White light interference three-dimensional reconstruction method and system based on modal decomposition and improved phase shift

Abstract

The application discloses a white light interference three-dimensional reconstruction method and system based on modal decomposition and improved phase shift, and relates to the technical field of three-dimensional reconstruction. The method comprises the following steps: acquiring an original white light interference signal sequence of each pixel point, decomposing the original white light interference signal sequence into a plurality of eigenmode function components, performing wavelet transform filtering and signal reconstruction on each component, acquiring a denoised interference signal sequence, calculating a modulation degree sequence of the denoised interference signal and performing fitting, acquiring an interference envelope curve, extracting a coarse positioning position of a coherent peak, calculating an initial phase by using a seven-step phase shift algorithm in a preset sampling window near the coarse positioning position, performing wave number domain orthogonal projection analysis on the denoised interference signal, fusing the reliable phase and the coarse positioning position of the coherent peak, acquiring a fine positioning coherent peak position, and reconstructing a three-dimensional surface topography of a sample to be measured. The application can significantly suppress the influence of external interference such as environmental vibration, light source fluctuation and detector noise on the quality of the interference signal, and realizes sub-nanometer three-dimensional topography reconstruction precision.

Description

Technical Field

[0001] This invention relates to the field of three-dimensional reconstruction technology, and in particular to a method and system for three-dimensional reconstruction based on mode decomposition and improved phase shift using white light interferometry. Background Technology

[0002] White light interferometry is a non-contact, high-precision method for measuring three-dimensional topography. Its basic working principle is as follows: using broadband white light as the illumination source, the interference light intensity signal of each pixel on the measured surface is recorded frame by frame (i.e., interferogram stacking) through vertical scanning along the optical axis. The white light interference signal reaches its maximum envelope amplitude at the zero optical path difference position. By locating the coherence peak positions corresponding to each pixel, the relative height information of the surface can be obtained, thereby reconstructing the three-dimensional topography of the sample. Compared to monochromatic interferometry, white light interferometry can effectively avoid phase ambiguity problems, and its vertical resolution can reach the sub-nanometer level. Therefore, it has significant application value in fields such as semiconductor wafer inspection, microelectromechanical system (MEMS) device characterization, and precision optical component evaluation.

[0003] A typical white-light interferometry 3D reconstruction process includes key steps such as interferometric signal preprocessing, coherence peak localization, and height calculation. For signal preprocessing, methods such as smoothing filtering and bandpass filtering are commonly used to suppress noise interference. For coherence peak localization, common algorithms include the extremum method, centroid method, envelope fitting method, and phase shift method. For height calculation, precise localization can be achieved using only envelope information or in combination with phase information. These methods can obtain good reconstruction results under ideal measurement conditions and with uniform reflection characteristics of the measured surface.

[0004] However, in practical measurement scenarios, existing technical solutions generally face the following four problems: First, it is difficult to effectively separate signal noise from effective information. Interference factors such as environmental vibration, non-uniform fluctuations in light source intensity, and CCD electronic noise can cause waveform distortion in the acquired interference signal. Especially when facing the measured surface with significant differences in reflectivity (such as a hydrophobic metal surface) or in working conditions with significant environmental vibration, the signal-to-noise ratio of the interference signal is greatly reduced. Traditional filtering methods struggle to balance the degree of noise reduction with signal fidelity. Insufficient filtering leaves residual noise that affects subsequent positioning accuracy, while excessive filtering can easily damage the detailed information of the interference fringes, leading to distortion of the effective signal.

[0005] Secondly, the positioning accuracy of coherence peaks is insufficient. Commonly used extreme value methods are sensitive to local noise, and a single abnormal sampling point can cause the positioning result to deviate. The centroid method and envelope fitting method perform better under ideal Gaussian envelopes, but when there is asymmetric distortion of the signal or background light intensity drift, there will be a systematic deviation between the fitted envelope center and the true zero optical path difference position, which directly affects the measurement accuracy of 3D reconstruction.

[0006] Third, the phase information is not fully utilized. White light interferometry signals carry both envelope intensity information and carrier phase information, the latter of which theoretically can achieve higher positioning sensitivity and resolution. However, many existing reconstruction methods only utilize the envelope information for high-level computation, failing to fully exploit the phase information to improve accuracy. Although some traditional phase-shift demodulation algorithms can achieve sub-nanometer resolution using phase information, they have stringent requirements for the uniformity of the sampling step size and coarse positioning prior accuracy, making it difficult to simultaneously achieve measurement accuracy and algorithm efficiency in real-world vibration environments.

[0007] Fourth, the algorithm has limited adaptability. Different measured objects vary significantly in terms of material, surface roughness, and reflectivity, resulting in different modulation, contrast, and envelope shapes of the interference signals. When dealing with objects of different materials or surface qualities, existing algorithms often require technicians to manually adjust filtering parameters, fitting ranges, or judgment thresholds based on experience. They lack the ability to adaptively adjust based on the characteristics of the signal itself, limiting the applicability and automation level of the measurement system under diverse and variable operating conditions.

[0008] In summary, there is an urgent need to develop a three-dimensional reconstruction method that can effectively suppress noise interference, accurately locate coherent peaks, fully utilize phase information, and possess good adaptive capabilities, in order to meet the application requirements of high-precision and high-robust surface morphology measurement. Summary of the Invention

[0009] This invention provides a white light interferometry three-dimensional reconstruction method based on mode decomposition and improved phase shift, including: Step S1: Perform a vertical scan on the surface of the sample to be tested, collect light intensity data of each pixel at different scanning heights, and obtain the original white light interference signal sequence corresponding to each pixel. Step S2: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence is decomposed into several intrinsic mode function components. Wavelet transform filtering and signal reconstruction are performed on each component to obtain the denoised interference signal sequence. Step S3: Based on the denoised interference signal sequence, calculate the modulation distribution pixel by pixel, filter out background areas without interference fringes, and obtain the effective measurement area; Step S4: Perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculate the modulation sequence of each pixel, fit it, and obtain the interference envelope curve; Step S5: Extract the peak position of the interference envelope curve as the coarse positioning position of the coherent peak of each pixel. Within a preset sampling window near the coarse positioning position of the coherent peak, the initial phase is obtained by solving the seven-step phase shift algorithm. Step S6: Perform wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area to quantify the reliability of the initial phase, and fuse the reliable phase with the coarse positioning position of the coherent peak to obtain the fine positioning position of the coherent peak. Step S7: Based on the precise location of the coherence peak of each pixel and combined with the system calibration parameters, calculate the surface height value corresponding to each pixel and reconstruct the three-dimensional surface morphology of the sample under test.

[0010] The white light interferometric 3D reconstruction method based on mode decomposition and improved phase shift, as described above, decomposes the original white light interferometric signal sequence into several intrinsic mode function components through adaptive noise complete set empirical mode decomposition. Wavelet transform filtering and signal reconstruction are then performed on each component to obtain a denoised interferometric signal sequence. This includes the following sub-steps: Step S21: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence of each pixel is decomposed into several intrinsic mode function components and a residual component; Step S22: Calculate the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screen out the effective intrinsic mode function components; Step S23: Perform wavelet transform filtering on each effective intrinsic mode function component, and superimpose the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence.

[0011] The white light interferometric 3D reconstruction method based on mode decomposition and improved phase shift described above includes the following sub-steps: performing seven-step phase shift demodulation on the denoised interferometric signal of each pixel within the effective measurement area, calculating the modulation sequence of each pixel, fitting it, and obtaining the interference envelope curve. Step S41: Based on the vertical scanning step size, perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, and calculate the modulation sequence of each pixel. Step S42: Fit the modulation sequence of each pixel using a Gaussian fitting function to obtain the interference envelope curve.

[0012] The white light interferometric 3D reconstruction method based on mode decomposition and improved phase shift, as described above, involves performing wavenumber domain orthogonal projection analysis on the denoised interferometric signal of each pixel within the effective measurement area to quantify the reliability of the initial phase. The reliable phase is then fused with the coarsely located coherent peak position to obtain the finely located coherent peak position. This process includes the following sub-steps: Step S61: Perform half-wave loss compensation on the initial phase to obtain the compensated phase; Step S62: Guided by the coherent peak coarse positioning position, the denoised interference signal is decoupled by wavenumber domain orthogonal projection, and the complex spectrum is decomposed into the main mode and the degradation mode to construct the non-equidistant optical path degradation index. Step S63: Based on the non-equidistant optical path degradation degree, determine the physical reliability of the compensated phase, selectively fuse it with the compensated phase, obtain the precise location of the coherent peak position, and perform edge protection; Step S64: Mark the positions of the coherent peaks in the fine positioning as reliable points, and use nearby reliable points to perform spatial interpolation repair for unreliable points to obtain complete fine positioning surface morphology data.

[0013] The white light interferometric 3D reconstruction method based on mode decomposition and improved phase shift, as described above, involves using the coarse localization position of the coherent peak as a guide to perform wavenumber domain orthogonal projection decoupling on the denoised interferometric signal, decomposing the complex spectrum into principal modes and degenerate modes, and constructing a non-equipathic degradation index, including the following sub-steps: Step S621: Using the coherent peak coarse location as the center, extract the local denoised interference signal segment, perform discrete Fourier transform, and obtain the complex spectrum that varies with the wavenumber distribution. Step S622: Construct the signal subspace spanned by the ideal linear phase model, and determine the optimal bias phase through optimization solution; Step S623: Perform orthogonal projection based on the optimal bias phase to decompose the complex spectrum into principal modes and degenerate modes; Step S624: Calculate the non-equipathic degradation degree based on the degradation mode and complex spectrum.

[0014] The white light interferometric 3D reconstruction method based on mode decomposition and improved phase shift, as described above, includes the following sub-steps: determining the physical reliability of the compensated phase based on the degree of non-equipathic degradation, selectively fusing the compensated phase to obtain the precisely located coherent peak position, and performing edge protection. Step S631: Determine the physical reliability of the compensated phase based on the non-equidistant optical path degradation degree. If reliable, fuse it with the coarse positioning position of the coherent peak. If unreliable, degrade it to pure envelope positioning to obtain the fine positioning position of the coherent peak. Step S632: Based on the precise location of the coherence peak of each pixel, perform pre-marking for the protection of the real physical edge and output edge protection marking information.

[0015] This invention also provides a white light interferometry three-dimensional reconstruction system based on mode decomposition and improved phase shift, comprising: The interference signal acquisition module performs a vertical scan on the surface of the sample to be tested, collects the light intensity data of each pixel at different scanning heights, and obtains the original white light interference signal sequence corresponding to each pixel. The interference signal denoising module decomposes the original white light interference signal sequence into several intrinsic mode function components through adaptive noise complete set empirical mode decomposition, and performs wavelet transform filtering and signal reconstruction on each component to obtain the denoised interference signal sequence. The effective measurement area acquisition module calculates the modulation distribution pixel by pixel based on the denoised interference signal sequence, filters out background areas without interference fringes, and obtains the effective measurement area. The interference envelope curve acquisition module performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculates the modulation sequence of each pixel, fits it, and obtains the interference envelope curve. The initial phase acquisition module extracts the peak position of the interference envelope curve as the coarse positioning position of the coherent peak of each pixel, and obtains the initial phase by solving the seven-step phase shift algorithm within a preset sampling window near the coarse positioning position of the coherent peak. The precise positioning coherent peak position acquisition module performs wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area, quantifies the reliability of the initial phase, and fuses the reliable phase with the coarse positioning position of the coherent peak to obtain the precise positioning coherent peak position. The three-dimensional surface morphology reconstruction module calculates the surface height value corresponding to each pixel based on the precise location of the coherence peak of each pixel and the system calibration parameters, thereby reconstructing the three-dimensional surface morphology of the sample under test.

[0016] The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, as described above, includes an interferometric signal denoising module that specifically comprises: The interference signal decomposition submodule decomposes the original white light interference signal sequence of each pixel into several intrinsic mode function components and a residual component through adaptive noise complete set empirical mode decomposition. The effective intrinsic mode function component screening submodule calculates the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screens out the effective intrinsic mode function components. The wavelet transform filtering submodule performs wavelet transform filtering on each effective intrinsic mode function component, and then superimposes the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence.

[0017] The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, as described above, includes an interference envelope curve acquisition module, which specifically comprises: The modulation sequence acquisition submodule performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area based on the vertical scanning step size, and calculates the modulation sequence of each pixel. The Gaussian fitting submodule uses a Gaussian fitting function to fit the modulation sequence of each pixel to obtain the interference envelope curve.

[0018] The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, as described above, includes a module for precisely locating and acquiring the coherence peak position, specifically comprising: The compensated phase acquisition submodule performs half-wave loss compensation on the initial phase to obtain the compensated phase; The non-equipathic degradation index construction submodule, guided by the coarse positioning of the coherent peak, performs wavenumber domain orthogonal projection decoupling on the denoised interference signal, decomposes the complex spectrum into principal mode and degradation mode, and constructs the non-equipathic degradation index. The fusion submodule determines the physical reliability of the compensated phase based on the degree of non-equidistant optical path degradation, selectively fuses it with the compensated phase, obtains the precise location of the coherent peak, and performs edge protection. The spatial interpolation and repair submodule marks the positions of the coherent peaks in the fine positioning as reliable points, and performs spatial interpolation and repair on unreliable points using nearby reliable points to obtain complete fine positioning surface morphology data.

[0019] The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, as described above, includes a sub-module for constructing non-equipathic degradation indices, specifically comprising: The complex spectrum acquisition submodule takes the coarse location of the coherent peak as the center, extracts a local denoised interference signal segment, performs discrete Fourier transform, and obtains the complex spectrum that varies with the wavenumber distribution. The optimal bias phase solution submodule constructs a signal subspace spanned by an ideal linear phase model and determines the optimal bias phase through optimization. The complex spectrum decomposition submodule performs orthogonal projection based on the optimal bias phase, decomposing the complex spectrum into principal modes and degenerate modes; The non-equipathic degradation calculation submodule calculates the non-equipathic degradation degree based on the degradation mode and complex spectrum.

[0020] The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, as described above, includes a fusion submodule that specifically comprises: The selective fusion decision submodule judges the physical reliability of the compensated phase based on the degree of non-equidistant optical path degradation. If reliable, it fuses with the coarse localization position of the coherent peak; if unreliable, it degenerates into pure envelope localization to obtain the fine localization position of the coherent peak. The edge protection marking submodule, based on the precise location of the coherence peak of each pixel, performs pre-marking for the protection of the real physical edge and outputs edge protection marking information.

[0021] The beneficial effects achieved by this invention are as follows: This invention can significantly suppress the influence of external interference such as environmental vibration, light source fluctuations, and detector noise on the quality of interference signals. Under low signal-to-noise ratio conditions, it reduces measurement errors by more than 40% compared to traditional single-filter methods, and the measurement repeatability is better than 0.5 nm, effectively improving stability and data reliability in practical measurement environments. Simultaneously, the accuracy of coherence peak localization is greatly improved, with the average measurement error reduced by 62% compared to traditional FFT demodulation methods, achieving sub-nanometer-level three-dimensional topography reconstruction accuracy. When the phase is disturbed by non-ideal factors such as material dispersion and thin-film effects, it can adaptively degenerate into robust envelope localization, avoiding measurement errors introduced by phase distortion. It exhibits good adaptability and robustness to various material surfaces such as metals, semiconductors, and optical components. Furthermore, the sharp edge features of real physical topography such as steps and grooves are completely preserved, and abnormal points in non-edge areas are effectively smoothed and repaired. The reconstruction results balance high accuracy, high integrity, and edge fidelity, obtaining stable and accurate three-dimensional topography measurement results even in high-noise environments and complex surface conditions. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0023] Figure 1 This is a flowchart of the white light interferometry three-dimensional reconstruction method based on modal decomposition and improved phase shift provided in Embodiment 1 of this application; Figure 2 This is a schematic diagram of a white light interferometry three-dimensional reconstruction system based on modal decomposition and improved phase shift provided in Embodiment 2 of this application. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example 1 like Figure 1 As shown, Embodiment 1 of this application provides a white light interferometry three-dimensional reconstruction method based on mode decomposition and improved phase shift. The method includes the following steps: Step S1: Perform a vertical scan on the surface of the sample to be tested, collect light intensity data of each pixel at different scanning heights, and obtain the original white light interference signal sequence corresponding to each pixel. Specifically, the sample to be tested is placed on the stage of the white light interferometry system. Based on the surface reflection characteristics and roughness range of the sample, the light source brightness, camera exposure time, and gain parameters of the interferometer are set. A vertical scan with equal steps is performed along the optical axis perpendicular to the sample surface. The scan path covers the maximum height undulation range of the sample surface and leaves upper and lower margins. The vertical scan step size is pre-calibrated according to the center wavelength of the light source, so that the nominal phase step between adjacent sampling frames is π / 2. At each scanning position, the camera is triggered to acquire an interferometric image, and the light intensity value of all pixels is recorded to form a three-dimensional interferogram stack. The light intensity change along the scanning direction is extracted pixel by pixel from the interferogram stack to form the original white light interferometric signal sequence of each pixel. All pixel sequences form the original interferogram stack data volume.

[0026] Step S2: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence is decomposed into several intrinsic mode function components. Wavelet transform filtering and signal reconstruction are performed on each component to obtain the denoised interference signal sequence. Furthermore, through adaptive noise-complete set empirical mode decomposition, the original white light interferometric signal sequence is decomposed into several intrinsic mode function components. Wavelet transform filtering and signal reconstruction are performed on each component to obtain the denoised interferometric signal sequence, including the following sub-steps: Step S21: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence of each pixel is decomposed into several intrinsic mode function components and a residual component; Specifically, the original white light interference signal sequence of each pixel is processed by CEEMDAN mode decomposition (adaptive noise complete set empirical mode decomposition) to obtain a set of intrinsic mode function components and residual components corresponding to each pixel.

[0027] In the first-level decomposition, multiple sets of Gaussian white noise sequences with different implementations are added to the original interference signal, resulting in multiple sets of noisy signals. Empirical Mode Decomposition (EMD) is performed on each set of noisy signals, and its first-order intrinsic mode function (IMF) components are extracted. A ensemble average is performed on the first-order IMF components of all noisy signals to obtain the final IMF components, and the residual quantity after removing the first-order components is calculated. In the second and subsequent levels of decomposition, adaptive noise is added to the residual quantity of the previous level. This adaptive noise consists of the first-order IMF components extracted from the white noise after EMD, and its amplitude is adaptively determined based on the standard deviation of the current level's residual quantity. EMD is performed on each set of noisy residual quantities, and its first-order components are extracted. A ensemble average is then performed to obtain the current-order IMF components, and the residual quantity is updated. This process is repeated until the number of extreme points in the residual quantity is insufficient to support further decomposition. At this point, the residual quantity cannot be further decomposed and is recorded as the final residual component. After the above decomposition, the original interference signal is decomposed into several eigenmode function components arranged from high frequency to low frequency and a residual component characterizing the overall trend of the signal change.

[0028] The CEEMDAN mode decomposition effectively counteracts the impact of noise on signal reconstruction accuracy by adaptively adding noise components and taking ensemble averages in each decomposition layer. At the same time, it assigns effective interference oscillation components and noise components at different frequency scales to different intrinsic mode function components, fundamentally overcoming the mode aliasing problem in traditional empirical mode decomposition, and providing a high-quality component basis for subsequent correlation screening and noise component separation.

[0029] Step S22: Calculate the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screen out the effective intrinsic mode function components; Specifically, through the correlation coefficient calculation formula The correlation coefficient between each intrinsic mode function component of each pixel and the original white light interference signal is calculated to quantify the degree of correlation between each intrinsic mode function component of each pixel and the original white light interference signal. For the first The correlation coefficients between each intrinsic mode function component and the original white light interference signal sequence. The value range of is [-1, 1]. The closer its absolute value is to 1, the higher the linear correlation between the intrinsic mode function component and the original white light interference signal, and the richer the effective interference information it carries. The closer its absolute value is to 0, the weaker the correlation between the intrinsic mode function component and the original white light interference signal, and the more likely it is a noise-dominated component or a spurious mode component. This represents the signal length of the corresponding pixel (i.e., the total number of sampling steps in the vertical scan). , This is the original white light interference signal sequence corresponding to each pixel. The arithmetic mean of the original white light interference signal sequence is given. For the first A sequence of eigenmode function components For the first The arithmetic mean of the sequence of intrinsic mode function components.

[0030] The correlation coefficients correlation threshold with preset If a comparison is made, , will the Each intrinsic mode function component is determined to be a valid intrinsic mode function component and is retained. , will the The intrinsic mode function components are determined to be noise-dominant components and are discarded; the preset correlation threshold τ is dynamically set according to the noise level of the measurement scenario, with a typical value range of 0.1 to 0.3. Under low signal-to-noise ratio conditions, the threshold is appropriately reduced to ensure that effective signal components are not mistakenly discarded.

[0031] Step S23: Perform wavelet transform filtering on each effective intrinsic mode function component, and superimpose the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence; Specifically, based on the local oscillation characteristics of the white light interference signal, a matching wavelet basis function is selected, and multi-scale wavelet decomposition is performed on each effective intrinsic mode function component to obtain the detail coefficients and approximation coefficients of the final layer at each scale. The wavelet basis function is selected from wavelet systems with tight support and symmetry to effectively capture the local transient characteristics of the interference signal near zero optical path difference. The number of decomposition layers is determined according to the signal length and the effective frequency range of the interference fringes, so that the approximation coefficients of the final layer reflect the baseline trend of the signal, and the detail coefficients of each layer correspond to the oscillation components and residual noise components at different frequency scales.

[0032] For each layer of detail coefficients, the absolute deviation of its median is calculated as a robust estimate of the noise standard deviation for that layer. The threshold for the corresponding layer is determined based on this noise standard deviation. When noise-dominated coefficients account for a high proportion in a layer of detail coefficients, a soft thresholding function is used to reconstruct a smoother signal. When signal-dominated coefficients account for a high proportion in a layer of detail coefficients and signal edge features need to be preserved, a hard thresholding function is used to retain local edge details of the interference fringes. Noise-dominated coefficients with absolute values ​​below the corresponding threshold are set to zero. Signal-dominated coefficients with absolute values ​​above the corresponding threshold are either retained (hard thresholding) or shrunk to zero (soft thresholding) according to the selected thresholding function rules. The final layer approximation coefficients are not thresholded and are directly retained.

[0033] Based on the detail coefficients of each layer after thresholding and the approximation coefficients of the final layer, the filtered intrinsic mode function components are obtained by reconstructing layer by layer through inverse wavelet transform. All effective intrinsic mode function components after filtering are then superimposed with the residual components of equal length to obtain the denoised interference signal sequence for the corresponding pixel.

[0034] Step S3: Based on the denoised interference signal sequence, calculate the modulation distribution pixel by pixel, filter out background areas without interference fringes, and obtain the effective measurement area; Specifically, for the denoised interference signal sequence, the maximum and minimum light intensity values ​​along the vertical scanning direction are extracted pixel by pixel, and then... Calculate the modulation level of the corresponding pixel, where, This represents the maximum light intensity in the denoised interference signal sequence corresponding to each pixel. This represents the minimum light intensity in the denoised interference signal sequence corresponding to each pixel. The modulation index is the value of the corresponding pixel, ranging from [0,1]. Its value reflects whether there is an effective interference fringe signal for the corresponding pixel. When the pixel is located in the effective area of ​​the surface being measured, there is obvious interference oscillation in its denoised interference signal sequence, the maximum and minimum light intensity values ​​are significantly different, and the modulation index is close to 1. When the pixel is located in the background area or out-of-focus area without the surface being measured, there is no obvious interference oscillation along the scanning direction, the maximum and minimum values ​​are close, and the modulation index is close to 0.

[0035] Adjust the pixel settings With preset modulation threshold When comparing, When the corresponding pixel is determined to be a pixel in the valid measurement area, its denoised interference signal data is retained. When the corresponding pixel is determined, it is identified as either a background area or an area without interference fringes.

[0036] The preset modulation threshold The threshold is determined by combining the noise level of the measurement scene and the reflectivity characteristics of the surface being measured. The typical value range is 0.05 to 0.15. Under low signal-to-noise ratio conditions, the threshold should be appropriately reduced to ensure that weak reflection areas are not mistakenly excluded.

[0037] Step S4: Perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculate the modulation sequence of each pixel, fit it, and obtain the interference envelope curve; Furthermore, a seven-step phase-shift demodulation process is performed on the denoised interference signal of each pixel within the effective measurement area. The modulation sequence of each pixel is calculated, fitted, and the interference envelope curve is obtained, including the following sub-steps: Step S41: Based on the vertical scanning step size, perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, and calculate the modulation sequence of each pixel. Specifically, based on the vertical scanning step size (the nominal phase step between adjacent sampling frames is π / 2), for the denoised interferometric signal sequence of each pixel in the effective measurement area, a calculation window is formed by using seven consecutive frames of interferometric light intensity data, i.e., the [number of frames]. Frame to the Frame constitutes the first A calculation window, N is the total number of sampling steps for the vertical scanning of this pixel. The window slides frame by frame along the scanning direction, and the center position of each window corresponds to the first frame. The scan height of the frame.

[0038] For each calculation window containing seven frames of interference light intensity values, the linear equations established by the seven-step phase-shifting algorithm, combined with the inter-frame phase step π / 2, are used to solve for the amplitudes of the sine and cosine components of the interference signal at the center position of the corresponding window through a linear combination of the seven frames of interference light intensity values. The modulation index at the center position of the corresponding window is obtained by taking the square root of the sum of the squares of the sine and cosine components. The solution is performed window by window along the scanning direction, and the modulation index values ​​of each window are arranged sequentially according to the scanning position corresponding to the center of the window, forming a modulation index sequence in which the modulation index of the corresponding pixel changes with the scanning position.

[0039] Step S42: Fit the modulation sequence of each pixel using a Gaussian fitting function to obtain the interference envelope curve; Specifically, construct a Gaussian fitting function. Where z is the scan position coordinate along the vertical scan direction. To fit the envelope curve at position The predicted value at the zero optical path difference position is A, where A is the peak amplitude of the envelope, representing the maximum modulation of the interference signal at the zero optical path difference position. This refers to the position of the coherent peak, i.e., the scan height corresponding to the envelope peak value. The envelope width parameter is related to the spectral width of the light source and the numerical aperture of the white light interferometry system.

[0040] The modulation sequence of each pixel is fitted through nonlinear least squares iteration, and the optimization objective is to iteratively adjust the parameters A. and This ensures that the predicted values ​​of the Gaussian fitting function at each scan position are... The sum of squared deviations between the measured values ​​and the modulation sequence is minimized. A nonlinear optimization algorithm iteratively searches around the preset initial parameter values. When the fitting residuals meet the preset convergence condition, output A. and The optimal estimate is obtained. Substituting the optimal parameters into the Gaussian function yields the continuous and smooth interference envelope curve for the corresponding pixel. .

[0041] Step S5: Extract the peak position of the interference envelope curve as the coarse positioning position of the coherent peak of each pixel. Within a preset sampling window near the coarse positioning position of the coherent peak, the initial phase is obtained by solving the seven-step phase shift algorithm. Specifically, for each pixel within the effective measurement area, the coherence peak position parameters are extracted. This is recorded as the coarse positioning position of the coherent peak of the corresponding pixel. Interference envelope curve The peak value corresponds to the scan position coordinates. The position is coarsely located using the coherent peak. A preset sampling window containing seven consecutive sampling frames is selected near the center of the window along the vertical scanning direction. The coverage of the sampling window ensures that the interference signal has sufficient contrast and a smooth envelope change. For the seven frames of interference light intensity values ​​within the corresponding sampling window, based on the linear combination relationship of the seven frames of equal-step phase-shifted interferograms, a seven-step phase-shifting algorithm is used to calculate the phase value at the center position of the corresponding window. The phase-shifting step size of the seven-step phase-shifting algorithm between adjacent frames is... Under the given conditions, the amplitudes of the sinusoidal and cosine components of the interference signal are solved by linear combination of the light intensity values ​​of seven frames, and the initial phase value at the center of the window is extracted by the arctangent relationship between the two.

[0042] Step S6: Perform wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area to quantify the reliability of the initial phase, and fuse the reliable phase with the coarse positioning position of the coherent peak to obtain the fine positioning position of the coherent peak. Furthermore, wavenumber domain orthogonal projection analysis is performed on the denoised interferometric signal of each pixel within the effective measurement area to quantify the reliability of the initial phase. The reliable phase is then fused with the coarsely located position of the coherent peak to obtain the finely located position of the coherent peak, including the following sub-steps: Step S61: Perform half-wave loss compensation on the initial phase to obtain the compensated phase; Specifically, based on the optical path structure of the white-light interferometry system, it is determined whether a half-wave loss exists when the measurement beam is reflected from the sample surface. This half-wave loss depends on the configuration of optical components such as the beam splitter and mirrors in the optical path, as well as the relative relationship between the refractive index of the sample surface material and the refractive index of the incident medium. If a half-wave loss exists, the actual phase difference between the measurement beam and the reference beam will be increased by π phase jumps based on the phase corresponding to the geometric optical path difference, compensating for the initial phase with π phase. If there is no half-wave loss, the initial phase is retained, and the compensated phase is obtained. .

[0043] Step S62: Guided by the coherent peak coarse positioning position, the denoised interference signal is decoupled by wavenumber domain orthogonal projection, and the complex spectrum is decomposed into the main mode and the degradation mode to construct the non-equidistant optical path degradation index. Furthermore, guided by the coarse location of the coherent peak, the denoised interference signal is decoupled by orthogonal projection in the wavenumber domain, and the complex spectrum is decomposed into principal modes and degenerate modes. The construction of the non-equipathic degradation index includes the following sub-steps: Step S621: Using the coherent peak coarse location as the center, extract the local denoised interference signal segment, perform discrete Fourier transform, and obtain the complex spectrum that varies with the wavenumber distribution. Specifically, the position is roughly determined by the coherent peak. Centered on the denoised interferometric signal sequence, cubic spline interpolation resampling is performed to obtain center-aligned results. The length is L, and the step size is the same as the original nominal scanning step size. Consistent local interference signal segment ,in, , The window covers the main oscillation region of the interference fringes.

[0044] Performing a discrete Fourier transform on the local interference signal segment yields the complex spectrum in the spatial angular frequency domain. ,in The spatial angular frequency, the discrete frequency grid points are determined by the sampling step size and the signal length. , .

[0045] Based on the round-trip optical path difference characteristics of white light interference systems, a one-to-one mapping relationship between spatial angular frequency and optical wavenumber is established: Where k is the angular wavenumber of light, satisfying , The wavelength is given. The spatial angular frequency axis is then mapped to the wavenumber axis of light according to this relationship, yielding the complex spectrum distributed with respect to wavenumber k. ,in, The complex amplitude of the interference signal at wavenumber k contains amplitude and phase information of the spectral components. The amplitude spectrum represents the magnitude of the spectral components in the interference signal at wavenumber k. The imaginary unit, The phase factor carries the phase information of the interference signal at wavenumber k. The measured phase represents the phase delay of the interference signal relative to the reference signal at wavenumber k. Based on the wavenumber range corresponding to the nominal half-width at half-maximum of the light source's spectrum, complex spectrum data within the effective wavenumber interval are extracted, and out-of-band noise is removed to finally obtain M effective wavenumber points.

[0046] Step S622: Construct the signal subspace spanned by the ideal linear phase model, and determine the optimal bias phase through optimization solution; Specifically, under ideal white light interference conditions, the phase of the interference signal in the wavenumber domain has a strictly linear relationship with the wavenumber, that is, there exists a nominal phase slope. and constant bias phase This causes the energy of the signal's complex spectrum to be concentrated in the basis vectors. Within the spanned one-dimensional signal subspace. For the complex spectrum. To compensate for the post-phase Centered on the preset search range Inside, through Iterate through each candidate bias phase For each candidate phase value, calculate the square of the inner product magnitude, and select the value that maximizes the square of the magnitude. As the optimal bias phase ,in, This is the optimal bias phase. The constant bias phase to be optimized. To compensate for the phase, The half-width of the search interval is determined based on a combination of phase compensation accuracy and system calibration error, typically ranging from π / 4 to π / 2. The total number of discrete wavenumber points within the effective wavenumber range. , The m-th discrete wavenumber value within the effective wavenumber range is determined by the frequency grid points of the discrete Fourier transform. For the complex spectrum at the m-th wavenumber point The complex amplitude at that point, including the amplitude of the spectral components. and measured phase basis vectors .

[0047] Step S623: Perform orthogonal projection based on the optimal bias phase to decompose the complex spectrum into principal modes and degenerate modes; Specifically, define the principal modal complex signal. For complex spectrum Orthogonal projection onto the signal subspace: Degenerate mode complex signal Defined as complex spectrum With the main mode complex signal difference.

[0048] By the mathematical properties of orthogonal projection, the principal mode and the degenerate mode satisfy a strict orthogonality relation, and the total energy satisfies... The principal mode carries linear phase information directly corresponding to the true height of the sample surface, and its phase spectrum strictly satisfies the following within the effective wavenumber range. The degradation mode includes wavelength-dependent phase distortion caused by non-ideal factors such as thin film effect and material dispersion.

[0049] Step S624: Calculate the non-equipathic degradation degree based on the degradation mode and complex spectrum; Specifically, based on degenerate modes and complex spectra, through Calculate the non-equipathic degradation degree The value range of NDD is [0,1]. An NDD close to 0 indicates that almost all the complex spectrum energy is concentrated in the principal mode, the phase spectrum is highly linear, and the phase after compensation is... It is physically highly reliable; NDD approaches 1, indicating that the complex spectrum energy is mainly contributed by degenerate modes, and the phase spectrum exhibits severe wavelength-dependent distortion. It is severely affected by the physical properties of the material.

[0050] Step S63: Based on the non-equidistant optical path degradation degree, determine the physical reliability of the compensated phase, selectively fuse it with the compensated phase, obtain the precise location of the coherent peak position, and perform edge protection; Furthermore, based on the non-equipathic degradation degree, the physical reliability of the compensated phase is determined, and selective fusion is performed with the compensated phase to obtain the precisely located coherent peak position. Edge protection is then performed, including the following sub-steps: Step S631: Determine the physical reliability of the compensated phase based on the non-equidistant optical path degradation degree. If reliable, fuse it with the coarse positioning position of the coherent peak. If unreliable, degrade it to pure envelope positioning to obtain the fine positioning position of the coherent peak. Specifically, the non-equipathic optical path degradation degree (NDD) is compared with a preset degradation degree threshold. Compare and selectively fuse: like This indicates that the proportion of degenerate mode energy is low, the phase spectrum linearity is high, and the compensated phase is physically reliable. The compensated phase... Coarse positioning of coherent peaks pass By performing fusion, the positions of the coherent peaks can be precisely located. To accurately locate the position of the coherent peak, To roughly locate the position of the coherent peak. The center wavelength of the light source To compensate for the phase.

[0051] like This indicates that the proportion of degradation mode energy is too high, and the phase spectrum is severely disturbed by the material's physical properties. After compensation, the phase... This method cannot be used for fine positioning correction. In this case, the phase term should be discarded, and the position should be coarsely located using the coherence peak. As a precise location of the coherent peak, i.e. .

[0052] The degradation threshold The systemic background distribution of NDD is determined by measuring it on a standard sample beforehand, and the high quantile of the background distribution is used as a general threshold, typically ranging from 0.05 to 0.15.

[0053] Step S632: Based on the precise location of the coherence peak of each pixel, perform pre-marking for the protection of the real physical edge and output edge protection marking information; Specifically, for each pixel, the modulation gradient magnitude of its interference signal envelope is calculated. At the edges of real physical steps, diffraction effects cause a local decrease in modulation, while anomalies caused by computational noise typically do not exhibit modulation changes. Based on the precisely located coherence peak positions of each pixel, the height difference between each pixel and its neighboring pixels is calculated. If the height difference exceeds a preset jump threshold, but its modulation gradient magnitude is simultaneously higher than an edge gradient threshold, the height jump is determined to originate from real surface features. The corresponding pixel is pre-marked as an edge protection point and will not participate in subsequent interpolation repair to preserve the sharp edge features of the real morphology. The edge protection pre-marking information for each pixel is output.

[0054] Step S64: Mark the positions of the coherent peaks in the fine positioning as reliable points, and use nearby reliable points to perform spatial interpolation repair on unreliable points to obtain complete fine positioning surface morphology data; Specifically, for pixels not marked for edge protection, the difference magnitude of the height of the coherence peak position between each pixel and its neighboring pixels is calculated. Pixels with height difference magnitudes less than a preset jump threshold are marked as reliable points, while those with height difference magnitudes greater than or equal to the preset jump threshold are marked as unreliable points. For pixels marked as unreliable, spatial interpolation is performed based on the precise positioning height values ​​of their neighboring reliable points. Inverse distance weighted interpolation or linear interpolation methods can be used, with the height values ​​of surrounding reliable points and their spatial distances to unreliable points as weights, to estimate a reasonable height value for the unreliable point. After interpolation repair, the abnormal height values ​​of unreliable points are replaced with estimated values ​​that are consistent with the surrounding topography, while the sharp features of the real physical edges are fully preserved due to the edge protection mechanism exemption. Ultimately, complete, continuous, and edge-fidelity precise positioning surface topography data within the effective measurement area are obtained.

[0055] Step S7: Based on the precise location of the coherence peak of each pixel and combined with the system calibration parameters, calculate the surface height value corresponding to each pixel and reconstruct the three-dimensional surface morphology of the sample under test. Specifically, the precise positioning coherence peak positions of each pixel in the complete, precisely positioned surface topography data are converted into surface height values ​​with physical units. This conversion is based on the pre-calibration parameters of the white light interferometry system, including the actual calibrated value of the vertical scan step, the nonlinear compensation curve of the piezoelectric scanner, and the lateral sampling resolution corresponding to the camera pixel size and objective lens magnification. For each pixel, its physical surface height value is obtained by multiplying the precise positioning coherence peak position of that pixel by the calibrated vertical scan step, and then adding the nonlinear compensation amount of the piezoelectric scanner at the corresponding position. The physical surface height values ​​of all pixels are integrated to generate the complete three-dimensional surface topography of the sample under test, output as a height map or a three-dimensional point cloud. The lateral position of each pixel is determined by the camera pixel coordinates and the lateral sampling resolution, while the vertical value is the corresponding physical surface height value.

[0056] Example 2 like Figure 2 As shown, Embodiment 2 of this application provides a white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, including: Interference signal acquisition module 21 performs vertical scanning on the surface of the sample to be tested, collects light intensity data of each pixel at different scanning heights, and obtains the original white light interference signal sequence corresponding to each pixel. The interference signal denoising module 22 decomposes the original white light interference signal sequence into several intrinsic mode function components through adaptive noise complete set empirical mode decomposition, performs wavelet transform filtering and signal reconstruction on each component, and obtains the denoised interference signal sequence. Furthermore, the interference signal denoising module 22 includes the following sub-modules: The interference signal decomposition submodule decomposes the original white light interference signal sequence of each pixel into several intrinsic mode function components and a residual component through adaptive noise complete set empirical mode decomposition. The effective intrinsic mode function component screening submodule calculates the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screens out the effective intrinsic mode function components. The wavelet transform filtering submodule performs wavelet transform filtering on each effective intrinsic mode function component, and superimposes the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence. The effective measurement area acquisition module 23 calculates the modulation distribution pixel by pixel based on the denoised interference signal sequence, filters out the background area without interference fringes, and acquires the effective measurement area. The interference envelope curve acquisition module 24 performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculates the modulation sequence of each pixel, fits it, and obtains the interference envelope curve. Furthermore, the interference envelope curve acquisition module 24 includes the following sub-modules: The modulation sequence acquisition submodule performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area based on the vertical scanning step size, and calculates the modulation sequence of each pixel. The Gaussian fitting submodule uses a Gaussian fitting function to fit the modulation sequence of each pixel to obtain the interference envelope curve. The initial phase acquisition module 25 extracts the peak position of the interference envelope curve as the coherent peak coarse positioning position of each pixel, and obtains the initial phase by solving the seven-step phase shift algorithm within a preset sampling window near the coherent peak coarse positioning position. The fine-position coherent peak acquisition module 26 performs wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area, quantifies the reliability of the initial phase, and fuses the reliable phase with the coarse-position coherent peak to obtain the fine-position coherent peak. Furthermore, the precise positioning coherent peak location acquisition module 26 includes the following sub-modules: The compensated phase acquisition submodule performs half-wave loss compensation on the initial phase to obtain the compensated phase; The non-equipathic degradation index construction submodule, guided by the coarse positioning of the coherent peak, performs wavenumber domain orthogonal projection decoupling on the denoised interference signal, decomposes the complex spectrum into principal mode and degradation mode, and constructs the non-equipathic degradation index. Furthermore, the submodule for constructing the non-equipathic degradation index includes the following submodules: The complex spectrum acquisition submodule takes the coarse location of the coherent peak as the center, extracts a local denoised interference signal segment, performs discrete Fourier transform, and obtains the complex spectrum that varies with the wavenumber distribution. The optimal bias phase solution submodule constructs a signal subspace spanned by an ideal linear phase model and determines the optimal bias phase through optimization. The complex spectrum decomposition submodule performs orthogonal projection based on the optimal bias phase, decomposing the complex spectrum into principal modes and degenerate modes; The non-equipathic degradation calculation submodule calculates the non-equipathic degradation degree based on the degradation mode and complex spectrum; The fusion submodule determines the physical reliability of the compensated phase based on the degree of non-equidistant optical path degradation, selectively fuses it with the compensated phase, obtains the precise location of the coherent peak, and performs edge protection. Furthermore, the fusion submodule includes the following submodules: The selective fusion decision submodule judges the physical reliability of the compensated phase based on the degree of non-equidistant optical path degradation. If reliable, it fuses with the coarse localization position of the coherent peak; if unreliable, it degenerates into pure envelope localization to obtain the fine localization position of the coherent peak. The edge protection marking submodule, based on the precise location of the coherence peak of each pixel, performs protection pre-marking on the real physical edge and outputs edge protection marking information; The spatial interpolation and repair submodule marks the positions of the coherent peaks in the fine positioning as reliable points, and performs spatial interpolation and repair on unreliable points using nearby reliable points to obtain complete fine positioning surface topography data. The three-dimensional surface morphology reconstruction module 27, based on the precise location of the coherence peak of each pixel and combined with the system calibration parameters, calculates the surface height value corresponding to each pixel and reconstructs the three-dimensional surface morphology of the sample under test. The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A three-dimensional reconstruction method based on mode decomposition and improved phase shift using white light interferometry, characterized in that, include: Step S1: Perform a vertical scan on the surface of the sample to be tested, collect light intensity data of each pixel at different scanning heights, and obtain the original white light interference signal sequence corresponding to each pixel. Step S2: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence is decomposed into several intrinsic mode function components. Wavelet transform filtering and signal reconstruction are performed on each component to obtain the denoised interference signal sequence. Step S3: Based on the denoised interference signal sequence, calculate the modulation distribution pixel by pixel, filter out background areas without interference fringes, and obtain the effective measurement area; Step S4: Perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculate the modulation sequence of each pixel, fit it, and obtain the interference envelope curve; Step S5: Extract the peak position of the interference envelope curve as the coarse positioning position of the coherent peak of each pixel. Within a preset sampling window near the coarse positioning position of the coherent peak, the initial phase is obtained by solving the seven-step phase shift algorithm. Step S6: Perform wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area to quantify the reliability of the initial phase, and fuse the reliable phase with the coarse positioning position of the coherent peak to obtain the fine positioning position of the coherent peak. Step S7: Based on the precise location of the coherence peak of each pixel and combined with the system calibration parameters, calculate the surface height value corresponding to each pixel and reconstruct the three-dimensional surface morphology of the sample under test.

2. The white light interferometry 3D reconstruction method based on mode decomposition and improved phase shift as described in claim 1, characterized in that, The original white light interferometric signal sequence is decomposed into several intrinsic mode function components by adaptive noise complete set empirical mode decomposition. Wavelet transform filtering and signal reconstruction are then performed on each component to obtain the denoised interferometric signal sequence, including the following sub-steps: Step S21: Through adaptive noise complete set empirical mode decomposition, the original white light interference signal sequence of each pixel is decomposed into several intrinsic mode function components and a residual component; Step S22: Calculate the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screen out the effective intrinsic mode function components; Step S23: Perform wavelet transform filtering on each effective intrinsic mode function component, and superimpose the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence.

3. The white light interferometry 3D reconstruction method based on mode decomposition and improved phase shift as described in claim 1, characterized in that, The denoised interference signal of each pixel within the effective measurement area is demodulated using a seven-step phase-shift demodulation process. The modulation sequence of each pixel is calculated, fitted, and the interference envelope curve is obtained. This process includes the following sub-steps: Step S41: Based on the vertical scanning step size, perform seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, and calculate the modulation sequence of each pixel. Step S42: Fit the modulation sequence of each pixel using a Gaussian fitting function to obtain the interference envelope curve.

4. The white light interferometry three-dimensional reconstruction method based on mode decomposition and improved phase shift as described in claim 1, characterized in that, Wavenumber domain orthogonal projection analysis is performed on the denoised interferometric signal of each pixel within the effective measurement area to quantify the reliability of the initial phase. The reliable phase is then fused with the coarsely located position of the coherent peak to obtain the finely located position of the coherent peak. This process includes the following sub-steps: Step S61: Perform half-wave loss compensation on the initial phase to obtain the compensated phase; Step S62: Guided by the coherent peak coarse positioning position, the denoised interference signal is decoupled by wavenumber domain orthogonal projection, and the complex spectrum is decomposed into the main mode and the degradation mode to construct the non-equidistant optical path degradation index. Step S63: Based on the non-equidistant optical path degradation degree, determine the physical reliability of the compensated phase, selectively fuse it with the compensated phase, obtain the precise location of the coherent peak position, and perform edge protection; Step S64: Mark the positions of the coherent peaks in the fine positioning as reliable points, and use nearby reliable points to perform spatial interpolation repair for unreliable points to obtain complete fine positioning surface morphology data.

5. The white light interferometry three-dimensional reconstruction method based on mode decomposition and improved phase shift as described in claim 4, characterized in that, Guided by the coarse location of the coherent peak, the denoised interferometric signal is decoupled by orthogonal projection in the wavenumber domain. The complex spectrum is decomposed into principal modes and degenerate modes, and a non-equidistant optical path degradation index is constructed, including the following sub-steps: Step S621: Using the coherent peak coarse location as the center, extract the local denoised interference signal segment, perform discrete Fourier transform, and obtain the complex spectrum that varies with the wavenumber distribution. Step S622: Construct the signal subspace spanned by the ideal linear phase model, and determine the optimal bias phase through optimization solution; Step S623: Perform orthogonal projection based on the optimal bias phase to decompose the complex spectrum into principal modes and degenerate modes; Step S624: Calculate the non-equipathic degradation degree based on the degradation mode and complex spectrum.

6. A white-light interferometry 3D reconstruction system based on mode decomposition and improved phase shift, characterized in that, include: The interference signal acquisition module performs a vertical scan on the surface of the sample to be tested, collects the light intensity data of each pixel at different scanning heights, and obtains the original white light interference signal sequence corresponding to each pixel. The interference signal denoising module decomposes the original white light interference signal sequence into several intrinsic mode function components through adaptive noise complete set empirical mode decomposition, and performs wavelet transform filtering and signal reconstruction on each component to obtain the denoised interference signal sequence. The effective measurement area acquisition module calculates the modulation distribution pixel by pixel based on the denoised interference signal sequence, filters out background areas without interference fringes, and obtains the effective measurement area. The interference envelope curve acquisition module performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area, calculates the modulation sequence of each pixel, fits it, and obtains the interference envelope curve. The initial phase acquisition module extracts the peak position of the interference envelope curve as the coarse positioning position of the coherent peak of each pixel, and obtains the initial phase by solving the seven-step phase shift algorithm within a preset sampling window near the coarse positioning position of the coherent peak. The precise positioning coherent peak position acquisition module performs wavenumber domain orthogonal projection analysis on the denoised interference signal of each pixel in the effective measurement area, quantifies the reliability of the initial phase, and fuses the reliable phase with the coarse positioning position of the coherent peak to obtain the precise positioning coherent peak position. The three-dimensional surface morphology reconstruction module calculates the surface height value corresponding to each pixel based on the precise location of the coherence peak of each pixel and the system calibration parameters, thereby reconstructing the three-dimensional surface morphology of the sample under test.

7. The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift as described in claim 6, characterized in that, The interference signal denoising module specifically includes: The interference signal decomposition submodule decomposes the original white light interference signal sequence of each pixel into several intrinsic mode function components and a residual component through adaptive noise complete set empirical mode decomposition. The effective intrinsic mode function component screening submodule calculates the correlation coefficient between each intrinsic mode function component and the original white light interference signal sequence, and screens out the effective intrinsic mode function components. The wavelet transform filtering submodule performs wavelet transform filtering on each effective intrinsic mode function component, and then superimposes the filtered intrinsic mode function components with the residual components to obtain a denoised interference signal sequence.

8. The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift as described in claim 6, characterized in that, The interference envelope curve acquisition module specifically includes: The modulation sequence acquisition submodule performs seven-step phase-shift demodulation on the denoised interference signal of each pixel in the effective measurement area based on the vertical scanning step size, and calculates the modulation sequence of each pixel. The Gaussian fitting submodule uses a Gaussian fitting function to fit the modulation sequence of each pixel to obtain the interference envelope curve.

9. The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift as described in claim 6, characterized in that, The precise positioning and coherence peak location acquisition module specifically includes: The compensated phase acquisition submodule performs half-wave loss compensation on the initial phase to obtain the compensated phase; The non-equipathic degradation index construction submodule, guided by the coarse positioning of the coherent peak, performs wavenumber domain orthogonal projection decoupling on the denoised interference signal, decomposes the complex spectrum into principal mode and degradation mode, and constructs the non-equipathic degradation index. The fusion submodule determines the physical reliability of the compensated phase based on the degree of non-equidistant optical path degradation, selectively fuses it with the compensated phase, obtains the precise location of the coherent peak, and performs edge protection. The spatial interpolation and repair submodule marks the positions of the coherent peaks in the fine positioning as reliable points, and performs spatial interpolation and repair on unreliable points using nearby reliable points to obtain complete fine positioning surface morphology data.

10. The white light interferometry 3D reconstruction system based on mode decomposition and improved phase shift as described in claim 9, characterized in that, The submodule for constructing non-equipathic degradation indices specifically includes: The complex spectrum acquisition submodule takes the coarse location of the coherent peak as the center, extracts a local denoised interference signal segment, performs discrete Fourier transform, and obtains the complex spectrum that varies with the wavenumber distribution. The optimal bias phase solution submodule constructs a signal subspace spanned by an ideal linear phase model and determines the optimal bias phase through optimization. The complex spectrum decomposition submodule performs orthogonal projection based on the optimal bias phase, decomposing the complex spectrum into principal modes and degenerate modes; The non-equipathic degradation calculation submodule calculates the non-equipathic degradation degree based on the degradation mode and complex spectrum.

Images