A three-dimensional reconstruction method, system and device based on white light interference

Abstract

The application discloses a three-dimensional reconstruction method, system and device based on white light interference, and the method comprises the following steps: in an LED light source environment, scanning a to-be-measured object to obtain an original interference signal data matrix of the to-be-measured object; wherein the original interference signal data matrix comprises a plurality of original interference signals; preprocessing the original interference signals to obtain target interference signals; performing envelope processing based on the target interference signals to obtain initial envelope signals of the target interference signals; performing data fitting processing on the initial envelope signals to obtain target envelope signals; and performing height smoothing processing based on the plurality of target envelope signals to obtain target reconstructed three-dimensional data of the to-be-measured object. The application solves the problem that, when a three-dimensional reconstruction is performed in an LED light source environment by using an existing method, the tail part of the interference intensity stripe obtained by scanning decays slowly, measurement errors are introduced, and the accuracy of the measured three-dimensional data is low.

Description

Technical Field

[0001] This application relates to the field of three-dimensional measurement technology, and in particular to a three-dimensional reconstruction method, system and device based on white light interferometry. Background Technology

[0002] Vertical scanning interferometry (VSI) is one of the core devices for high-precision surface topography measurement. Based on the principle of white light interference, it acquires the interference signal intensity of various points on the surface of the measured object under different optical path differences through an axial (Z-axis) scanning reference mirror, and then calculates the three-dimensional height information of the surface to complete the three-dimensional data reconstruction. In the traditional VSI three-dimensional data reconstruction process, when using an LED with bimodal spectral characteristics as the white light source, the tail of the scanned interference intensity fringes decays slowly, which introduces measurement errors and leads to lower accuracy of the measured three-dimensional data. Summary of the Invention

[0003] To overcome the problem that the slow attenuation of the tail of the interference intensity fringes obtained by scanning in an LED light source environment when performing three-dimensional reconstruction using existing methods introduces measurement errors and leads to low accuracy of the measured three-dimensional data, this application provides a three-dimensional reconstruction method, system and device based on white light interferometry.

[0004] Firstly, in order to solve the aforementioned technical problems, this application provides a three-dimensional reconstruction method based on white light interferometry, comprising:

[0005] In an LED light source environment, a white light interferometer vertical scanning interferometer is used to scan the object under test to obtain the original interference signal data matrix for the object under test. The original interference signal data matrix includes multiple original interference signals. The rows of the original interference signal data matrix correspond to each scanning position, and the columns correspond to each pixel in the scanning plane of the white light interferometer vertical scanning interferometer.

[0006] The original interference signal is preprocessed to obtain the target interference signal;

[0007] The initial envelope signal of the target interference signal is obtained by performing envelope processing based on the target interference signal;

[0008] The initial envelope signal is fitted to obtain the target envelope signal.

[0009] The target reconstructed three-dimensional data of the object under test is obtained by highly smoothing multiple target envelope signals corresponding to the original interference signal data matrix.

[0010] Furthermore, the original interference signal is preprocessed to obtain the target interference signal, including:

[0011] The original interference signal is filtered using a preset median filtering method to obtain the filtered signal;

[0012] The filtered signal is detrended using a pre-defined polynomial fitting method to obtain the target interference signal.

[0013] Furthermore, envelope processing is performed on the target interferometric signal to obtain the initial envelope signal of the target interferometric signal, including:

[0014] Energy analysis of the target interference signal is performed to obtain the energy distribution curve of the target interference signal;

[0015] The energy distribution curve is truncated according to the preset truncation rule to obtain a stable energy curve;

[0016] The envelope of the portion of the interference signal corresponding to the stable energy curve in the target interference signal is extracted to obtain the initial envelope signal of the target interference signal.

[0017] Furthermore, the envelope of the portion of the interference signal corresponding to the stable energy curve in the target interference signal is extracted to obtain the initial envelope signal of the target interference signal, including:

[0018] Using the pre-defined Hilbert transform method, a fast Fourier transform is performed on the portion of the interference signal corresponding to the stable energy curve in the target interference signal to obtain the frequency domain signal;

[0019] The frequency domain signal is filtered to obtain the frequency domain filtered signal;

[0020] Perform a fast inverse Fourier transform on the frequency domain filtered signal to obtain a complex analytic signal;

[0021] The initial envelope signal of the target interference signal is obtained by calculating the modulus of the analytical signal.

[0022] Furthermore, the initial envelope signal is subjected to data fitting processing to obtain the target envelope signal, including:

[0023] The initial value of the peak position of the initial envelope signal is determined by performing Gaussian fitting on the initial envelope signal using a preset Gaussian function.

[0024] Using the initial value of the peak position, the initial envelope signal is fitted based on the LED light source environment to obtain the target envelope signal.

[0025] Furthermore, using the initial peak position value, the initial envelope signal is fitted based on the LED light source environment to obtain the target envelope signal. This includes:

[0026] Obtain the measured light source spectrum curve corresponding to the LED light source environment;

[0027] Numerical integration simulation was performed on the measured light source spectrum curve based on the white light interference principle to obtain the LED interference fringe distribution curve;

[0028] A cubic spline interpolation model of LED interference fringes is obtained by performing cubic spline interpolation on the LED interference fringe distribution curve.

[0029] Find the fringe curve that matches the initial value of the peak position in the cubic spline interpolation model of LED interference fringes;

[0030] The target envelope signal is obtained by fitting the initial envelope signal with the stripe curve.

[0031] Furthermore, based on the multiple target envelope signals corresponding to the original interferometric signal data matrix, highly smoothed processing is performed to obtain the target reconstruction 3D data of the object under test, including:

[0032] The height value of the pixel position of the target envelope signal is obtained by calculating based on the peak value in the target envelope signal and the acquisition interval; where the acquisition interval is the scanning interval distance when scanning the object under test.

[0033] The median filtering method is used to perform global filtering on multiple height values ​​corresponding to the original interferometric signal data matrix to obtain the initial reconstructed three-dimensional data of the object under test.

[0034] The initial reconstructed 3D data is subjected to global filtering using a preset Gaussian filter kernel to obtain the target reconstructed 3D data of the object under test.

[0035] Furthermore, the methods also include:

[0036] Based on the number of rows and columns of the original interferometric signal data matrix, a thread network is constructed for the object under test, and the thread network includes multiple threads;

[0037] Using a pre-defined parallel computing technique, the original interference signals in the original interference signal data matrix are processed in parallel using a thread network with pre-defined steps. The threads in the thread network correspond one-to-one with the original interference signals in the original interference signal data matrix.

[0038] The preset steps for each original interference signal include preprocessing, envelope processing, and data fitting processing.

[0039] Secondly, this application also provides a three-dimensional reconstruction system based on white light interferometry, comprising:

[0040] The scanning module is used to scan the object under test using a white light interferometer vertical scanning interferometer to obtain the original interference signal data matrix for the object under test. The original interference signal data matrix includes multiple original interference signals. The rows of the original interference signal data matrix correspond to each scanning position, and the columns correspond to each pixel in the scanning plane of the white light interferometer vertical scanning interferometer.

[0041] The preprocessing module is used to preprocess the original interference signal to obtain the target interference signal;

[0042] The envelope processing module is used to perform envelope processing on the target interference signal to obtain the initial envelope signal of the target interference signal;

[0043] The fitting module is used to perform data fitting processing on the initial envelope signal to obtain the target envelope signal;

[0044] The 3D reconstruction module is used to perform highly smoothing processing on multiple target envelope signals corresponding to the original interference signal data matrix to obtain the target reconstruction 3D data of the object under test.

[0045] Thirdly, this application also provides a computing device, including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the three-dimensional reconstruction method based on white light interferometry as described above.

[0046] Fourthly, this application also provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform steps of a three-dimensional reconstruction method based on white light interferometry.

[0047] The beneficial effects of this application are as follows: First, a white light interferometer is used to scan the object under test in an LED light source environment to obtain the original interference signal data matrix for the object under test. Each original interference signal in the original interference signal data matrix is ​​preprocessed to remove interference information and improve the accuracy of the obtained target interference signal. Second, for each target interference signal, envelope processing and data fitting are performed sequentially based on that target interference signal. This reduces signal errors caused by the slow attenuation of the tail of the interference intensity fringes due to white light interference generated by the LED light source, thus improving the accuracy of the obtained target envelope signal. Finally, high-smoothing processing is performed on multiple target envelope signals corresponding to the original interference signal data matrix. This smoothing process further reduces the height error between different target envelope signals, ensuring that the target reconstructed 3D data obtained through high-smoothing processing contains the position and height parameters of the object under test that meet accuracy requirements. In this way, by reducing signal and height errors during signal processing, the measurement errors introduced into the 3D detection of the object under test can be reduced, thereby improving the accuracy of the obtained target reconstructed 3D data. Attached Figure Description

[0048] Figure 1 This is a schematic flowchart illustrating a three-dimensional reconstruction method based on white light interferometry, as shown in an exemplary embodiment of this application.

[0049] Figure 2 This is an exemplary embodiment of the present application, showing the original interference signal data matrix acquired based on the principle of white light interference;

[0050] Figure 3 This is a diagram illustrating the difference between a Gaussian light source and an LED light source in an exemplary embodiment of this application.

[0051] Figure 4 This is a schematic diagram of signal fitting in an exemplary embodiment of this application;

[0052] Figure 5 This is a diagram illustrating the reconstructed 3D data of the target in an exemplary embodiment of this application;

[0053] Figure 6 This is a schematic flowchart illustrating the application of the provided three-dimensional reconstruction method based on white light interferometry in an exemplary embodiment of this application.

[0054] Figure 7 This is a schematic diagram of a three-dimensional reconstruction system based on white light interferometry, which is an exemplary embodiment of this application. Detailed Implementation

[0055] The following embodiments are further explanations and supplements to this application and do not constitute any limitation on this application.

[0056] In traditional VSI 3D data reconstruction workflows, the acquired interference signals are typically analyzed and heights calculated pixel-by-pixel sequentially. However, this traditional method faces multiple challenges, especially when using LEDs as white light sources. The inherent bimodal spectral characteristics of LEDs become a key limiting factor, causing the model to deviate from the assumption of a narrow-band Gaussian spectral distribution in traditional white light source measurements, introducing measurement errors. Although the interference fringes of monochromatic light are even functions at the zero optical path difference position, and the intensity of the interference fringes remains symmetrical about the zero optical path difference position even with bimodal characteristics in the LED spectrum, the shape of its envelope curve has significantly deviated from a Gaussian distribution. For noisy, symmetrical, non-Gaussian, single-peak distributions, the accuracy of traditional Gaussian function fitting in determining peak positions has significant room for improvement. To further improve the measurement accuracy of white light interferometry methods, addressing the bimodal nature of LED spectral curves, the following issues need to be considered:

[0057] First, there are issues with signal quality and accuracy. The spectrum of LED light sources exhibits bimodality, resulting in a slower decay at the tail of the interference intensity fringes. The envelope of the LED white light interference signal shows oscillating decay, significantly deviating from the ideal Gaussian distribution. Its overall distribution is wider, and this tail data is more likely and more susceptible to noise. Furthermore, the white light interference signal itself is easily affected by environmental noise, light source fluctuations, and uneven sample reflectivity, leading to significant random noise and low-frequency drift in the original signal. This makes it difficult to accurately extract the zero optical path difference position (corresponding to surface height) from the contaminated, non-Gaussian envelope signal, severely limiting measurement accuracy. For example, in the roughness measurement of micro / nano structures on semiconductor chip surfaces, noise and distortion in the envelope can easily obscure the true morphological details or introduce misjudgments.

[0058] Secondly, there are issues with model adaptability and reconstruction error. Traditional envelope extraction and data fitting methods (such as simple polynomial fitting or a single Gaussian model) are particularly inadequate when dealing with complex non-Gaussian envelopes caused by LED bimodalities. For surfaces with steep edges, minute protrusions / depressions, or complex textures, these methods cannot accurately capture the true envelope shape, leading to increased fitting bias. Ultimately, the reconstructed 3D topography data contains errors compared to the actual surface features, failing to meet the stringent requirements of high-precision measurements (such as requiring measurement accuracy or repeatability on the order of nanometers). There is an urgent need to propose a new mathematical model to characterize the interference intensity distribution of the actual light source in the non-Gaussian spectral curve, while ensuring that the number of parameters is sufficiently small, thereby achieving a more accurate fit.

[0059] Finally, the algorithm also faces a bottleneck in computational efficiency; the pixel-by-pixel sequential processing mode is inefficient. As the size of the sample being tested increases and the measurement resolution (number of XY pixels, Z sampling points) improves, the amount of data grows exponentially. For large-area, high-resolution applications (such as comprehensive inspection of large-scale integrated circuit wafers), the processing time of traditional methods can be as long as several hours, making it difficult to meet the timeliness requirements of modern industrial online or rapid inspection.

[0060] To address the aforementioned issues, embodiments of this application provide a three-dimensional reconstruction method, system, and device based on white light interferometry, which will be described in detail below.

[0061] The three-dimensional reconstruction method based on white light interferometry provided in this application can be specifically executed by a server. It should be noted that the server can be an independent server, or it can be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms. No limitation is imposed here.

[0062] Please see Figure 1 , Figure 1 An exemplary embodiment of this application illustrates a three-dimensional reconstruction method based on white light interferometry, such as... Figure 1 As shown, this application provides a three-dimensional reconstruction method based on white light interferometry, including:

[0063] S11, In an LED light source environment, a white light interferometer is used to scan the object under test to obtain the original interference signal data matrix for the object under test; wherein, the original interference signal data matrix includes multiple original interference signals, the rows of the original interference signal data matrix correspond to each scanning position, and the columns correspond to each pixel in the scanning plane of the white light interferometer.

[0064] S12, preprocess the original interference signal to obtain the target interference signal;

[0065] S13, perform envelope processing based on the target interference signal to obtain the initial envelope signal of the target interference signal;

[0066] S14, Perform data fitting processing on the initial envelope signal to obtain the target envelope signal;

[0067] S15. Based on the original interference signal data matrix, multiple target envelope signals are highly smoothed to obtain the target reconstruction three-dimensional data of the object under test.

[0068] The white light interferometry-based 3D reconstruction method provided in this application firstly scans the object under test in an LED light source environment using a white light interferometer, obtaining an original interference signal data matrix for the object. Each original interference signal in the original interference signal data matrix is ​​then preprocessed to obtain the corresponding target interference signal, achieving format uniformity for all target interference signals, reducing the difficulty of subsequent processing, and thus improving the overall efficiency of the 3D measurement. Secondly, for each target interference signal, envelope processing and data fitting processing are performed sequentially based on the target interference signal. This reduces signal errors caused by the slow attenuation of the tail of the interference intensity fringes due to white light interference from the LED light source, improving the accuracy of the obtained target envelope signal. Finally, high-smoothing processing is performed on multiple target envelope signals corresponding to the original interference signal data matrix to further reduce height errors between different target envelope signals. This ensures that the target reconstruction 3D data obtained through high-smoothing processing contains the position and height parameters of the object under test that meet accuracy requirements. Thus, by reducing signal and height errors during signal processing, the measurement errors introduced into the 3D detection of the object under test can be reduced, thereby improving the accuracy of the obtained target reconstruction 3D data. In this context, the envelope refers to the "limit boundary curve" of a set of curves (or surfaces). For example, given a set of parameter-dependent curves, the envelope is the curve that is tangent to every one of these curves.

[0069] In an exemplary embodiment provided in this application, the experimental equipment and environment for implementing the three-dimensional reconstruction method based on white light interferometry of this application are as follows: The experiment uses a white light interferometer with a vertical scanning interferometer equipped with an LED white light source, with a scanning range of 0-100μm, a sampling interval of 0.05μm during scanning, and a lateral resolution of 1μm×1μm. The GPU device selected is an NVIDIA GeForce RTX 4060 Ti (a type of graphics card) supporting CUDA (Compute Unified Device Architecture), with a video memory capacity of 8GB. The software development environment is Visual Studio 2022, combined with CUDA Toolkit 12.6 (a toolkit) for programming implementation, and the programming language is C++.

[0070] When implementing the white light interferometry-based 3D reconstruction method of this application, the object to be measured is placed on the stage of a white light interferometer vertical scanning interferometer, and the white light interferometer vertical scanning interferometer is started for scanning. During the scanning process, the white light interferometer vertical scanning interferometer acquires the original interference signal of each pixel in the scanning plane area at different scanning positions according to the set scanning parameters (such as scanning range, acquisition interval, lateral resolution, etc.), forming a signal as shown in the figure. Figure 2The original interference signal data matrix is ​​shown. This original interference signal data matrix is ​​then transferred to the GPU device's memory via the PCIe (Peripheral Component Interconnect Express) bus to prepare for subsequent CUDA parallel computing. The rows of the original interference signal data matrix correspond to different scanning positions in the Z-axis direction of the white light interferometer's vertical scanning direction, and the columns correspond to different pixels in the scanning plane (XY plane).

[0071] Figure 2 In the diagram, each index corresponds to a different scanning position. Each index diagram represents the original interference signal group obtained by scanning the object under test at different scanning positions of the scanning plane. The original interference signal group corresponding to each scanning position forms the original interference signal data matrix (vertical scanning interferometer data).

[0072] Optionally, the original interferometric signal is preprocessed to obtain the target interferometric signal, including:

[0073] The original interference signal is filtered using a preset median filtering method to obtain the filtered signal;

[0074] The filtered signal is detrended using a pre-defined polynomial fitting method to obtain the target interference signal.

[0075] In the embodiment provided in this application, a preset median filtering method is used to filter the original interference signal to remove salt-and-pepper noise, resulting in a filtered signal. A preset polynomial fitting method is then used to de-trend the filtered signal to remove low-frequency drift components, making the signal more stable and improving the signal stability of the obtained target interference signal. Thus, by removing salt-and-pepper noise and low-frequency drift components from the original interference signal, interference information in the target interference signal can be reduced, improving the accuracy of the target interference signal. This, in turn, improves the accuracy of the subsequent 3D reconstruction data obtained based on the target interference signal.

[0076] In an exemplary embodiment provided in this application, when filtering the original interference signal using a preset median filtering method, the current pixel value is replaced by the median of the data within a sliding window, effectively suppressing random pulse interference and obtaining a filtered signal. For example, in the original signal preprocessing stage, the median filtering method uses a 3×1 sliding window to process the Z-axis original interference signal of each pixel in the scanning plane point by point. For an original interference signal of length N, from the 2nd data point to the (N-1)th data point, three consecutive data points centered on the current data point are taken, and the median of the current data point is replaced by its value, thereby removing salt-and-pepper noise and obtaining a filtered signal.

[0077] When detrending a filtered signal using a pre-defined polynomial fitting method, the order of the fitting polynomial is dynamically adjusted based on the signal's fluctuations (typically 3rd to 5th order). This removes low-frequency drift components from the signal, making it more stable and improving signal quality to obtain the target interference signal. For example, during detrending, a suitable polynomial order is first selected based on the fluctuations of the filtered signal (e.g., if the fluctuation amplitude of the filtered signal is less than or equal to a preset amplitude, indicating relatively smooth fluctuations, a 3rd order polynomial is chosen; if the fluctuation amplitude of the filtered signal is greater than the preset amplitude, indicating larger fluctuations, a 5th order polynomial is chosen). Then, the signal is fitted using the least squares method to obtain the coefficients of the fitted polynomial. The original signal is then subtracted from the value of the fitted polynomial to achieve detrending and obtain the target interference signal.

[0078] Optionally, envelope processing is performed based on the target interferometric signal to obtain the initial envelope signal of the target interferometric signal, including:

[0079] Energy analysis of the target interference signal is performed to obtain the energy distribution curve of the target interference signal;

[0080] The energy distribution curve is truncated according to the preset truncation rule to obtain a stable energy curve;

[0081] The envelope of the portion of the interference signal corresponding to the stable energy curve in the target interference signal is extracted to obtain the initial envelope signal of the target interference signal.

[0082] In the embodiment provided in this application, firstly, energy analysis is performed on the target interference signal to obtain an energy distribution curve that reflects the energy fluctuations of the target interference signal. This allows for intuitive truncation of the energy distribution curve according to a preset truncation rule, reducing the difficulty of truncation, removing curves with unstable energy fluctuations, and obtaining a stable energy curve. Secondly, the envelope of the interference signal corresponding to the stable energy curve in the target interference signal is extracted to obtain the initial envelope signal of the target interference signal. This reduces interference signals with unstable energy fluctuations in the initial envelope signal, improves the signal stability of the initial envelope signal, and thus improves the accuracy and stability of subsequent 3D reconstruction based on the initial envelope signal. This, in turn, improves the accuracy of the target reconstruction 3D data obtained from the 3D reconstruction. The truncation rule involves truncating the unstable portions at both ends of the energy distribution curve (the portion of the curve other than the gently fluctuating portion).

[0083] In an exemplary embodiment provided in this application, the motion state of the white light interferometer vertical scanning interferometer may be unstable during the start and end phases of scanning, leading to significant errors in the acquired raw interference signal. Therefore, the unstable portions at both ends of each interference signal are truncated. By analyzing the energy distribution or signal-to-noise ratio of the target interference signal, a suitable truncation position is determined, retaining a portion of the interference signal in the stable region for subsequent processing, thus avoiding the influence of unstable signals on the measurement results.

[0084] To determine the signal truncation point, energy analysis is performed on the target interferometric signal to obtain its energy distribution curve. The calculation formula is as follows:

[0085]

[0086] in, Represents the energy distribution curve. This represents the value of the target interference signal at the i-th sampling point, and N is the signal length of the target interference signal.

[0087] By analyzing the energy distribution curve, the position where the energy begins to stabilize is found according to the preset truncation rule as the truncation start point, and the position where the energy begins to decrease is found as the truncation end point. The part of the interference signal corresponding to the stable energy curve is retained as the stable signal for subsequent processing.

[0088] Optionally, the envelope of the portion of the interferometric signal corresponding to the stable energy curve in the target interferometric signal is extracted to obtain the initial envelope signal of the target interferometric signal, including:

[0089] Using the pre-defined Hilbert transform method, a fast Fourier transform is performed on the portion of the interference signal corresponding to the stable energy curve in the target interference signal to obtain the frequency domain signal;

[0090] The frequency domain signal is filtered to obtain the frequency domain filtered signal;

[0091] Perform a fast inverse Fourier transform on the frequency domain filtered signal to obtain a complex analytic signal;

[0092] The initial envelope signal of the target interference signal is obtained by calculating the modulus of the analytical signal.

[0093] In the embodiment provided in this application, since the Hilbert transform method can accurately reflect the instantaneous amplitude and phase information of a signal, performing the Hilbert transform operation on the frequency domain signal in the frequency domain can modify the phase information of the signal according to the magnitude and phase signal of the frequency domain signal, thereby presenting the intensity change trend of the signal and providing key data support for subsequent height calculation. Therefore, this embodiment uses a preset Hilbert transform method to perform a Fast Fourier Transform (FFT) on the part of the interference signal corresponding to the stable energy curve in the target interference signal to convert the part of the interference signal to the frequency domain, obtain a frequency domain signal, filter the frequency domain signal to obtain a frequency domain filtered signal, and then perform an Inverse Fast Fourier Transform (IFFT) on the frequency domain filtered signal to convert the transformed signal back to the time domain, obtaining an analytical signal that can clearly present the intensity change trend of the target interference signal. Then, the analytical signal is subjected to modulus calculation to quantify the intensity change trend of the target interference signal, obtaining the initial envelope signal of the target interference signal. Among them, the complex analytical signal obtained by processing the part of the interference signal corresponding to the stable energy curve in the target interference signal through the Hilbert transform method has the following characteristics: the negative frequency signal is set to 0, the positive frequency signal is doubled, and the zero frequency signal remains unchanged.

[0094] In an exemplary embodiment provided in this application, the envelope signal is a smooth curve describing the overall amplitude variation trend of a fast oscillating signal. It "wraps" around the original signal, reflecting its macroscopic variation characteristics while ignoring high-frequency details. The Hilbert transform method is used in the envelope extraction process: a Fast Fourier Transform is performed on the portion of the interference signal corresponding to the truncated stable energy curve using the CUFFT library (CUDA FFT library), converting the signal to the frequency domain to obtain a frequency domain signal. In the frequency domain, the frequency domain signal is filtered to obtain a frequency domain filtered signal. Then, a Fast Inverse Fourier Transform is used to convert the frequency domain filtered signal back to the time domain. After constructing the analytic signal, its magnitude is calculated to obtain the signal envelope, thus obtaining the initial envelope signal of the target interference signal.

[0095] Optionally, the initial envelope signal is subjected to data fitting processing to obtain the target envelope signal, including:

[0096] The initial value of the peak position of the initial envelope signal is determined by performing Gaussian fitting on the initial envelope signal using a preset Gaussian function.

[0097] Using the initial value of the peak position, the initial envelope signal is fitted based on the LED light source environment to obtain the target envelope signal.

[0098] In the embodiment provided in this application, the purpose of Gaussian fitting is to obtain the initial solution of the center position (i.e., peak position) of the actual data envelope, which is the b-value in the subsequent cubic spline model. In other words, the purpose is to initially determine the peak position of the initial envelope signal. Specifically, a preset Gaussian function is used to perform Gaussian fitting on the initial envelope signal to determine the initial value of the peak position at the signal peak. However, since Gaussian fitting may have certain errors at signal edges or complex surfaces, based on Gaussian fitting, the initial peak position value is used to fit the initial envelope signal according to the LED light source environment to obtain the target envelope signal. This reduces the errors existing at signal edges or complex surfaces, thereby reducing the signal error caused by the slow attenuation of the tail of the interference intensity fringes due to white light interference generated by the LED light source. This further optimizes the signal and improves the accuracy of the obtained target envelope signal.

[0099] The expression for the Gaussian function is:

[0100]

[0101] in, Indicates the initial value of the peak position, and A represents the peak intensity. Indicates the initial envelope signal. This is the peak position. Standard deviation, This represents the baseline offset. The parameters of the Gaussian function are optimized using the least squares method and nonlinear optimization to make the fitted curve match the envelope signal as closely as possible, thereby initially determining the peak position of the interference signal.

[0102] In an exemplary embodiment provided in this application, during the data fitting process, both Gaussian fitting and cubic spline interpolation fitting employ CUDA-parallelized least squares methods. In each thread, based on the data points of the envelope signal, a least squares objective function is constructed, and the parameters of the Gaussian function and cubic spline curve model are solved through iterative optimization.

[0103] Optionally, the target envelope signal is obtained by fitting the initial envelope signal based on the LED light source environment using the initial peak position value. This includes:

[0104] Obtain the measured light source spectrum curve corresponding to the LED light source environment;

[0105] Numerical integration simulation was performed on the measured light source spectrum curve based on the white light interference principle to obtain the LED interference fringe distribution curve;

[0106] A cubic spline interpolation model of LED interference fringes is obtained by performing cubic spline interpolation on the LED interference fringe distribution curve.

[0107] Find the fringe curve that matches the initial value of the peak position in the cubic spline interpolation model of LED interference fringes;

[0108] The target envelope signal is obtained by fitting the initial envelope signal with the stripe curve.

[0109] In the embodiment provided in this application, the Gaussian fitting curve is fitted by constructing a cubic spline interpolation model of LED interference fringes corresponding to the LED light source environment. This ensures that the fitted interpolation curve has continuous first and second derivatives at the nodes, making the interpolation curve smoother. At the same time, due to its high degree of freedom, it can accurately describe the subtle changes in the envelope signal, improving the accuracy of subsequent height calculation. Therefore, in this embodiment, the LED interference fringe distribution curve obtained by numerical integration simulation of the measured light source spectrum curve corresponding to the LED light source environment based on the white light interference principle is first subjected to cubic spline interpolation to obtain the LED interference fringe cubic spline interpolation model. Then, the initial envelope signal is fitted using the fringe curve that matches the initial value of the peak position in the LED interference fringe cubic spline interpolation model. This can take into account the three undetermined factors of scaling, translation, and offset in the LED light source environment, so as to highlight the subtle changes of the fitted target envelope signal and reduce the errors existing at the signal edges or complex surface shapes. This can reduce the signal error caused by the slow attenuation of the tail of the interference intensity fringes due to the white light interference generated by the LED light source, thereby improving the accuracy of the obtained target envelope signal. In turn, it can improve the accuracy of the target reconstructed three-dimensional data obtained by high smoothing processing based on the target envelope signal.

[0110] In an exemplary embodiment provided in this application, the cubic spline interpolation model is divided into an LED interference fringe envelope spline curve modeling network and a measured data envelope fitting network.

[0111] In the LED interference fringe envelope spline curve modeling network, the first step is to use the measured light source spectrum curve corresponding to the LED light source environment. The ideal LED interference fringe distribution curve was obtained using the numerical integration method. The calculation formula is as follows:

[0112]

[0113] in, express The position coordinates of the x-th stripe in the middle. Indicates wavelength. This represents the derivative of the wavelength, which is achieved by uniformly sampling the spectrum and accumulating the cosine fringes of each wavelength.

[0114] Cubic spline interpolation is performed on the LED interference fringe distribution curve to obtain the cubic spline interpolation model of the LED interference fringes. The expression of the cubic spline interpolation model of the LED interference fringes is:

[0115]

[0116] in, The scaling factor represents the magnification factor of the actual data; it is a scaling parameter. This represents the offset of the measured data and is a translation parameter. This represents the scaling of the width, and is a scaling parameter. The function model has only three undetermined parameters: a, b, and c, which are derived from spline curves. It retains sufficient feature representation capabilities and uses function scaling formulas. This ensures high stability in the fitting process. In this embodiment, the mean parameter of the Gaussian fitting can be used as the translation parameter. initial value, scaling parameter The initial value is The scaling parameter c was initially set to 1.0, and a fitting algorithm was used in conjunction with the result. The obtained parameters... The value represents the center position (i.e., peak position) of the actual data network.

[0117] Then, in the cubic spline interpolation model of LED interference fringes, a fringe curve matching the initial value of the peak position is found. This fringe curve is then used to fit the initial envelope signal to obtain the target envelope signal. The specific steps are as follows: Calculate the fringe curve matching the initial value of the peak position based on the calculated I(x). This fringe curve is also called the envelope curve E(x). Then, the envelope curve... The spline curve is obtained by fitting. Then, for the spline curve Scaling is performed to obtain the target envelope signal. .

[0118] The envelope curve E(x) that matches the initial value of the peak position is calculated based on the calculated I(x). The corresponding calculation formula is as follows:

[0119]

[0120] in, This indicates that the Hilbert transform of I(x) can be implemented using the Fourier transform.

[0121] Please see Figure 3 , Figure 3 This is a diagram illustrating the difference between a Gaussian light source and an LED light source in an exemplary embodiment of this application, as shown below. Figure 3The diagrams shown include differences in spectral curves, interference intensity (white light interference fringes), and envelope curves between Gaussian light sources (narrow-band white light sources) and LED light sources.

[0122] Please see Figure 4 , Figure 4 This is a schematic diagram of signal fitting in an exemplary embodiment of this application, as shown below. Figure 4 As shown, Figure 4 (a) shows the shape of the spline curve under different parameters. Figure 4 (b) shows the noise-infused interference intensity distribution data in the measured original interference signal. Figure 4 (c) is the envelope curve of the measured original interference signal. Figure 4 (d) shows the fitting results using Gaussian functions and B-spline models (cubic spline functions), based on... Figure 4 (c) and Figure 4 (d) shows that the B-spline fitting result is more accurate than the Gaussian function fitting, and the center offset is closer to the true value.

[0123] Optionally, based on multiple target envelope signals corresponding to the original interferometric signal data matrix, highly smoothed processing is performed to obtain the target reconstruction 3D data of the object under test, including:

[0124] The height value of the pixel position of the target envelope signal is obtained by calculating based on the peak value in the target envelope signal and the acquisition interval; where the acquisition interval is the scanning interval distance when scanning the object under test.

[0125] The median filtering method is used to perform global filtering on multiple height values ​​corresponding to the original interferometric signal data matrix to obtain the initial reconstructed three-dimensional data of the object under test.

[0126] The initial reconstructed 3D data is subjected to global filtering using a preset Gaussian filter kernel to obtain the target reconstructed 3D data of the object under test.

[0127] In the embodiment provided in this application, the height value of the pixel position of the target envelope signal is calculated based on the peak value in the target envelope signal and the acquisition interval when scanning the object under test. Then, the median filtering method and the preset Gaussian filter kernel are used to perform global filtering processing on the multiple height values ​​corresponding to the original interference signal data matrix in sequence. The median filtering method is used to remove isolated noise points in each height value, and the Gaussian filter kernel is used to further smooth the height value, making the reconstructed three-dimensional surface more continuous and smooth, while retaining the main morphological features of the surface, thereby improving the accuracy of the final target reconstructed three-dimensional data.

[0128] In an exemplary embodiment provided in this application, firstly, the height value of the pixel location of the target envelope signal is calculated based on the peak value and the acquisition interval in the target envelope signal. The specific steps are as follows:

[0129] The height value of the pixel is obtained by multiplying the x-coordinate (peak position) of the highest point in the fitted target envelope signal by the acquisition interval, and the result is stored in a specified location in the GPU memory. The acquisition interval is determined by the scanning parameters of the white light interferometer vertical scanning interferometer and is a key parameter for converting the pixel position information into actual physical height. Through the above calculation, the conversion from interference signal to surface height information is achieved.

[0130] Secondly, the median filtering method and a preset Gaussian filter kernel are used to perform global smoothing on all height values ​​corresponding to the original interferometric signal data matrix to obtain the target reconstructed 3D data. The specific steps are as follows:

[0131] During global smoothing, a 3×3 median filter window is first used to filter the height values ​​of all pixels to remove isolated noise points. Then, a Gaussian filter kernel with a standard deviation of 1 is used for a second filtering to further smooth the height data. Finally, the processed target reconstruction 3D data is transmitted back to CPU memory for subsequent 2D and 3D display, analysis, and storage operations. The target reconstruction 3D data can be displayed as follows: Figure 5 The two-dimensional (2D) image shown in the left figure is displayed as follows: Figure 5 The right image in the figure shows a three-dimensional (3D) image.

[0132] Optionally, the method further includes:

[0133] Based on the number of rows and columns of the original interferometric signal data matrix, a thread network is constructed for the object under test, and the thread network includes multiple threads;

[0134] Using a pre-defined parallel computing technique, the original interference signals in the original interference signal data matrix are processed in parallel using a thread network with pre-defined steps. The threads in the thread network correspond one-to-one with the original interference signals in the original interference signal data matrix.

[0135] The preset steps for each original interference signal include preprocessing, envelope processing, and data fitting processing.

[0136] In the embodiment provided in this application, a thread network is constructed for the object under test, so that the number of threads in the thread network is the same as the number of signals in the original interference signal data matrix. This facilitates the use of preset parallel computing techniques to perform preset steps of parallel processing on the original interference signals in the original interference signal data matrix based on the thread network, thereby achieving parallel processing of all original interference signals and improving the efficiency of global three-dimensional measurement of the object under test.

[0137] In an exemplary embodiment provided in this application, before implementing the three-dimensional reconstruction method based on white light interferometry, preparations for CUDA parallel computing are made. The specific steps are as follows:

[0138] The raw interferometric signal data matrix is ​​transferred to the memory of a CUDA-enabled GPU. Based on the number of GPU cores and memory capacity (also known as computing power and data scale), a kernel function is defined to rationally divide the data into thread blocks and a thread grid, forming a thread network with the same number of threads as the number of signals in the raw interferometric signal data matrix. Each thread in the thread network is responsible for processing one raw interferometric signal (Z-axis data at one pixel location). Parallel computation enables synchronous processing of the raw interferometric signals for all pixels, fully leveraging the GPU's parallel computing advantages and significantly improving data processing efficiency. For example, for a 1024×1024 pixel scan plane region, the thread grid can be set to (32,32), with each thread block containing 32×32 threads, meaning each thread is responsible for processing one corresponding raw interferometric signal (Z-axis data at one pixel location). Inside the kernel function, each thread independently performs preprocessing, signal truncation, envelope extraction, data fitting, and height calculation operations on the raw interferometric signal, achieving parallel processing of the raw interferometric signals at all pixel locations.

[0139] Please see Figure 6 , Figure 6 This is a schematic flowchart illustrating the application of the provided three-dimensional reconstruction method based on white light interferometry in an exemplary embodiment of this application, as shown below. Figure 6 As shown, it includes the following steps:

[0140] Data acquisition: The white light interferometer vertical scanning interferometer scans and acquires the original interference signal data matrix of the object under test. The original interference signal data matrix includes multiple original interference signals.

[0141] CUDA parallel computing preparation: Data is transferred to the GPU and the network is divided into threads so that multiple threads in the network can process multiple raw interference signals in parallel; the processing of each raw interference signal during parallel processing includes raw signal preprocessing, signal truncation, envelope extraction, data fitting, and height value calculation.

[0142] Original signal preprocessing: median filtering + polynomial detrending processing;

[0143] Signal truncation: Removing unstable regions at both ends:

[0144] Envelope extraction: Hilbert transform is used to obtain the signal envelope;

[0145] Data fitting: Gaussian function fitting + cubic spline interpolation optimization;

[0146] Height calculation: Peak position x acquisition interval to convert to height value;

[0147] Global smoothing: Median filtering + Gaussian filtering optimize the surface to obtain the target reconstructed 3D data.

[0148] In summary, the white light interferometry-based 3D reconstruction method of this application aims to overcome the problems of poor signal processing, low computational efficiency, and insufficient 3D data reconstruction accuracy in existing LED white light interferometry VSI 3D data reconstruction technologies. It innovatively solves the technical challenge of the interference signal envelope curve deviating from the Gaussian distribution due to the bimodal characteristics of the LED white light spectrum. Through advanced signal processing and analysis algorithms, it achieves rapid and accurate 3D topography reconstruction. This method is particularly suitable for surface contour detection of precision devices such as semiconductor chips, optical components, and microelectromechanical systems (MEMS). Compared to traditional methods, this technical solution significantly reduces the stringent requirements for white light sources, achieving high-quality measurement results without the need for filters. This not only improves system energy utilization efficiency but also provides a more economical and efficient solution for industrial inspection. Meanwhile, by leveraging the powerful processing capabilities of CUDA parallel computing, it enables rapid parallel processing of the original interference signal in the Z direction for each pixel. Combined with various signal processing and data fitting algorithms, it effectively improves signal quality, accurately extracts surface height information, and achieves high-precision, high-efficiency 3D data reconstruction, meeting the growing demand for precision surface topography measurement in fields such as semiconductors, displays, optics, and biomedicine.

[0149] The three-dimensional reconstruction method based on white light interferometry proposed in this application can improve computational efficiency, enhance signal quality, improve reconstruction accuracy, and has wide applicability.

[0150] The improved computational efficiency is achieved by utilizing CUDA parallel computing technology to transform traditional sequential processing into parallel computing. By leveraging the GPU's numerous computing cores to simultaneously process Z-axis data from multiple pixels, data processing speed is significantly increased compared to traditional methods. Actual testing shows that processing time for the same scale of 3D data can be reduced by more than 80%, significantly improving measurement efficiency and meeting the demands of rapid inspection in industrial production. For example, in semiconductor wafer surface inspection, the time for a single measurement can be reduced from several hours to tens of minutes, greatly improving the inspection efficiency of the production line.

[0151] The improved signal quality is reflected in the following aspects: Preprocessing operations such as filtering, detrending, and truncation of the original interference signal effectively remove noise, low-frequency drift, and unstable components, significantly improving the signal-to-noise ratio. Simultaneously, the Hilbert transform accurately extracts the signal envelope, and combined with Gaussian and cubic spline fitting models, it precisely describes the characteristics of the interference signal, providing an accurate data foundation for subsequent height calculations. This ensures that the reconstructed 3D data truly reflects the morphological features of the sample surface.

[0152] The enhanced reconstruction accuracy is reflected in the use of a Gaussian fitting combined with a cubic spline function fitting method. Compared to traditional single fitting methods, this approach is applicable to scenarios where the light source spectral curve does not meet the requirements of a narrow-band Gaussian function approximation, and it can better adapt to the morphological characteristics of complex surfaces. When processing surfaces with steep edges, minute protrusions, or depressions, it can accurately capture detailed surface information, effectively reducing fitting deviations and improving the accuracy of height calculation. Experimental verification shows that, in terms of measurement accuracy, the 3D data reconstruction error of this invention is reduced by more than 60% compared to traditional methods, meeting the high-precision requirements for surface morphology measurement of precision devices such as semiconductor chips and optical components.

[0153] The broad applicability is reflected in the fact that this method is suitable for measuring the surface morphology of various types of samples, whether it is the surface of smooth optical components, the surface of microelectromechanical systems (MEMS) devices with complex structures, or the cell surface in the biomedical field. High-precision three-dimensional data reconstruction can be achieved using this method. Furthermore, this method has good compatibility with LED white light interferometers of different specifications. By simply adjusting parameters such as the acquisition interval according to the scanning parameters of the equipment and importing the spectral curve data of the actual LED light source, it can be applied to different measurement scenarios, demonstrating strong versatility and promotional value.

[0154] Please see Figure 7 , Figure 7 An exemplary embodiment of this application illustrates a three-dimensional reconstruction system based on white light interferometry, such as... Figure 7As shown, this application provides a three-dimensional reconstruction system 700 based on white light interferometry, comprising:

[0155] The scanning module 701 is used to scan the object under test using a white light interferometer vertical scanning interferometer to obtain the original interference signal data matrix for the object under test; wherein, the original interference signal data matrix includes multiple original interference signals, the rows of the original interference signal data matrix correspond to each scanning position, and the columns correspond to each pixel in the scanning plane of the white light interferometer vertical scanning interferometer.

[0156] Preprocessing module 702 is used to preprocess the original interference signal to obtain the target interference signal;

[0157] The envelope processing module 703 is used to perform envelope processing based on the target interference signal to obtain the initial envelope signal of the target interference signal;

[0158] The fitting module 704 is used to perform data fitting processing on the initial envelope signal to obtain the target envelope signal;

[0159] The 3D reconstruction module 705 is used to perform highly smoothing processing on multiple target envelope signals corresponding to the original interference signal data matrix to obtain the target reconstruction 3D data of the object under test.

[0160] The white light interferometry-based 3D reconstruction system 700 provided in this application firstly uses a scanning module 701 to scan the object under test in an LED light source environment using a white light interferometer vertical scanning interferometer, obtaining an original interference signal data matrix for the object under test. A preprocessing module 702 then preprocesses each original interference signal in the original interference signal data matrix to obtain the corresponding target interference signal, thus unifying the format of each target interference signal, reducing the difficulty of subsequent processing, and improving the overall efficiency of the 3D measurement. Secondly, for each target interference signal, an envelope processing module 703 and a fitting module 704 sequentially perform envelope processing and data fitting processing based on the target interference signal. This reduces signal errors caused by the slow attenuation of the tail of the interference intensity fringes due to white light interference generated by the LED light source, improving the accuracy of the obtained target envelope signal. Finally, a 3D reconstruction module 705 performs height smoothing processing on multiple target envelope signals corresponding to the original interference signal data matrix. This smoothing further reduces the height error between different target envelope signals, ensuring that the target reconstruction 3D data obtained through height smoothing contains the position and height parameters of the object under test that meet accuracy requirements. In this way, by reducing signal and height errors during signal processing, the measurement errors introduced into the three-dimensional detection of the object under test can be reduced, thereby improving the accuracy of the obtained target reconstruction three-dimensional data.

[0161] Optionally, the preprocessing module 702 is specifically used for:

[0162] The original interference signal is filtered using a preset median filtering method to obtain the filtered signal;

[0163] The filtered signal is detrended using a pre-defined polynomial fitting method to obtain the target interference signal.

[0164] Optionally, the envelope processing module 703 is specifically used for:

[0165] Energy analysis of the target interference signal is performed to obtain the energy distribution curve of the target interference signal;

[0166] The energy distribution curve is truncated according to the preset truncation rule to obtain a stable energy curve;

[0167] The envelope of the portion of the interference signal corresponding to the stable energy curve in the target interference signal is extracted to obtain the initial envelope signal of the target interference signal.

[0168] Optionally, the envelope processing module 703 is specifically used for:

[0169] Using the pre-defined Hilbert transform method, a fast Fourier transform is performed on the portion of the interference signal corresponding to the stable energy curve in the target interference signal to obtain the frequency domain signal;

[0170] The frequency domain signal is filtered to obtain the frequency domain filtered signal;

[0171] Perform a fast inverse Fourier transform on the frequency domain filtered signal to obtain a complex analytic signal;

[0172] The initial envelope signal of the target interference signal is obtained by calculating the modulus of the analytical signal.

[0173] Optionally, the fitting module 704 is specifically used for:

[0174] The initial value of the peak position of the initial envelope signal is determined by performing Gaussian fitting on the initial envelope signal using a preset Gaussian function.

[0175] Using the initial value of the peak position, the initial envelope signal is fitted based on the LED light source environment to obtain the target envelope signal.

[0176] Optionally, the fitting module 704 is specifically used for:

[0177] Obtain the measured light source spectrum curve corresponding to the LED light source environment;

[0178] Numerical integration simulation was performed on the measured light source spectrum curve based on the white light interference principle to obtain the LED interference fringe distribution curve;

[0179] A cubic spline interpolation model of LED interference fringes is obtained by performing cubic spline interpolation on the LED interference fringe distribution curve.

[0180] Find the fringe curve that matches the initial value of the peak position in the cubic spline interpolation model of LED interference fringes;

[0181] The target envelope signal is obtained by fitting the initial envelope signal with the stripe curve.

[0182] Optionally, the 3D reconstruction module 705 is specifically used for:

[0183] The height value of the pixel position of the target envelope signal is obtained by calculating based on the peak value in the target envelope signal and the acquisition interval; where the acquisition interval is the scanning interval distance when scanning the object under test.

[0184] The median filtering method is used to perform global filtering on multiple height values ​​corresponding to the original interferometric signal data matrix to obtain the initial reconstructed three-dimensional data of the object under test.

[0185] The initial reconstructed 3D data is subjected to global filtering using a preset Gaussian filter kernel to obtain the target reconstructed 3D data of the object under test.

[0186] Optionally, the system 700 also includes a parallel processing module, which is specifically used for:

[0187] Based on the number of rows and columns of the original interferometric signal data matrix, a thread network is constructed for the object under test, and the thread network includes multiple threads;

[0188] Using a pre-defined parallel computing technique, the original interference signals in the original interference signal data matrix are processed in parallel using a thread network with pre-defined steps. The threads in the thread network correspond one-to-one with the original interference signals in the original interference signal data matrix.

[0189] The preset steps for each original interference signal include preprocessing, envelope processing, and data fitting processing.

[0190] It should be noted that the white light interferometry-based 3D reconstruction system and the white light interferometry-based 3D reconstruction method provided in the above embodiments belong to the same concept. The specific operation methods of each module and unit have been described in detail in the method embodiments and will not be repeated here. In practical applications, the white light interferometry-based 3D reconstruction system provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above. This is not a limitation here.

[0191] A computing device according to an embodiment of this application includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements some or all of the steps of the above-described three-dimensional reconstruction method based on white light interferometry.

[0192] The computing device can be a computer, and the corresponding program is computer software. The parameters and steps of the computing device described above can be referred to the parameters and steps in the embodiment of the three-dimensional reconstruction method based on white light interference mentioned above, and will not be repeated here.

[0193] This application embodiment provides a computer-readable storage medium storing instructions that, when executed, perform the steps of the aforementioned three-dimensional reconstruction method based on white light interferometry.

[0194] The computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0195] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of this disclosure. The aforementioned computer-readable storage medium can be a non-transitory computer-readable storage medium, including: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, and other media capable of storing program code; it can also be a transient computer-readable storage medium.

[0196] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0197] Those skilled in the art will recognize that this application can be implemented as a system, method, or computer program product. Therefore, this disclosure can be implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "module" or "system." Furthermore, in some embodiments, this application can also be implemented as a computer program product contained in one or more computer-readable media, which contains computer-readable program code. Computer-readable storage media can be, for example, but not limited to—electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof.

[0198] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0199] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for three-dimensional reconstruction based on white light interferometry, characterized in that, The application relates to a method for reconstructing three-dimensional data of an object in an LED light source environment. The method comprises the following steps: scanning an object to be measured by using a white light vertical scanning interferometer in an LED light source environment, to obtain an original interference signal data matrix of the object to be measured; wherein the original interference signal data matrix comprises a plurality of original interference signals, and the rows of the original interference signal data matrix correspond to scanning positions, and the columns correspond to each pixel point in the scanning plane of the white light vertical scanning interferometer; preprocessing the original interference signals to obtain target interference signals; performing envelope processing on the target interference signals to obtain initial envelope signals of the target interference signals; performing data fitting processing on the initial envelope signals to obtain target envelope signals; 2. The method of claim 1, wherein, performing height smoothing processing on a plurality of target envelope signals corresponding to the original interference signal data matrix to obtain target reconstructed three-dimensional data of the object to be measured. The preprocessing of the original interference signals to obtain target interference signals comprises the following steps: filtering the original interference signals by using a preset median filtering method to obtain filtered signals; 3. The method of claim 1, wherein, performing detrending processing on the filtered signals by using a preset polynomial fitting method to obtain target interference signals. The envelope processing on the target interference signals to obtain initial envelope signals of the target interference signals comprises the following steps: performing energy analysis on the target interference signals to obtain an energy distribution curve of the target interference signals; performing truncation on the energy distribution curve according to a preset truncation rule to obtain a stable energy curve; 4. The method of claim 3, wherein, performing envelope extraction on the part of the target interference signals corresponding to the stable energy curve to obtain initial envelope signals of the target interference signals. The envelope extraction on the part of the target interference signals corresponding to the stable energy curve to obtain initial envelope signals of the target interference signals comprises the following steps: performing fast Fourier transform on the part of the target interference signals corresponding to the stable energy curve by using a preset Hilbert transform method to obtain a frequency domain signal; performing filtering processing on the frequency domain signal to obtain a frequency domain filtered signal; performing fast inverse Fourier transform on the frequency domain filtered signal to obtain an analytic signal of a complex number; 5. The method of claim 1, wherein, taking a modulus value of the analytic signal to obtain initial envelope signals of the target interference signals. The data fitting processing on the initial envelope signals to obtain target envelope signals comprises the following steps: performing Gaussian fitting on the initial envelope signals by using a preset Gaussian function to determine an initial value of a peak position of the initial envelope signals; 6. The method of claim 5, wherein, performing fitting on the initial envelope signals based on the LED light source environment by using the initial value of the peak position to obtain target envelope signals. The fitting on the initial envelope signals based on the LED light source environment by using the initial value of the peak position to obtain target envelope signals comprises the following steps: obtaining an actually measured light source spectrum curve corresponding to the LED light source environment; performing numerical integral simulation on the actually measured light source spectrum curve based on a white light interference principle to obtain an LED interference fringe distribution curve; The LED interference fringe distribution curve is subjected to cubic spline interpolation to obtain a cubic spline interpolation model of the LED interference fringe; A fringe curve matching the initial value of the peak position is found in the cubic spline interpolation model of the LED interference fringe; The initial envelope signal is fitted using the fringe curve to obtain a target envelope signal.

7. The method of claim 1, wherein, The plurality of target envelope signals corresponding to the original interference signal data matrix are subjected to high smoothness processing to obtain target reconstructed three-dimensional data of the object under test, including: The height value of the pixel position where the target envelope signal is located is calculated based on the peak value in the target envelope signal and the acquisition interval, wherein the acquisition interval is the scanning interval distance when the object under test is scanned; The plurality of height values corresponding to the original interference signal data matrix are subjected to global filtering processing using a preset median filtering method to obtain initial reconstructed three-dimensional data of the object under test; The initial reconstructed three-dimensional data is subjected to global filtering processing using a preset Gaussian filtering kernel to obtain target reconstructed three-dimensional data of the object under test.

8. The method according to any one of claims 1 to 7, characterized in that, The method further includes: Based on the number of rows and columns of the original interference signal data matrix, a thread network for the object under test is constructed, including a plurality of threads; The original interference signals in the original interference signal data matrix are subjected to preset step parallel processing based on the thread network using a preset parallel computing technology, and the threads in the thread network correspond one-to-one to the original interference signals in the original interference signal data matrix; Wherein, the preset steps for each original interference signal include the preprocessing, the envelope processing and the data fitting processing.

9. A three-dimensional reconstruction system based on white light interferometry, characterized in that It includes: The scanning module is used for scanning the object under test using a white light interference vertical scanning interferometer to obtain an original interference signal data matrix for the object under test; wherein the original interference signal data matrix includes a plurality of original interference signals, and the rows of the original interference signal data matrix correspond to each scanning position, and the columns correspond to each pixel point in the scanning plane of the white light interference vertical scanning interferometer; The preprocessing module is used for preprocessing the original interference signal to obtain a target interference signal; The envelope processing module is used for envelope processing based on the target interference signal to obtain an initial envelope signal of the target interference signal; The fitting module is used for data fitting processing of the initial envelope signal to obtain a target envelope signal; The three-dimensional reconstruction module is used for high smoothness processing of a plurality of target envelope signals corresponding to the original interference signal data matrix to obtain target reconstructed three-dimensional data of the object under test.

10. A computing device comprising a memory, a processor, and a program stored on the memory and running on the processor, wherein, The processor executes the program to implement the steps of the white light interference-based three-dimensional reconstruction method according to any one of claims 1 to 8.

Images