Calibration method of dispersive confocal measurement device based on principal component analysis and neural network

Abstract

The present invention provides a method for calibrating a dispersion confocal measuring device based on principal component analysis and neural network, and relates to the field of confocal measurement technology. The method comprises: performing principal component analysis on the acquired light intensity matrix to obtain a matching degree matrix; calculating the peak wavelength of the light intensity matrix to obtain a peak wavelength vector; taking a standard distance vector as a fitting target and a peak wavelength vector as an independent variable, fitting the functional relationship between the two to obtain a displacement vector and an error vector; inputting the matching degree matrix and the error vector into a neural network model for training to obtain a back-propagation neural network model; obtaining a current light intensity sequence for preprocessing to obtain a current matching degree vector and a fitting displacement; inputting the current matching degree vector into a back-propagation neural network model to obtain a predicted error value; and calibrating the dispersion confocal measuring device to be calibrated based on the fitting displacement and the predicted error value. The present invention solves the problem of inaccurate calibration of the dispersion confocal measuring device.

Description

Technical Field

[0001] The invention relates to the technical field of confocal measurement, and in particular to a dispersion confocal measurement device calibration method based on principal component analysis and neural network. Background Art

[0002] With the continuous development and progress of science and technology, higher requirements are placed on the measurement accuracy of precision instruments, especially the precision measurement of optical components has always been an important research topic in the field of measurement. In order to avoid damage to the surface of optical components during the measurement process, non-contact high-precision measurement technology is usually used, including dispersion confocal method, white light interferometry, multi-wavelength interferometry, etc.

[0003] Dispersive confocal microscopy is mainly used for micro-nano surface structure measurement, fast coordinate measurement, transparent medium thickness measurement and online process detection. In dispersive confocal microscopy, a point light source is focused and imaged on the image plane behind it after passing through a semi-transparent and semi-reflective mirror and a dispersive lens. The dispersive lens can focus different incident light waves at different axial positions. The dispersive confocal measurement device is a type of optical test device manufactured based on this principle, and it usually needs to be calibrated in practical applications.

[0004] The existing calibration methods mainly use the peak extraction algorithm to extract the peak wavelength corresponding to the peak of the spectral response signal and establish a mapping relationship between the peak wavelength and the measurement position. The mapping relationship mainly includes linear relationship, nonlinear relationship and piecewise linear relationship.

[0005] However, in the actual calibration process, due to errors in the design, processing, and assembly of the measuring device; or due to factors such as the noise level, signal level, and light intensity non-uniformity in the actual measurement; all of these will directly affect the accuracy of the mapping relationship, thereby leading to inaccurate calibration of the dispersive confocal measurement device. Summary of the invention

[0006] The invention provides a dispersion confocal measurement device calibration method based on principal component analysis and neural network, which is used to solve the problem of inaccurate calibration of the dispersion confocal measurement device.

[0007] The present invention provides a method for calibrating a dispersion confocal measurement device based on principal component analysis and a neural network, the method comprising:

[0008] A dual-frequency laser interferometer is used to measure multiple standard positions of the plane reflector to obtain a standard distance vector; a dispersive confocal measurement device is used to obtain spectral response signals corresponding to the multiple standard positions to obtain a light intensity matrix;

[0009] Performing principal component analysis on the light intensity matrix to obtain a matching degree matrix;

[0010] Calculating the peak wavelength of the light intensity matrix to obtain a peak wavelength vector;

[0011] Taking the standard distance vector as the fitting target and the peak wavelength vector as the independent variable, fitting the functional relationship between the two to obtain a displacement vector; and calculating the error value between each element in the displacement vector and each element in the corresponding standard distance vector to obtain an error vector;

[0012] Inputting the matching degree matrix and the error vector into a neural network model for training to obtain a back propagation neural network model;

[0013] Acquire a current light intensity sequence of the dispersive confocal measurement device to be calibrated; preprocess the current light intensity sequence to obtain a current matching degree vector; and input the current matching degree vector into a back propagation neural network model to obtain a prediction error value;

[0014] Obtaining a current peak wavelength based on the current light intensity sequence, and substituting the current peak wavelength as an independent variable into the functional relationship to obtain a fitting displacement;

[0015] The dispersive confocal measurement device to be calibrated is calibrated based on the fitting displacement and the prediction error value.

[0016] Furthermore, the light intensity matrix is ​​subjected to principal component analysis to obtain a matching degree matrix including:

[0017] Performing principal component analysis on the light intensity matrix to obtain a calibration matrix and a dimension reduction matrix;

[0018] A matching degree matrix is ​​obtained based on the calibration matrix.

[0019] Furthermore, the method for obtaining the matching degree matrix includes:

[0020] Performing principal component analysis on the light intensity matrix to obtain a first matrix, a second matrix, a first vector, and a second vector;

[0021] Obtaining the calibration matrix based on the first vector and the first matrix;

[0022] Performing K-means clustering calculation on the calibration matrix to obtain a center matrix;

[0023] The Euclidean distance between each row vector in the calibration matrix and each row vector in the center matrix is ​​calculated, and after activation by a Gaussian function, the matching degree matrix is ​​obtained.

[0024] Furthermore, the matching matrix The way to get is:

[0025] ;

[0026] in, represents the Gaussian function; represents the calibration matrix The row vector and the center matrix The Euclidean distance of row vectors; Represents the standard deviation of the Gaussian function, characterizing the width of the Gaussian function. ; Representation calculation The maximum Euclidean distance value among the Euclidean distances between the cluster centers.

[0027] Furthermore, the calibration matrix is ​​obtained by: The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first matrix The columns are used as the calibration matrix.

[0028] Furthermore, the method for obtaining the peak wavelength vector includes:

[0029] Obtaining the monochromatic light wavelength corresponding to each light collection point in each light intensity sequence in the light intensity matrix to obtain a wavelength vector;

[0030] Based on the wavelength value in the wavelength vector and the light intensity value in the light intensity sequence, the peak wavelength of each light intensity sequence is calculated;

[0031] A peak wavelength vector of the light intensity matrix is ​​obtained based on the peak wavelengths of a plurality of light intensity sequences.

[0032] Furthermore, the peak wavelength The calculation formula is:

[0033] ;

[0034] in, Indicates the wavelength vector with The corresponding monochromatic light wavelength value; Indicated in the light intensity sequence Greater than or equal to the set light intensity threshold The light intensity value, , Indicated in the light intensity sequence Greater than or equal to the set light intensity threshold The number of light intensity values, ,express The number of sampling points, is a positive integer; Indicates setting the light intensity threshold; represents the light intensity sequence, , the length is .

[0035] Furthermore, the current light intensity sequence is preprocessed to obtain a current matching degree vector, including:

[0036] Subtracting the corresponding element in the second vector from each element in the current light intensity sequence to obtain a third vector; and performing matrix multiplication processing on the third vector and the reduced-dimensional matrix to obtain a reduced-dimensional light intensity matrix;

[0037] The Euclidean distance between each row vector in the reduced-dimensional light intensity matrix and each row vector in the central matrix is ​​calculated, and after activation by a Gaussian function, a matching degree vector is obtained.

[0038] Furthermore, the dimension reduction matrix is ​​obtained by: The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first The columns serve as the dimension reduction matrix.

[0039] Furthermore, the neural network model includes an input layer, a hidden layer and an output layer;

[0040] The operational relationship of the neural network model is:

[0041] ;

[0042] in, Representing a weight coefficient matrix between the input layer and the hidden layer; Representing a weight coefficient matrix between the hidden layer and the output layer; and Represent the bias vectors of the hidden layer and the output layer respectively; Represents the input value; Indicates the output value.

[0043] In general, the present invention provides a method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network. The technical solution conceived by the present invention can achieve the following beneficial effects compared with the prior art:

[0044] The present invention combines principal component analysis with a neural network, uses principal component analysis to obtain a matching degree matrix, and inputs the matching degree matrix and the obtained error vector into a neural network model for training, so that the function relationship fitted by taking the standard distance vector as the fitting target and the peak wavelength vector as the independent variable is more accurate, thereby greatly improving the calibration accuracy of the dispersive confocal measurement device. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 It is a schematic diagram of a method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network provided by the present invention;

[0047] Figure 2 It is a data logic schematic diagram of a dispersion confocal measurement device calibration method based on principal component analysis and neural network provided by the present invention;

[0048] Figure 3 It is a schematic diagram of a calibration device of a dispersion confocal measurement device calibration method based on principal component analysis and neural network provided by the present invention;

[0049] Figure 4 It is a schematic diagram of a filtering method for a dispersion confocal measurement device calibration method based on principal component analysis and neural network provided by the present invention;

[0050] 1-dual-frequency laser interferometer; 2-corner cone prism; 3-plane mirror; 4-dispersion confocal measurement device; 5-high-precision motion platform; 6-optical platform. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings and embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0052] It should be noted that, in the description of the embodiments of the present invention, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a method, step or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such method, step or device. In the absence of further restrictions, the elements defined by the sentence "include a ..." do not exclude the existence of other identical elements in the method, step or device including the elements.

[0053] In order to solve the problem of inaccurate calibration of a dispersive confocal measurement device, the present invention provides a method for calibrating a dispersive confocal measurement device based on principal component analysis and a neural network, such as Figure 1 and 2 As shown, the method includes:

[0054] S101: using a dual-frequency laser interferometer to measure multiple standard positions of a plane reflector to obtain a standard distance vector; obtaining spectral response signals corresponding to the multiple standard positions through a dispersive confocal measurement device to obtain a light intensity matrix.

[0055] Measuring the plane mirror using dual-frequency laser interferometer standard positions, including As an embodiment, the measurement of the standard distance vector can utilize a calibration device, such as Figure 3 As shown, the calibration device includes a dual-frequency laser interferometer 1, a corner cube prism 2, a plane mirror 3 and a dispersion confocal measuring device 4 to be calibrated, which are arranged in sequence along the optical path; the corner cube prism 2 and the plane mirror 3 are arranged on a high-precision motion platform 5, and the dispersion confocal measuring device 4 to be calibrated and the high-precision motion platform 5 are arranged on an optical platform 6.

[0056] The spectral response signals corresponding to a plurality of standard positions are acquired through a dispersive confocal measurement device to obtain a light intensity matrix.

[0057] Specifically, the light intensity matrix is ​​generally obtained by filtering and normalizing the light intensity sequence of each spectral response signal in sequence.

[0058] It should be noted that multiple standard positions correspond to multiple spectral response signals; one spectral response signal corresponds to one light intensity sequence, and multiple spectral response signals correspond to multiple light intensity sequences. ,in, express sampling points.

[0059] As an example, Figure 4 As shown, the filtering process includes:

[0060] Slide and take values ​​from the light intensity sequence, each time taking light intensity values. That is, for each For values, the range of each value is: ;in, Indicates the sliding value position, ; Is a positive integer.

[0061] Will Arrange the light intensity values ​​in ascending order, discarding the smallest values ​​and the maximum value, get light intensity values; and calculate The average value of the light intensity .

[0062] Slide value, when are equal to When , multiple average values ​​are obtained , and then get the light intensity sequence after filtering .

[0063] For example: When When , the value range is: , get the average value ;when When , the value range is: , get the average value ;when When , the value range is: , get the average value , and so on, when When , the value range is: , get the average value .

[0064] As an embodiment, the normalization process includes:

[0065] Get the filtered light intensity sequence The maximum value in and minimum value ;

[0066] right Each light intensity value in is normalized in turn, and the corresponding normalized light intensity value is obtained. ; Then we get the normalized light intensity sequence .

[0067] As an example, the normalized light intensity value The calculation method is:

[0068] , .

[0069] It should be noted that a standard position obtains a spectral response signal; a spectral response signal is filtered and normalized to obtain a light intensity sequence , the size is .

[0070] A standard position obtains a light intensity sequence; The light intensity sequence of the spectral response signal corresponding to each standard position is filtered and normalized to obtain A light intensity matrix , the size is , corresponding to the standard distance vector elements, .

[0071] S102: Performing principal component analysis on the light intensity matrix to obtain a matching degree matrix.

[0072] As an embodiment, performing principal component analysis on the light intensity matrix to obtain a matching degree matrix includes: first performing principal component analysis on the light intensity matrix to obtain a calibration matrix and the dimension reduction matrix ; Then, the matching matrix is ​​obtained based on the calibration matrix .

[0073] More specifically, a calibration matrix is ​​obtained based on the first vector and the first matrix; a dimension reduction matrix is ​​obtained based on the first vector and the second matrix; a center matrix is ​​obtained based on the calibration matrix; and a matching degree matrix is ​​obtained based on the calibration matrix and the center matrix.

[0074] Furthermore, the method for obtaining the matching degree matrix includes:

[0075] S201: Perform principal component analysis on the light intensity matrix to obtain a first matrix score, a second matrix coeff, a first vector eigen, and a second vector avr.

[0076] It should be noted that the first matrix is ​​the light intensity matrix The projection of the light intensity sequence on each principal component vector; the size is The second matrix is ​​to convert the light intensity matrix The first vector is the matrix obtained by arranging the eigenvectors according to their importance. The second vector is the vector obtained by averaging the columns of the light intensity matrix, and its length is .

[0077] S202: Obtain a calibration matrix based on the first vector and the first matrix.

[0078] As an embodiment, the calibration matrix is ​​obtained by: The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first matrix The columns are used as calibration matrix.

[0079] S203: Perform K-means clustering calculation on the calibration matrix to obtain a center matrix.

[0080] It should be noted that the size of the center matrix is , Represents the number of target center vectors of the K-means clustering algorithm, which is also the calibration matrix Cluster center.

[0081] S204: Calculate the Euclidean distance between each row vector in the calibration matrix and each row vector in the center matrix, and obtain a matching degree matrix after activation by a Gaussian function.

[0082] The size of the matching matrix is , No. Row, No. The elements of the columns are the first The row vector and the center matrix The matching degree of the row vector.

[0083] As an example, the matching matrix The way to get is:

[0084] ;

[0085] in, represents the Gaussian function; represents the calibration matrix The row vector and the center matrix The Euclidean distance of row vectors; Represents the standard deviation of the Gaussian function, characterizing the width of the Gaussian function. ; Representation calculation The maximum Euclidean distance value among the Euclidean distances between the cluster centers.

[0086] S103: Calculate the peak wavelength of the light intensity matrix to obtain a peak wavelength vector.

[0087] In other words, the light intensity matrix is ​​calculated Each light intensity sequence in The peak wavelength in , multiple light intensity sequences correspond to multiple peak wavelengths, and then the peak wavelength vector is obtained.

[0088] The peak wavelength is calculated based on the wavelength vector and the light intensity sequence The light intensity sequence calculated from the light intensity values ​​in The wavelength value corresponding to the peak light intensity; the wavelength vector is the light intensity sequence The monochromatic light wavelength value corresponding to each light intensity value.

[0089] As an embodiment, obtaining the peak wavelength vector includes:

[0090] S301: Obtain the monochromatic light wavelength value corresponding to each lighting point in each light intensity sequence in the light intensity matrix to obtain a wavelength vector.

[0091] The length of the wavelength vector is , one wavelength value corresponds to one light intensity value; ; represents the wavelength vector, Represents the wavelength values ​​in the wavelength vector.

[0092] S302: Calculate the peak wavelength of each light intensity sequence based on the wavelength value in the wavelength vector and the light intensity value in the light intensity sequence;

[0093] Light intensity sequence , the length is , Represents the light intensity value in the light intensity sequence. The wavelength vector is the wavelength value corresponding to these light intensity values.

[0094] As an embodiment, the peak wavelength is calculated using an error-corrected centroid method.

[0095] Furthermore, the light intensity threshold is set to , for a length of The intensity sequence , take out the light intensity sequence Greater than and equal to The light intensity value is From the wavelength vector Take out The light intensity value corresponds to wavelength values; based on The light intensity value and The peak wavelength is calculated from the wavelength values.

[0096] Specifically, based on The light intensity value and The peak wavelength is calculated from the wavelength values The calculation formula is:

[0097] ;

[0098] in, Represents the wavelength vector with The corresponding monochromatic light wavelength value; Indicates that in the light intensity sequence Greater than or equal to the set light intensity threshold The light intensity value, , Indicates that in the light intensity sequence Greater than or equal to the set light intensity threshold The number of light intensity values, ,express The number of sampling points, is a positive integer; Indicates setting the light intensity threshold; represents the light intensity sequence, , the length is .

[0099] S303: Obtaining a peak wavelength vector of a light intensity matrix based on peak wavelengths of a plurality of light intensity sequences.

[0100] That is to say, a light intensity sequence corresponds to a peak wavelength; then Positions obtained A light intensity sequence, The light intensity sequence corresponds to The peak wavelength, and thus the length is The peak wavelength vector of .

[0101] S104: Taking the standard distance vector as the fitting target and the peak wavelength vector as the independent variable, fitting the functional relationship between the two to obtain a displacement vector; and calculating the error value between each element in the displacement vector and each element in the corresponding standard distance vector to obtain an error vector.

[0102] As an embodiment, the method of obtaining the displacement vector includes: using a multi-order polynomial to fit the functional relationship between the standard distance vector and the peak wavelength vector, and obtaining multiple fitting results through the functional relationship and the peak wavelength vector calculation, and using the fitting results as the displacement vector.

[0103] The multi-order polynomial is preferably a tenth-order polynomial.

[0104] As an embodiment, the error vector is calculated as follows:

[0105] ;

[0106] in, represents the error vector; represents the displacement vector; represents the standard distance vector; , Represents the number of elements in the displacement vector and the standard distance vector.

[0107] S105: Input the matching degree matrix and the error vector into the neural network model for training to obtain a back propagation neural network model.

[0108] As an embodiment, the neural network model includes an input layer, a hidden layer and an output layer. The matching degree matrix and the error vector are input into the neural network model for training to obtain a back propagation neural network model.

[0109] Among them, the operation relationship of the neural network model is:

[0110] ;

[0111] in, Represents the weight coefficient matrix between the input layer and the hidden layer; Represents the weight coefficient matrix between the hidden layer and the output layer; and Represent the bias vectors of the hidden layer and the output layer respectively; Represents the input value; Indicates the output value.

[0112] More specifically, when training the neural network model, the input layer is the first row vector, the output layer is The prediction error value of the polynomial fitting function corresponding to the row vector The target value of the output layer is the target error vector elements, namely , The peak wavelength is calculated based on the light intensity sequence and wavelength vector, and then the target fitting displacement is obtained based on the peak wavelength and polynomial fitting function, and then the target error value is obtained. .

[0113] In order to make the prediction error value Infinitely close to the target error value That is, the output error is infinitely close to 0 to continuously train the neural network model, so that the displacement value of the output of the dispersive confocal measurement device to be calibrated is more accurate.

[0114] Furthermore, the output error ;

[0115] According to the output error Adjust the weight coefficient matrix , and the bias vector , The method of adjusting each coefficient is as follows: Expressed as a function of the coefficients in the weight coefficient matrix and the bias vector, and then solving the error The partial derivatives of each coefficient in the weight coefficient matrix and the bias vector are obtained, and the changes of each coefficient in the weight coefficient matrix and the bias vector are obtained based on the partial derivatives.

[0116] S106: Obtain the current light intensity sequence of the dispersive confocal measurement device to be calibrated; preprocess the current light intensity sequence to obtain a current matching degree vector; and input the current matching degree vector into the back propagation neural network model to obtain a prediction error value; obtain the current peak wavelength based on the current light intensity sequence, and substitute the current peak wavelength as an independent variable into the above functional relationship to obtain a fitting displacement; calibrate the dispersive confocal measurement device to be calibrated based on the fitting displacement and the prediction error value.

[0117] It should be noted that the current method for acquiring the light intensity sequence may be: firstly acquiring a spectral response signal through the dispersion confocal measurement device to be calibrated, and then filtering and normalizing the light intensity sequence of the spectral response signal to obtain the result.

[0118] As an embodiment, the current light intensity sequence is preprocessed to obtain a current matching degree vector, including:

[0119] Subtract the corresponding element in the second vector from each element in the current light intensity sequence to obtain a third vector; and perform matrix multiplication on the third vector and the reduced-dimensional matrix to obtain a reduced-dimensional light intensity matrix. It should be noted that the reduced-dimensional matrix is ​​obtained by: The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first The columns are used as the dimension reduction matrix.

[0120] The Euclidean distance between each row vector in the reduced-dimensional light intensity matrix and each row vector in the center matrix is ​​calculated, and after activation by the Gaussian function, a matching degree vector is obtained.

[0121] The current matching degree vector is input into the back propagation neural network model to obtain the predicted error value of the displacement error of the polynomial fitting. In other words, the input is the current matching degree vector composed of the matching degree between the reduced-dimensional light intensity sequence and each center vector in the center matrix; the output is the predicted error value of the displacement error of the polynomial fitting.

[0122] Finally, the current peak wavelength is obtained based on the current light intensity sequence, and the current peak wavelength is substituted into the above functional relationship as an independent variable to obtain the fitting displacement; the fitting displacement value with error compensation is obtained based on the fitting displacement and the predicted error value, thereby realizing the calibration of the dispersive confocal measurement device to be calibrated.

[0123] In summary, the present invention provides a dispersion confocal measurement device calibration method based on principal component analysis and neural network, which greatly improves the calibration accuracy of the dispersion confocal measurement device.

[0124] It should be noted that, for the above-mentioned various embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application.

[0125] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0126] In the several embodiments provided in this application, it should be understood that the disclosed method or system can be implemented in other ways. For example, the above-described embodiments are only schematic, and the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed.

[0127] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0128] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0129] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable memory. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory, including a number of instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.

[0130] Those skilled in the art will appreciate that all or part of the various circuits in the above embodiments may be completed by entering a program to instruct related hardware, and the program may be stored in a computer-readable memory, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0131] The above is only an exemplary embodiment of the present disclosure, and the scope of the present disclosure cannot be limited thereto. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the specification and practicing the disclosure here, those skilled in the art will easily think of the implementation scheme of the present disclosure. This application is intended to cover any modification, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the technical field not recorded in the present disclosure. The description and examples are regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

[0132] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. As long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] It will be easily understood by those skilled in the art that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network, characterized in that: The method comprises: A dual-frequency laser interferometer is used to measure multiple standard positions of the plane reflector to obtain a standard distance vector; a dispersive confocal measurement device is used to obtain spectral response signals corresponding to the multiple standard positions to obtain a light intensity matrix; Performing principal component analysis on the light intensity matrix to obtain a matching degree matrix; Calculating the peak wavelength of the light intensity matrix to obtain a peak wavelength vector; Taking the standard distance vector as the fitting target and the peak wavelength vector as the independent variable, fitting the functional relationship between the two to obtain a displacement vector; and calculating the error value between each element in the displacement vector and each element in the corresponding standard distance vector to obtain an error vector; Inputting the matching degree matrix and the error vector into a neural network model for training to obtain a back propagation neural network model; Acquire a current light intensity sequence of the dispersive confocal measurement device to be calibrated; preprocess the current light intensity sequence to obtain a current matching degree vector; and input the current matching degree vector into a back propagation neural network model to obtain a prediction error value; Obtaining a current peak wavelength based on the current light intensity sequence, and substituting the current peak wavelength as an independent variable into the functional relationship to obtain a fitting displacement; The dispersive confocal measurement device to be calibrated is calibrated based on the fitting displacement and the prediction error value.

2. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 1, characterized in that: The light intensity matrix is ​​subjected to principal component analysis to obtain a matching degree matrix including: Performing principal component analysis on the light intensity matrix to obtain a calibration matrix and a dimension reduction matrix; A matching degree matrix is ​​obtained based on the calibration matrix.

3. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 2, characterized in that: The method for obtaining the matching degree matrix includes: Performing principal component analysis on the light intensity matrix to obtain a first matrix, a second matrix, a first vector, and a second vector; Obtaining the calibration matrix based on the first vector and the first matrix; Performing K-means clustering calculation on the calibration matrix to obtain a center matrix; The Euclidean distance between each row vector in the calibration matrix and each row vector in the center matrix is ​​calculated, and after activation by a Gaussian function, the matching degree matrix is ​​obtained.

4. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 3, characterized in that: The matching matrix The way to get is: ; in, represents the Gaussian function; represents the calibration matrix The row vector and the center matrix The Euclidean distance of row vectors; Represents the standard deviation of the Gaussian function, characterizing the width of the Gaussian function. ; Representation calculation The maximum Euclidean distance value among the Euclidean distances between the cluster centers.

5. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 2, characterized in that: The calibration matrix is ​​obtained by: setting the first vector The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first matrix The columns are used as the calibration matrix.

6. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 1, characterized in that: The method for obtaining the peak wavelength vector includes: Obtaining the monochromatic light wavelength corresponding to each light collection point in each light intensity sequence in the light intensity matrix to obtain a wavelength vector; Based on the wavelength value in the wavelength vector and the light intensity value in the light intensity sequence, the peak wavelength of each light intensity sequence is calculated; A peak wavelength vector of the light intensity matrix is ​​obtained based on the peak wavelengths of a plurality of light intensity sequences.

7. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 6, characterized in that: The peak wavelength The calculation formula is: ; in, Indicates the wavelength vector with The corresponding monochromatic light wavelength value; Indicated in the light intensity sequence Greater than or equal to the set light intensity threshold The light intensity value, , Indicated in the light intensity sequence Greater than or equal to the set light intensity threshold The number of light intensity values, ,express The number of sampling points, is a positive integer; Indicates setting the light intensity threshold; represents the light intensity sequence, , the length is .

8. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 3, characterized in that: Preprocessing the current light intensity sequence to obtain a current matching degree vector includes: Subtracting the corresponding element in the second vector from each element in the current light intensity sequence to obtain a third vector; and performing matrix multiplication processing on the third vector and the reduced-dimensional matrix to obtain a reduced-dimensional light intensity matrix; The Euclidean distance between each row vector in the reduced-dimensional light intensity matrix and each row vector in the central matrix is ​​calculated, and after activation by a Gaussian function, a matching degree vector is obtained.

9. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 8, characterized in that: The method of obtaining the dimension reduction matrix includes: setting the first vector The proportion of the sum of eigenvalues ​​in the total sum of all eigenvalues ​​is , get Minimum value greater than 0.99 value ; Take the first The columns serve as the dimension reduction matrix.

10. The method for calibrating a dispersion confocal measurement device based on principal component analysis and neural network according to claim 1, characterized in that: The neural network model includes an input layer, a hidden layer and an output layer; The operational relationship of the neural network model is: ； in, Representing a weight coefficient matrix between the input layer and the hidden layer; Representing a weight coefficient matrix between the hidden layer and the output layer; and Represent the bias vectors of the hidden layer and the output layer respectively; Represents the input value; Indicates the output value.

Images