Contour extraction method and device of SEM image, computer device and storage medium
By combining Bézier curves and gradient algorithms, the problems of large data volume and information loss in SEM image contour extraction are solved, achieving more accurate and smoother contour extraction, which is suitable for SEM image processing in semiconductor manufacturing.
Patent Information
- Application Number
- CN202211291246.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-10-20
- Publication Date
- 2026-02-17
- Estimated Expiration
- 2042-10-20
AI Technical Summary
In existing technologies, SEM image contour extraction suffers from problems such as large data volume and easy loss of contour information.
The method of Bézier curves is used to extract M discrete contour points from the SEM image, randomly select N initial control points, and iteratively calibrate the initial control points through a preset loss function and gradient algorithm so that the Bézier curve gradually fits the contour points, and generate the calibrated Bézier curve as the contour information output.
It achieves more complete and accurate extraction of SEM image contours, reduces the amount of data, and the extracted contours are continuous and smooth, enabling fast and accurate extraction of contour points even without GDS layout files.
Smart Images
Figure CN115841496B_ABST
Abstract
Description
[Technical Field]
[0001] This invention relates to the fields of semiconductor manufacturing and computational lithography, and specifically to a method, apparatus, computer equipment, and storage medium for contour extraction of SEM images. [Background Technology]
[0002] SEM (Scanning Electron Microscopy) is used to scan photoresist to obtain SEM images. SEM images are an important technique in semiconductor manufacturing, used for processes such as measuring critical dimensions of patterns. Therefore, it is necessary to accurately extract and efficiently store the contours of the patterns in the SEM images. Common SEM image contour extraction techniques extract and save the coordinates of a series of discrete points on the contour, resulting in large amounts of data and failing to cover the entire contour line. In other words, existing SEM image contour extraction techniques suffer from problems such as large data volumes and loss of contour information. [Summary of the Invention]
[0003] To address the technical problems of large data volume and easy loss of contour information in existing SEM image contour extraction technologies, this invention provides a contour extraction method and storage medium for SEM images.
[0004] To solve the above-mentioned technical problems, the technical solution of the present invention is to provide a contour extraction method for SEM images, the contour extraction method for SEM images comprising the following steps:
[0005] M discrete contour points are extracted from the SEM image, where M is an integer;
[0006] N initial control points are randomly selected, and an initial Bézier curve is obtained based on the N initial control points, where N is an integer and N is less than or equal to M;
[0007] The N initial control points are calibrated to gradually fit the initial Bézier curve with the M contour points to obtain the calibrated Bézier curve.
[0008] The calibrated Bézier curve is output as the contour information extracted from the SEM image.
[0009] Preferably, calibrating the N initial control points to gradually align the Bézier curve with the contour points specifically includes the following steps:
[0010] A preset loss function is generated based on the M contour points and the Bézier curve;
[0011] The initial control point is iteratively moved based on a preset gradient algorithm until the loss function satisfies a preset condition;
[0012] Obtain the final position information of the current control point after the preset conditions are met;
[0013] The calibrated Bézier curve is obtained based on the final position information.
[0014] Preferably, the preset loss function satisfies the following: the preset loss function is established based on the sum of the squares of the distances between the M contour points and the Bézier curve.
[0015] Preferably, the preset gradient algorithm is the gradient descent method.
[0016] Preferably, the direction of iterative movement is determined based on the direction information and distance information of the initial control point relative to the control point after the next iteration.
[0017] Preferably, the preset condition is that the change in the degree of optimization of the loss function is less than a preset threshold.
[0018] Preferably, extracting M discrete contour points from the SEM image specifically involves:
[0019] The SEM image is pixelated to obtain the original matrix;
[0020] After performing Gaussian filtering on the original matrix, an image matrix is obtained, and the gradient values of all pixels in the image matrix are calculated.
[0021] Obtain all pixels whose gradient values are within a preset range, and find the peak point along the gradient direction among all selected pixels;
[0022] The peak points are the M discrete contour points extracted from the SEM image.
[0023] Another solution of the present invention to solve the above-mentioned technical problem is to provide a contour extraction device for SEM images, applied to the above-mentioned contour extraction method for SEM images, wherein the contour extraction device for SEM images includes:
[0024] Extraction module: used to extract M discrete contour points from the SEM image, where M is an integer;
[0025] Calculation module: used to obtain the initial Bézier curve based on N initial control points, where N is an integer and N is less than or equal to M;
[0026] Calibration module: used to calibrate the N initial control points so that the initial Bézier curve gradually fits the M contour points to obtain the calibrated Bézier curve;
[0027] Output module: Used to output the calibrated Bézier curve.
[0028] Another solution of the present invention to solve the above-mentioned technical problems is to provide a computer device, including a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the contour extraction method for SEM images as described in any of the preceding claims.
[0029] Another solution of the present invention to solve the above-mentioned technical problems is to provide a storage medium including a processor, wherein when the storage medium is executed by the processor, it implements a contour extraction method for SEM images as described in any of the preceding claims.
[0030] Compared with existing technologies, the contour extraction method for SEM images provided by this invention has the following advantages:
[0031] 1. The contour extraction method for SEM images provided in the embodiments of the present invention utilizes the fact that a Bézier curve is a differentiable mathematical curve, thus it can be described as a continuous and smooth curve using a finite number of initial control points. Therefore, the present invention first extracts M discrete contour points from the SEM image, then randomly selects N initial control points to calculate an initial Bézier curve; then, it calibrates the N initial control points based on a preset calibration method, thereby making the initial Bézier curve fit the M contour points to obtain a calibrated Bézier curve. This design allows for a better and more accurate representation of the contour information of the SEM image. Furthermore, since the present invention represents the calibrated Bézier curve as the contour in the SEM image, the amount of data obtained from the contour extraction result is relatively small. Therefore, the contour extracted by this method is more complete and accurate, the amount of extracted data is smaller, and the extracted contour is continuous and smooth.
[0032] 2. The contour extraction method for SEM images provided in the embodiments of the present invention generates a preset loss function through M contour points and Bézier curves, and iteratively moves the initial control points based on a preset gradient algorithm until the loss function meets the preset conditions. Since the randomly selected initial control points cannot directly obtain the contour of the SEM image and the obtained contour curve is not smooth, it is necessary to iteratively move the initial control points in a preset direction based on the preset gradient algorithm, so that the subsequently obtained Bézier curves are more continuous and smooth. That is, the Bézier curves calibrated by the preset calibration method of the present invention are smoother and more continuous, thereby better recovering the complete contour information in the SEM image.
[0033] 3. The contour extraction method for SEM images provided in the embodiments of the present invention, by designing a preset loss function based on the sum of the squares of the distances between M contour points and the Bézier curve, enables more accurate calibration.
[0034] 4. The contour extraction method for SEM images provided in the embodiments of the present invention can confirm the movement direction information and distance information of the control points after iteration by using the gradient descent method to iterate the design of the initial control points.
[0035] 5. The contour extraction method for SEM images provided in the embodiments of the present invention determines the direction of iterative movement based on the direction and distance information of the initial control point relative to the control point after the next iteration. This design allows the control point to be moved according to the determined position information to complete the next iteration process, reducing the computational load and enabling more accurate and rapid determination of the contour information of the SEM image.
[0036] 6. The contour extraction method for SEM images provided in the embodiments of the present invention, through the design of the method to extract contour points of SEM images, enables the rapid and accurate extraction of M contour points based solely on the SEM image without the presence of a GDS layout file.
[0037] 7. The contour extraction device for SEM images provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method for SEM images of the present invention, and will not be described again here.
[0038] 8. The computer device provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method of SEM images of the present invention, and will not be described again here.
[0039] 9. The storage medium provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method of SEM images of the present invention, and will not be described again here. [Attached Image Description]
[0040] Figure 1 This is a schematic flowchart of the contour extraction method for SEM images provided in the first embodiment of the present invention;
[0041] Figure 2 This is a detailed flowchart illustrating the contour extraction method for SEM images provided in the first embodiment of the present invention. Figure 1 ;
[0042] Figure 3 This is a schematic diagram of the SEM image for the contour extraction method of SEM images provided in the first embodiment of the present invention. Figure 1 ;
[0043] Figure 4 This is a schematic diagram of the SEM image contour extraction method provided in the first embodiment of the present invention, which extracts M contour points from the SEM image.
[0044] Figure 5(a) to (c) are schematic diagrams of the state during the initial control point calibration process of the contour extraction method for SEM images provided in the first embodiment of the present invention;
[0045] Figure 6 This is a detailed flowchart illustrating the contour extraction method for SEM images provided in the first embodiment of the present invention. Figure 1 ;
[0046] Figure 7 This is a schematic diagram of the SEM image for the contour extraction method of SEM images provided in the first embodiment of the present invention. Figure 2 ;
[0047] Figure 8 (a) to (c) are schematic diagrams of state changes when extracting M contour points of a SEM image using the contour extraction method of the first embodiment of the present invention.
[0048] Figure 9 This is a system block diagram of the contour extraction device for SEM images provided in the second embodiment of the present invention.
[0049] Explanation of reference numerals in the attached diagram:
[0050] 1. Contour extraction method for SEM images; 2. Contour extraction device for SEM images;
[0051] 21. Extraction module; 22. Calculation module; 23. Calibration module; 24. Output module.
Detailed Implementation Methods
[0052] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0053] Please see Figure 1 The first embodiment of the present invention provides a contour extraction method 1 for SEM images, which includes the following steps:
[0054] S1: Extract M discrete contour points from the SEM image, where M is an integer;
[0055] S2: Randomly select N initial control points, and obtain the initial Bézier curve based on the N initial control points, where N is an integer and N is less than or equal to M;
[0056] S3: Calibrate the N initial control points to gradually fit the Bézier curve with the M contour points to obtain the calibrated Bézier curve;
[0057] S4: Output the calibrated Bézier curve as the contour information extracted from the SEM image.
[0058] Understandably, existing contour extraction methods store contour points on a curve and then connect these points in pairs to reconstruct a polygonal line approximating the original contour. This results in a large amount of data and potential loss of contour information. Therefore, this invention first extracts M discrete contour points from the SEM image and then randomly selects N initial control points to calculate an initial Bézier curve. Since a Bézier curve is a differentiable mathematical curve, it can describe a continuous, smooth curve using a finite number of initial control points. The N initial control points are then calibrated to ensure the initial Bézier curve aligns with the M contour points, resulting in a calibrated Bézier curve. Finally, the calibrated Bézier curve is output as the contour information extracted from the SEM image. This design provides a better and more accurate representation of the contour information in the SEM image. Furthermore, because this invention represents the calibrated Bézier curve as the contour output in the SEM image, the resulting data volume for contour extraction is smaller. Additionally, since the calculated contour curve is also determined once the final control points are established, both contain the same information. If the results are to be stored on disk, only the final control points need to be stored, which reduces the amount of extracted data. When the contour needs to be visualized, the Bézier curve only needs to be calculated based on the final control points. Therefore, it is evident that this method extracts more complete and accurate contours from SEM images, results in less data, and produces smoother, continuous contours.
[0059] It should be noted that the calibrated Bézier curve closely resembles the original profile, falling within an acceptable error range, but will not completely overlap. The degree of recovery depends on several preset parameter settings, such as the number of initial control points and the number of iterations in the initial control point calibration process. In practical applications, these parameters can be adjusted to improve calibration efficiency and the degree of recovery.
[0060] Further, please refer to Figure 2 The calibration of N initial control points to gradually align the Bézier curve with the contour points includes the following steps:
[0061] S31: Generate a preset loss function based on M contour points and the Bézier curve;
[0062] S32: Iteratively move the initial control point based on the preset gradient algorithm until the loss function satisfies the preset conditions;
[0063] S33: Obtain the final position information of the current control point after the preset conditions are met;
[0064] S34: The calibrated Bézier curve is obtained based on the final position information.
[0065] Understandably, the design involves generating a preset loss function using M contour points and a Bézier curve, and iteratively moving the initial control points based on a preset gradient algorithm until the loss function meets the preset conditions. Since randomly selected initial control points cannot directly obtain the contour of the SEM image and the obtained contour curve is not smooth, it is necessary to iteratively move the initial control points in a preset direction based on the preset gradient algorithm. This makes the subsequently obtained Bézier curve closer to the original contour, continuous and smooth. In other words, the Bézier curve calibrated by the preset calibration method in this invention is smoother and more continuous, thus enabling better recovery of the complete contour information in the SEM image.
[0066] Furthermore, the preset loss function satisfies the following: the preset loss function is established based on the sum of the squares of the distances between the M contour points and the Bézier curve.
[0067] Furthermore, the preset gradient algorithm is gradient descent.
[0068] Understandably, by using the gradient descent method to iterate the design of the initial control points, it is possible to confirm the direction and distance information of the control points after each iteration relative to the previous control points. This makes it easier to move the control points according to this direction and distance information in subsequent iterations, thereby further reducing the amount of computational data in the contour extraction results.
[0069] Furthermore, the direction of iterative movement is determined based on the direction and distance information of the initial control point relative to the control point after the next iteration.
[0070] Understandably, this design allows control points to be moved according to defined direction and distance information to complete the next iteration, reducing computational load and enabling more accurate and faster determination of the contour information of the SEM image.
[0071] Furthermore, the preset condition is specifically: the change in the degree of optimization of the loss function is less than a preset threshold.
[0072] It should be noted that the optimization degree of the loss function refers to the degree of change of the value in each iteration relative to the value in the previous iteration. Iteration can stop when the current value of the loss function no longer decreases significantly relative to the previous value after a certain number of consecutive iterations, thus meeting the preset condition. Furthermore, the preset threshold can be set according to actual needs.
[0073] For example, please refer to Figure 3 First, obtain a SEM image to be processed. Please refer to [link / reference]. Figure 4Then, several discrete contour points are extracted using a preset contour extraction method. Please refer to [link to relevant documentation]. Figure 5 ,and Figure 5 In the diagram, (a) to (c) represent the curves corresponding to the control points at the initial stage, during the iteration process, and after the iteration calibration, respectively. Specifically, N initial control points are iteratively moved using the gradient descent method. Based on the gradient descent method, the direction and distance information of each initial control point relative to each control point after the next iteration can be determined. In each iteration, each control point is moved according to the direction and distance information obtained in the previous iteration. The loss function value needs to be calculated during each iteration. When the loss function value no longer decreases significantly, the iteration can be stopped, and the Bézier curve obtained in this iteration is used as the contour information extracted from the SEM image. It can be seen that the contour information extracted by the contour extraction method of this invention is a continuous, smooth curve that closely resembles the original contour.
[0074] Further, please refer to Figure 6 Extracting M discrete contour points from a SEM image specifically involves:
[0075] S11: Perform pixelation on the SEM image to obtain the original matrix;
[0076] S12: After performing Gaussian filtering on the original matrix, the image matrix is obtained, and the gradient value of each pixel in the image matrix is calculated.
[0077] S13: Obtain all pixels with gradient values within a preset range, and find the peak point along the gradient direction among all selected pixels;
[0078] The peak points are the M discrete contour points extracted from the SEM image.
[0079] It should be noted that acquiring all pixels with gradient values within a preset range refers to acquiring pixels with larger gradient values. This design enables rapid identification, thereby improving the speed and accuracy of subsequent contour point extraction.
[0080] Understandably, this method for extracting contour points from SEM images allows for the rapid and accurate extraction of M contour points from a single SEM image, even without a GDS map file. Furthermore, by first acquiring all pixels within a preset range and then searching for peak points among the selected pixels, the extraction speed and accuracy of contour points are significantly improved, facilitating subsequent contour extraction.
[0081] Specifically, finding the peak point along the gradient direction among all pixels involves: dividing the image matrix into blocks of a preset size, finding the middle pixel with the largest gradient value in each block; filtering the middle pixels in all blocks, and when the gradient value of the middle pixel is greater than a preset gradient threshold, selecting the corresponding key pixel and recording its coordinates and gradient direction; starting from the coordinates of the key pixel in the image matrix, finding the pixel with the smallest gray value along the gradient direction of the key pixel within a preset search range, which is the peak point.
[0082] Understandably, dividing the image matrix into blocks and filtering out key pixels for reference ensures that the subsequent search for peak points starts from the edges of lines in the SEM image. The pixel with the lowest grayscale value is the brightest point within a certain range in the SEM image (it should be understood that the lines in the SEM image have a certain width, and the brightest point in the width direction best reflects the local position of the SEM image contour at that location; connecting these brightest points sequentially yields the SEM image contour that retains the most detail of the original image). Searching along the gradient direction allows for accurate and rapid identification of the desired peak points. This method is simple to implement, highly reliable, and the identified peak points accurately reflect the contour point information of the SEM image. Furthermore, the preset search range is 30-50 pixels.
[0083] For example, first randomly select a SEM image to be processed (e.g. Figure 7 As shown), and then the SEM image is pixelated (as shown). Figure 8 As shown in (a)), the original matrix is then subjected to Gaussian filtering to obtain the image matrix, and the gradient value of each pixel in the image matrix is calculated. Then, all pixels whose gradient values fall within a preset range are obtained (e.g., ...). Figure 8 (as shown in (b)) and finds peak points along the gradient direction among all selected pixels (e.g. Figure 8 (The pixels between the two lines in (c)). It can be seen that the contour points obtained in this way are more accurate, which facilitates subsequent contour extraction.
[0084] Please see Figure 9 The second embodiment of the present invention provides a contour extraction device 2 for SEM images, applied to a contour extraction method 1 for SEM images according to the first embodiment of the present invention. The contour extraction device 2 for SEM images includes:
[0085] Extraction module 21: used to extract M discrete contour points from the SEM image, where M is an integer;
[0086] Calculation module 22: used to obtain the initial Bézier curve based on N initial control points, where N is an integer and N is less than or equal to M;
[0087] Calibration module 23: used to calibrate N initial control points so that the initial Bézier curve gradually fits with M contour points to obtain the calibrated Bézier curve;
[0088] Output module 24: Used to output the calibrated Bézier curve.
[0089] It is understood that the contour extraction device for SEM images according to the second embodiment of the present invention has the same beneficial effects as the contour extraction method for SEM images according to the first embodiment of the present invention, and will not be described again here.
[0090] A third embodiment of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the contour extraction method for SEM images according to the first embodiment of the present invention.
[0091] It is understood that the computer device of the third embodiment of the present invention has the same beneficial effects as the contour extraction method of SEM image of the first embodiment of the present invention, and will not be described again here.
[0092] A fourth embodiment of the present invention provides a storage medium including a processor, wherein when the storage medium is executed by the processor, it implements a contour extraction method for SEM images as described in the first embodiment of the present invention.
[0093] It is understood that the storage medium of the fourth embodiment of the present invention has the same beneficial effects as the contour extraction method of SEM image of the first embodiment of the present invention, and will not be described again here.
[0094] Compared with existing technologies, the contour extraction method for SEM images provided by this invention has the following advantages:
[0095] 1. The contour extraction method for SEM images provided in the embodiments of the present invention utilizes the fact that a Bézier curve is a differentiable mathematical curve, thus it can be described as a continuous and smooth curve using a finite number of initial control points. Therefore, the present invention first extracts M discrete contour points from the SEM image, then randomly selects N initial control points to calculate an initial Bézier curve; then, it calibrates the N initial control points based on a preset calibration method, thereby making the initial Bézier curve fit the M contour points to obtain a calibrated Bézier curve. This design allows for a better and more accurate representation of the contour information of the SEM image. Furthermore, since the present invention represents the calibrated Bézier curve as the contour in the SEM image, the amount of data obtained from the contour extraction result is relatively small. Therefore, the contour extracted by this method is more complete and accurate, the amount of extracted data is smaller, and the extracted contour is continuous and smooth.
[0096] 2. The contour extraction method for SEM images provided in the embodiments of the present invention generates a preset loss function through M contour points and Bézier curves, and iteratively moves the initial control points based on a preset gradient algorithm until the loss function meets the preset conditions. Since the randomly selected initial control points cannot directly obtain the contour of the SEM image and the obtained contour curve is not smooth, it is necessary to iteratively move the initial control points in a preset direction based on the preset gradient algorithm, so that the subsequently obtained Bézier curves are more continuous and smooth. That is, the Bézier curves calibrated by the preset calibration method of the present invention are smoother and more continuous, thereby better recovering the complete contour information in the SEM image.
[0097] 3. The contour extraction method for SEM images provided in the embodiments of the present invention, by designing a preset loss function based on the sum of the squares of the distances between M contour points and the Bézier curve, enables more accurate calibration.
[0098] 4. The contour extraction method for SEM images provided in the embodiments of the present invention can confirm the movement direction information and distance information of the control points after iteration by using the gradient descent method to iterate the design of the initial control points.
[0099] 5. The contour extraction method for SEM images provided in the embodiments of the present invention determines the direction of iterative movement based on the direction and distance information of the initial control point relative to the control point after the next iteration. This design allows the control point to be moved according to the determined position information to complete the next iteration process, reducing the computational load and enabling more accurate and rapid determination of the contour information of the SEM image.
[0100] 6. The contour extraction method for SEM images provided in the embodiments of the present invention, through the design of the method to extract contour points of SEM images, enables the rapid and accurate extraction of M contour points based solely on the SEM image without the presence of a GDS layout file.
[0101] 7. The contour extraction device for SEM images provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method for SEM images of the present invention, and will not be described again here.
[0102] 8. The computer device provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method of SEM images of the present invention, and will not be described again here.
[0103] 9. The storage medium provided in the embodiments of the present invention has the same beneficial effects as the contour extraction method of SEM images of the present invention, and will not be described again here.
[0104] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of the invention. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Those skilled in the art should also recognize that the embodiments described in the specification are optional embodiments, and the actions and modules involved are not necessarily essential to the invention.
[0105] In various embodiments of the present invention, it should be understood that the sequence number of each process does not necessarily imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.
[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, or they may sometimes be executed in reverse order, depending on the functions involved. It is particularly important to note that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0107] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A method of contour extraction of a SEM image, characterized by: The profile extraction method of the SEM image comprises the following steps: extracting M discrete profile points from the SEM image, wherein M is an integer, and the profile points are peak points of the SEM image; randomly selecting N initial control points, and obtaining an initial Bezier curve based on the N initial control points, wherein N is an integer and N is less than or equal to M; calibrating the N initial control points to gradually fit the initial Bezier curve to the M profile points to obtain a calibrated Bezier curve; outputting the calibrated Bezier curve as the profile information extracted from the SEM image; wherein the calibration of the N initial control points to gradually fit the Bezier curve to the profile points specifically comprises the steps of: generating a preset loss function according to the M profile points and the Bezier curve; iteratively moving the initial control points based on a preset gradient algorithm until the loss function meets a preset condition; obtaining final position information of the current control points after the preset condition is met; obtaining the calibrated Bezier curve based on the final position information; the iteratively moving of the initial control points based on the preset gradient algorithm comprises: obtaining movement direction information and distance information of the control points after each iteration relative to the control points of the previous iteration; moving the control points according to the direction information and the distance information; the obtaining of the peak points specifically comprises the steps of: pixelizing the SEM image to obtain an original matrix; performing Gaussian filtering on the original matrix to obtain an image matrix and calculating gradient values of all pixel points in the image matrix; dividing the original matrix into blocks of a preset size, and determining an intermediate pixel point with the largest gradient value in each block; regarding the intermediate pixel points with gradient values greater than a preset gradient threshold as key pixel points, and recording coordinates and gradient directions of the key pixel points; in the original matrix, starting from the coordinates of the key pixel points, determining the brightest pixel point within a preset search range along the gradient direction, and regarding the pixel point as the peak point, the brightness of the pixel point being determined according to the gray value of the pixel point.
2. The profile extraction method of a SEM image according to claim 1, wherein: The preset loss function meets: the preset loss function is established according to the sum of squares of distances between the M profile points and the Bezier curve.
3. The profile extraction method of the SEM image according to claim 1, characterized by: The preset gradient algorithm is a gradient descent method.
4. The profile extraction method of the SEM image according to claim 1, characterized by: The direction of the iterative movement is determined according to the direction information and the distance information of the initial control points relative to the control points after the next iteration.
5. The profile extraction method of the SEM image according to claim 1, characterized by: The preset condition specifically is that a change value of an optimization degree of the loss function is less than a preset threshold.
6. The profile extraction method of a SEM image according to Claim 1, wherein: The extracting of the M discrete profile points from the SEM image specifically comprises: obtaining all pixel points with gradient values within a preset range, and finding peak points along the gradient direction in all the selected pixel points; the peak points are the M discrete profile points extracted from the SEM image.
7. A device for contour extraction of SEM images, applied to the method for contour extraction of SEM images according to any one of claims 1-6, characterized in that: The profile extraction device of the SEM image comprises: an extraction module: configured to extract M discrete profile points from the SEM image, wherein M is an integer, and the profile points are peak points of the SEM image; The computing module is configured to obtain an initial Bezier curve based on N initial control points, where N is an integer and N is less than or equal to M; The calibration module is configured to calibrate the N initial control points to gradually fit the initial Bezier curve to the M contour points to obtain a calibrated Bezier curve; The output module is configured to output the calibrated Bezier curve, and is further configured to calibrate the N initial control points to gradually fit the Bezier curve to the contour points, specifically comprising the steps of: generating a preset loss function based on the M contour points and the Bezier curve; moving the initial control points based on a preset gradient algorithm until the loss function meets a preset condition; obtaining final position information of the current control points after the preset condition is met; obtaining the calibrated Bezier curve based on the final position information; and obtaining movement direction information and distance information of the control points after each iteration relative to the control points of the previous iteration; and moving the control points based on the direction information and the distance information. The extraction module is further configured to: perform pixelization processing on the SEM image to obtain an original matrix; perform Gaussian filtering processing on the original matrix to obtain an image matrix and calculate gradient values of all pixel points in the image matrix; divide the original matrix into blocks of a preset size, and determine a middle pixel point with the largest gradient value in each block; regard the middle pixel point with a gradient value greater than a preset gradient threshold value as a key pixel point, and record coordinates and a gradient direction of the key pixel point; and in the original matrix, determine a pixel point with the largest gray value in a preset search range along the gradient direction starting from the coordinates of the key pixel point, regard the pixel point as the peak value point, and determine brightness of the pixel point based on a gray value of the pixel point.
8. A computer device, comprising: The computer program is stored in the memory and is executed by the processor to implement the steps of the contour extraction method of the SEM image according to any one of claims 1-6.
9. A storage medium characterized by: The computer program is stored in the memory and is executed by the processor to implement the steps of the contour extraction method of the SEM image according to any one of claims 1-6.