A moire fringe wrapped phase distribution extraction method, system and device based on unsupervised learning and a storage medium

Abstract

The application discloses a kind of based on the extraction method of moire fringe wrapped phase distribution of unsupervised learning, comprising: obtaining the moire fringe of measured flow field;The moire fringe is preprocessed, and first matrix is obtained, peak position determination is carried out to first matrix, and wave crest matrix is obtained;Wave crest matrix is input into pre-trained unsupervised learning cluster analysis model and is calculated, and first-order spectrum position is obtained;First-order spectrum position is post-processed, and the wrapped phase distribution of moire fringe is obtained.Relative to traditional human eye observation, the time of artificial determination is saved and the error caused by naked eye is reduced.At the same time, compared with the algorithm of convolutional neural network, through the cluster analysis of unsupervised learning, unlabeled data can be processed, the calculation cost is low, and the extraction efficiency is high.The wrapped phase calculation time of moire fringe can be saved, the data processing steps are simplified, and more accurate and reliable basis can be provided for later phase unwrapping.

Description

Technical Field

[0001] This invention relates to a method, system, apparatus, and storage medium for extracting the phase distribution of moiré fringe wrapping based on unsupervised learning, belonging to the field of optical measurement technology. Background Technology

[0002] Moiré fringes are most widely used in grating displacement measurement. Based on the principle of moiré fringes, static and dynamic measurements of linear and angular displacements can be achieved. Precision displacement measurement can be realized based on the relationship between the number of moiré fringes and displacement. This allows for measurement and control feedback to meet various needs, including contact, non-contact, small-range, large-range, one-dimensional, and multi-dimensional measurements. It is widely used in many fields such as CNC machine tools and coordinate measuring machines, precision measurement and positioning, ultra-precision machining, microelectronics IC manufacturing, earthquake prediction, quality inspection, nanomaterials, robotics, MEMS, and vibration detection. Understanding the relationship between the number of moiré fringes and displacement requires accurately extracting the first-order spectrum of the moiré fringes to obtain the wrapping phase. However, current technologies often use Fourier analysis to extract the wrapping phase of moiré fringes. After the fringe image undergoes Fourier transform, the first-order spectrum of the fringes needs to be manually extracted, which is cumbersome and time-consuming. Currently, there is no systematic method to automatically extract the first-order spectrum of moiré fringes, and improving the efficiency of information extraction from moiré fringes is a problem that urgently needs to be solved. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method, system, device and storage medium for extracting the phase distribution of moiré fringe wrapping based on unsupervised learning. This can save the calculation time of the moiré fringe wrapping phase, simplify the data processing steps, and provide a more accurate and reliable basis for subsequent phase unwrapping.

[0004] To achieve the above objectives, the present invention is implemented using the following technical solution:

[0005] In a first aspect, the present invention provides a method for extracting the phase distribution of moiré fringes based on unsupervised learning, comprising:

[0006] Obtain the moiré fringes of the flow field to be measured;

[0007] The moiré fringes are preprocessed to obtain the first matrix;

[0008] Determine the peak position of the first matrix to obtain the peak matrix;

[0009] Cluster analysis of the peak matrix yields the first-order spectral positions; the unsupervised learning cluster analysis model includes...

[0010] Post-processing of the first-order spectral positions yields the phase distribution of the moiré fringes.

[0011] Furthermore, the training method for the unsupervised learning clustering analysis model includes:

[0012] Obtain the matrix containing peak location information and its initial cluster centers;

[0013] Based on the matrix containing peak position information and its initial cluster centers, the distance from each element in the matrix containing peak position information to each initial cluster center is calculated. Elements with the same distance are grouped into the same category to obtain the clusters of the matrix containing peak position information. The number of clusters matches the number of initial cluster centers.

[0014] Based on the obtained clusters, each initial cluster center is updated to obtain the updated cluster centers;

[0015] Calculate the error function between the updated cluster centers and the corresponding initialized cluster centers until the error function converges, thus completing the training of the cluster centers in the cluster analysis, and obtaining the first-level spectral positions based on the trained cluster centers;

[0016] The formula for calculating the error function is as follows:

[0017]

[0018] In the formula, L is the error function value; k is the number of initial cluster centers; an is the initial cluster center; and an' is the updated cluster center.

[0019] Furthermore, the preprocessing of the moiré fringes to obtain the first matrix specifically includes:

[0020] First, select the calculation region of the moiré fringes of the flow field to be measured. The calculation region is an m×n matrix A.

[0021] Perform a Fourier transform on matrix A to obtain matrix B;

[0022] The first matrix C is obtained by averaging all elements in each row of matrix B.

[0023] Furthermore, the peak position determination of the first matrix to obtain the peak matrix specifically includes:

[0024] Peak determination is performed on the spectrum corresponding to the first matrix C. The peak determination formula is x. i-1 <x i >x i+1 And x that does not satisfy the peak determination formula i Setting it to 0 yields the peak matrix D.

[0025] Furthermore, the post-processing of the first-order spectral positions yields the enveloping phase distribution of the moiré fringes; specifically, this includes:

[0026] The first-order spectrum position is shifted to the center of the moiré fringe calculation region to obtain the spectrum E;

[0027] Performing an inverse Fourier transform on the spectrum E yields the matrix H;

[0028] The phase distribution of the moiré fringes is calculated using the arctangent function based on matrix H.

[0029] Secondly, the present invention provides a system for extracting the phase distribution of moiré fringe wrapping based on unsupervised learning, characterized in that it includes:

[0030] Acquisition module: used to acquire the moiré fringes of the flow field under test;

[0031] Preprocessing module: preprocesses the moiré fringes to obtain the first matrix;

[0032] Peak determination module: Determines the peak position of the first matrix to obtain the peak matrix;

[0033] Clustering analysis module: Performs clustering analysis on the peak matrix to obtain the first-level spectral positions;

[0034] Post-processing module: Post-processes the first-level spectral positions to obtain the phase distribution of the moiré fringes.

[0035] Thirdly, an extraction device for moiré fringe wrapping phase distribution based on unsupervised learning, including a processor and a storage medium;

[0036] The storage medium is used to store instructions;

[0037] The processor is configured to operate according to the instructions to perform the steps of the method according to any of the foregoing.

[0038] Fourthly, a computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of any of the methods described above.

[0039] Compared with the prior art, the beneficial effects achieved by the present invention are as follows:

[0040] This invention preprocesses moiré fringes, then determines the peak positions of the preprocessed fringes, and rapidly extracts the first-order spectral positions through unsupervised learning clustering analysis, thereby obtaining the wrapping phase distribution of the moiré fringes. Compared to traditional human observation, this saves time on manual determination and reduces errors introduced by the human eye. Furthermore, compared to convolutional neural network algorithms, unsupervised learning clustering analysis can handle unlabeled data, resulting in lower computational costs and higher extraction efficiency. Additionally, the use of Fourier transform as a preprocessing method in both stages, and inverse Fourier transform in the post-processing, improves the accuracy of moiré fringe wrapping phase extraction. Attached Figure Description

[0041] Figure 1 This is a flowchart of the extraction method in Example 1.

[0042] Figure 2 The moiré pattern of the flow field to be tested was obtained from the propane combustion field experiment in Test Example 1.

[0043] Figure 3 yes Figure 2 The computational region map of the moiré fringes of the flow field under test;

[0044] Figure 4 It is the graph corresponding to matrix B in test example 1;

[0045] Figure 5 This is a graph of the first matrix C in Test Example 1;

[0046] Figure 6 This is the spectrum of the peak matrix D in Test Example 1;

[0047] Figure 7 This is the image of spectrum E in test example 1.

[0048] Figure 8 This is the phase distribution diagram of the moiré fringes obtained in Test Example 1. Detailed Implementation

[0049] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0050] Example 1:

[0051] To simplify the extraction process of moiré fringe-wrapped phase and improve extraction efficiency, this invention provides a method, system, apparatus, and storage medium for extracting moiré fringe-wrapped phase distribution based on unsupervised learning. The extraction method mainly consists of three stages: a preprocessing stage, a cluster analysis stage, and a post-processing stage. Figure 1 As shown, it specifically includes:

[0052] S1: Obtain the moiré fringes of the flow field to be measured;

[0053] S2: Preprocess the moiré fringes to obtain the first matrix;

[0054] S3: Determine the peak position of the first matrix to obtain the peak matrix;

[0055] S4: Perform cluster analysis on the peak matrix to obtain the first-order spectral positions;

[0056] S5: Post-process the first-order spectral positions to obtain the phase distribution of the moiré fringes.

[0057] (1) Pre-processing stage

[0058] The preprocessing stage includes steps S2 and S3. Step S2 specifically includes the following steps:

[0059] S21: First, select the calculation region of the moiré fringes of the flow field to be measured. The calculation region is an m×n matrix A.

[0060] S22: Perform a Fourier transform on matrix A to obtain matrix B. The calculation formula is as follows:

[0061]

[0062] In the formula, Fourier transform;

[0063] S23: Take the average of all elements in each row of matrix B to obtain the first matrix C. The calculation formula is:

[0064]

[0065] In the formula, i is the row number of matrix B; j is the column number of matrix B; N is the total number of columns in matrix B; B ij x is the element in the i-th row and j-th column of matrix B; i Let be the vectors in the first matrix C, that is, the average vector of the row elements of matrix B;

[0066] Based on the spectrum corresponding to matrix B obtained after Fourier transform, it can be intuitively seen that the value of the first-order spectrum is very prominent relative to the vicinity, which facilitates the determination of the position of the first-order spectrum in subsequent steps, simplifies the extraction method, and improves the accuracy of extraction.

[0067] For step S3, the formula for determining the peak value is:

[0068] x i-1 <x i >x i+1 ,

[0069] The vectors that satisfy the judgment formula retain their original values, while the vectors that do not satisfy the judgment formula are assigned a value of 0, thus obtaining the peak matrix D.

[0070] (2) Cluster analysis stage

[0071] The clustering analysis stage includes step S4, which performs clustering analysis on the peak matrix D based on unsupervised learning to quickly extract the first-level spectral positions. Compared with traditional human observation, this saves time for manual determination and reduces errors caused by the human eye. At the same time, compared with convolutional neural network algorithms, clustering analysis through unsupervised learning can process unlabeled data, with low computational cost and high extraction efficiency.

[0072] The general method of cluster analysis in this embodiment specifically includes:

[0073] S41: Obtain the matrix containing peak position information and its initialized cluster centers. There are k cluster centers in total, where k is a positive integer, denoted as (a1, a2, a3, ..., a...). k )express;

[0074] S42: Based on the matrix containing peak position information and its initial cluster centers, calculate the distance from each element in the matrix containing peak position information to each initial cluster center, and classify the clusters according to the different distances. Specifically, groups the elements with the same distance into the same category to obtain the clusters of the matrix containing peak position information. The calculation formula is as follows:

[0075]

[0076] In the formula, x i a is the i-th element in the matrix containing peak location information; j For the j-th cluster center, r i For element x i To cluster center a j The j value returned when the distance is minimized.

[0077] r i Those that are similarly classified into the same category, that is... Indicates r i Identical clusters; obtain clusters containing a matrix of peak position information, wherein the clusters match the number of initial cluster centers.

[0078] S43: Based on the obtained clusters, update each initial cluster center to obtain the updated cluster centers. The calculation formula is as follows:

[0079]

[0080] In the formula, a′n For the updated cluster centers; C n For clusters; p n C n The number of data points in the middle;

[0081] S44: Calculate the error function between the updated cluster centers and the corresponding initialized cluster centers until the error function converges, thus completing the training of the unsupervised learning cluster analysis model;

[0082] The formula for calculating the error function is as follows:

[0083]

[0084] In the formula, L is the error function value; k is the number of initial cluster centers; an is the initial cluster center; and an' is the updated cluster center.

[0085] After training is complete, simply input the peak matrix from step S3 into the trained unsupervised learning clustering analysis model for calculation. This will divide all data in the peak matrix into k classes and obtain the class of the first-order spectrum. This will make it easy and effective to determine the position of the first-order spectrum in the moiré fringes in subsequent steps.

[0086] For step S5, the specific steps include:

[0087] S51: Shift the position of the first-order spectrum to the center of the moiré fringe calculation region to obtain the spectrum E;

[0088] S52: Perform an inverse Fourier transform on the spectrum E to obtain matrix H, calculated using the following formula:

[0089]

[0090] S53: Based on matrix H, the phase distribution of the moiré fringes is calculated using the arctangent function. The calculation formula is as follows:

[0091]

[0092] Test Example 1

[0093] In this test example, a propane combustion field experiment was selected to obtain the moiré fringes of the flow field under test. The moiré fringes pattern is shown below. Figure 2 As shown, the operating environment specifically includes a 16-inch 2019 MacBook Pro with an Intel Core i7-9750H CPU (six cores, 2.6GHz) and 16GB of memory (2667MHz, DDR4).

[0094] For the preprocessing stage, the computational region of the moiré fringes of the flow field to be measured is selected as follows: Figure 3As shown, the region to be calculated is matrix A, which is m×n. Then, a Fourier transform is performed on matrix A to obtain matrix B. The graph corresponding to matrix B is shown below. Figure 4 As shown, some data in matrix B are as follows: Figure 4 As shown in the data table on the right, the values ​​of the first-order spectrum are relatively prominent compared to the surrounding area. Then, the average of all elements in each row of matrix B is taken to obtain the first matrix C. The curve of the first matrix C is shown below. Figure 5 As shown, it can be observed that both the zeroth and first orders are located at "peaks." The maximum peak of the average value vector x represents the zeroth-order spectrum, and the second-maximum peak represents the first-order spectrum. Finally, peak position determination is performed on the first matrix C to obtain the peak matrix D. The peak matrix D is a vector group containing peak position information, and its image is shown below. Figure 6 As shown, this completes the preprocessing steps for the moiré fringes of the flow field under test. The peak matrix D is then input into the unsupervised learning clustering analysis model for calculation, ultimately obtaining the position of the first-order spectrum.

[0095] In the post-processing stage, the first-order spectral position is shifted to the center of the moiré fringe calculation region to obtain spectrum E. The image of spectrum E is as follows. Figure 7 As shown, some data are as follows: Figure 7 The data table on the right is shown. Then, an inverse Fourier transform is performed on the spectrum E to obtain matrix H. Finally, based on matrix H, the enveloping phase distribution of the moiré fringes is calculated using the arctangent function. The final enveloping phase distribution diagram of the moiré fringes is shown below. Figure 8 As shown.

[0096] On a 16-inch 2019 MacBook Pro with an Intel Core i7-9750H CPU (six-core, 2.6GHz) and 16GB of RAM (2667MHz, DDR4), the entire extraction process took 0.0782 seconds. Compared to traditional human observation, this saves time on manual determination and reduces errors caused by the human eye. Furthermore, compared to convolutional neural network algorithms, it has lower computational costs and is faster.

[0097] Example 2:

[0098] A system for extracting the phase distribution of moiré fringes based on unsupervised learning, comprising:

[0099] Acquisition module: used to acquire the moiré fringes of the flow field under test;

[0100] Preprocessing module: preprocesses the moiré fringes to obtain the first matrix;

[0101] Peak determination module: Determines the peak position of the first matrix to obtain the peak matrix;

[0102] Clustering analysis module: Input the peak matrix into the pre-trained unsupervised learning clustering analysis model for calculation to obtain the first-level spectrum position;

[0103] Post-processing module: Post-processes the first-level spectral positions to obtain the phase distribution of the moiré fringes.

[0104] Example 3:

[0105] This invention also provides an extraction device for the phase distribution of moiré fringe wrapping based on unsupervised learning, including a processor and a storage medium;

[0106] The storage medium is used to store instructions;

[0107] The processor is configured to operate according to the instructions to perform the steps of the following method:

[0108] Obtain the moiré fringes of the flow field to be measured;

[0109] The moiré fringes are preprocessed to obtain the first matrix;

[0110] Determine the peak position of the first matrix to obtain the peak matrix;

[0111] The peak matrix is ​​input into a pre-trained unsupervised learning clustering analysis model for calculation to obtain the first-level spectral positions;

[0112] Post-processing of the first-order spectral positions yields the phase distribution of the moiré fringes.

[0113] Example 4:

[0114] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the following method:

[0115] Obtain the moiré fringes of the flow field to be measured;

[0116] The moiré fringes are preprocessed to obtain the first matrix;

[0117] Determine the peak position of the first matrix to obtain the peak matrix;

[0118] The peak matrix is ​​input into a pre-trained unsupervised learning clustering analysis model for calculation to obtain the first-level spectral positions;

[0119] Post-processing of the first-order spectral positions yields the phase distribution of the moiré fringes.

[0120] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0122] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0123] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0124] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for extracting Moire fringe wrapped phase distribution based on unsupervised learning, characterized in that, The method comprises the following steps: obtaining a moire fringe of a flow field to be measured; preprocessing the moire fringe to obtain a first matrix; The method comprises the following steps: First, select the calculation region of the Moire fringe of the flow field to be measured. The matrix A of the to-be-calculated region is ​ performing Fourier transform on the matrix A to obtain a matrix B; averaging all elements in each row of the matrix B to obtain a first matrix C; performing peak position determination on the first matrix to obtain a wave peak matrix; the method comprises the following steps: Peak determination is performed on the spectrum corresponding to the first matrix C. The peak determination formula is as follows: And x that does not satisfy the peak determination formula i Setting it to 0 yields the peak matrix D; performing cluster analysis on the wave peak matrix to obtain a primary spectrum position; performing post-processing on the primary spectrum position to obtain a wrapped phase distribution of the moire fringe.

2. The method of claim 1, wherein, The cluster analysis method comprises the following steps: obtaining a matrix containing peak position information and an initial cluster center thereof; calculating the distance from each element in the matrix containing peak position information to each initial cluster center according to the matrix containing peak position information and the initial cluster center thereof, and classifying the elements with the same distance into one class to obtain a class cluster of the matrix containing peak position information, wherein the number of the class cluster matches the number of the initial cluster center; updating each initial cluster center according to the obtained class cluster to obtain an updated cluster center; calculating an error function of the updated cluster center and the corresponding initial cluster center until the error function converges, completing the training of the cluster center in the cluster analysis, and obtaining the primary spectrum position according to the trained cluster center; the calculation formula of the error function is: ， In the formula, L is an error function value; k is the number of initialized clustering centers; a n is the initialized clustering center; a n is the updated clustering center.

3. The method of claim 1, wherein, The post-processing of the primary spectrum position to obtain the wrapped phase distribution of the moire fringe; the method comprises the following steps: shifting the primary spectrum position to the center of a calculation region of the moire fringe to obtain a spectrum E; performing inverse Fourier transform on the spectrum E to obtain a matrix H; calculating the wrapped phase distribution of the moire fringe by using an inverse tangent function according to the matrix H.

4. A system for extracting Moire fringe wrapped phase distribution based on unsupervised learning, characterized in that, The method comprises the following steps: an acquisition module for obtaining a moire fringe of a flow field to be measured; a preprocessing module for preprocessing the moire fringe to obtain a first matrix; The method comprises the following steps: First, a calculation region of the Moire fringe of the flow field to be measured is selected, and the matrix A of the to-be-calculated region is ​ performing Fourier transform on the matrix A to obtain a matrix B; averaging all elements in each row of the matrix B to obtain a first matrix C; a peak determination module for performing peak position determination on the first matrix to obtain a wave peak matrix; the method comprises the following steps: Peak value determination formula for the atlas corresponding to the first matrix C is x i Set to 0, and get the wave peak matrix D; a cluster analysis module for performing cluster analysis on the wave peak matrix to obtain a primary spectrum position; a post-processing module for performing post-processing on the primary spectrum position to obtain a wrapped phase distribution of the moire fringe.

5. An apparatus for extracting a Moire fringe wrapped phase distribution based on unsupervised learning, characterized by, The method comprises a processor and a storage medium; The storage medium is used to store instructions; The processor is used to operate according to the instructions to perform the steps of the method according to any one of claims 1-3.

6. A computer readable storage medium having stored thereon a computer program, characterized in that The program is executed by the processor to implement the steps of the method according to any one of claims 1-3.

Images