A three-dimensional topography determination method, system, electronic device, and medium
By combining convolutional neural networks and coherent peak detection algorithms with the cross method, the bat wing effect in white light interferometry is identified and corrected, solving the problem of inaccurate measurement in existing technologies and achieving high-precision 3D topography restoration without prior knowledge.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV
- Filing Date
- 2023-09-25
- Publication Date
- 2026-07-24
Smart Images
Figure CN117288118B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of white light interferometry three-dimensional topography measurement technology, and in particular to a three-dimensional topography determination method, system, electronic device and medium. Background Technology
[0002] White light scanning interferometry is characterized by high precision, high resolution, and non-contact operation. It is primarily used to measure surface morphology, steps, and grooves, and has wide applications in high-precision measurement. The coherence length of white light interference is extremely short; interference fringes only appear near the zero optical path difference position, where the light intensity reaches a maximum. By determining the height of this maximum value at each point on the sample through longitudinal scanning, the sample morphology can be measured. However, when the measured step height is less than the coherence length of the light source, a "batwing" effect occurs at the height abrupt change at the step edge, resulting in inaccurate height measurements, which appear as burrs in the image.
[0003] To address the batwing effect in white light interferometry, previous studies, such as Roy et al., have proposed using achromatic phase shifters to reduce its influence. However, this method alters the optical path, making it more complex. Harasaki et al. removed the abnormal height of the batwing effect by setting a threshold, but this method requires prior knowledge of the surface being measured to determine the threshold beforehand. For unknown surfaces, the location of the batwing effect cannot be determined, potentially resulting in the removal of the true height of the sample surface and leading to inaccurate 3D topography.
[0004] Therefore, there is a need for a method to determine the three-dimensional morphology that can accurately eliminate the influence of the bat wing effect and improve accuracy without changing the optical path or knowing any prior knowledge of the surface to be tested before restoring the morphology. Summary of the Invention
[0005] The purpose of this invention is to provide a method, system, electronic device, and medium for determining three-dimensional morphology, which can determine the location of the bat wing effect in advance without prior knowledge of the surface to be tested before morphology restoration, thereby accurately eliminating the influence of the bat wing effect and improving the accuracy of three-dimensional morphology detection.
[0006] To achieve the above objectives, the present invention provides the following solution:
[0007] A method for determining three-dimensional topography, comprising:
[0008] Obtain the interferogram of the sample to be tested;
[0009] The interference signal sequence of each pixel in the interferogram of the sample under test is input into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interferogram of the sample under test; the category is a normal interference signal sequence or an interference signal sequence affected by the bat wing effect;
[0010] The height of each pixel in the interference pattern of the sample under test is obtained by processing the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm.
[0011] The height of each pixel in the pixel set is corrected based on the height of each pixel in the interferogram of the sample under test using the cross method, to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect.
[0012] The three-dimensional morphology of the sample under test is determined based on the height of each pixel in the corrected interferogram of the sample under test.
[0013] Optionally, the height of each pixel in the interferogram of the sample under test is obtained by processing the interference signal sequence of each pixel in the interferogram using a coherent peak detection algorithm, specifically including:
[0014] The envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test is obtained by performing a Hilbert transform on the interference signal sequence of each pixel in the interference pattern of the sample under test.
[0015] The height of each pixel in the interference pattern of the sample under test is obtained by processing the envelope of the interference signal sequence of each pixel using the centroid method.
[0016] Optionally, the height of each pixel in the pixel set is corrected using the cross method based on the height of each pixel in the interference pattern of the sample under test, to obtain the corrected height of each pixel in the interference pattern of the sample under test. Specifically, this includes:
[0017] For any pixel in the pixel set, the height of the target pixel is determined as the height of the pixel; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
[0018] A three-dimensional topography determination system, comprising:
[0019] The acquisition module is used to acquire the interferogram of the sample to be tested;
[0020] The classification module is used to input the interference signal sequence of each pixel in the interferogram of the sample under test into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interferogram of the sample under test; the category is a normal interference signal sequence or an interference signal sequence affected by the bat wing effect;
[0021] The height calculation module is used to process the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm to obtain the height of each pixel in the interference pattern of the sample under test.
[0022] The cross-method processing module is used to correct the height of each pixel in the pixel set according to the height of each pixel in the interferogram of the sample under test, so as to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect.
[0023] The three-dimensional morphology determination module is used to determine the three-dimensional morphology of the sample under test based on the height of each pixel in the corrected interferogram of the sample under test.
[0024] Optionally, the height calculation module specifically includes:
[0025] The Hilbert transform unit is used to perform Hilbert transform on the interference signal sequence of each pixel in the interferogram of the sample under test to obtain the envelope of the interference signal sequence of each pixel in the interferogram of the sample under test.
[0026] The centroid processing unit is used to process the envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test using the centroid method to obtain the height of each pixel in the interference pattern of the sample under test.
[0027] Optionally, the cross-method processing module specifically includes:
[0028] The cross-method processing unit is used to determine the height of a target pixel as the height of any pixel in the pixel set; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
[0029] An electronic device, comprising:
[0030] The electronic device includes a memory and a processor, wherein the memory stores a computer program and the processor runs the computer program to enable the electronic device to perform the three-dimensional topography determination method described above.
[0031] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the three-dimensional topography determination method described above.
[0032] According to specific embodiments provided by the present invention, the present invention discloses the following technical effects:
[0033] This invention classifies interference signal sequences based on convolutional neural networks. Before restoring the morphology, it can identify the location of the bat wing effect in advance without prior knowledge of the surface to be tested, and accurately correct the height error at that location, thereby precisely eliminating the influence of the bat wing effect and obtaining a more accurate three-dimensional morphology of the sample to be tested, thus improving the accuracy of three-dimensional morphology detection. Attached Figure Description
[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0035] Figure 1 This is a schematic diagram of the apparatus for determining the three-dimensional topography of the present invention;
[0036] Figure 2 This is a flowchart of the three-dimensional topography determination method of the present invention;
[0037] Figure 3 This is a diagram showing the result of a normal interference signal sequence identified by the convolutional neural network proposed in an embodiment of the present invention.
[0038] Figure 4 This is a diagram showing the result of the convolutional neural network proposed in this embodiment of the invention identifying the interference signal sequence affected by the bat wing effect;
[0039] Figure 5 The image shows a 3D topography without bat wing effect after processing using existing 3D topography determination methods.
[0040] Figure 6 A three-dimensional topography image processed using the three-dimensional topography determination method provided by this invention;
[0041] Figure 7 A cross-sectional view of the three-dimensional topography without removing the bat wing effect, processed using the three-dimensional topography determination method provided by the present invention;
[0042] Figure 8 A cross-sectional view of a three-dimensional topography processed using the three-dimensional topography determination method provided by this invention.
[0043] Symbol explanation:
[0044] 1-White light source, 2-Illumination objective, 3-Beam splitter, 4-Piezoelectric ceramic, 5-Miller objective, 6-Tube mirror, 7-CCD, 8-Computer, 9-Sample to be tested. Detailed Implementation
[0045] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0046] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0047] This invention discloses a method for determining three-dimensional morphology. It uses a convolutional neural network to pre-identify the region where the bat wing effect occurs, then uses the coherence peak detection method to obtain the three-dimensional morphology of the sample to be tested, and finally uses the cross method to perform precise filtering and correction of the bat wing effect.
[0048] This invention provides a method for determining three-dimensional topography, including:
[0049] Obtain the interference pattern of the sample to be tested.
[0050] The interference signal sequence of each pixel in the interference pattern of the sample under test is input into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interference pattern of the sample under test; the category is a normal interference signal sequence or an interference signal sequence affected by the bat wing effect.
[0051] The height of each pixel in the interference pattern of the sample under test is obtained by processing the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm.
[0052] The height of each pixel in the pixel set is corrected based on the height of each pixel in the interferogram of the sample under test using the cross method, to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect.
[0053] The three-dimensional morphology of the sample under test is determined based on the height of each pixel in the corrected interferogram of the sample under test.
[0054] In practical applications, the coherent peak detection algorithm is used to process the interference signal sequence of each pixel in the interference pattern of the sample under test to obtain the height of each pixel in the interference pattern of the sample under test. Specifically, this includes:
[0055] The envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test is obtained by performing a Hilbert transform on the interference signal sequence of each pixel in the interference pattern of the sample under test.
[0056] The height of each pixel in the interference pattern of the sample under test is obtained by processing the envelope of the interference signal sequence of each pixel using the centroid method.
[0057] In practical applications, the cross method is used to correct the height of each pixel in the pixel set based on the height of each pixel in the interference pattern of the sample under test, resulting in the corrected height of each pixel in the interference pattern of the sample under test. Specifically, this includes:
[0058] For any pixel in the pixel set, the height of the target pixel is determined as the height of the pixel; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
[0059] In practical applications, the method of inputting the interference signal sequence of each pixel in the interference map of the sample under test into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interference map of the sample under test is as follows: input the interference signal sequence of all pixels into the trained convolutional neural network model, and determine whether it is a normal interference signal sequence or an interference signal sequence affected by the bat wing effect based on the feedback result value, using 0 and 1 as labels to distinguish them, with 1 corresponding to normal signal and 0 corresponding to signal affected by bat wing.
[0060] In practical applications, the envelope of the interference signal sequence of each pixel in the interference image of the sample under test is obtained by performing a Hilbert transform on the interference signal sequence of each pixel in the interference image of the sample under test, specifically calculated according to formulas (1) and (2):
[0061]
[0062] Where z represents the distance moved during the scanning process, I(τ) represents the interference signal sequence during the white light scanning process, and I′(i) represents the result of I(τ) after Hilbert transform when the number of moves during the scanning process is i, where i represents the number of moves during the scanning process.
[0063] Then the envelope of the signal is extracted, where τ is the integral sign:
[0064]
[0065] Where M(i) represents the envelope of signal I(τ) when the number of moves during the scanning process is i, and I(i) represents the interference signal sequence when the number of moves during the white light scanning process is i.
[0066] In practical applications, the centroid method is used to process the envelope of the interference signal sequence of each pixel in the interference image of the sample under test to obtain the height of each pixel in the interference image of the sample under test. Specifically, it is calculated according to formula (3):
[0067]
[0068] Where H represents the height, M(i) represents the envelope of signal I(τ) when the number of moves during the scanning process is i, and Δz represents the scanning interval during the scanning process.
[0069] In practical applications, for any pixel in the pixel set, determining the height of the target pixel as the height of the pixel specifically includes:
[0070] For pixels affected by the bat wing effect, the height of adjacent normal pixels is searched by looking in the front, back, left, and right directions. The height of the pixel with the smallest height difference from its own is selected as the corrected height.
[0071] This invention provides a more specific embodiment to describe in detail the above-described three-dimensional topography determination method:
[0072] 1) First, collect several normal interference signal sequences and interference signal sequences affected by the bat wing effect as training sets for training the convolutional neural network.
[0073] The convolutional neural network model is an existing network, consisting of: two one-dimensional convolutional layers, two max-pooling layers, two dropout layers, and two fully connected layers. Each one-dimensional convolutional layer contains 5 convolutional kernels and 64 filters, using the same padding strategy; the max-pooling layers all have a pooling size of 1; the dropout rate is 0.5 for all layers; the fully connected layers have 64 and 2 neurons respectively, the former used to perform high-level representation of the features extracted by the convolutional layers, and the latter corresponding to two classification labels. Additionally, it includes ReLU activation layers and Softmax layers, the former using the ReLU function for non-linear transformation, and the latter converting the output of the fully connected layers into a probability distribution for class prediction.
[0074] The ReLU function expression is shown below:
[0075] F(x) = max(0,x)
[0076] If x is greater than 0, the function outputs x itself; otherwise, the output is 0.
[0077] 2) Test using a trained convolutional neural network. Divide the interference signal sequence of each pixel into normal interference signal and signal sequence affected by bat wing effect, and use 0 and 1 as labels to distinguish them. 1 corresponds to normal signal and 0 corresponds to signal affected by bat wing.
[0078] 3) The height of each pixel is obtained through a coherent peak detection algorithm, which is based on Hilbert transform and centroid method.
[0079] 4) After recovering the height of each pixel, the height of pixels affected by the bat wing is corrected using the cross method. The cross method process is as follows:
[0080] Search for the height of normal pixels in all directions (front, back, left, right), and select the height of the pixel with the smallest difference from its own height as the corrected height.
[0081] The present invention provides an embodiment to verify the effectiveness of the above method:
[0082] The verification experiment of this invention tests a 460nm high stepped block, and the experimental system setup is as follows. Figure 1 As shown, the light emitted from the white light source 1 is collimated by the illumination objective lens 2. After being reflected by the beam splitter 3, the beam is reflected again by the Miller objective lens 5 onto the sample 9. The reflected light interferes with the reference light inside the Miller objective lens 5, and the image is formed on the CCD 7 through the tube lens 6. The piezoelectric ceramic 4 is moved in equal steps by the computer 8. Each time it moves, the CCD 7 obtains an interference pattern, which is stored on the computer 8. The interference pattern is then calculated using the method described in the above embodiment to finally recover the three-dimensional morphology of the sample and eliminate the influence of the batwing effect. Figure 3 This is a diagram showing the result of a normal interference signal sequence identified by the convolutional neural network proposed in an embodiment of the present invention. Figure 4 This is a diagram showing the result of the convolutional neural network proposed in this embodiment of the invention identifying the interference signal sequence affected by the bat wing effect. According to... Figures 3 to 8 It is evident that the algorithm of this invention can effectively reduce the impact of the bat wing effect.
[0083] like Figure 2 As shown, the bat wing effect is first identified based on a convolutional neural network. Then, the morphology of the sample is recovered using the coherent peak detection method (Hilbert transform and centroid method). Finally, the bat wing effect is corrected using the cross method (filtering). The specific implementation steps of this method are as follows:
[0084] After training the model by collecting several interference signal sequences, the interference signal sequences of all pixels are tested first. 0 and 1 are used as labels to distinguish them. 1 corresponds to normal signal and 0 corresponds to signal affected by bat wing. The pixels affected by bat wing effect are found.
[0085] The height of each pixel is obtained through a coherent peak detection algorithm, which is based on Hilbert transform and centroid method.
[0086] After the height of each pixel is restored, the height of the pixels affected by the bat wings is corrected using the "cross method".
[0087] This invention is based on convolutional neural network processing of interference signal sequences, which can identify the location of the bat wing effect in advance, thereby cleanly eliminating the influence of the bat wing effect and obtaining a more accurate three-dimensional morphology of the sample to be tested.
[0088] In view of the above method, this embodiment of the invention provides a three-dimensional topography determination system, including:
[0089] The acquisition module is used to acquire the interferogram of the sample to be tested.
[0090] The classification module is used to input the interference signal sequence of each pixel in the interferogram of the sample under test into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interferogram of the sample under test; the category is a normal interference signal sequence or an interference signal sequence affected by the bat wing effect.
[0091] The height calculation module is used to process the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm to obtain the height of each pixel in the interference pattern of the sample under test.
[0092] The cross-method processing module is used to correct the height of each pixel in the pixel set according to the height of each pixel in the interferogram of the sample under test, so as to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect.
[0093] The three-dimensional morphology determination module is used to determine the three-dimensional morphology of the sample under test based on the height of each pixel in the corrected interferogram of the sample under test.
[0094] As an optional implementation, the height calculation module specifically includes:
[0095] The Hilbert transform unit is used to perform a Hilbert transform on the interference signal sequence of each pixel in the interferogram of the sample under test to obtain the envelope of the interference signal sequence of each pixel in the interferogram of the sample under test.
[0096] The centroid processing unit is used to process the envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test using the centroid method to obtain the height of each pixel in the interference pattern of the sample under test.
[0097] As an optional implementation, the cross-method processing module specifically includes:
[0098] The cross-method processing unit is used to determine the height of a target pixel as the height of any pixel in the pixel set; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
[0099] This invention also provides an electronic device, comprising:
[0100] A memory and a processor, the memory being used to store a computer program, the processor running the computer program to cause the electronic device to perform the three-dimensional topography determination method according to the above embodiments.
[0101] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the three-dimensional topography determination method described in the above embodiments.
[0102] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.
[0103] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.
Claims
1. A method for determining three-dimensional topography, characterized in that, include: Obtain the interferogram of the sample to be tested; The interference signal sequence of each pixel in the interference pattern of the sample under test is input into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interference pattern of the sample under test. The category is either a normal interference signal sequence or an interference signal sequence affected by the bat wing effect; The height of each pixel in the interference pattern of the sample under test is obtained by processing the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm. The height of each pixel in the pixel set is corrected based on the height of each pixel in the interferogram of the sample under test using the cross method, to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect. The three-dimensional morphology of the sample under test is determined based on the height of each pixel in the corrected interferogram of the sample under test.
2. The three-dimensional topography determination method according to claim 1, characterized in that, The height of each pixel in the interferogram of the sample under test is obtained by processing the interference signal sequence of each pixel in the interferogram using a coherent peak detection algorithm. Specifically, this includes: The envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test is obtained by performing a Hilbert transform on the interference signal sequence of each pixel in the interference pattern of the sample under test. The height of each pixel in the interference pattern of the sample under test is obtained by processing the envelope of the interference signal sequence of each pixel using the centroid method.
3. The three-dimensional topography determination method according to claim 1, characterized in that, The height of each pixel in the pixel set is corrected based on the height of each pixel in the interference pattern of the sample under test using the cross method, resulting in the corrected height of each pixel in the interference pattern of the sample under test. Specifically, this includes: For any pixel in the pixel set, the height of the target pixel is determined as the height of the pixel; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
4. A three-dimensional topography determination system, characterized in that, include: The acquisition module is used to acquire the interferogram of the sample to be tested; The classification module is used to input the interference signal sequence of each pixel in the interferogram of the sample under test into a trained convolutional neural network to determine the category of the interference signal sequence of each pixel in the interferogram of the sample under test. The category is either a normal interference signal sequence or an interference signal sequence affected by the bat wing effect; The height calculation module is used to process the interference signal sequence of each pixel in the interference pattern of the sample under test using a coherent peak detection algorithm to obtain the height of each pixel in the interference pattern of the sample under test. The cross-method processing module is used to correct the height of each pixel in the pixel set according to the height of each pixel in the interferogram of the sample under test, so as to obtain the corrected height of each pixel in the interferogram of the sample under test; the pixel set includes all pixels in the interferogram of the sample under test that correspond to the interference signal sequence of the category affected by the bat wing effect. The three-dimensional morphology determination module is used to determine the three-dimensional morphology of the sample under test based on the height of each pixel in the corrected interferogram of the sample under test.
5. The three-dimensional topography determination system according to claim 4, characterized in that, The height calculation module specifically includes: The Hilbert transform unit is used to perform Hilbert transform on the interference signal sequence of each pixel in the interferogram of the sample under test to obtain the envelope of the interference signal sequence of each pixel in the interferogram of the sample under test. The centroid processing unit is used to process the envelope of the interference signal sequence of each pixel in the interference pattern of the sample under test using the centroid method to obtain the height of each pixel in the interference pattern of the sample under test.
6. The three-dimensional topography determination system according to claim 4, characterized in that, The cross-method processing module specifically includes: The cross-method processing unit is used to determine the height of a target pixel as the height of any pixel in the pixel set; the target pixel is the pixel with the smallest height difference from the pixel in the neighboring pixels; the neighboring pixels are all pixels in the category of normal interference signal sequence among all the pixels adjacent to the pixel.
7. An electronic device, characterized in that, include: A memory and a processor, the memory being used to store a computer program, the processor running the computer program to cause the electronic device to perform the three-dimensional topography determination method according to any one of claims 1 to 3.
8. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the three-dimensional topography determination method as described in any one of claims 1 to 3.