Multi-magnification topography data fusion method based on frequency screening of interferometric objective lens

Abstract

The present invention discloses a multi-rate morphology data fusion method based on frequency screening of an interference objective lens, which belongs to the field of optical detection. The multi-aperture algorithm of wavelet transform is adopted, and based on the cutoff frequency parameter of the selected interference objective lens, the original morphology data matrix independent of each magnification is positively transformed to obtain multiple spatial domain sub-data matrices, which respectively correspond to the three frequency components of low, medium and high. The multiple spatial domain data matrices of two magnifications are fused according to the screening fusion strategy described in the present invention, and then the wavelet inverse transform is performed to realize the multi-rate fusion of morphology data. The fused morphology data not only covers the globality of a large field of view, but also provides detailed data for the area of ​​interest.

Description

Technical Field

[0001] The invention belongs to the field of optical detection, and in particular relates to a multi-magnification morphology data fusion method based on interference objective lens frequency screening. Background Art

[0002] As mathematical models are developed, the need for multiscale surface analysis continues to grow, leading to the development of multiscale surfaces that exhibit specific properties at different scales to achieve specific purposes. Many surfaces with different aspect ratios require different resolutions at different locations, which poses a challenge to current white light interferometers.

[0003] Although white light interferometers can achieve measurements of different resolutions at different positions by switching interferometer lenses of different magnifications, the data lacks the ability to represent multiple resolutions. Existing instruments are unable to match and fuse measurement data from multiple lenses for display. Common data matching and fusion algorithms all use field of view stitching at a single magnification. Although they can obtain high-resolution morphological data for large areas, they are extremely time-consuming and some areas are forced to participate in high-resolution measurements.

[0004] In response to the problem that existing white light interferometers lack multi-resolution characterization capabilities, some scholars have studied multi-resolution fusion strategies. Ramasamy proposed a wavelet fusion method for morphological data of different magnifications in "Data Fusion Strategy for Multiscale Surface Measurements", but this method does not take into account the frequency differences of morphological data of each magnification. Summary of the invention

[0005] The purpose of the present invention is to provide a multi-magnification morphology data fusion method based on interference objective frequency screening to solve the problem of large-scale high-resolution surface morphology detection.

[0006] The technical solution to achieve the purpose of the present invention is: a multi-magnification morphology data fusion method based on interference objective frequency screening, the steps are as follows:

[0007] Step 1: Measure the sample using a white light interferometer and two interference objective lenses with different magnifications, where the ratio of the magnification of the first interference objective lens to the magnification of the second interference objective lens is Γ, Γ>1, the first interference objective lens measures to obtain a first morphology data matrix A, and the second interference objective lens measures to obtain a second morphology data matrix B.

[0008] Step 2: According to the numerical aperture NA and the corresponding cutoff frequency f of the two interference objective lenses with different magnifications s, respectively obtain the frequency range of the first morphology data matrix A and the frequency range of the second morphology data matrix B, and calculate the ratio S of the cutoff frequencies of the first morphology data matrix A and the second morphology data matrix B.

[0009] Step 3: Perform filter preprocessing on the first morphology data matrix A to remove outliers to obtain a third morphology data matrix P. Perform filter preprocessing on the second morphology data matrix B to remove outliers to obtain a fourth morphology data matrix Q. The sizes of the third morphology data matrix P and the fourth morphology data matrix Q are both M×N.

[0010] Step 4: Perform bilinear interpolation on the third shape data matrix P to adjust the size of the third shape data matrix P to The fifth topography data matrix P' is obtained, and the fifth topography data matrix P' and the fourth topography data matrix Q are registered:

[0011] First, a grayscale-based normalized correlation matching algorithm is used for rough matching. The fifth shape data matrix P' and the fourth shape data matrix Q are searched and compared pixel by pixel according to the similarity measurement method to match them and obtain the initial translation vector t0. The fifth shape data matrix P' and the fourth shape data matrix Q are precisely matched using the normal vector iteration nearest point. The normal vector of the fifth shape data matrix P' and the normal vector of the fourth shape data matrix Q are calculated. In the fifth shape data matrix P' and the fourth shape data matrix Q to be matched, the corresponding nearest points are found according to the constraints. The singular value decomposition method is used to obtain the rotation matrix and displacement vector in the optimization process. After continuous iteration, the optimal rotation matrix R under the optimal matching is obtained. * and the optimal displacement vector T * .

[0012] Step 5: interpolate the matching area Q1 of the fourth topography data matrix Q with the third topography data matrix P to obtain a sixth topography data matrix Q', so that the sixth topography data matrix Q' and the third topography data matrix P have the same amount of data.

[0013] Step 6: Using the optimized multi-scale fusion strategy adapted to non-planar surfaces, the sixth morphology data matrix Q' and the third morphology data matrix P are processed to obtain α sub-data matrices:

[0014] The sixth morphological data matrix Q' and the third morphological data matrix P are decomposed by wavelet using the porous algorithm to obtain two groups of spatial sub-data matrices. Each group of spatial sub-data matrices contains α wavelet decomposition sub-data matrices, denoted as p1~p α and q1~q α , each group of sub-data matrices contains different frequency components, where p1~p α / 3 and q1~q α / 3It belongs to the low-frequency component, indicating the shape, p (α / 3)+1 ~p 2α / 3 and q (α / 3)+1 ~q 2α / 3 Belongs to the intermediate frequency component, indicating the ripple, p (2α / 3)+1 ~p α and q (2α / 3)+1 ~q α It is a high-frequency component and represents roughness.

[0015] Step 7: Construct α data matrices f1~f α :

[0016] [f1,…,f α / 3 ]=[q1,…,q α / 3 ],[f (2α / 3)+1 , …, f α ]=[p (2α / 3)+1 ,…,p α ], f (α / 3)+1 ~f 2α / 3 According to the value of S, the corresponding p and q sub-data matrices are selected for fusion, and the regional edge intensity method is used to obtain α data matrices f1~f α .

[0017] Step 8: For the data matrix f1~f α Perform inverse wavelet transform to obtain fused data matrix F, embed the fused data matrix F into the second morphology data matrix B at the matching area position with the first morphology data matrix A and display it, so as to obtain the fusion effect of multi-magnification morphology data.

[0018] Compared with the prior art, the present invention has the following significant advantages:

[0019] (1) Compared with the stitching of the entire field of view, the present invention proposes to selectively fuse the area of ​​interest to ensure high-resolution morphological information at the key structure, achieve effective multi-scale characterization, and greatly shorten the measurement time.

[0020] (2) An optimized multi-scale data fusion strategy that adapts to non-planar surfaces and reduces the edge effects produced during the fusion process is developed.

[0021] (3) According to the cutoff frequency of the interference objective lens, a frequency screening strategy for the morphology data matrix in the fusion process is provided, which is suitable for the fusion of morphology data with arbitrary magnification differences. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is the flow chart of multi-rate fusion of white light interferometer.

[0023] Figure 2 Schematic diagram of the fourth morphology data matrix Q and the third morphology data matrix P.

[0024] Figure 3 Schematic diagram of normal vectors of the fourth morphology data matrix Q and the fifth morphology data matrix P'.

[0025] Figure 4 Schematic diagram of the six-level decomposition and fusion of the sixth morphology data matrix Q' and the third morphology data matrix P using the porous algorithm.

[0026] Figure 5 A magnified schematic diagram of the fused data matrix F embedding.

[0027] Figure 6 The fourth morphological data comparison matrix Comparison matrix with the third morphological data Schematic diagram.

[0028] Figure 7 The fourth morphological data comparison matrix And the fifth morphological data comparison matrix Schematic diagram of normal vector.

[0029] Figure 8 The sixth morphological data comparison matrix Comparison matrix with the third morphological data Schematic diagram of the three-level decomposition and fusion of the porous algorithm.

[0030] Fig. 9 is the fusion data comparison matrix Embed schematic. DETAILED DESCRIPTION

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0032] In view of the defect of white light interferometer lacking multi-scale characterization capability, the present invention proposes for the first time a multi-rate fusion method based on frequency screening, which is applicable to interference lenses of any magnification, reduces the edge effect of fused data, performs data fusion on the area of ​​interest, effectively reduces data redundancy, and improves the efficiency of instrument measurement.

[0033] Combination Figure 1 , a multi-magnification morphology data fusion method based on interference objective frequency screening, the steps are as follows:

[0034] Step 1, measure the sample using a white light interferometer and two interference objective lenses with different magnifications, wherein the ratio of the magnification of the first interference objective lens to the magnification of the second interference objective lens is Γ, Γ>1, the first interference objective lens measures to obtain a first morphology data matrix A, which mainly focuses on the local area that requires high-precision measurement, and the second interference objective lens measures to obtain a second morphology data matrix B, which mainly focuses on the global morphology.

[0035] Step 2: According to the numerical aperture NA and the corresponding cutoff frequency f of the two interference objective lenses with different magnifications s , respectively obtain the frequency range of the first morphology data matrix A and the frequency range of the second morphology data matrix B, and calculate the ratio S of the cutoff frequencies of the first morphology data matrix A and the second morphology data matrix B.

[0036] Cut-off frequency f s The calculation formula is as follows:

[0037]

[0038] Where NA is the numerical aperture of the interference lens, and λ is the central wavelength of the light source.

[0039] Step 3: Perform filter preprocessing on the first morphology data matrix A to remove outliers to obtain a third morphology data matrix P. Perform filter preprocessing on the second morphology data matrix B to remove outliers to obtain a fourth morphology data matrix Q. The sizes of the third morphology data matrix P and the fourth morphology data matrix Q are both M×N.

[0040] Step 4: Perform bilinear interpolation on the third shape data matrix P to adjust the size of the third shape data matrix P to The fifth topography data matrix P' is obtained, and the fifth topography data matrix P' and the fourth topography data matrix Q are registered:

[0041] First, a grayscale-based normalized correlation matching algorithm is used for rough matching. The fifth shape data matrix P' and the fourth shape data matrix Q are searched and compared pixel by pixel according to the similarity measurement method to match them and obtain the initial translation vector t0. The fifth shape data matrix P' and the fourth shape data matrix Q are precisely matched using the normal vector iteration nearest point. The normal vector of the fifth shape data matrix P' and the normal vector of the fourth shape data matrix Q are calculated. In the fifth shape data matrix P' and the fourth shape data matrix Q to be matched, the corresponding nearest points are found according to the constraints. The singular value decomposition method is used to obtain the rotation matrix and displacement vector in the optimization process. After continuous iteration, the optimal rotation matrix R under the optimal matching is obtained. * and the optimal displacement vector T * .

[0042] The calculation formula of the grayscale-based normalized correlation matching algorithm is as follows:

[0043]

[0044] In the formula, the size of the fifth topography data matrix P' is m×n, the size of the fourth topography data matrix Q is M×N, NCC(x,y) represents the normalized correlation parameter with (x,y) as the vertex in the fourth topography data matrix Q, P'(i,j) is the corresponding value at the coordinate (i,j) in the fifth topography data matrix P', and Q(x+i,y+j) is the corresponding value at the coordinate (x+i,y+j) in the fourth topography data matrix Q.

[0045] The normal vector calculation and curvature calculation process are as follows:

[0046] Calculate the mean μ a :

[0047]

[0048] In the formula, is any point in the third morphological data matrix P, So The set of all points within a radius of r, It is a point set Hit the point.

[0049] Calculate the covariance Σ a :

[0050]

[0051] Where T represents transpose.

[0052] Perform singular value decomposition on the covariance matrix to obtain the eigenvalues ​​λ1, λ2, and λ3 in ascending order, where λ1 is the normal vector.

[0053] Step 5: interpolate the matching area Q1 of the fourth topography data matrix Q with the third topography data matrix P to obtain a sixth topography data matrix Q', so that the sixth topography data matrix Q' and the third topography data matrix P have the same amount of data.

[0054] Step 6: Using the optimized multi-scale fusion strategy adapted to non-planar surfaces, the sixth morphology data matrix Q' and the third morphology data matrix P are processed to obtain the α sub-data matrix:

[0055] The sixth morphology data matrix Q' and the third morphology data matrix P are subjected to wavelet decomposition using the porous algorithm, and the B3 spline is selected as the wavelet basis to obtain two sets of spatial domain sub-data matrices.

[0056] The value of α is determined by the value of the ratio S of the cutoff frequencies of the first morphological data matrix A and the second morphological data matrix B. When α=9, When α=6, , α=3.

[0057] Each set of spatial sub-data matrices contains α wavelet decomposition sub-data matrices, denoted as p1~p α and q1~q α ,when When α=9, When α=6, When α=3, each group of sub-data matrices contains different frequency components, where p1~p α / 3 and q1~q α / 3 It belongs to the low-frequency component, indicating the shape, p (α / 3)+1 ~p 2α / 3 and q (α / 3)+1 ~q 2α / 3 Belongs to the intermediate frequency component, indicating the ripple, p (2α / 3)+1 ~p α and q (2α / 3)+1 ~q α It is a high-frequency component and represents roughness.

[0058] Step 7: Construct α data matrices f1~f α :

[0059] [f1,…,f α / 3 ]=[q1,…,q α / 3 ],[f (2α / 3)+1 , …, f α ]=[p (2α / 3)+1 ,…,p α ], because the low-frequency sub-data part f1~f α / 3 No fusion is performed, so that the fused data matrix F reduces the edge effect after embedding and display, f (α / 3)+1 ~f 2α / 3 According to the value of S, the corresponding p and q sub-data matrices are selected for fusion, and the regional edge intensity method is used to obtain α data matrices f1~f α .

[0060] The fusion symbol of two groups of sub-data matrices is denoted by ⊕.

[0061] when , the third morphology data matrix P and the sixth morphology data matrix Q' are both decomposed into nine levels, p4 and q4 are fused, and p5 and q5 are fused, that is, f1=q1, f2=q2, f3=q3, f4=p4⊕q4, f5=p5⊕q5, f6=p6, f7=p7, f8=p8, f9=p9.

[0062] when , the third morphology data matrix P and the sixth morphology data matrix Q' are both decomposed into nine levels, p4 is fused with q4, p5 is fused with q5, and p6 is fused with q6, that is, f1=q1, f2=q2, f3=q3, f4=p4⊕q4, f5=p5⊕q5, f6=p6⊕q6, f7=p7, f8=p8, and f9=p9.

[0063] when When the third morphology data matrix P and the sixth morphology data matrix Q' are both decomposed to the sixth level, p3 and q3 are merged, that is, f1 = q1, f2 = q2, f3 = p3⊕q3, f4 = p4, f5 = p5, f6 = p6.

[0064] when When the third morphology data matrix P and the sixth morphology data matrix Q' are both decomposed into six levels, p3 and q3 are fused, and p4 and q4 are fused, that is, f1=q1, f2=q2, f3=p3⊕q3, f4=p4⊕q4, f5=p5, and f6=p6.

[0065] when When the third morphology data matrix P and the sixth morphology data matrix Q' are both decomposed at the third level, p2 and q2 are fused, that is, f1=q1, f2=p2⊕q2, f3=p3.

[0066] Step 8: For the data matrix f1~f α Perform inverse wavelet transform to obtain fused data matrix F, embed the fused data matrix F into the second morphology data matrix B at the matching area position with the first morphology data matrix A and display it, so as to obtain the fusion effect of multi-magnification morphology data.

[0067] Embodiment 1:

[0068] Step 1: Use 5X and 10X interference objective lenses to measure the sample on a white light interferometer. The sample is a microlens array, focusing on high-precision measurement of the concave area. Use a 5X interference objective lens to measure a large range of the sample to obtain a second morphology data matrix B. Use a 10X interference objective lens to measure the concave area to obtain a high-precision first morphology data matrix A. The first morphology data matrix A and the second morphology data matrix B have a Γ of 2, where the central wavelength of the white light interferometer λ = 0.638um, and the NA of the 5X interference objective lens is 5X =0.13, 10X interference objective NA 10X =0.3.

[0069] Step 2: According to the above formula, the cutoff frequency of the first morphology data matrix A is calculated to be 0.4075, and the cutoff frequency of the second morphology data matrix B is calculated to be 0.9404, and the ratio of the cutoff frequencies of the first morphology data matrix A and the second morphology data matrix B is obtained to be S=0.4333.

[0070] Step 3: Filter preprocess the first shape data matrix A to remove outliers to obtain a third shape data matrix P. Filter preprocess the second shape data matrix B to remove outliers to obtain a fourth shape data matrix Q. The sizes of the third shape data matrix P and the fourth shape data matrix Q are both 1024×1024. The third shape data matrix P and the fourth shape data matrix Q are as follows: Figure 2 shown.

[0071] Step 4: Perform bilinear interpolation on the third topography data matrix P, adjust the size of the third topography data matrix P to 512×512, obtain the fifth topography data matrix P', and align the fifth topography data matrix P' with the fourth topography data matrix Q:

[0072] First, a grayscale-based normalized correlation matching algorithm is used for rough matching to obtain the initial translation vector Use the normal vector iteration to accurately match the nearest point to the data and calculate the normal vectors of the third shape data matrix P and the fourth shape data matrix Q, as shown in Figure 3 As shown, in the third shape data matrix P and the fourth shape data matrix Q to be matched, according to certain constraints, the corresponding nearest points are found, and the optimal matching parameter rotation matrix is ​​obtained by continuous iteration. and displacement vector

[0073] Step 5. After the matching is completed, the matching area in the fourth morphology data matrix Q is interpolated to obtain the sixth morphology data matrix Q'. The sixth morphology data matrix Q' has the same amount of data as the third morphology data matrix P. According to S = 0.6667, α = 6, so the porous algorithm is used to perform six-level wavelet decomposition on the sixth morphology data matrix Q' and the third morphology data matrix P to obtain two groups of data, each group of data contains six wavelet decomposition sub-data, recorded as p1~p6 and q1~q6.

[0074] Step 6: To obtain f1-f6, select the following fusion strategy according to the value of S: fuse p3 with q3, and fuse p4 with q4, that is, f1 = q1, f2 = q2, f3 = p3⊕q3, f4 = p4⊕q4, f5 = p5, f6 = p6, as shown in Figure 4 shown.

[0075] Step 7: Perform an inverse transformation on f1-f6 to obtain a fused data matrix F, embed the fused data matrix F into the second shape data matrix B at the matching area position with the first shape data matrix A and display it, combined with Figure 5 As shown, the multi-magnification morphology data fusion method based on interference objective frequency screening proposed in the present invention can effectively reduce the edge effect, reduce the noise in the morphology data matrix obtained by 5X interference objective detection, effectively realize multi-scale characterization, and reduce data redundancy.

[0076] Example 2 Comparative Test

[0077] Combination Figures 1 to 9 , a multi-magnification morphology data fusion method based on interference objective frequency screening, the steps are as follows:

[0078] Step 1: Use 10X and 20X interference objective lenses to measure the sample on a white light interferometer. This sample is a microlens array, which focuses on high-precision measurement of the concave area. Use a 10X interference objective lens to measure a large range of the sample to obtain the second morphology data comparison matrix. Use a 20X interference objective lens to measure the concave area and obtain a high-precision first morphology data comparison matrix The first morphological data comparison matrix Compare the matrix with the second topographic data of The central wavelength of the white light interferometer is λ = 0.638um, and the NA of the 10X interference objective lens is 10X =0.3, 20X interference objective NA 20X =0.4.

[0079] Step 2: According to the above formula, the cutoff frequency of the second morphological data comparison matrix B is calculated to be 0.9404, and the cutoff frequency of the first morphological data comparison matrix The cutoff frequency is 1.2539, and the first morphological data comparison matrix is ​​obtained Compare the second morphological data matrix The ratio of the cut-off frequencies

[0080] Step 3: Compare the first morphological data matrix Perform filtering preprocessing to remove outliers and obtain the third morphological data comparison matrix Compare the second morphological data matrix Perform filtering preprocessing to remove outliers and obtain the fourth morphological data comparison matrix The third morphological data comparison matrix And the fourth morphological data comparison matrix The size of the third morphological data comparison matrix is ​​1024×1024. And the fourth morphological data comparison matrix like Figure 6 shown.

[0081] Step 4: Compare the third morphological data matrix Perform bilinear interpolation to compare the third morphological data matrix The size is adjusted to 512×512 to obtain the new fifth morphological data comparison matrix Comparison matrix for the fifth morphological data And the fourth morphological data comparison matrix Perform the registration:

[0082] First, a grayscale-based normalized correlation matching algorithm is used for rough matching to obtain the initial translation vector Use normal vector iteration to accurately match the nearest point to the data and calculate the third shape data comparison matrix And the fourth morphological data comparison matrix The normal vector of Figure 7 As shown, in the third morphological data comparison matrix to be matched And the fourth morphological data comparison matrix According to certain constraints, find the corresponding nearest point, iterate continuously, and get the optimal matching parameter rotation matrix and displacement vector

[0083] Step 5: After the matching is completed, compare the fourth morphological data matrix The sixth morphological data comparison matrix is ​​obtained by interpolation of the matching area Sixth morphological data comparison matrix Comparison matrix with the third morphological data The amount of data is the same, according to We can get α = 3, so we use the multi-pore algorithm to compare the sixth morphological data matrix Comparison matrix with the third morphological data Perform three-level wavelet decomposition to obtain two sets of data, each of which contains three wavelet decomposition sub-data, recorded as and

[0084] Step 6: To get according to The value of selects the following fusion strategy: and To merge, like Figure 8 shown.

[0085] Step 7: Construct a data comparison matrix

[0086] Step 8: Perform inverse transformation to obtain the fusion data comparison matrix Compare the fused data to the matrix Embed the second topographic data comparison matrix Comparison matrix with the first morphological data The matching area position is displayed and combined with Fig. 9 As shown, the multi-magnification morphology data fusion method based on interference objective frequency screening proposed in the present invention can effectively reduce the edge effect, is suitable for the fusion of morphology data with arbitrary magnification differences, and effectively realizes multi-scale characterization.

Claims

1. A multi-magnification morphology data fusion method based on interference objective frequency screening, characterized in that: Here are the steps: Step 1, measuring the sample by a white light interferometer and two interference objective lenses with different magnifications, wherein the ratio of the magnification of the first interference objective lens to the magnification of the second interference objective lens is Γ, Γ>1, the first interference objective lens measures to obtain a first morphology data matrix A, and the second interference objective lens measures to obtain a second morphology data matrix B; Step 2: According to the numerical aperture NA and the corresponding cutoff frequency f of the two interference objective lenses with different magnifications s , respectively obtain the frequency range of the first topography data matrix A and the frequency range of the second topography data matrix B, and calculate the ratio S of the cutoff frequencies of the first topography data matrix A and the second topography data matrix B; Step 3, filter preprocessing the first shape data matrix A to remove outliers to obtain a third shape data matrix P, filter preprocessing the second shape data matrix B to remove outliers to obtain a fourth shape data matrix Q, the sizes of the third shape data matrix P and the fourth shape data matrix Q are both M×N; Step 4: Perform bilinear interpolation on the third morphological data matrix P to adjust the size of the third morphological data matrix P to The fifth topography data matrix P' is obtained, and the fifth topography data matrix P' and the fourth topography data matrix Q are registered: First, a grayscale-based normalized correlation matching algorithm is used for rough matching. The fifth shape data matrix P' and the fourth shape data matrix Q are searched and compared pixel by pixel according to the similarity measurement method to match them and obtain the initial translation vector t0. The fifth shape data matrix P' and the fourth shape data matrix Q are precisely matched using the normal vector iteration nearest point. The normal vector of the fifth shape data matrix P' and the normal vector of the fourth shape data matrix Q are calculated. In the fifth shape data matrix P' and the fourth shape data matrix Q to be matched, the corresponding nearest points are found according to the constraints. The singular value decomposition method is used to obtain the rotation matrix and displacement vector in the optimization process. After continuous iteration, the optimal rotation matrix R under the optimal matching is obtained. * and the optimal displacement vector T * ; Step 5, interpolating the matching area Q1 of the fourth topography data matrix Q with the third topography data matrix P to obtain a sixth topography data matrix Q', so that the sixth topography data matrix Q' has the same amount of data as the third topography data matrix P; Step 6: Using the optimized multi-scale fusion strategy adapted to non-planar surfaces, the sixth morphology data matrix Q' and the third morphology data matrix P are processed to obtain α sub-data matrices: The sixth morphological data matrix Q' and the third morphological data matrix P are decomposed by wavelet using the porous algorithm to obtain two groups of spatial sub-data matrices. Each group of spatial sub-data matrices contains α wavelet decomposition sub-data matrices, denoted as p1~p α and q1~q α , each group of sub-data matrices contains different frequency components, where p1~p α / 3 and q1~q α / 3 It belongs to the low-frequency component, indicating the shape, p (α / 3)+1 ~p 2α / 3 and q (α / 3)+1 ~q 2α / 3 Belongs to the intermediate frequency component, indicating the ripple, p (2α / 3)+1 ~p α and q (2α / 3)+1 ~q α It is a high-frequency component, indicating roughness; Step 7: Construct α data matrices f1~f α : [f1,…,f α / 3 ]=[q1,…,q α / 3 ],[f (2α / 3)+1 , …, f α ]=[p (2α / 3)+1 ,…,p α ], f (α / 3)+1 ~f 2α / 3 According to the value of S, the corresponding p and q sub-data matrices are selected for fusion, and the regional edge intensity method is used to obtain α data matrices f1~f α ; Step 8: For the data matrix f1~f α Perform inverse wavelet transform to obtain fused data matrix F, embed the fused data matrix F into the second morphology data matrix B at the matching area position with the first morphology data matrix A and display it, so as to obtain the fusion effect of multi-magnification morphology data.

2. The multi-rate morphology data fusion method based on interference objective lens frequency screening according to claim 1 is characterized in that: The cutoff frequency f in step 2 s The calculation formula is as follows: Where NA is the numerical aperture of the interference lens, and λ is the central wavelength of the light source.

3. The multi-rate morphology data fusion method based on interference objective lens frequency screening according to claim 1 is characterized in that: In step 4, the calculation formula of the grayscale-based normalized correlation matching algorithm is as follows: In the formula, the size of the fifth topography data matrix P' is m×n, the size of the fourth topography data matrix Q is M×N, NCC(x,y) represents the normalized correlation parameter with (x,y) as the vertex in the fourth topography data matrix Q, P'(i,j) is the corresponding value at the coordinate (i,j) in the fifth topography data matrix P', and Q(x+i,y+j) is the corresponding value at the coordinate (x+i,y+j) in the fourth topography data matrix Q.

4. The multi-magnification morphology data fusion method based on interference objective lens frequency screening according to claim 1 is characterized in that: In step 4, the normal vector calculation and curvature calculation process are as follows: Calculate the mean μ a : In the formula, is any point in the third morphological data matrix P, So The set of all points within a radius of r, It is a point set center point; Calculate the covariance Σ a : In the formula, T represents transpose; Perform singular value decomposition on the covariance matrix to obtain the eigenvalues ​​λ1, λ2, and λ3 in ascending order, where λ1 is the normal vector.

5. The multi-rate morphology data fusion method based on interference objective lens frequency screening according to claim 1 is characterized in that: In step 6, the sixth morphology data matrix Q' and the third morphology data matrix P are processed using an optimized multi-scale fusion strategy adapted to non-planar surfaces to obtain α sub-data matrices, as follows: The registered third morphological data matrix P and the sixth morphological data matrix Q' are subjected to wavelet decomposition using a porous algorithm, and the B3 spline is selected as the wavelet basis to obtain α sub-data matrices.

6. The multi-rate morphology data fusion method based on interference objective lens frequency screening according to claim 5 is characterized in that: In step 6, the value of α is determined according to the value of the ratio S of the cutoff frequencies of the first morphological data matrix A and the second morphological data matrix B. When α=9, When α=6, , α=3.

7. The method for multi-rate morphology data fusion based on interference objective lens frequency screening according to claim 1, characterized in that: In step 7, construct α data matrices f1~f α , as follows: The fusion symbol of two groups of sub-data matrices is denoted by when When , the third morphological data matrix P and the sixth morphological data matrix Q' are both decomposed to nine levels, p4 is fused with q4, and p5 is fused with q5, that is, f1=q1, f2=q2, f3=q3, f4=p4⊕q4, f5=p5⊕q5, f6=p6, f7=p7, f8=p8, f9=p9; when When the third morphological data matrix P and the sixth morphological data matrix Q' are both decomposed to nine levels, p4 is fused with q4, p5 is fused with q5, and p6 is fused with q6, that is, f1 = q1, f2 = q2, f3 = q3, f7=p7, f8=p8, f9=p9; when When the third morphological data matrix P and the sixth morphological data matrix Q' are both decomposed to level 6, p3 and q3 are merged, that is, f1 = q1, f2 = q2, f4=p4, f5=p5, f6=p6; when When the third morphological data matrix P and the sixth morphological data matrix Q' are both decomposed to six levels, p3 and q3 are fused, and p4 and q4 are fused, that is, f1 = q1, f2 = q2, f5=p5,f6=p6; when When the third morphological data matrix P and the sixth morphological data matrix Q' are both decomposed at the third level, p2 and q2 are fused, that is, f1 = q1. f3=p3.

Images