect based on a preset deep learning network includes: When the captured image does not contain any annotation, inputting the captured image into a preset deep learning network, performing feature extraction on the defect in the captured image, and obtaining defect information; Based on the defect information, the type and parameter information of the defect are obtained.
4. The method according to claim 1, characterized in that When the captured image contains a label, incremental training is performed on the preset deep learning network using the captured image to obtain a new preset deep learning network, and the type and parameter information of the defect are obtained, including: Obtaining a target training set based on the captured images containing annotations; Inputting the target training set into the preset deep learning network for training, adjusting the network parameters in the preset deep learning network, and obtaining a new preset deep learning network; The annotation includes a type annotation and a parameter information annotation of the defect, the type annotation is used as the type of the defect, and the parameter information annotation is used as the parameter information of the defect.
5. The method according to any one of claims 1 to 4, characterized in that: The detection range includes a shape range and an area range, and the parameter information includes a defect shape and a defect area; and determining the detection condition of the defect based on the detection range corresponding to the type and the parameter information includes: When the defect shape is within the shape range and the defect area is within the area range, determining that the detection status of the defect is to be detected; When the defect shape is not within the shape range, and / or the defect area is not within the area range, it is determined that the detection status of the defect is not detected.
6. The method according to any one of claims 1 to 4, characterized in that: The step of determining the detected defects of the wafer to be tested based on the detected conditions includes: When the detection condition is to be detected, taking the defect as a detected defect of the wafer to be tested; When the detection condition is no detection, it is determined that the detection defect of the wafer to be tested is empty.
7. A wafer defect detection device, characterized in that: A wafer defect detection method as described in any one of claims 1 to 6 is applied.
8. A wafer defect detection system, characterized in that: include: An acquisition module, used for acquiring a photographed image containing defects of the wafer to be tested; An identification module, used for obtaining the type and parameter information of the defect based on a preset deep learning network when no annotation is included in the captured image; An incremental training module, used to perform incremental training on the preset deep learning network using the captured image when the captured image contains annotations, to obtain a new preset deep learning network, and to obtain the type and parameter information of the defect; A determination module, used to determine the detection status of the defect based on the detection range corresponding to the type and the parameter information; The detection module is used to determine the detection defects of the wafer to be tested based on the detection situation.
9. A computing device comprising a memory, a processor, and a program stored in the memory and running on the processor, characterized in that: When the processor executes the program, the steps of a wafer defect detection method as described in any one of claims 1 to 6 are implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed on a terminal device, the terminal device executes the steps of a wafer defect detection method as described in any one of claims 1 to 6.
Citation Information
Cited By
Wafer labeling method and device
CN120909480A