Electron beam measuring equipment template pattern positioning method and system based on deep learning

Abstract

The invention discloses an electron beam measurement equipment template graph positioning method and system based on deep learning, and the method comprises the steps: 1) obtaining a measurement template picture to construct a target image set, and marking the graph edge of the picture in the target image set; 2) training a deep neural network by using the marked target image set to obtain a graph edge detection inference model; 3) carrying out edge detection on the measurement template picture by utilizing the graph edge detection inference model obtained by training, intercepting a measurement point region, and storing the measurement point region as a template image; 4) carrying out feature extraction on the graphic edge contained in the stored template image to construct a feature vector of the measurement point region; and 5) performing edge detection on the to-be-measured picture by using the graphic edge detection inference model, and positioning the measurement point region of the to-be-measured picture by using the feature vector of the measurement point region of the template image. The method has the advantages of high measurement precision, high measurement efficiency and the like.

Description

Technical Field

[0001] The present invention mainly relates to the field of semiconductor equipment technology, and specifically to a template graphic positioning method and system for electron beam measurement equipment based on deep learning. Background Art

[0002] The vigorous development of the information industry has greatly driven the development of the integrated circuit industry. Electron beam measurement equipment can be used to measure the feature patterns of wafers and masks in the integrated circuit manufacturing process, and is one of the most important process equipment on the integrated circuit production line. The positioning accuracy and positioning robustness of electron beam measurement equipment for feature patterns are its most critical indicators, and the diversity of feature patterns on the wafer and the noise and feature pattern deformation caused by electron beam scanning imaging will lead to the failure of feature pattern positioning. Therefore, in the electron beam measurement process, a stable and high-precision feature pattern positioning algorithm must be designed to ensure the feature pattern positioning accuracy of the electron beam measurement equipment. Summary of the invention

[0003] In view of the technical problems existing in the prior art, the present invention provides a template graphic positioning method and system for electron beam measurement equipment based on deep learning with high positioning accuracy.

[0004] In order to solve the above technical problems, the technical solution proposed by the present invention is:

[0005] A deep learning-based template pattern positioning method for electron beam measurement equipment comprises the following steps:

[0006] 1) Obtain a measurement template image to construct a target image set, and mark the graphic edges of the images in the target image set;

[0007] 2) Use the labeled target image set to train the deep neural network to obtain a graphic edge detection inference model;

[0008] 3) Use the trained graphic edge detection inference model to perform edge detection on the measurement template image, intercept the measurement point area, and save the measurement point area as a template image;

[0009] 4) Extracting features from the edges of the graphics contained in the saved template image to construct a feature vector of the measurement point area;

[0010] 5) Use the graphic edge detection inference model to perform edge detection on the image to be measured, and use the feature vector of the measurement point area of ​​the template image to locate the measurement point area of ​​the image to be measured.

[0011] Preferably, the specific process of step 4) is:

[0012] The contour of the template image is Fourier transformed to obtain its corresponding Fourier shape descriptor vector, and then the contour edge points are downsampled, and the gradient value of each edge point after downsampling is calculated as the gradient feature vector of the edge. Finally, the Fourier shape descriptor vector and the gradient feature vector are used as the global feature description vector and local feature description vector of the template image, respectively.

[0013] Preferably, the gradient magnitude and direction are calculated for each pixel point in an 8*8 neighborhood of the contour edge point, and a gradient histogram in the neighborhood is obtained as a local feature description vector of the contour edge point.

[0014] Preferably, the specific process of step 5) is:

[0015] Firstly, the graphic edge detection inference model is used to perform global edge detection on the image to be measured, and the global contour map of the image to be measured is obtained. Then the template image is slid on the global contour map of the image to be measured. Finally, the measurement point area of ​​the graphic to be measured is located based on the local and global deformable diversity similarity by calculating the feature description vector of the template image and the sliding overlapping area.

[0016] Preferably, the formula for calculating the feature description vector of the template image and the sliding overlap area based on the local and global deformable diversity similarity is:

[0017]

[0018] in is the template image contour point, is the image contour point of the sliding overlapping area, K(NN α (q j ,P)) is the number of points with similar features between the template image contour points and the sliding overlap area image contour points, f p is the Fourier shape descriptor vector of the template image contour, f q is the Fourier shape descriptor vector of the image contour of the sliding overlap area, r j is the distance between the contour points of the sliding overlapping area image and the contour points of the template image, exp is the exponential function, and N is the number of contour points of the sliding overlapping area image.

[0019] Preferably, before step 2), the training set images are denoised using a Neighbor2Blind deep neural network.

[0020] Preferably, the objective function of the Neighbor2Blind deep neural network is:

[0021]

[0022] Among them, f θis the trained denoising network, g1 and g2 are sub-sampled images, y is the noise image, h is the sub-sampled image of the denoised image, Ω y is the blind spot mask, γ and β are weighting coefficients respectively.

[0023] The present invention also discloses a computer program product, comprising a computer program, wherein the computer program executes the steps of the method described above when executed by a processor.

[0024] The present invention further discloses a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the method described above are executed.

[0025] The present invention also discloses a deep learning-based template graphic positioning system for electron beam measurement equipment, comprising a memory and a processor connected to each other, wherein a computer program is stored on the memory, and when the computer program is run by the processor, the steps of the method described above are executed.

[0026] Compared with the prior art, the advantages of the present invention are:

[0027] The electron beam measurement equipment template pattern positioning method based on deep learning of the present invention can not only meet the positioning accuracy of the characteristic pattern on the wafer, but also eliminate the error caused by the noise and deformation generated by electron beam imaging to the characteristic pattern positioning, and realize the online intelligent learning of the measurement equipment through the deep learning method, making the electron beam measurement equipment more intelligent and automated.

[0028] The positioning method and device of the present invention have the advantages of strong anti-interference ability, easy operation, high positioning accuracy, etc., which well meet the needs of electron beam measurement equipment for accurate positioning of measurement point graphics and significantly improve the measurement efficiency of electron beam measurement equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a structural block diagram of the graphic positioning device in an embodiment of the present invention.

[0030] Figure 2 The figure is a flowchart of an embodiment of the graphic positioning method of the present invention.

[0031] Figure 3 The figure is a flowchart of online training reasoning of the graphic positioning device of the present invention in a specific application.

[0032] Figure 4 This is a block diagram of the Neighbor2Blind deep neural network denoising model in the present invention. DETAILED DESCRIPTION

[0033] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.

[0034] like Figure 2 As shown, the electron beam measurement equipment template graphic positioning method based on deep learning provided by an embodiment of the present invention comprises the steps of:

[0035] 1) Obtaining a measurement template image to construct a target image set, and marking the graphic edges of the images in the target image set;

[0036] 2) Use the HED (Holistically-Nested Edge Detection) deep neural network to train the target image set to obtain a graphic edge detection inference model;

[0037] 3) Use the trained graphic edge detection inference model to perform edge detection on the measurement template image, intercept the measurement point area, and save the measurement point area as a template image;

[0038] 4) Extracting features from the edges of the graphics contained in the saved template image to construct a feature vector of the measurement point area;

[0039] 5) Use the graphic edge detection inference model to perform edge detection on the image to be measured, and use the template image to locate the measurement point area.

[0040] The electron beam measurement equipment template graphic positioning method based on deep learning of the present invention adopts the HED deep neural network, which can process the entire image and detect the nested edge structure, thereby realizing high-precision graphic edge detection; through automatic edge detection and image capture, the traditional manual template making process is eliminated, thereby improving efficiency and consistency; through feature extraction of the template image edge, richer information is provided for graphic positioning, thereby improving the accuracy of positioning; using the deep learning model to perform edge detection and positioning on the image to be measured, it realizes fast and accurate mapping from the image to the actual measurement point, thereby improving the efficiency of measurement.

[0041] The electron beam measurement equipment template pattern positioning method based on deep learning of the present invention can not only meet the positioning accuracy of the characteristic pattern on the wafer, but also eliminate the error caused by the noise and deformation generated by electron beam imaging to the characteristic pattern positioning, and realize the online intelligent learning of the measurement equipment through the deep learning method, making the electron beam measurement equipment more intelligent and automated.

[0042] like Figure 1-Figure 3 As shown, an embodiment of the present invention further provides a template graphic positioning system for electron beam measurement equipment based on deep learning, including:

[0043] An acquisition module, used to collect and process secondary electron signals on the wafer to generate a measurement template image;

[0044] Annotation module, used to realize human-computer interaction for edge annotation of target images;

[0045] A training module is used to implement real-time training of the image edge detection inference model of the target image set;

[0046] The reasoning module is electrically connected to the acquisition module, the training module and the annotation module, and is used to realize the positioning of the measurement point area in the measurement image in the above graphic positioning method.

[0047] The graphic positioning system of the present invention is used to execute the positioning method as described above, and also has the advantages of the positioning method as described above.

[0048] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0049] like Figure 1 and Figure 2 As shown, the template pattern positioning method of electron beam measurement equipment based on deep learning includes the following steps:

[0050] 1) Obtaining a measurement template image to construct a target image set, and marking the edges of the graphics contained in the images in the target image set;

[0051] Specifically, the electron beam scanning image is acquired by the acquisition module of the device, and the edge of the pattern is annotated by the annotation module. The edge annotation of the pattern of the target image set may include historical measurement images and measurement positioning failure images.

[0052] 2) Use the HED (Holistically-Nested Edge Detection) deep neural network to train the target image set and obtain the graphic edge detection inference model;

[0053] Specifically, the training set images are first denoised using the Neighbor2Blind deep neural network, and then the labeled target image set is trained using the training module of the device to obtain a graphic edge detection inference model based on the HED deep neural network.

[0054] like Figure 4 As shown in the figure, the Neighbor2Blind deep neural network is trained by collecting noisy pictures to improve the robustness of the denoising network; the objective function of the Neighbor2Blind deep neural network is:

[0055]

[0056] Among them, f θis the trained denoising network, g1 and g2 are sub-sampled images, y is the noise image, h is the sub-sampled image of the denoised image, Ω y is the blind spot mask, γ and β are weighting coefficients respectively.

[0057] 3) Use the trained graphic edge detection inference model to perform edge detection on the measurement template image, and intercept the measurement point area and save it as a template image;

[0058] Specifically, the measurement template image is subjected to global edge detection through the inference module, and then the global contour map of the measurement template image is obtained, and finally the local contour map of the measurement point area is intercepted and saved as the template image.

[0059] 4) Extracting features from the edges of the graphics contained in the saved template image to construct a feature vector of the measurement point area;

[0060] Specifically, the inference module first performs Fourier transform on the contour of the obtained template image to obtain its corresponding Fourier shape descriptor vector, then downsamples the contour edge points, and calculates the gradient value of each edge point after downsampling as the gradient feature vector of the edge, and finally uses the Fourier shape descriptor vector and the gradient feature vector as the global feature description vector and local feature description vector of the template image, respectively.

[0061] The gradient magnitude and direction are calculated for each pixel in the 8*8 neighborhood of the contour edge point, and the gradient histogram in the neighborhood is obtained as the local feature vector of the contour edge point;

[0062] 5) Use the graphic edge detection inference model to perform edge detection on the image to be measured, and use the template image to locate the measurement point area;

[0063] Specifically, the graphic edge detection inference model is first used to perform global edge detection on the image to be measured, and the global contour map of the image to be measured is obtained. Then the template image is slid on the global contour map of the image to be measured. Finally, the measurement point area of ​​the graphic to be measured is located by calculating the similarity between the template image and the sliding overlapping area.

[0064] Specifically, the inference module performs Fourier transform on the contour of the above-mentioned sliding overlapping area to obtain its corresponding Fourier shape descriptor vector, then downsamples the contour edge points, and calculates the gradient histogram of each edge point after downsampling as the gradient feature vector of the edge point, and finally uses the Fourier shape descriptor vector and the gradient feature vector as the global feature description vector and local feature description vector of the sliding overlapping area, respectively.

[0065] Then, the similarity evaluation value between the sliding overlapping area and the template image is evaluated by calculating the local and global deformable diversity similarity between the feature description vector of the sliding overlapping area and the feature description vector of the template image.

[0066] The similarity between the sliding overlap area and the template image is evaluated based on the local and global deformable diversity similarity, and the calculation formula is as follows:

[0067]

[0068] in is the template image contour point, is the image contour point of the sliding overlapping area, K(NN α (q j ,P)) is the number of points with similar features between the template image contour points and the sliding overlap area image contour points, f p is the Fourier shape descriptor vector of the template image contour, f q is the Fourier shape descriptor vector of the image contour of the sliding overlap area, r j is the distance between the contour points of the sliding overlapping area image and the contour points of the template image, exp is the exponential function, and N is the number of contour points of the sliding overlapping area image.

[0069] 6) The image to be measured that failed to be positioned is added back to the target image set, and the graphic edges contained therein are annotated, and then the graphic edge detection reasoning model is retrained, and finally the measurement point area positioning of the image to be measured is completed using the new graphic edge detection reasoning model.

[0070] The present invention has the advantages of strong anti-interference ability, easy operation, high positioning accuracy, etc., which well meets the needs of electron beam measurement equipment for accurate positioning of measurement point graphics and significantly improves the measurement efficiency of electron beam measurement equipment.

[0071] The embodiment of the present invention further discloses a computer program product, including a computer program, which executes the steps of the above method when executed by a processor. The embodiment of the present invention further discloses a computer-readable storage medium, on which a computer program is stored, and the computer program executes the steps of the above method when executed by a processor. The embodiment of the present invention further discloses a template graphic positioning system for electron beam measurement equipment based on deep learning, including a memory and a processor connected to each other, the memory storing a computer program, and the computer program executes the steps of the above method when executed by the processor. The product, medium and system of the present invention correspond to the above method and also have the advantages described in the above method.

[0072] The present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by hardware related to computer program instructions. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiment when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable storage media include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium. The memory is used to store computer programs and / or modules, and the processor implements various functions by running or executing computer programs and / or modules stored in the memory, and calling data stored in the memory. The memory may include high-speed random access memory and may also include non-volatile memory, such as a hard disk, an internal memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0073] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. A template pattern positioning method for electron beam measurement equipment based on deep learning, characterized in that: Includes steps: 1) Obtain the measurement template image to construct a target image set, and mark the graphic edges of the images in the target image set; 2) Use the labeled target image set to train the deep neural network to obtain a graphic edge detection inference model; 3) Use the trained graphic edge detection inference model to perform edge detection on the measurement template image, intercept the measurement point area, and save the measurement point area as a template image; 4) Extracting features from the edges of the graphics contained in the saved template image to construct a feature vector of the measurement point area; 5) Use the graphic edge detection inference model to perform edge detection on the image to be measured, and use the feature vector of the measurement point area of ​​the template image to locate the measurement point area of ​​the image to be measured.

2. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 1, characterized in that: The specific process of step 4) is: The contour of the template image is Fourier transformed to obtain its corresponding Fourier shape descriptor vector, and then the contour edge points are downsampled, and the gradient value of each edge point after downsampling is calculated as the gradient feature vector of the edge. Finally, the Fourier shape descriptor vector and the gradient feature vector are used as the global feature description vector and local feature description vector of the template image, respectively.

3. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 2, characterized in that: The gradient magnitude and direction are calculated for each pixel in the 8*8 neighborhood of the contour edge point, and the gradient histogram in the neighborhood is obtained as the local feature description vector of the contour edge point.

4. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 2 or 3, characterized in that: The specific process of step 5) is: Firstly, the graphic edge detection inference model is used to perform global edge detection on the image to be measured, and the global contour map of the image to be measured is obtained. Then the template image is slid on the global contour map of the image to be measured. Finally, the measurement point area of ​​the graphic to be measured is located based on the local and global deformable diversity similarity by calculating the feature description vector of the template image and the sliding overlapping area.

5. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 4, characterized in that: The formula for calculating the feature description vector of the template image and the sliding overlap area based on the local and global deformable diversity similarity is: in is the template image contour point, is the image contour point of the sliding overlapping area, K(NN α (q j ,P)) is the number of points with similar features between the template image contour points and the sliding overlap area image contour points, fp is the Fourier shape descriptor vector of the template image contour, f q is the Fourier shape descriptor vector of the image contour of the sliding overlap area, r j is the distance between the contour points of the sliding overlapping area image and the contour points of the template image, exp is the exponential function, and N is the number of contour points of the sliding overlapping area image.

6. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 1, 2 or 3, characterized in that: Before step 2), the training set images are denoised using the Neighbor2Blind deep neural network.

7. The electron beam measurement equipment template pattern positioning method based on deep learning according to claim 6, characterized in that: The objective function of the Neighbor2Blind deep neural network is: Among them, f θ is the trained denoising network, g1 and g2 are sub-sampled images, y is the noise image, h is the sub-sampled image of the denoised image, Ω y is the blind spot mask, γ and β are weighting coefficients respectively.

8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are performed.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

10. A template pattern positioning system for electron beam measurement equipment based on deep learning, comprising a memory and a processor connected to each other, wherein a computer program is stored in the memory, characterized in that: When the computer program is executed by a processor, the computer program performs the steps of the method according to any one of claims 1 to 7.

Images