A method for automatically and accurately eliminating noise in white light interferometer detection data

Abstract

The present invention discloses a method for automatically and accurately eliminating noise in white light interferometer detection data, which relates to the field of ultra-precision machining detection technology. The method comprises: collecting 2D cross-sectional data of the surface of a microstructure sample, obtaining data within a single cycle range in the 2D cross-sectional data; detecting the data within the single cycle range, selecting a valid interval, and aligning the center of the detection data with the origin; determining the surface of the sample, and selecting a curve analysis method according to the surface of the sample; calculating the detection data according to the curve analysis method, finding and deleting noise; and outputting noise-free data. The present invention automatically and accurately finds and deletes noise points, avoiding the errors and costs caused by manual operation, and greatly improving the work efficiency of test analysis; the present invention is based on actual application scenarios, does not use complex and time-consuming means such as high-pass filtering, avoids the tedious calculation process of signal conversion, and accurately removes noise points while retaining the original real data to the maximum extent.

Description

Technical Field

[0001] The present invention belongs to the technical field of ultra-precision machining detection, and in particular relates to a method for automatically and accurately eliminating noise in detection data of a white light interferometer. Background Art

[0002] White-light interferometers are the most commonly used and highly accurate testing equipment in ultra-precision inspections, making them irreplaceable in optical component quality testing. However, during the inspection process, unrealistic noise may occasionally be generated due to light source variations, random and subtle environmental factors, as well as sample surfaces that are too narrow, sample heights that are too high, and recesses that are too narrow. If this noise is not removed and directly included in the calculation, it will significantly affect the calculated surface roughness and PV values.

[0003] These noise points do not actually exist on the sample surface and can be clearly seen on the 2D curve of the surface. Manually deleting noise points is the most direct way, but manual deletion requires human intervention, which reduces detection efficiency and increases detection costs.

[0004] This noise can be classified as salt and pepper noise. Median filtering, a common filtering method, can reduce the impact of this type of noise. However, the filter kernel size of the median filter varies for different samples and is difficult to determine. In addition, the median filter will filter out data in normal areas, which will reduce the calculated value of the sample surface roughness and undermine the accuracy of the white light interferometer detection data. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for automatically and accurately eliminating noise in white light interferometer detection data. By writing a software program, after the user inputs the detection data and surface shape, the program automatically eliminates the noise in the detection data while retaining most of the real detection data, thereby improving the reliability of the detection data and reducing the cost of manual measurement and manual calibration.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] The present application provides a method for automatically and accurately eliminating noise in white light interferometer detection data, comprising the steps of:

[0008] Collecting 2D cross-sectional data of the surface of the microstructure sample, and obtaining data within a single period in the 2D cross-sectional data;

[0009] Detecting data within the single cycle range, selecting a valid interval, and aligning the center of the detected data within the valid interval with the origin;

[0010] Determine the sample surface shape, and select a curve analysis method according to the sample surface shape;

[0011] Calculating the detection data according to the curve analysis method, finding and deleting noise points;

[0012] Output noise-free data.

[0013] As a preferred technical solution of the present invention, the sample surface type includes a plane and a curved surface; the curve analysis method includes a slope curve analysis method and an error curve analysis method.

[0014] As a preferred technical solution of the present invention, the planar surface type includes a plane and an inclined surface, and the curved surface type includes a spherical surface and an aspherical surface.

[0015] As a preferred technical solution of the present invention, the slope curve analysis method is applied to the plane, specifically comprising the following steps:

[0016] S1, calculating the slope curve of the detection data, and obtaining the average value and standard deviation of the slope curve;

[0017] S2, searching and deleting the leading noise point in the slope curve, and recalculating the slope curve;

[0018] S3, determining whether the average value of the slope curve converges to a fixed value, if so, executing step S4; otherwise, executing step S2;

[0019] S4, end the loop.

[0020] As a preferred technical solution of the present invention, the error curve analysis method is applied to the curved surface, specifically comprising the following steps:

[0021] S1, fitting and calculating the polynomial closest to the original curve to obtain the fitting curve;

[0022] S2, obtaining an error curve between the original curve and the fitting curve, and analyzing error curve data;

[0023] S3, calculating an error curve based on the error curve data, finding and deleting the leading noise point, and recalculating the error curve;

[0024] S4, determining the integral area of the error curve in the interval [0,1] after linear normalization; if the integral area is greater than 0.1, executing step S5; otherwise, executing step S3;

[0025] S5, ends the loop.

[0026] As a preferred technical solution of the present invention, the 2D cross-sectional data includes sample area data and partial area data outside the sample boundary.

[0027] As a preferred technical solution of the present invention, the method for obtaining data within the single cycle range includes the following steps:

[0028] Determine the opening direction of the curve;

[0029] Find the maximum value coordinate or the minimum value coordinate, and translate the maximum value coordinate or the minimum value coordinate to the origin;

[0030] The coordinates of the extreme points on the left and right sides of the origin are obtained and determined as the starting point and the ending point of the data within the single cycle range.

[0031] As a preferred technical solution of the present invention, aligning the data within the single cycle specifically includes the following steps:

[0032] Analyze the 2D cross-section data, select the data length as N, and find the maximum Y value index MaxY and the minimum Y value index MinY;

[0033] Calculate the smaller value of MaxY-0 and N-MaxY and set it as MaxY; calculate the smaller value of MinY-0 and N-MinY and set it as MinY;

[0034] Compare the sizes of MaxY and MinY. If MaxY>MinY, the maximum value is in the middle and opens downward; if MinY>MaxY, the minimum value is in the middle and opens upward.

[0035] Searching the middle extreme point of the 2D cross-sectional data to the left, selecting the extreme point opposite to the middle extreme point as the left boundary, and deleting the area to the left of the left boundary;

[0036] Searching rightward for the middle extreme point of the 2D cross-section data, selecting the extreme point opposite to the middle extreme point as the right boundary, and deleting the area to the right of the right boundary;

[0037] The middle extreme point is translated to the origin to obtain the analysis data curve V; the side where VX<0 is set as the left curve, recorded as LV; the side where VX>0 is set as the right curve, recorded as RV; wherein the analysis data curve V is left-right symmetrical and the data within a single cycle range is aligned left-right.

[0038] As a preferred technical solution of the present invention, calculating the slope curve and removing noise specifically includes the following steps:

[0039] The slopes of the analysis data curve V are set to 0;

[0040] By formula Get the coordinate k of each data point in the slope curve i ;

[0041] Calculate the mean value m and the standard deviation std of the slope curve Ky, and save the mean value m to the mean value list mlist;

[0042] Traverse the slope list and loop to delete noise points; among them, if the current slope value Kiy of the data point is greater than std + m or Kiy is less than m - std, the noise point corresponds to the data point Kix;

[0043] Judge the mean value list mlist to determine that the values in it start from the i-th and converge stably to a fixed value.

[0044] As a preferred technical solution of the present invention, use a fifth-order polynomial to fit the original curve; calculate the error curve according to the original curve and the polynomial coefficients, and perform linear normalization processing in the x and y directions.

[0045] The beneficial effects of the present invention are:

[0046] (1) Automatically and accurately find and delete noise points, avoiding errors and costs caused by manual operations, greatly improving the efficiency of test analysis work, and can be integrated into the software system of the white light interferometer device, or can also be provided to users as a secondary development program of the white light interferometer alone.

[0047] (2) Starting from the actual application scenario, relatively complex and time-consuming technical means such as high-pass filtering are not used, avoiding the cumbersome calculation process of signal conversion, accurately removing noise points while maximizing the retention of original real data. Brief Description of the Drawings

[0048] For better understanding and implementation, the technical solutions of this application will be described in detail below with reference to the accompanying drawings.

[0049] Figure 1 It is a step flow chart of a method for automatically and accurately eliminating noise points in the detection data of a white light interferometer provided by an embodiment of this application;

[0050] Figure 2 It is a step flow chart of obtaining data within a single cycle provided by an embodiment of this application;

[0051] Figure 3 It is an algorithm flow chart of selecting detection data within a single cycle and aligning with the origin provided by an embodiment of this application;

[0052] Figure 4 It is a step flow chart of the slope curve analysis method provided by an embodiment of this application;

[0053] Figure 5 It is an algorithm flow chart of the slope curve analysis method provided by an embodiment of this application;

[0054] Figure 6 A flowchart of the steps of the error curve analysis method provided in an embodiment of the present application;

[0055] Figure 7 This is a flowchart of the error curve analysis method provided in the embodiment of the present application;

[0056] Figure 8 A flowchart of the steps for removing noise from a triangular grating surface using a slope curve analysis method provided in an embodiment of the present application;

[0057] Figure 9 A comparison diagram of the triangular grating surface curve provided in an embodiment of the present application including noise points and the surface shape after the noise points are removed;

[0058] Figure 10 A flowchart of the steps for removing noise using the error curve analysis method for an aspheric surface provided in an embodiment of the present application;

[0059] Figure 11 The aspheric curve provided in the embodiment of the present application includes a comparison diagram of the surface shape after noise points are removed and the noise points are removed. DETAILED DESCRIPTION

[0060] To further illustrate the technical means and effectiveness of the present invention in achieving its intended purpose, exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all possible embodiments consistent with the present application. Rather, they are merely examples of methods and systems consistent with certain aspects of the present application, as detailed in the appended claims.

[0061] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. As used in this application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items.

[0062] The following describes in detail the specific implementation methods, features and effects of the present invention in conjunction with the accompanying drawings and preferred embodiments.

[0063] Example 1

[0064] like Figure 1 As shown, the embodiment of the present application provides a method for automatically and accurately eliminating noise in white light interferometer detection data, including the following steps:

[0065] S101, collecting 2D cross-sectional data of the surface of the microstructure sample, and obtaining data within a single period in the 2D cross-sectional data;

[0066] S102, detecting data within a single cycle, selecting a valid interval, and aligning the center of the detected data within the valid interval with the origin;

[0067] S103, determining the sample surface shape, and selecting a curve analysis method according to the sample surface shape;

[0068] S104, calculating the detection data according to the curve analysis method, finding and deleting noise points;

[0069] S105: Output noise-free data.

[0070] The above steps S101 to S105 will be described in detail below.

[0071] Regarding steps S101 to 102, the hardware equipment of the white light interferometer is used to collect 2D cross-sectional data of the surface of the microstructure sample, and ensure that the collected data is within the effective range of a single cycle. The data within the effective range of a single cycle is detected, and the effective interval is selected, and the center of the detection data within the effective interval is aligned with the origin.

[0072] In one embodiment provided herein, the 2D cross-sectional data includes the complete sample area and a portion of the area outside the sample boundary. When selecting data, only data within a single cycle range should be retained, and data outside the boundary and data outside the valid area should be removed.

[0073] like Figure 2 As shown, in one embodiment provided in the present application, a method for obtaining data within a single cycle range includes the following steps:

[0074] S201, determining the opening direction of the curve. The curve with the maximum value in the middle is the curve opening downward; the curve with the minimum value in the middle is the curve opening upward.

[0075] S202, finding the maximum value coordinate or the minimum value coordinate, and translating the maximum value coordinate or the minimum value coordinate to the origin.

[0076] Find the maximum value coordinate, translate it to the origin, and then search for the minimum and maximum value coordinates to the left and right respectively. Similarly, find the minimum value coordinate, translate it to the origin, and then search for the minimum and maximum value coordinates to the left and right respectively.

[0077] S203, obtaining the coordinates of the extreme points on the left and right sides of the origin, and determining them as the starting point and the ending point of the data within a single cycle.

[0078] like Figure 3As shown, specifically, the algorithm process of selecting data within a single cycle range from the original detection data and aligning the origin mainly includes the following:

[0079] Minimum value analysis: In the 2D surface data being tested, when the curve opens upward, it indicates the minimum is in the center. This minimum is then translated to align with the origin. Starting from this minimum position, the search is performed left and right along the x-axis to find the maximum value. After determining the minimum and maximum values, data points in the original curve with x values less than the left and greater than the right are deleted, thereby removing the redundant left and right portions. Finally, valid analysis data aligned within a single cycle is obtained.

[0080] Analyze from the maximum value: In the 2D surface data, when the curve opens downward, it indicates that the maximum value is in the center. This maximum value is then translated to align with the origin. Starting from the maximum position, the search is performed left and right along the x-axis to find the minimum value. After determining the minimum and maximum values, the data points in the original curve with x values less than the left and greater than the right are deleted, thereby eliminating the redundant parts on the left and right. Finally, valid analysis data aligned within a single cycle is obtained.

[0081] Regarding step S103, the user determines the detection sample surface type of the white light interferometer. In the embodiment of the present application, the sample surface type includes a plane and a curved surface, wherein the plane sample surface type includes a plane and an inclined surface, and the curved surface sample surface type includes a spherical surface and an aspherical surface.

[0082] Select different curve analysis methods according to different sample surface shapes.

[0083] In the embodiment of the present application, the curve analysis method includes a slope curve analysis method and an error curve analysis method.

[0084] Because plane surface types such as planes and inclined planes are theoretical surface types with fixed slopes, the slope curve analysis method can be used. However, for curved surface types such as spheres and aspherical surfaces, the slope varies greatly depending on the convexity of the curve because their slopes are different at different positions. Therefore, the error curve analysis method is used.

[0085] Regarding step S104, the detection data is calculated according to different selected curve analysis methods, noise points are found, and the noise points are deleted.

[0086] like Figure 4 As shown, in one embodiment provided in the present application, the slope curve analysis method is applied to planes and curved surfaces, specifically comprising the following steps:

[0087] S401, calculating the slope curve of the detection data, and obtaining the average value and standard deviation of the slope curve;

[0088] S402, searching and deleting the first noise point in the slope curve, and recalculating the slope curve;

[0089] S403, determining whether the average value of the slope curve converges to a fixed value, if so, executing step S404; otherwise, repeating step S402;

[0090] S404, end the loop.

[0091] Specifically, the slope curve analysis method involves repeatedly calculating the slope of each point on the curve, calculating the mean and standard deviation of the slope curve, searching for noise points in the slope curve, and deleting the first unqualified noise point found. After several cycles of deleting noise points, the average value of the slope curve corresponding to the test data curve begins to change less dramatically until it stabilizes and converges to a fixed value.

[0092] like Figure 5 As shown, the embodiment of the present application provides the main algorithm flow of the slope curve analysis method, which specifically includes the following contents:

[0093] Calculate the slope curve corresponding to the test data curve, calculate the mean m and standard deviation std of the slope curve; record the current slope mean in the mean list mlist. Determine the number of elements in the mlist list and the convergence status. If the number of elements in the mlist list is greater than or equal to 3 and converges, the first noise point that meets the requirements is removed and the current algorithm process ends. If the number of elements in the mlist list is less than 3 and does not converge, traverse the slope curve, loop through the slope curve, and delete noise points several times until the mean value of the slope curve no longer changes dramatically and converges to a fixed value. Determine the current curve slope k. If k>m+std, delete the original test point corresponding to the current curve slope point in the test data; otherwise, traverse the slope curve.

[0094] like Figure 6 As shown, in one embodiment provided in the present application, the error curve analysis method is applied to spherical and aspherical surfaces, specifically including the following steps:

[0095] S601, fitting and calculating a polynomial closest to the original curve to obtain a fitting curve;

[0096] S602, obtaining an error curve between the original curve and the fitting curve, and analyzing the error curve data;

[0097] S603, calculating the error curve based on the error curve data, finding and deleting the leading noise point, and recalculating the error curve;

[0098] S604, determining the integral area of the error curve after linear normalization in the interval [0,1]; if the integral area is greater than 0.1, executing step S605; otherwise, repeating step S603;

[0099] S605, end the loop.

[0100] Specifically, the error curve analysis method first requires fitting and calculating the polynomial that best fits the original curve. The error curve between the original curve and the fitted curve is used as the analysis data. The error curve is then repeatedly calculated within the error curve data, and noise points are searched for on the error curve. The first unsatisfactory noise point is then deleted. After several iterations of removing noise points, the error curve is recalculated within the loop. After linearly normalizing the error curve in both the x and y directions, the integral area of the error curve within the [0, 1] interval is examined until the integral area exceeds 0.1, indicating that the noise points in the original curve have been completely removed.

[0101] like Figure 7 As shown, the embodiment of the present application provides the main algorithm flow of the error curve analysis method, which specifically includes the following contents:

[0102] Perform polynomial fitting on the detection data, obtain a detection data curve based on the polynomial parameters, and calculate the detection data curve;

[0103] Then calculate the error curve and calculate the integral area of the normalized error curve in the interval [0,1]. When the integral area is greater than or equal to 0.1, delete the noise points and end the algorithm process. Otherwise, the above steps of calculating the error curve will be looped, and the noise points will be deleted in a loop until the integral area is greater than or equal to 0.1.

[0104] Regarding step S105, after all the above steps, the data without noise interference is finally output, thereby completing the operation of automatically and accurately eliminating the noise generated by the white light interferometer when monitoring the data.

[0105] This application automatically and accurately finds and eliminates noise points, avoiding a series of problems caused by manual operation and greatly improving the efficiency of test and analysis work. It can be integrated into the white light interferometer equipment software system, or provided to users as a secondary development program for the white light interferometer.

[0106] At the same time, this application starts from the actual application scenario and does not use more complex and time-consuming technical means such as high-pass filtering, avoiding the tedious calculation process of signal conversion, while accurately removing noise and retaining the original real data to the maximum extent.

[0107] Example 2

[0108] Since this application is applicable to various face types, a relatively complex triangular face is taken as an example in this application to elaborate on the method of eliminating noise by using the slope curve analysis method.

[0109] The triangular face is formed by splicing two inclined planes with fixed slopes on the left and right, and the left and right faces need to be split into two curves for separate analysis. If it is a plane, the splitting step can be omitted.

[0110] As Figure 8 shown, the triangular face in this embodiment is a triangular grating, and the slope curve analysis method is used to remove noise. The specific steps are as follows:

[0111] S801, align the effective analysis data within a single period;

[0112] Specifically: Analyze the original 2D data P with a data length of N, find the index MaxY of the maximum value of the Y value and the index MinY of the minimum value, calculate the smaller value of the two values MaxY - 0 and N - MaxY as MaxY; calculate the smaller value of the two values MinY - 0 and N - MinY as MinY; if MaxY > MinY, it means the maximum value is in the middle and the opening is downward; if MinY > MaxY, it means the minimum value is in the middle and the opening is upward; the extreme point on the left of the middle extreme point of the original data P is taken as the left boundary with the opposite extreme point to the middle extreme point, and the area more to the left is deleted; the extreme point on the right of the middle extreme point of the original data P is taken as the right boundary with the opposite extreme point to the middle extreme point, and the area more to the right is deleted; the middle extreme point is translated to the origin to obtain the effective analysis data curve V that is left and right aligned within a single period and is basically symmetric left and right. The part where VX < 0 is the left line, denoted as LV; the part where VX > 0 is the right line, denoted as RV.

[0113] S802, traverse the left curve LV obtained in step S801 and calculate the slope curve K;

[0114] Specifically: Set the slope values at the beginning and end to 0. In the slope curve, each data point can be obtained from the definition of the slope

[0115]

[0116] And calculate the average value m of Ky and the standard deviation std of Ky. Save m to the average value list mlist.

[0117] S803, traverse the slope list and delete noise;

[0118] Specifically: When the current slope value Kiy > std + m or Kiy < m - std, it is considered that the original data point corresponding to this point Kix is noise, and it is deleted in the original data curve LV, and the traversal ends.

[0119] S804, repeat steps S802 to S803 several times, which can be set to 20 times (based on the fact that the number of sample noise points is generally no more than 20), calculate the slope curve of the original data curve V, calculate the average slope value and save it to the average value list mlist.

[0120] S805, judging the slope average value list mlist, and determining whether the value in the list, starting from the i-th one, tends to be stable and converges to a certain value;

[0121] Specifically, it can be determined by linearly normalizing the derivative list of the mlist list. Let it be kmlist. The values after the i-th value in kmlist are the minimum values that tend to 0. From the last value in kmlist, when the i-th value is greater than 0, its index is the starting point of convergence of the mlist slope average list, which means the number of times N the noise point is deleted in the loop.

[0122] S806, re-obtain the valid analysis data curve LV aligned left and right within the original single cycle, and cyclically delete the noise points according to the number of cyclic noise removal times N obtained in step S805 to obtain the curve GLV that removes all noise points and retains the original measurement data to the maximum extent.

[0123] S807 , repeating steps S802 to S806 , analyzing the right curve RV , and also obtaining a curve GRV that removes noise and retains the original measurement data.

[0124] like Figure 9 As shown, an embodiment of the present application provides a comparison diagram of the triangular grating surface (slant) containing noise points and the surface after automatic noise removal. In the comparison diagram, the upper left and lower left figures are original detection data diagrams containing noise points, and the upper right and lower right figures are data diagrams after noise removal. It can be clearly seen from the figure that in the upper left figure, between the data 0 and 10 of the horizontal axis, and between the data 60 and 70, the curve appears as a wavy curve, and these wavy curved data are noise points; and the upper right figure is a comparison diagram of the upper left figure. In the upper right figure, after the noise points are removed by the method provided by the present invention, it can be clearly seen that the noise points originally in the upper left figure have been eliminated, and a smooth curve is obtained. Similarly, the analysis of the lower left figure and the lower right figure is as described above.

[0125] It can be seen from this that the present invention can automatically and accurately find and delete noise points, and can precisely remove noise points while retaining the original real data to the greatest extent possible.

[0126] Example 3

[0127] Because the present application is applicable to various surface shapes, the present application takes a more complex aspheric surface as an example and explains in detail the method of eliminating noise using the error curve analysis method.

[0128] Aspheric surfaces are a more general case than other surface types. Spherical surfaces can be regarded as special aspheric surfaces with an aspheric coefficient of 0. The error curves of other common surface types can be calculated by fitting polynomials to the closest theoretical surface type.

[0129] like Figure 10 As shown, in this embodiment, the error curve analysis method is used to remove noise from the aspheric surface detection data. Specifically, the following steps are included:

[0130] S901, aligning valid analysis data within a single cycle;

[0131] Specifically: analyze the original 2D data P with a data length of N, find the maximum Y value index MaxY and the minimum Y value index MinY, calculate the smaller of the two values MaxY-0 and N-MaxY as MaxY; calculate the smaller of the two values MinY-0 and N-MinY as MinY; if MaxY>MinY, it means that the maximum value is in the middle and opens downward; if MinY>MaxY, it means that the minimum value is in the middle and opens upward; the middle extreme point of the original data P is taken to the left as the extreme point opposite to the middle extreme point as the left boundary, and the area on the left is deleted; the middle extreme point of the original data P is taken to the right as the extreme point opposite to the middle extreme point as the right boundary, and the area on the right is deleted; the middle extreme point is translated to the origin to obtain an effective analysis data curve V that is basically symmetrical on the left and aligned on the left within a single cycle.

[0132] S902 , performing high-order polynomial fitting on the data curve V in step S901 , typically using fifth-order fitting, and the curve corresponding to the polynomial expression is a smooth curve that is closest to the curve V.

[0133] S903, calculating an error curve based on the curve V and the polynomial coefficients, and performing linear normalization calculations in both the x and y directions;

[0134] Specifically: calculate the integral area of the E curve in the interval [0,1], recorded as S. When S<0.1, continue with the subsequent steps. When s>0.1, it means that there is no noise on the aspheric surface.

[0135] S904, traverse the curve V obtained in step S901 and calculate the slope curve K;

[0136] Specifically: the first and last slope values are set to 0, and each data point in the slope curve can be obtained by the definition of the slope

[0137] S905, traverse the slope list;

[0138] Specifically, when the current slope value Kiy > std + m or Kiy < m - std, it is considered that the original data point corresponding to this point Kix is a noise point, which is deleted from the original data curve V, and the traversal ends.

[0139] S906. Repeat steps S903 - S905 several times until the integral area S of the error curve E within the interval [0, 1] after a certain linear normalization is greater than 0.1, indicating that the noise points have been deleted, and the loop ends.

[0140] As Figure 11 shown, the embodiment of the present application provides a comparison diagram of an aspherical curve with noise points and the surface shape after automatically removing the noise points. In this comparison diagram, the upper left and lower left figures on the left side are the original detection data containing noise points, and the upper right and lower right figures on the right side are the data after removing the noise points. It can be clearly seen from the figure that in the upper left figure, between the data 200 - 400 on the horizontal axis, there is a vertical line on the curve with a slope close to 90 degrees; between the data 0 - 200 and the data 600 - 800, the slope of the data is close to 0 degrees, and these data are the noise points; while the upper right figure is a comparison diagram of the upper left figure. In the upper right figure, after removing the noise points by the method provided by the present invention, it can be clearly seen that the original noise points in the upper left figure have been removed, and a smooth curve is obtained. Similarly, the analysis of the lower left figure and the lower right figure is as described above.

[0141] It can be seen from this that the present invention can automatically and accurately find the noise points and delete them, and while accurately removing the noise points, it maximally retains the original true data.

[0142] The above is only a preferred embodiment of the present invention, and it does not impose any form of limitation on the present invention. Although the present invention has been disclosed above with a preferred embodiment, it is not intended to limit the present invention. Any person skilled in the art can make some changes or modifications to equivalent embodiments by using the disclosed technical content within the scope of the technical solution of the present invention. However, as long as it does not depart from the content of the technical solution of the present invention, any brief modification, equivalent change, and modification made to the above embodiments based on the technical essence of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for automatically and accurately eliminating noise from white light interferometer detection data, characterized by: The following steps are involved: Collecting 2D cross-sectional data of the surface of the microstructure sample, and obtaining data within a single period in the 2D cross-sectional data; Detecting data within the single cycle range, selecting a valid interval, and aligning the center of the detected data within the valid interval with the origin; Determine the sample surface shape, and select a curve analysis method according to the sample surface shape; Calculating the detection data according to the curve analysis method, finding and deleting noise points; Output noise-free data; The sample surface type includes a flat surface type and a curved surface type; the curve analysis method includes a slope curve analysis method and an error curve analysis method; The slope curve analysis method is applied to the plane surface, specifically comprising the following steps: S1, calculating the slope curve of the detection data, and obtaining the average value and standard deviation of the slope curve; S2, searching and deleting the leading noise point in the slope curve, and recalculating the slope curve; S3, determining whether the average value of the slope curve converges to a fixed value, if so, executing step S4; otherwise, repeating step S2; S4, end the loop; The error curve analysis method is applied to the curved surface profile, specifically comprising the following steps: S1, fitting and calculating the polynomial closest to the original curve to obtain the fitting curve; S2, obtaining an error curve between the original curve and the fitting curve, and analyzing error curve data; S3, calculating an error curve based on the error curve data, finding and deleting the leading noise point, and recalculating the error curve; S4, determining the integral area of the error curve in the interval [0,1] after linear normalization; if the integral area is greater than 0.1, executing step S5; otherwise, repeating step S3; S5, ends the loop.

2. The method for automatically and accurately eliminating noise in white light interferometer detection data according to claim 1, characterized in that: The plane surface type includes a plane and an inclined surface, and the curved surface type includes a spherical surface and an aspherical surface.

3. The method for automatically and accurately eliminating noise in white light interferometer detection data according to claim 1, characterized in that: The 2D cross-sectional data includes sample area data and partial area data outside the sample boundary.

4. The method for automatically and accurately eliminating noise in white light interferometer detection data according to claim 1, characterized in that: The method for obtaining data within the single cycle range comprises the following steps: Determine the opening direction of the curve; Find the maximum value coordinate or the minimum value coordinate, and translate the maximum value coordinate or the minimum value coordinate to the origin; The coordinates of the extreme points on the left and right sides of the origin are obtained and determined as the starting point and the ending point of the data within the single cycle range.

5. The method for automatically and accurately eliminating noise in white light interferometer detection data according to claim 1, characterized in that: The original curve is fitted using a fifth-order polynomial; the error curve is calculated based on the original curve and the polynomial coefficients, and linear normalization processing is performed in both the x and y directions.

Images