Wafer detection method, system and device, storage medium and computer program product

Abstract

The invention discloses a wafer appearance defect detection method, which is used for solving the problem that the existing wafer detection scheme is low in detection efficiency and poor in detection accuracy due to the fact that the existing wafer appearance defect detection mainly depends on manual observation and judgment. The method comprises the following steps: carrying out preset times of image acquisition on a to-be-detected wafer material under multiple light sources to obtain multiple groups of to-be-detected wafer images corresponding to the to-be-detected material; determining defect data and normal data corresponding to each group of to-be-detected wafer images according to the defect detection model and the positioning model; respectively merging the defect data and the normal data corresponding to each group of to-be-detected wafer images to obtain a single image detection result corresponding to each group of to-be-detected wafer images; and combining the single image detection results corresponding to each group of to-be-detected wafer images to obtain detection data corresponding to the to-be-detected wafer materials, and determining whether the to-be-detected wafer materials have appearance defects or not according to the detection data.

Description

Technical Field

[0001] The present application relates to the technical field of image detection, and in particular, to a method, a system, a device, a storage medium, and a computer program product for detecting wafer appearance defects. Background Art

[0002] As chips play an increasingly crucial role in the development of the country's high-tech industries, wafers, as their main raw material, have become extremely important. A wafer is essentially made of high-purity polysilicon through specific processes. This process involves dissolving the polysilicon and doping it with silicon crystal seeds, followed by a carefully controlled stretching process to form a cylindrical single-crystal silicon. The single-crystal silicon ingot is then subjected to fine grinding, polishing, and slicing processes to finally produce silicon wafer slices, which are what the industry calls wafers.

[0003] As a basic component in integrated circuit manufacturing, the surface quality of a wafer directly affects the performance of the final product. Therefore, wafer appearance defect detection has become a crucial part of the semiconductor manufacturing process. Wafer appearance defect detection aims to accurately identify and locate potential defects, such as dirt, scratches, flips, dark cracks, and chipping, through a comprehensive inspection of the wafer surface, so as to ensure that the quality and stability of the wafer meet high standards. These appearance defects may originate from any link in the manufacturing process. If not discovered and processed in time, they will easily lead to a decline in circuit performance and even failure, posing a certain threat to the overall quality and reliability of the product.

[0004] Common wafer appearance defect detection methods include observing the wafer surface through an optical microscope, observing the surface topography with the help of high-resolution imaging technology, and detecting surface defects and contamination by measuring the optical reflection characteristics of the wafer surface.

[0005] However, due to the tiny size of the die in the wafer, the defect features are often at the single-digit pixel level. Traditional wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and accuracy, and being easily affected by the subjective factors of the operator.

[0006] Therefore, there is an urgent need for a method that can efficiently detect wafer appearance defects while ensuring the accuracy of wafer appearance defect detection. Summary of the Invention

[0007] Embodiments of the present application provide a method for detecting wafer appearance defects, which is used to solve the problems that due to the tiny size of the die in the wafer, the defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in the existing wafer detection solutions.

[0008] The embodiments of the present application also provide a wafer appearance defect detection system, which is used to solve the problems that since the die size in the wafer is tiny and its defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in the existing wafer detection solutions.

[0009] The embodiments of the present application also provide a wafer appearance defect detection device, which is used to solve the problems that since the die size in the wafer is tiny and its defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in the existing wafer detection solutions.

[0010] The embodiments of the present application also provide a computer-readable storage medium, which is used to solve the problems that since the die size in the wafer is tiny and its defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in the existing wafer detection solutions.

[0011] A computer program product, which is used to solve the problems that since the die size in the wafer is tiny and its defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in the existing wafer detection solutions.

[0012] The embodiments of the present application adopt the following technical solutions: A wafer appearance defect detection method includes: performing image acquisition on a to-be-detected wafer material a preset number of times under multiple light sources to obtain multiple groups of to-be-detected wafer images corresponding to the to-be-detected material; sequentially performing target detection on the multiple groups of to-be-detected images according to a pre-trained defect detection model and a positioning model to determine the defect data and normal data corresponding to each group of to-be-detected wafer images; respectively merging the defect data and the normal data corresponding to each group of to-be-detected wafer images to obtain a single-image detection result corresponding to each group of to-be-detected wafer images; merging the single-image detection results corresponding to each group of to-be-detected wafer images to obtain the detection data corresponding to the to-be-detected wafer material, and determining whether there are appearance defects on the to-be-detected wafer material according to the detection data.

[0013] A wafer appearance defect detection system, comprising: an image acquisition unit, configured to perform image acquisition on a wafer material to be detected a preset number of times under multiple light sources, so as to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; a detection unit, configured to perform object detection on the multiple groups of images to be detected in sequence according to a pre-trained defect detection model and a positioning model, so as to determine defect data and normal data corresponding to each group of the wafer images to be detected; a single-image detection unit, configured to respectively merge the defect data and the normal data corresponding to each group of the wafer images to be detected, so as to obtain a single-image detection result corresponding to each group of the wafer images to be detected; and an appearance defect recognition unit, configured to merge the single-image detection results corresponding to each group of the wafer images to be detected, so as to obtain detection data corresponding to the wafer material to be detected, and determine whether there is an appearance defect in the wafer material to be detected according to the detection data.

[0014] A wafer appearance defect detection device, comprising: a processor; and a memory arranged to store computer-executable instructions, the executable instructions, when executed, causing the processor to perform the following operations: performing image acquisition on a wafer material to be detected a preset number of times under multiple light sources, so as to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; performing object detection on the multiple groups of images to be detected in sequence according to a pre-trained defect detection model and a positioning model, so as to determine defect data and normal data corresponding to each group of the wafer images to be detected; respectively merging the defect data and the normal data corresponding to each group of the wafer images to be detected, so as to obtain a single-image detection result corresponding to each group of the wafer images to be detected; and merging the single-image detection results corresponding to each group of the wafer images to be detected, so as to obtain detection data corresponding to the wafer material to be detected, and determining whether there is an appearance defect in the wafer material to be detected according to the detection data.

[0015] A computer-readable storage medium stores one or more programs. When the one or more programs are executed by an electronic device including a plurality of application programs, the electronic device is caused to perform the following operations: perform image acquisition on a wafer material to be detected a preset number of times under multiple light sources to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; sequentially perform object detection on the multiple groups of images to be detected according to a pre-trained defect detection model and a positioning model to determine defect data and normal data corresponding to each group of the wafer images to be detected; respectively merge the defect data and the normal data corresponding to each group of the wafer images to be detected to obtain a single-image detection result corresponding to each group of the wafer images to be detected; merge the single-image detection results corresponding to each group of the wafer images to be detected to obtain detection data corresponding to the wafer material to be detected, and determine whether there are appearance defects in the wafer material to be detected according to the detection data.

[0016] A computer program product includes a computer program which, when executed by a processor, implements: performing image acquisition on a wafer material to be detected a preset number of times under multiple light sources to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; sequentially performing object detection on the multiple groups of images to be detected according to a pre-trained defect detection model and a positioning model to determine defect data and normal data corresponding to each group of the wafer images to be detected; respectively merge the defect data and the normal data corresponding to each group of the wafer images to be detected to obtain a single-image detection result corresponding to each group of the wafer images to be detected; merge the single-image detection results corresponding to each group of the wafer images to be detected to obtain detection data corresponding to the wafer material to be detected, and determine whether there are appearance defects in the wafer material to be detected according to the detection data.

[0017] The above at least one technical solution adopted in the embodiments of the present application can achieve the following beneficial effects: Using the wafer appearance defect detection method provided by the embodiments of the present application, the wafer defect detection system can collect images of the to-be-detected wafer material a preset number of times under multiple light sources, obtaining multiple groups of to-be-detected wafer images corresponding to the to-be-detected material. Then, according to the pre-trained defect detection model and positioning model, target detection is sequentially performed on the multiple groups of to-be-detected images to determine the defect data and normal data corresponding to each group of to-be-detected wafer images. By respectively merging the defect data and the normal data corresponding to each group of the to-be-detected wafer images, a single-image detection result corresponding to each group of to-be-detected wafer images is obtained. Finally, by merging the single-image detection results corresponding to each group of to-be-detected wafer images, the detection data corresponding to the to-be-detected wafer material is obtained, and it is determined whether there are appearance defects in the to-be-detected wafer material according to the detection data. Using the wafer appearance defect detection method provided by the embodiments of the present application, on the one hand, by collecting images under multiple light sources, it can ensure that the collected images have sufficient clarity and uniform lighting, and thus can improve the accuracy of subsequent image detection to a certain extent; in addition, by performing target detection on the images collected under multiple light sources and merging the target detection results, it is possible to supplement the missed detection targets and further improve the accuracy of image detection; on the other hand, the embodiments of the present application can automatically perform wafer defect detection through the pre-trained defect detection model and positioning model, avoiding the problems that the traditional wafer detection scheme is easily affected by human factors, resulting in low accuracy of the wafer detection result and low detection efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings: Figure 1 is a specific flowchart of a wafer appearance defect detection method provided by an embodiment of the present application; Figure 2 is a specific structural diagram of a wafer appearance defect detection system provided by an embodiment of the present application; Figure 3 is a specific structural diagram of a wafer appearance defect detection device provided by an embodiment of the present application; Figure 4 is a specific structural diagram of a wafer appearance defect detection device provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments of this application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of them. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in this application without creative efforts belong to the scope of protection of this application.

[0020] An embodiment of this application provides a method for detecting wafer appearance defects, which is used to solve the problems that in existing wafer appearance defect detection, due to the tiny die size in the wafer and its defect features often being at the single-digit pixel level, and the existing wafer appearance defect detection mainly relying on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy in existing wafer detection solutions.

[0021] The execution subject of the wafer appearance defect detection method provided in the embodiment of this application can be, but is not limited to, at least one of a quality inspection system, a wafer detection system, or an appearance defect detection system, etc.

[0022] For ease of description, in the following, the execution subject of this method is taken as an example of a wafer appearance defect detection system to introduce the implementation manner of this method. It can be understood that taking the execution subject of this method as a wafer appearance defect detection system is only an exemplary illustration and should not be construed as a limitation to this method.

[0023] The schematic diagram of the specific implementation process of the wafer appearance defect detection method provided in this application is as Figure 1 shown, and mainly includes the following steps: Step 11, perform image acquisition on the wafer material to be detected a preset number of times under multiple light sources to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; In the embodiment of this application, the multiple light sources may refer to a ring light source and a dot light source. In one implementation manner, the wafer appearance defect detection system can, when the ring light source and the dot light source are respectively turned on, use a high-resolution industrial intelligent camera to collect image data of the wafer material to be detected placed on a disk, and then respectively obtain a group of ring light source images and dot light source images corresponding to the wafer material to be detected.

[0024] In one implementation manner, the layout structure among the multiple light source illumination device, the high-resolution industrial intelligent camera, and the wafer material to be detected used in the embodiment of this application is as Figure 2As shown in the figure, the high-resolution industrial intelligent camera consists of two parts: a high-resolution camera 201 and a lens 202. The distance between the lens 202 and the wafer material 205 to be detected can be 210 ± 5 mm; the multi-light source illumination device consists of two parts: a dot light source 203 and an annular light source 204. The distance between the multi-light source illumination device and the wafer material to be detected can be 40 ± 5 mm.

[0025] In the embodiment of the present application, the wafer appearance defect detection system can collect wafer images through the following method: collect an image of the wafer material to be detected under the annular light source to obtain an annular light source image corresponding to the material to be detected; collect an image of the wafer material to be detected under the dot light source to obtain a dot light source image corresponding to the material to be detected, and use the annular light source image and the dot light source image as a set of wafer images to be detected; after completing the collection of a set of wafer images to be detected, move the wafer material to be detected to the next detection point, and respectively collect images of the wafer material to be detected under the annular light source and the dot light source; move the wafer material to be detected along a preset trajectory for a preset number of times, and respectively collect images of the wafer material to be detected under the annular light source and the dot light source at each detection point to obtain multiple sets of wafer images to be detected corresponding to the material to be detected.

[0026] In one implementation, the wafer appearance defect detection system can control the high-resolution industrial intelligent camera to collect images of the wafer material to be detected respectively when the annular light source and the dot light source are turned on, and obtain a set of wafer images to be detected composed of an annular light source image and a dot light source image (in one implementation, the resolution of the two images can be 5120××5120). After completing the collection of a set of wafer images to be detected through two image acquisitions, the wafer appearance defect detection system can control the disk on which the wafer material to be detected is placed to move to the next detection point (i.e., the next image acquisition point) according to a preset movement path (such as an S-shaped path), and repeat turning on the annular light source and the dot light source, and then complete the image acquisition of the wafer material to be detected at the second detection point. The above operations are cyclically executed until all areas of the surface of the wafer material to be detected are covered by the completed detection points, so as to ensure that every part of the wafer material to be detected is detected, and at the same time reduce the possibility of repeated detection and missed detection.

[0027] It should be noted here that, in one implementation, the wafer appearance defect detection system can control the high-resolution industrial intelligent camera to take 36 S-shaped cyclic photos for combined imaging with two lights, and collect 72 images with a resolution of 5120×5120 pixels under the annular light source and the dot light source, so as to provide high-quality visual image data for the subsequent defect detection process.

[0028] In one embodiment, an STC-CL25M industrial camera with 25 million pixels from Sentech can be used to collect images of the wafers to be detected. In the embodiments of the present application, there are no specific limitations on what kind of industrial camera is used, what kind of moving path is followed, and the number of times of image collection.

[0029] In addition, it should be noted here that in order to improve the model processing efficiency and reduce the waiting period of the model, in the embodiments of the present application, the wafer appearance defect detection system can crop the collected images. Specifically, the annular light source image used for the subsequent positioning model can be cropped from 5120×5120 pixels to 640×640 pixels, and the annular light source image used for the annular light source defect detection model can be cropped from 5120×5120 pixels to 2696×2696 pixels, with an overlap of 136; while the dot light source image used for the dot light source defect detection model can be cropped from 5120×5120 pixels to 2696×2696 pixels, with an overlap of 136.

[0030] Step 12: According to the pre-trained defect detection model and positioning model, perform target detection on multiple groups of images to be detected collected in Step 11 in sequence, and determine the defect data and normal data corresponding to each group of wafer images to be detected. It should be noted here that before performing wafer appearance defect detection, the wafer appearance defect detection system can pre-train the positioning model and the defect detection model through an offline training method. In one embodiment, the wafer appearance defect detection system can specifically perform model offline training according to the following steps, including: Sub-step 1201: Construction and storage of the data set; The wafer appearance defect detection system can control the high-resolution industrial intelligent camera to collect images of the wafers for training a preset number of times respectively under the condition of turning on the annular light source and the dot light source.

[0031] Sub-step 1202: Image cropping; The wafer appearance defect detection system crops the multi-light source images collected by performing Sub-step 1201. Specifically, the annular light source image used for the subsequent positioning model can be cropped from 5120×5120 pixels to 640×640 pixels, and the annular light source image used for the annular light source defect detection model can be cropped from 5120×5120 pixels to 2696×2696 pixels, with an overlap of 136; while the dot light source image used for the dot light source defect detection model can be cropped from 5120×5120 pixels to 2696×2696 pixels, with an overlap of 136.

[0032] Step 1203: Image annotation; The wafer appearance defect detection system can use data annotation software to perform rectangular box annotation on the defective wafer particles in the annular light source image and the dot light source image after cropping by executing sub-step 1202, convert the annotation data into a fixed format, and divide it into a training set and a validation set.

[0033] Meanwhile, the wafer normal particle positioning detection system can use data annotation software to perform rectangular box annotation on the similar normal wafer particles in the annular light source image after cropping by executing sub-step 1202, convert the annotation data into a fixed format, and divide it into a training set and a validation set.

[0034] Sub-step 1204: Train a positioning model according to the training set obtained by executing step 1203; It should be noted here that since in the annular light source image, the wafer die features are similar and the boundaries are clear, which is conducive to positioning detection. Therefore, in the embodiment of the present application, the wafer appearance defect detection system can use the annular light source images in the training set to train the positioning model.

[0035] In addition, it should be noted here that in the initial model training stage, the learning rate of the model can be set to 0.01, the batch size can be set to 8, and a training plan including 10,000 training steps is adopted to improve the training efficiency of the model and lay a solid foundation for the optimization of the model performance. At the same time, during this training process, data enhancement strategies can also be performed on the collected image data through methods such as scaling, translation, and flipping, so as to improve the model's recognition ability for targets under variable background conditions, thereby enhancing its generalization performance.

[0036] Specifically, in one implementation, the wafer appearance defect detection system can train the positioning model according to the following method, including: inputting the annular light source images in the training image set into the original positioning model for training, and outputting multiple candidate boxes; sequentially determining the loss value between each candidate box and the target box, and updating the parameters of the original positioning model according to the loss value to obtain the positioning model.

[0037] It should be noted here that the embodiments of the present application do not limit the specific calculation methods for calculating the loss value and how to update the parameters of the original positioning model according to the loss value.

[0038] Sub-step 1205: Train a defect detection model according to the training set obtained by executing step 1203; It should be noted that in the embodiments of the present application, for the multi-light source images collected (ring light source images and dot light source images), the embodiments of the present application can pre-train defect detection models for different light source images. For example, in the embodiments of the present application, a ring light source defect detection model for ring light source images and a dot light source defect detection model for dot light source images can be pre-trained.

[0039] In one implementation, the wafer appearance defect detection system can use the ring light source images in the training set to train the ring light source defect detection model. In the embodiments of the present application, the ring light source defect detection model is trained according to the following method, including: performing image acquisition on the training wafer material a preset number of times under the ring light source to obtain a set of ring light source images corresponding to the training wafer material; performing target defect box annotation on each ring light source image in the set of ring light source images, and dividing the set of ring light source images into a ring training image set and a ring verification image set; inputting the ring light source images in the ring training image set into the original ring light source defect detection model for training to output a plurality of candidate defect boxes; sequentially determining the second loss value between each candidate defect box and the target defect box; and updating the parameters of the original ring light source defect detection model according to the second loss value to obtain the ring light source defect detection model.

[0040] In addition, in one implementation, the wafer appearance defect detection system can also use the dot light source images in the training set to train the dot light source defect detection model. Specifically, in the embodiments of the present application, the dot light source defect detection model is trained according to the following method, including: inputting the dot light source images in the training image set into the original dot light source defect detection model for training to output a plurality of second candidate defect boxes; sequentially determining the third loss value between each candidate defect box and the target defect box; and updating the parameters of the original dot light source defect detection model according to the third loss value to obtain the defect detection model.

[0041] It should be noted that the embodiments of the present application do not limit the specific calculation methods for calculating the loss value and how to update the parameters of the original defect detection model according to the loss value.

[0042] After completing the training of the positioning model and the defect detection model through the above sub-steps, the wafer appearance defect detection system can sequentially input the multiple groups of images to be detected collected by performing step 11 into the above models, and use the above models to detect the ring light source images and dot light source images in each group of images to be detected, and then output the defect data and normal data corresponding to each group of images to be detected.

[0043] Specifically, in the embodiments of the present application, the circular appearance defect detection system can perform object detection on multiple groups of images to be detected collected by executing step 11 according to the following method, including: respectively detecting the annular light source image and the dot light source image in a group of wafers to be detected images according to the pre-trained defect detection model, to obtain the first defect data corresponding to the annular light source image and the second defect data corresponding to the dot light source image, wherein the defect data includes the position information of the defective wafer particles in the wafer image to be detected and the defect type corresponding to the defective wafer particles; detecting the annular light source image in the wafer image to be detected according to the pre-trained positioning model, to obtain the normal data corresponding to the annular light source image, wherein the normal data includes the position information of the normal wafer particles in the wafer image to be detected and the type corresponding to the normal wafer particles.

[0044] Specifically, in the embodiments of the present application, the wafer appearance defect detection system can use the trained defect detection model to accurately detect the annular light source image and the dot light source image in each group of images to be detected collected by executing step 11, and then output the position information (such as the center point position of the defective wafer particle) and defect category of the defective wafer particles identified in each group of images to be detected, and record them respectively as follows: Defect detection result of the annular light source image: S1{center point position: [(x1, y1), (x2, y2),...]; defect category: [scratch,... ]}.

[0045] Defect detection result of the dot light source image: S2{center point position: [(x1, y1), (x2, y2),...]; defect category: [scratch,... ]}.

[0046] At the same time, the circular appearance defect detection system can use the trained positioning model to accurately detect the annular light source image in each group of images to be detected collected by executing step 11, and then output the position information (such as the center point position of the wafer particle) and object category of the normal wafer particles identified in each group of images to be detected, and record them respectively as follows: Positioning result: S3{center point position: [(x1, y1), (x2, y2),...]; category: [tap,... ]}.

[0047] Step 13, respectively merge the defect data and the normal data corresponding to each group of the wafers to be detected images obtained by executing step 12, to obtain the single image detection result corresponding to each group of the wafers to be detected images; In an embodiment of the present application, the wafer appearance defect detection system may first merge the defect detection results output by the defect detection model for each group of images to be detected to obtain a defect merging result, and fuse the defect merging result with the positioning result output by the positioning model for the group of images to be detected, so as to obtain a single-image detection result corresponding to the group of images to be detected.

[0048] In an embodiment of the present application, the wafer appearance defect detection system may obtain a single-image detection result corresponding to each group of wafers to be detected according to the following method, including: merging the first defect data and the second defect data corresponding to a group of wafer images to be detected according to the non-maximum suppression algorithm to obtain third defect data corresponding to the wafer images to be detected; merging the third defect data with the normal data corresponding to the wafer images to be detected according to the non-maximum suppression algorithm to obtain a single-image detection result corresponding to each group of the wafers to be detected.

[0049] It should be noted here that there are detection items for the same type of defects in the pre-trained ring light source defect detection model and the dot light source defect detection model. The purpose is to complement the missed detections of the two defect models to improve the detection ability. In this case, after using the above two defect models to detect the same group of wafer images to be detected respectively, there may be a situation where there are multiple detection frames at the same chip position. Therefore, in an embodiment of the present application, the wafer appearance defect detection system may screen and merge the repeated defective wafer particles in S1 and S2 output by the ring light source defect detection model and the dot light source defect detection model for the same group of images to be detected according to the non-maximum suppression algorithm (Non-Maximum Suppression, NMS), and select the best defective wafer particle S4 {center point positions: [(x1, y1), (x2, y2),...]; categories: [scratch,..special]} through priority to improve the accuracy and efficiency of target detection.

[0050] Subsequently, the wafer appearance defect detection system may screen and merge S3 and S4 according to the NMS algorithm to obtain a single-image detection result corresponding to each group of the wafers to be detected, denoted as S5 {center point positions: [(x1, y1), (x2, y2),...]; categories: [scratch, tap,...special]}. It should be noted here that if the wafer particle is detected as the non-defective category tap, the tap result is retained. If the wafer particle is detected as both tap and the defective category at the same time, the defective result is retained.

[0051] Step 14: Combine the single-image detection results corresponding to each group of wafers to be detected obtained by performing the above Step 13 to obtain the detection data corresponding to the wafers to be detected, and determine whether there are appearance defects in the wafers to be detected according to the detection data.

[0052] Specifically, in the embodiment of the present application, the wafer appearance defect detection system can use the coordinate relationship mapping and reduction method to sequentially combine and restore the single-image detection results corresponding to each group of wafers to be detected, and locate and restore the detected defective wafer particles onto the wafers to be detected. Furthermore, the detection data corresponding to the wafers to be detected can be determined by counting the number of normal wafer particles and defective wafer particles on the wafers to be detected.

[0053] When the wafer appearance defect detection system determines that the number of defective wafer particles in the wafers to be detected is within the preset threshold according to the detection data, a crystal picking machine can be used to remove the defective wafer particles in the wafers to be detected by top-down picking, to obtain the second wafers to be detected, and the defect re-inspection of the second wafers to be detected can be completed by repeatedly executing the above detection scheme.

[0054] When the wafer appearance defect detection system determines that the number of defective wafer particles in the wafers to be detected is greater than the preset threshold according to the detection data, the wafers to be detected will be transferred to manual processing.

[0055] Using the wafer appearance defect detection method provided by the embodiments of the present application, the wafer defect detection system can collect images of the to-be-detected wafer material a preset number of times under multiple light sources, obtain multiple groups of to-be-detected wafer images corresponding to the to-be-detected material, and then, according to the pre-trained defect detection model and positioning model, perform object detection on multiple groups of to-be-detected images in sequence, determine the defect data and normal data corresponding to each group of to-be-detected wafer images, obtain a single-image detection result corresponding to each group of to-be-detected wafer images by respectively merging the defect data and the normal data corresponding to each group of the to-be-detected wafer images, and finally, obtain the detection data corresponding to the to-be-detected wafer material by merging the single-image detection results corresponding to each group of to-be-detected wafer images, and determine whether there are appearance defects on the to-be-detected wafer material according to the detection data. Using the wafer appearance defect detection method provided by the embodiments of the present application, on the one hand, by collecting images under multiple light sources, it can ensure that the collected images have sufficient clarity and uniform lighting, and thus can improve the accuracy of subsequent image detection to a certain extent; in addition, by performing object detection on the images collected under multiple light sources and merging the object detection results, it can achieve the supplement of undetected objects and further improve the accuracy of image detection; on the other hand, the embodiments of the present application can automatically perform wafer defect detection through the pre-trained defect detection model and positioning model, avoiding the problems that the traditional wafer detection scheme is easily affected by human factors, resulting in low accuracy of wafer detection results and low detection efficiency.

[0056] In one implementation manner, the embodiments of the present application further provide a wafer appearance defect detection system, which is used to solve the problems that due to the tiny size of the die in the wafer, its defect features are often at the single-digit pixel level, and the existing wafer appearance defect detection mainly relies on manual observation and judgment, resulting in low detection efficiency and poor detection accuracy of the existing wafer detection scheme. The specific structural schematic diagram of the wafer appearance defect detection system is as Figure 3 shown, and includes: an image acquisition unit 31, a detection unit 32, a single-image detection unit 33, and an appearance defect recognition unit 34.

[0057] Among them, the image acquisition unit 31 is configured to collect images of the to-be-detected wafer material a preset number of times under multiple light sources, and obtain multiple groups of to-be-detected wafer images corresponding to the to-be-detected material; The detection unit 32 is configured to perform object detection on multiple groups of the to-be-detected images in sequence according to the pre-trained defect detection model and positioning model, and determine the defect data and normal data corresponding to each group of the to-be-detected wafer images; The single-image detection unit 33 is configured to merge the defect data and the normal data corresponding to each group of the to-be-detected wafer images respectively, so as to obtain the single-image detection results corresponding to each group of the to-be-detected wafer materials; The appearance defect recognition unit 34 is configured to merge the single-image detection results corresponding to each group of the to-be-detected wafer materials, so as to obtain the detection data corresponding to the to-be-detected wafer materials, and determine whether there are appearance defects on the to-be-detected wafer materials according to the detection data.

[0058] In an implementation manner, the multi-light sources include a ring light source and a dot light source. The image acquisition unit 31 is specifically configured to: acquire an image of the to-be-detected wafer material under the ring light source to obtain a ring light source image corresponding to the to-be-detected material; acquire an image of the to-be-detected wafer material under the dot light source to obtain a dot light source image corresponding to the to-be-detected material, and use the ring light source image and the dot light source image as a group of to-be-detected wafer images; after completing the acquisition of a group of to-be-detected wafer images, move the to-be-detected wafer material to the next detection point, and respectively acquire images of the to-be-detected wafer material under the ring light source and the dot light source; move the to-be-detected wafer material along a preset trajectory for a preset number of times, and respectively acquire images of the to-be-detected wafer material under the ring light source and the dot light source at each detection point, so as to obtain multiple groups of to-be-detected wafer images corresponding to the to-be-detected material.

[0059] In an implementation manner, the detection unit 32 is specifically configured to: respectively detect the ring light source image and the dot light source image in a group of to-be-detected wafer images according to a pre-trained defect detection model, so as to obtain first defect data corresponding to the ring light source image and second defect data corresponding to the dot light source image, where the defect data includes the position information of the defective wafer particles in the to-be-detected wafer image and the defect type corresponding to the defective wafer particles; detect the ring light source image in the to-be-detected wafer image according to a pre-trained positioning model, so as to obtain normal data corresponding to the ring light source image, where the normal data includes the position information of the normal wafer particles in the to-be-detected wafer image and the type corresponding to the normal wafer particles.

[0060] In an implementation manner, the single-image detection unit 33 is specifically configured to: merge the first defect data and the second defect data corresponding to a group of to-be-detected wafer images according to the non-maximum suppression algorithm, so as to obtain third defect data corresponding to the to-be-detected wafer image; merge the third defect data with the normal data corresponding to the to-be-detected wafer image according to the non-maximum suppression algorithm, so as to obtain the single-image detection results corresponding to each group of the to-be-detected wafer materials.

[0061] In one embodiment, the appearance defect recognition unit 34 is specifically configured to: by means of a coordinate relationship mapping and restoration method, sequentially merge and restore the single-image detection results corresponding to each group of the wafers to be detected, and locate and restore the detected defective wafer particles onto the wafers to be detected; and determine the detection data corresponding to the wafers to be detected by counting the number of normal wafer particles and defective wafer particles on the wafers to be detected.

[0062] In one embodiment, when it is determined according to the detection data that the wafers to be detected have appearance defects, the appearance defect recognition unit 34 is further configured to: when it is determined according to the detection data that the number of defective wafer particles in the wafers to be detected is within a preset threshold, remove the defective wafer particles from the wafers to be detected to obtain second wafers to be detected, and perform defect re-inspection on the second wafers to be detected; when it is determined according to the detection data that the number of defective wafer particles in the wafers to be detected is greater than the preset threshold, perform manual processing on the wafers to be detected.

[0063] In one embodiment, it further includes a model training unit, which is specifically configured to: collect images of the training wafers a preset number of times under a ring light source to obtain a ring light source image set corresponding to the training wafers; perform target box annotation on each ring light source image in the ring light source image set, and divide the ring light source image set into a training image set and a validation image set; input the ring light source images in the training image set into an original positioning model for training to output a plurality of candidate boxes; sequentially determine the loss value between each candidate box and the target box; and update the parameters of the original positioning model according to the loss value to obtain the positioning model.

[0064] In one embodiment, the model training unit is specifically configured to: collect images of the training wafers a preset number of times under a ring light source to obtain a ring light source image set corresponding to the training wafers; perform target defect box annotation on each ring light source image in the ring light source image set, and divide the ring light source image set into a ring training image set and a ring validation image set; input the ring light source images in the ring training image set into an original ring light source defect detection model for training to output a plurality of candidate defect boxes; sequentially determine the second loss value between each second candidate defect box and the target defect box; and update the parameters of the original ring light source defect detection model according to the second loss value to obtain the ring light source defect detection model.

[0065] In one implementation, the model training unit is specifically configured to: collect images of the training wafer material a preset number of times under a dot light source to obtain a dot light source image set corresponding to the training wafer material; label target defect boxes for each dot light source image in the dot light source image set, and divide the dot light source image set into a training image set and a validation image set; input the dot light source images in the training image set into the original dot light source defect detection model for training, and output a plurality of second candidate defect boxes; sequentially determine a third loss value between each candidate defect box and the target defect box; and update the parameters of the original dot light source defect detection model according to the third loss value to obtain the dot light source defect detection model.

[0066] Using the wafer appearance defect detection system provided by the embodiments of the present application, the wafer defect detection system can collect images of the wafer material to be detected a preset number of times under multiple light sources to obtain multiple groups of wafer images to be detected corresponding to the material to be detected. Then, according to the pre-trained defect detection model and positioning model, target detection is sequentially performed on the multiple groups of images to be detected to determine the defect data and normal data corresponding to each group of wafer images to be detected. By respectively merging the defect data and the normal data corresponding to each group of the wafer images to be detected, a single image detection result corresponding to each group of the wafer images to be detected is obtained. Finally, by merging the single image detection results corresponding to each group of the wafer images to be detected, detection data corresponding to the wafer material to be detected is obtained, and it is determined whether there are appearance defects in the wafer material to be detected according to the detection data. Using the wafer appearance defect detection method provided by the embodiments of the present application, on the one hand, by collecting images under multiple light sources, it can ensure that the collected images have sufficient clarity and uniform lighting, and thus can improve the accuracy of subsequent image detection to a certain extent; in addition, by performing target detection on the images collected under multiple light sources and merging the target detection results, it is possible to supplement the missed detection targets and further improve the accuracy of image detection; on the other hand, the embodiments of the present application can automatically perform defect detection of the wafer through the pre-trained defect detection model and positioning model, avoiding the problems that the traditional wafer detection scheme is easily affected by human factors, resulting in low accuracy of the wafer detection result and low detection efficiency.

[0067] Figure 4 It is a schematic structural diagram of an electronic device according to an embodiment of the present application. Please refer to Figure 4, at the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and a memory. Among them, the memory may include memory, such as high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk memory, etc. Of course, the electronic device may also include other hardware required for other services.

[0068] The processor, network interface, and memory can be interconnected through an internal bus, and the internal bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the sake of convenience of representation, Figure 4 only a two-way arrow is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0069] The memory is used to store programs. Specifically, the program may include program code, and the program code includes computer operation instructions. The memory can include memory and non-volatile memory, and provide instructions and data to the processor.

[0070] The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it, forming a wafer appearance defect detection device at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations: Performing image acquisition on the wafer material to be detected a preset number of times under multiple light sources to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; according to the pre-trained defect detection model and positioning model, sequentially performing target detection on multiple groups of the images to be detected to determine the defect data and normal data corresponding to each group of the wafer images to be detected; respectively merging the defect data and the normal data corresponding to each group of the wafer images to be detected to obtain a single image detection result corresponding to each group of the wafer images to be detected; merging the single image detection results corresponding to each group of the wafer images to be detected to obtain the detection data corresponding to the wafer material to be detected, and determining whether there are appearance defects in the wafer material to be detected according to the detection data.

[0071] The above is as in this application Figure 4The method executed by the electronic device for detecting wafer appearance defects disclosed in the illustrated embodiments can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in the processor or the instructions in the form of software. The above processor may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps, and logic block diagrams disclosed in the embodiments of the present application. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present application can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by a combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method.

[0072] Of course, in addition to the software implementation, the electronic device of the present application does not exclude other implementation manners, such as a logic device or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, and may also be hardware or a logic device.

[0073] The embodiments of the present application also propose a computer-readable storage medium that stores one or more programs. The one or more programs include instructions that, when executed by a portable electronic device including a plurality of application programs, can cause the portable electronic device to execute Figure 1 the method of the illustrated embodiment, and specifically used to perform the following operations: Under multiple light sources, perform image acquisition on the wafer material to be detected for a preset number of times to obtain multiple groups of wafer images to be detected corresponding to the material to be detected; according to the defect detection model and the positioning model obtained by pre-training, perform target detection on multiple groups of the images to be detected in sequence to determine the defect data and normal data corresponding to each group of the wafer images to be detected; respectively merge the defect data and the normal data corresponding to each group of the wafer images to be detected to obtain a single image detection result corresponding to each group of the wafer images to be detected; merge the single image detection results corresponding to each group of the wafer images to be detected to obtain the detection data corresponding to the wafer material to be detected, and determine whether there are appearance defects in the wafer material to be detected according to the detection data.

[0074] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0075] The present invention is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for realizing the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0076] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device realizes the functions specified in Figure 1 one process or multiple processes and / or blocks Figure 1 one block or multiple blocks.

[0077] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are executed on the computer or other programmable apparatus to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process Figure 1 one process or multiple processes and / or blocks Figure 1 steps of the functions specified in one block or multiple blocks.

[0078] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0079] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM) and / or non-volatile memory such as read-only memory (ROM) or flash RAM. The memory is an example of computer-readable media.

[0080] Computer-readable media includes permanent and non-permanent, removable and non-removable media and can store information by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0081] It should also be noted that the term "comprising", "including" or any other variation thereof is intended to cover non-exclusive inclusion, so that a process, method, commodity or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, commodity or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, commodity or device comprising the element.

[0082] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] The above are only the embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the scope of the claims of the present application.

Claims

1. A wafer appearance defect detection method, characterized in that: include: Performing a preset number of image acquisitions on a wafer material to be inspected under multiple light sources to obtain multiple groups of wafer images to be inspected corresponding to the material to be inspected; According to the defect detection model and the positioning model obtained by pre-training, target detection is performed on the multiple groups of images to be detected in turn, and defect data and normal data corresponding to each group of wafer images to be detected are determined; Merging the defect data and the normal data corresponding to each group of wafer images to be inspected respectively to obtain a single image inspection result corresponding to each group of wafer images to be inspected; The single image detection results corresponding to each group of the wafer images to be detected are combined to obtain the detection data corresponding to the wafer material to be detected, and it is determined whether the wafer material to be detected has appearance defects based on the detection data.

2. The method according to claim 1, characterized in that The multiple light sources include: an annular light source and a point light source, and the image acquisition of the wafer material to be inspected is performed a preset number of times under the multiple light sources to obtain multiple groups of wafer images to be inspected corresponding to the material to be inspected, specifically including: Capturing an image of the wafer material to be inspected under an annular light source to obtain an annular light source image corresponding to the material to be inspected; Capturing an image of the wafer material to be inspected under a point light source to obtain a point light source image corresponding to the material to be inspected, and using the annular light source image and the point light source image as a group of wafer images to be inspected; After completing a set of image acquisition of the wafer to be inspected, the wafer material to be inspected is moved to the next inspection point, and the image of the wafer material to be inspected is acquired under a ring light source and a point light source respectively; The wafer material to be inspected is moved a preset number of times along a preset trajectory, and images of the wafer material to be inspected are collected under a ring light source and a point light source at each inspection point to obtain multiple groups of wafer images to be inspected corresponding to the material to be inspected.

3. The method according to claim 1, characterized in that The method of performing target detection on the plurality of groups of images to be detected in sequence according to the defect detection model and the positioning model obtained by pre-training, and determining defect data and normal data corresponding to each group of wafer images to be detected specifically includes: According to the defect detection model obtained by pre-training, a ring light source image and a point light source image in a group of wafer images to be detected are detected respectively to obtain first defect data corresponding to the ring light source image and second defect data corresponding to the point light source image, wherein the defect data includes position information of defective wafer particles in the wafer images to be detected and defect types corresponding to the defective wafer particles; According to the pre-trained positioning model, the annular light source image in the wafer image to be inspected is detected to obtain normal data corresponding to the annular light source image, wherein the normal data includes position information of normal wafer particles in the wafer image to be inspected and the types corresponding to the normal wafer particles.

4. The method according to claim 3, characterized in that The defect data and the normal data corresponding to each group of wafer images to be inspected are merged to obtain a single image inspection result corresponding to each group of wafer images to be inspected, specifically including: According to a non-maximum suppression algorithm, the first defect data and the second defect data corresponding to a group of wafer images to be inspected are merged to obtain third defect data corresponding to the wafer images to be inspected; According to the non-maximum suppression algorithm, the third defect data is merged with the normal data corresponding to the wafer image to be inspected, so as to obtain the single image detection results corresponding to each group of the wafer images to be inspected.

5. The method according to claim 1, characterized in that The step of merging the single image detection results corresponding to each group of wafer images to be detected to obtain the detection data corresponding to the wafer material to be detected specifically includes: By means of a coordinate relationship mapping and restoration method, the single image detection results corresponding to each group of wafer images to be detected are merged and restored in turn, and the detected defective wafer particles are positioned and restored to the wafer material to be detected; The detection data corresponding to the wafer material to be detected is determined by counting the number of normal wafer particles and defective wafer particles on the wafer material to be detected.

6. The method according to claim 1, characterized in that When it is determined according to the inspection data that the wafer material to be inspected has appearance defects, the method further includes: When it is determined according to the detection data that the number of defective wafer particles in the wafer material to be detected is within a preset threshold, the defective wafer particles in the wafer material to be detected are removed to obtain a second wafer material to be detected, and the second wafer material to be detected is re-inspected for defects; When it is determined according to the detection data that the number of defective wafer particles in the wafer material to be detected is greater than a preset threshold, the wafer material to be detected is manually processed.

7. The method according to claim 1, characterized in that Pre-training the positioning model specifically includes: Capturing images of a training wafer material under an annular light source for a preset number of times to obtain an annular light source image set corresponding to the training wafer material; Performing target frame annotation on each annular light source image in the annular light source image set, and dividing the annular light source image set into a training image set and a verification image set; Inputting the annular light source image in the training image set into the original positioning model for training, and outputting a plurality of candidate frames; Determine the loss value between each candidate box and the target box in turn; According to the loss value, the parameters of the original positioning model are updated to obtain the positioning model.

8. The method according to claim 1, characterized in that The defective mode detection model includes a ring light source defective mode detection model and a point light source defective mode detection model, and pre-training the ring light source defective mode detection model specifically includes: Capturing images of a training wafer material under an annular light source for a preset number of times to obtain an annular light source image set corresponding to the training wafer material; Annotating target defect frames for each annular light source image in the annular light source image set, and dividing the annular light source image set into an annular training image set and an annular verification image set; Inputting the annular light source image in the annular training image set into the original annular light source defect detection model for training, and outputting a plurality of candidate defect frames; Determining in sequence a second loss value between each of the candidate defect frames and the target defect frame; According to the second loss value, the parameters of the original annular light source defect detection model are updated to obtain the annular light source defect mode detection model.

9. The method according to claim 8, characterized in that Pre-training the point light source defect mode detection model specifically includes: Performing a preset number of image acquisitions on a training wafer material under a point light source to obtain a point light source image set corresponding to the training wafer material; Performing target defect frame annotation on each point light source image in the point light source image set, and dividing the point light source image set into a training image set and a verification image set; Inputting the point light source images in the training image set into the original point light source defect detection model for training, and outputting a plurality of second candidate defect frames; Determining in sequence a third loss value between each of the candidate defect frames and the target defect frame; According to the third loss value, the parameters of the original point light source defect detection model are updated to obtain the point light source defect model detection model.

10. A wafer appearance defect detection system, characterized in that: include: An image acquisition unit is used to acquire images of a wafer material to be inspected for a preset number of times under multiple light sources to obtain multiple groups of wafer images to be inspected corresponding to the material to be inspected; A detection unit, used to perform target detection on the plurality of groups of images to be detected in sequence according to a defect detection model and a positioning model obtained by pre-training, and determine defect data and normal data corresponding to each group of wafer images to be detected; A single image detection unit, used for merging the defect data and the normal data corresponding to each group of wafer images to be detected, to obtain a single image detection result corresponding to each group of wafer images to be detected; The appearance defect recognition unit is used to merge the single image detection results corresponding to each group of the wafer images to be detected to obtain the detection data corresponding to the wafer material to be detected, and determine whether the wafer material to be detected has appearance defects based on the detection data.

11. A wafer appearance defect detection device, comprising: processor; as well as a memory arranged to store computer executable instructions which, when executed, cause the processor to: Performing a preset number of image acquisitions on a wafer material to be inspected under multiple light sources to obtain multiple groups of wafer images to be inspected corresponding to the material to be inspected; According to the defect detection model and the positioning model obtained by pre-training, target detection is performed on the multiple groups of images to be detected in turn, and defect data and normal data corresponding to each group of wafer images to be detected are determined; Merging the defect data and the normal data corresponding to each group of wafer images to be inspected respectively to obtain a single image inspection result corresponding to each group of wafer images to be inspected; The single image detection results corresponding to each group of the wafer images to be detected are combined to obtain the detection data corresponding to the wafer material to be detected, and it is determined whether the wafer material to be detected has appearance defects based on the detection data.

12. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of application programs, enables the electronic device to execute the wafer appearance defect detection method as described in any one of claims 1 to 9.

13. A computer program product, characterized in that It comprises a computer program, which, when executed by a processor, implements the wafer appearance defect detection method as described in any one of claims 1 to 9.

Images