A novel single-spectrum confocal sensor precise thickness measurement calibration and evaluation method

Abstract

The application relates to the field of optical measurement, in particular to a novel single-spectrum confocal sensor precise thickness measurement calibration and evaluation method, which comprises the steps of spectrum data preprocessing, spectrum stability evaluation, algorithm error verification, displacement-wavelength calibration, exit angle-wavelength calibration and the like, noise is eliminated through normalization processing by acquiring dark current noise signals and the like, spectrum flatness and stability are evaluated by using a specific area filter, the accuracy and precision of an algorithm are calculated by using an odd-even grouping method, and the like, and the application also relates to a spectrum confocal thickness measurement system which stores a calibrated relationship for real-time thickness calculation.The application achieves the technical effects of eliminating light source modulation, dark noise and optical path crosstalk noise, accurately evaluating spectrum stability, accurately verifying algorithm error, realizing accurate calibration of displacement-wavelength and exit angle-wavelength, and further realizing accurate thickness measurement.

Description

Technical Field

[0001] This application relates to the field of sensor thickness measurement technology, and in particular to a novel method for accurate thickness measurement calibration and evaluation of a single spectral confocal sensor. Background Technology

[0002] In the field of industrial measurement and inspection, thickness measurement is a crucial technology, widely used in various stages such as material production, parts manufacturing, and quality control. Accurate thickness measurement ensures that products meet design requirements, improving product quality and performance. With continuous technological advancements, the demands for precision and accuracy in thickness measurement are constantly increasing. Spectral confocal sensors, as an advanced measurement device, have found widespread application in thickness measurement due to their high precision and high resolution. They utilize spectral characteristics to achieve precise measurement of object thickness, providing strong support for industrial production and scientific research.

[0003] In traditional spectral confocal sensor thickness measurement technology, to achieve accurate thickness measurement with a single-channel sensor, three important relationships typically need to be determined: the relationship between displacement S and the spectral center wavelength λ, the relationship between the emission angle θ and the spectral center wavelength λ, and the relationship between the refractive index and the spectral center wavelength λ (i.e., the Sellmeier equation). For the relationship between displacement S and the spectral center wavelength λ, the conventional approach is to use standard instruments such as an interferometer to perform displacement testing, simultaneously acquiring the spectral center wavelength from the controller, and then performing polynomial fitting. Before processing the spectrometer data, there is generally a lack of effective preprocessing strategies; often, an ideal light source signal is directly used as the normalized reference signal. There is also no systematic method for evaluating spectral stability. When verifying the error of the controller's spectral center wavelength extraction algorithm, there is a lack of scientific evaluation indicators and strategies. Furthermore, no effective method is used to calibrate the relationship between the probe's emission angle θ and the spectral center wavelength λ.

[0004] However, these traditional techniques have significant drawbacks. In the calibration process of the relationship between displacement and spectral center wavelength, inaccurate and poor-quality raw data, coupled with the poor robustness of the fitting method, lead to large testing errors. Regarding spectral data processing, the original light source intensity distribution may change after passing through the system. Using an ideal light source signal as a normalized reference signal will affect the signal quality and accuracy, failing to effectively deduct the modulation effect of light source intensity on the spectral signal, or remove dark noise, crosstalk noise, and lens backscattering noise. Furthermore, traditional methods lack effective means to evaluate spectral stability and scientific verification strategies for the error of the controller's spectral center wavelength extraction algorithm, all of which seriously affect the accuracy and precision of thickness measurement using spectral confocal sensors. Summary of the Invention

[0005] The purpose of this application is to overcome the above-mentioned technical problems and provide a novel method for accurate thickness measurement calibration and evaluation using a single spectral confocal sensor.

[0006] A novel method for accurate thickness measurement calibration and evaluation using a single spectral confocal sensor includes the following steps:

[0007] Spectral data preprocessing: The system dark current noise signal dark0 is obtained under no light conditions, and the dark1 signal containing dark current and lens reflection noise is obtained when the probe is blocked by the blackboard when the light source is on. The light source input signal λ0=K*λ+dark0, where λ is the light source spectrum of the input probe, and the detection signal λ1=S0+dark1 are normalized to obtain the normalized signal (λ1-dark1) / (λ0-dark0), which eliminates light source modulation, dark noise and optical path crosstalk noise.

[0008] Spectral stability evaluation: Using a 2×2 region filter on the PD detector, with corresponding center wavelengths of 420nm, 450nm, 600nm, and 680nm, a bandwidth of 5nm, and intensity values ​​of I1-I4, we calculated R1=I2 / I1, K1=I3 / I2, R2=I3 / I4, and K2=I4 / I1. R1, R2, K1, and K2 were used to evaluate the spectral flatness and stability of the 420-680nm band, where R1 characterizes the spectral jitter of the blue light excitation source.

[0009] Algorithm error verification: N×M sets of spectral data were collected by piezoelectric ceramic step displacement, corresponding to the step size STEP. Adjacent displacement straight lines were constructed using the odd-even grouping method. The expected root mean square (RMS) (μ0) and standard deviation (RMS) (Δ) of the displacement deviation of the wavelength extraction algorithm were calculated. The accuracy of the algorithm was evaluated by RMS (μ0), the precision was evaluated by RMS (Δ), and the minimum step size STEP (min) was used as the system's limiting resolution.

[0010] Displacement-wavelength calibration: Polynomial fitting is performed on the odd and even grouped data to obtain S1(λ) and S2(λ). By iteratively correcting the weights k1, k2 and the linear term C=a*λ+b, the residual δ=∑∆(j) approaches zero, and the final calibration relationship S(λ)=k1*S1(λ)+k2*S2(λ)+C is output.

[0011] Emission angle-wavelength calibration: The reference plate with known thickness H and Sellmeier equation is continuously moved perpendicular to the optical axis of the probe. Wavelengths λ1 and λ2 on the upper / lower surface at different positions and displacement difference S(λ2)-S(λ1) are collected. The emission angle θ(λ2) is inverted by combining the Sellmeier equation n(λ2), and the θ-λ relationship is obtained by alternating fitting correction.

[0012] By adopting the above technical solutions, in the spectral data preprocessing step, the system dark current noise signal and lens reflection noise signal are acquired through a blackboard, and normalized using the light source input signal and detection signal. This effectively eliminates the influence of light source modulation, dark noise, and optical path crosstalk noise, ensuring that the acquired normalized signal truly reflects the light source signal after passing through the system. In the spectral stability evaluation step, intensity values ​​are collected using filters with specific center wavelengths and bandwidths on the PD detector. The correlation ratio is calculated to evaluate the flatness and stability of the 420-680nm band spectrum, where R1 characterizes the spectral jitter of the blue light excitation source, helping to understand the stability of the light source spectrum. In the algorithm error verification step, spectral data is acquired through piezoelectric ceramic stepping displacement. Adjacent displacement lines are constructed using the odd-even grouping method, and the expected root mean square and standard deviation root mean square of the displacement deviation of the wavelength extraction algorithm are calculated. This allows for the evaluation of the algorithm's accuracy and precision, and the minimum step size can be used as the system's limiting resolution, effectively assessing the algorithm's performance. In the wavelength calibration step, polynomial fitting is performed on the odd and even grouped data respectively, and then the weights and linear terms are iteratively corrected to make the residuals approach zero, thereby outputting a more accurate final calibration relationship. In the emission angle-wavelength calibration step, the reference plate with known thickness is continuously moved perpendicular to the optical axis of the probe, relevant data is collected, and the emission angle is inverted by combining the Sellmeier equation. Then, the relationship between the emission angle and wavelength is obtained through alternating fitting correction, which provides reliable calibration data for subsequent accurate thickness measurement.

[0013] Preferably, in the spectral data preprocessing:

[0014] dark0 is obtained by blocking the blackboard with the light source off, and includes dark current and system noise;

[0015] dark1 is obtained by turning on the light source to block the blackboard, and includes dark current and reflection noise between lenses;

[0016] Normalized signals eliminate light source intensity modulation and backscattering noise.

[0017] By adopting the above technical solution, dark0, which contains dark current and system noise, and dark1, which contains dark current and inter-lens reflection noise, can be accurately obtained. Furthermore, the normalized signal can effectively eliminate light source intensity modulation and backscattering noise, thereby improving the quality of spectral data.

[0018] Preferably, in the spectral stability evaluation:

[0019] When K1=K2 and R1=R2, the spectrum in the 420-680nm band is determined to have a trapezoidal flat top parallel to the horizontal axis.

[0020] The change in R1 reflects the power drift in the 450nm / 420nm band caused by fluctuations in the blue light source current.

[0021] By adopting the above technical solution, it is possible to determine whether the 420-680nm band spectrum has a trapezoidal flat top parallel to the horizontal axis, and it can also reflect the power drift in the 450nm / 420nm band caused by the fluctuation of the blue light source current, thereby evaluating the spectral stability.

[0022] Preferably, the algorithm error verification includes:

[0023] Calculate the displacement deviation μ1 of even-numbered points by constructing a straight line using odd-numbered index points, and calculate the displacement deviation μ2 of odd-numbered points by constructing a straight line using even-numbered index points;

[0024] The mean deviation μ0 = (μ1 + μ2) / N, where N is the total number of displacement points;

[0025] Standard deviation Δ= ;

[0026] RMS(μ0) = RMS(Δ) = .

[0027] By adopting the above technical solution, the displacement deviation of even points can be calculated by constructing a straight line with odd index points and the displacement deviation of odd points can be calculated by constructing a straight line with even index points. The displacement deviation can be effectively calculated. By calculating the expected value of the deviation, the standard deviation, the root mean square of the expected value of the deviation, and the root mean square of the standard deviation, the accuracy and precision of the spectral center wavelength extraction algorithm can be accurately evaluated. Combined with the steps of acquiring spectral data by the medium-voltage ceramic step displacement, the algorithm error can be completely verified.

[0028] Preferably, the limiting resolution STEP(min) is defined as the minimum step size STEP when RMS(μ0) and RMS(Δ) reach a set threshold.

[0029] By adopting the above technical solution, the system's limit resolution can be clearly defined. The minimum step size STEP when RMS(μ0) and RMS(Δ) reach the set threshold can be used as the limit resolution. This allows for a more accurate measurement of the algorithm's performance at different step sizes, providing a more accurate reference standard for the system's high-precision measurement.

[0030] Preferably, the polynomial fitting uses a 5-term polynomial.

[0031] By adopting the above technical solution, five polynomials are fitted to the odd and even grouped data in the displacement-wavelength calibration. Combined with iterative correction of weights and linear terms, the residuals can be made close to zero, thereby obtaining a more accurate displacement-wavelength calibration relationship.

[0032] Preferably, the iterative correction in the displacement-wavelength calibration includes:

[0033] (a) Initially set the correction term C=0, and calculate the fitting error:

[0034] ;

[0035] Where Si(λ) = k1⋅S1(λ) + k2⋅S2(λ), ⋅S1(λ) is the curve fitted using odd points, and S2(λ) is the curve fitted using even points;

[0036] (b) Calculate the residuals according to the Malikoff principle:

[0037] Divide all N data points into two equal segments according to their index order.

[0038] First half: Residuals Δ(j) from index j=1 to ⌊N / 2⌋;

[0039] Second half: The residual Δ(j) from index j = ⌊N / 2⌋ + 1 to N;

[0040] Calculate the difference between the sums of the residuals:

[0041] ;

[0042] (c) If |δ|>ε, where ε is the threshold for approaching zero, update the correction term C=a⋅λ+b, recalculate Si(λ)=k1⋅S1(λ)+k2⋅S2(λ)+C and iteratively execute steps (a)-(b) until |δ|≤ε.

[0043] By adopting the above technical solution, in displacement-wavelength calibration, an iterative correction method is used. Initially, the correction term C is set to 0, the fitting error is calculated, and the residuals are calculated according to the Malikov principle. When the absolute value of the difference between the sum of residuals is greater than the threshold for approaching zero, the correction term is updated and the fitting function is recalculated until the absolute value of the difference between the sum of residuals is less than or equal to the threshold. This can make the fitting error approach zero and improve the accuracy of the calibration of the relationship between displacement and the center wavelength of the spectrum.

[0044] Preferably, in the emission angle-wavelength calibration:

[0045] The thickness H of the reference plate must be greater than the thickness corresponding to the minimum identifiable spectral interval of the sensor; the Sellmeier equation of the reference plate is known.

[0046] Verticality is determined by the maximum intensity of the spectral signal on the upper surface;

[0047] The emission angle θ(λ2) is obtained by inversion from the formula:

[0048] .

[0049] By adopting the above technical solution, the thickness of the reference plate is made greater than the thickness corresponding to the minimum recognizable spectral interval of the sensor, which ensures that the calibration process can be carried out effectively. The perpendicularity can be determined by using the maximum value of the spectral signal intensity on the upper surface, and the reference plate can be accurately adjusted to be perpendicular to the optical axis of the probe. The emission angle can be obtained by formula inversion, and the relationship between the emission angle and the center wavelength of the spectrum can be obtained by combining relevant data. This can be used as prior data in the actual detection process, and can be used for the calibration of the refractive index and thickness testing of unknown materials.

[0050] Preferably, the θ-λ relationship is optimized using an alternating fitting correction method.

[0051] By adopting the above technical solution, when calibrating the emission angle-wavelength relationship of the spectral confocal sensor, the alternating fitting correction method is used to optimize the θ-λ relationship, which makes the calibration of the relationship more accurate, makes the residual approach zero, improves the accuracy of the emission angle-wavelength calibration, and thus improves the accuracy of the entire spectral confocal sensor's precise thickness measurement calibration and evaluation.

[0052] A spectral confocal thickness measurement system is used to perform the above method, and the controller stores:

[0053] The displacement-wavelength relationship S(λ) calibrated by the method;

[0054] The emission angle-wavelength relationship θ(λ) calibrated by the method;

[0055] When used for real-time thickness calculation, the refractive index n(λ) of the measured material and the formula are combined:

[0056] .

[0057] By adopting the above technical solution, the controller of the spectral confocal thickness measurement system stores the calibrated displacement-wavelength relationship and emission angle-wavelength relationship. Combined with the refractive index of the measured material and the formula, real-time thickness calculation can be achieved.

[0058] In summary, this application includes at least one of the following beneficial technical effects:

[0059] 1. By preprocessing spectral data, a normalized signal is obtained to eliminate light source modulation, dark noise, optical path crosstalk noise, and backscattering noise, thus accurately reflecting the light source signal after passing through the system.

[0060] 2. R1, R2, K1, and K2 are used to evaluate the spectral flatness and stability of the 420-680nm band, which can effectively characterize the spectral jitter of the light source and determine whether the spectrum has a trapezoidal flat top parallel to the horizontal axis.

[0061] 3. Calculate the expected root mean square and standard deviation root mean square of the displacement deviation of the wavelength extraction algorithm using the odd-even grouping method, evaluate the accuracy and precision of the algorithm, and determine the system's limiting resolution;

[0062] 4. Perform polynomial fitting and iterative correction on the odd-even grouped data to make the residuals approach zero and improve the accuracy of displacement-wavelength calibration;

[0063] 5. The reflection angle-wavelength relationship is reflected by continuously moving the reference plate perpendicular to the optical axis of the probe, and this relationship is optimized to provide prior data for the refractive index calibration and thickness testing of unknown materials. Attached Figure Description

[0064] Figure 1 This is a block diagram illustrating the working principle of the probe and PD controller in the embodiments of this application.

[0065] Figure 2 This is a PD spectrum diagram in an embodiment of this application.

[0066] Figure 3 This is a spectral curve diagram from an embodiment of this application.

[0067] Figure 4 This is a schematic diagram of the displacement and wavelength test in an embodiment of this application.

[0068] Figure 5 This is a graph of S1(λ) and S2(λ) in the embodiments of this application.

[0069] Figure 6 This is a schematic diagram illustrating the determination of the emission angle θ and wavelength λ in this application. Detailed Implementation

[0070] The following is in conjunction with the appendix Figures 1-6 This application will be described in further detail.

[0071] Example 1:

[0072] This application provides a novel method for accurate thickness measurement calibration and evaluation of a single spectral confocal sensor, comprising a spectral data preprocessing module, a spectral stability evaluation module, an algorithm error verification module, a displacement-wavelength calibration module, and an emission angle-wavelength calibration module. These modules cooperate to complete the accurate thickness measurement calibration and evaluation of the spectral confocal sensor, effectively improving the precision and accuracy of thickness measurement.

[0073] Specifically, the spectral data preprocessing module includes a light source input signal acquisition device, a blackboard blocking device, and a normalization calculation unit.

[0074] The light source input signal acquisition device is used to acquire the light source information from the input probe. It is approximated as having a spectral signal distribution similar to the original light source, used as a normalized reference signal, expressed as λ0 = K * λ + dark0. Here, K is the modulation coefficient of the system after inputting the probe to the original spectral information, and dark0 is dark noise. Alternatively, the light source input signal acquisition device can also be a high-precision spectral acquisition instrument, capable of acquiring light source information more accurately. The blackboard device can acquire dark0 (containing dark current and system noise) by turning off the light source blackboard; and acquire dark1 (containing dark current and inter-lens reflection noise) by turning on the light source blackboard.

[0075] The blackboard device can be a flat, light-absorbing black board or a specially designed light-absorbing device. The normalization calculation unit performs normalization processing on the acquired λ0, dark0, λ1, and dark1 to obtain the normalized signal (λ1 - dark1) / (λ0 - dark0). This normalized signal can eliminate the effects of light source modulation, dark noise, optical path crosstalk noise, and light source intensity modulation on the spectral signal, while removing the more significant dark noise, crosstalk noise, and lens backscattering noise.

[0076] Furthermore, the normalization calculation unit can be a software algorithm module installed in the controller or a stand-alone computing chip. The combination of these components effectively eliminates the modulation effect of light source intensity on the spectral signal, while also removing noise interference, thus improving signal quality and fidelity.

[0077] Reference Figure 1 and Figure 2 Specifically, the spectral stability evaluation module includes a PD detector and an evaluation calculation unit.

[0078] The PD detector is divided into 2×2 regions, each with a filter, and data is collected independently for each region. The center wavelengths of the filters are 420nm, 450nm, 600nm, and 680nm, with a bandwidth of 5nm, corresponding to intensities I1, I2, I3, and I4, respectively. The filters can be made of optical glass with special coatings to achieve the required center wavelength and bandwidth. Alternatively, they can be filters manufactured using micro-nano optical structures, offering higher precision and stability.

[0079] Furthermore, the light output from the optical fiber is homogenized and then irradiated onto the PD. The evaluation calculation unit calculates R1 = I2 / I1, K1 = I3 / I2, R2 = I3 / I4, and K2 = I4 / I1 based on the collected intensity values. R1, R2, K1, and K2 are used to evaluate the spectral flatness and stability of the 420-680nm band, where R1 characterizes the spectral jitter of the blue light excitation source.

[0080] Reference Figure 3 When K1 = K2 and R1 = R2, the spectrum in the 420-680nm band is determined to have a trapezoidal flat top parallel to the horizontal axis; the change in R1 reflects the power drift in the 450nm / 420nm band caused by the fluctuation of the blue light source current.

[0081] Meanwhile, the evaluation calculation unit can be a software program installed in the controller or a dedicated data analysis device. In this way, the stability of the spectrum can be accurately evaluated, providing a reliable basis for subsequent measurements.

[0082] Specifically, the algorithm error verification module includes a piezoelectric ceramic stepper displacement device, a data acquisition unit, and an error calculation unit.

[0083] Piezoelectric ceramic stepper displacement devices can move objects according to a set step distance, where the step distance is STEP. Piezoelectric ceramics are characterized by high precision and high response speed, enabling precise control of displacement. Linear motors can also be used as stepper displacement devices, offering greater stroke and load capacity.

[0084] Furthermore, taking the process of the piezoelectric ceramic moving from 0 to 100 μm as an example, the data acquisition unit records the output spectral confocal center wavelength signal of the controller at intervals of step size STEP (e.g., 1 μm, which can be reduced according to the required resolution of the system). N frames are acquired at each position, resulting in a total of 101 displacement data points and N×101 sets of spectral confocal signal data. The data acquisition unit can be a high-speed data acquisition card or a controller with acquisition capabilities.

[0085] Furthermore, the error calculation unit constructs a straight line using odd-numbered index points to calculate the displacement deviation μ1 of even-numbered points, and constructs a straight line using even-numbered index points to calculate the displacement deviation μ2 of odd-numbered points; the mean deviation μ0 = (μ1 + μ2) / 101, where 101 is the total number of displacement points; the standard deviation Δ = RMS(μ0) = RMS(Δ) = .

[0086] The accuracy of the algorithm is evaluated by RMS(μ0), the precision is evaluated by RMS(Δ), and the minimum step size STEP(min) is used as the system's limiting resolution. The limiting resolution STEP(min) is defined as the minimum step size STEP when RMS(μ0) and RMS(Δ) reach a set threshold.

[0087] The error calculation unit can be a high-performance computer or an embedded computing module. Through the collaborative work of these components, the error of the wavelength extraction algorithm can be scientifically verified, and the system's limiting resolution can be determined.

[0088] Reference Figure 4 and Figure 5 Specifically, the displacement-wavelength calibration module includes a data grouping unit, a polynomial fitting unit, and an iterative correction unit.

[0089] The data grouping unit groups the collected data into odd and even groups. This unit can be software, which determines the grouping based on the data's index. The polynomial fitting unit performs 5-term polynomial fitting on the odd and even grouped data to obtain S1(λ) and S2(λ). This unit can be specialized fitting software or a fitting algorithm built into the controller. The iterative correction unit performs iterative correction.

[0090] (a) Initially set the correction term C = 0, and calculate the fitting error:

[0091] ;

[0092] Among them, Si(λ)=k1⋅S1(λ)+k2⋅S2(λ);

[0093] (b) Calculate the residuals according to the Malikoff principle:

[0094] Divide all N data points into two equal segments according to their index order.

[0095] First half: Residual Δ(j) from index j = 1 to ⌊N / 2⌋;

[0096] Second half: The residual Δ(j) from index j = ⌊N / 2⌋ + 1 to N;

[0097] Calculate the difference between the sums of the residuals:

[0098] ;

[0099] (c) If |δ| > ε (ε is the threshold for approaching zero), update the correction term C = a⋅λ + b, recalculate Si(λ) = k1⋅S1(λ) + k2⋅S2(λ) + C and iteratively execute steps (a)-(b) until |δ| ≤ ε.

[0100] The iterative correction unit can be an intelligent optimization algorithm that continuously adjusts parameters to achieve the best fit. Through these operations, a more accurate displacement-wavelength calibration relationship can be obtained.

[0101] Reference Figure 6 Specifically, the emission angle-wavelength calibration module includes a reference plate, a position adjustment device, and an angle inversion calculation unit.

[0102] The thickness H of the reference plate must be greater than the thickness corresponding to the minimum recognizable spectral interval of the sensor, and the Sellmeier of the reference plate must be known. The reference plate can be a high-precision optical flat plate with good flatness and uniformity. The position adjustment device is used to continuously move the reference plate perpendicular to the optical axis of the probe; the perpendicularity is determined by the maximum value of the spectral signal intensity on the upper surface. The position adjustment device can be a high-precision displacement platform capable of precisely controlling the position of the reference plate.

[0103] In addition, the angle inversion calculation unit collects the wavelengths λ1 and λ2 of the upper / lower surface and the displacement difference S(λ2) - S(λ1) at different positions, and combines them with the Sellmeier equation of the reference plate, n(λ2), from the formula

[0104] ;

[0105] The emission angle θ(λ2) is inverted, and the θ-λ relationship is obtained through alternating fitting correction (i.e., the iterative correction method in the displacement-wavelength calibration module mentioned above). The angle inversion calculation unit can be a powerful computing device capable of performing angle inversion and relationship fitting quickly and accurately. Through the cooperation of these components, the relationship between the probe's emission angle and the spectral center wavelength can be calibrated.

[0106] The implementation principle of this embodiment is as follows: Through the collaborative work of multiple modules, this embodiment preprocesses the spectral data of the spectral confocal sensor, thereby improving signal quality; evaluates spectral stability to provide a reliable basis for subsequent measurements; verifies algorithm errors and determines the system's limiting resolution; and calibrates the displacement-wavelength relationship and the emission angle-wavelength relationship, effectively improving the accuracy and precision of thickness measurement by the spectral confocal sensor. This overcomes the problems of inaccurate raw data, poor data quality, and poor robustness of fitting methods in traditional technologies, providing more reliable technical support for industrial measurement and inspection.

[0107] Example 2:

[0108] This application provides a spectral confocal thickness measurement system, which includes a controller and a data storage unit.

[0109] The controller is used to execute the method as described in Example 1, processing, calibrating, and evaluating spectral data. The controller can be a high-performance industrial computer or a specially designed embedded controller. The data storage unit is used to store the displacement-wavelength relationship S(λ) and emission angle-wavelength relationship θ(λ) calibrated by the method, and, when used for real-time thickness calculation, to combine the refractive index n(λ) of the measured material with the formula...

[0110] ;

[0111] The calculations are performed. The data storage unit can be a high-capacity hard drive or a high-speed flash memory.

[0112] The implementation principle of this embodiment is as follows: the controller executes a precise thickness measurement calibration and evaluation method, and stores the key relationships obtained from the calibration in the data storage unit. During real-time thickness calculation, the thickness of the object being measured can be accurately calculated by combining the refractive index of the material being measured, thereby improving the accuracy and reliability of the thickness measurement system and meeting the demand for high-precision thickness measurement in industrial measurement and inspection.

[0113] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor, characterized in that, Includes the following steps: Spectral data preprocessing: The system dark current noise signal dark0 is obtained under no light conditions, and the dark1 signal containing dark current and lens reflection noise is obtained when the probe is blocked by the blackboard when the light source is on. The light source input signal λ0=K*λ+dark0, where λ is the light source spectrum of the input probe, and the detection signal λ1=S0+dark1 are normalized to obtain the normalized signal (λ1-dark1) / (λ0-dark0), which eliminates light source modulation, dark noise and optical path crosstalk noise. Spectral stability evaluation: Using a 2×2 region filter on the PD detector, with corresponding center wavelengths of 420nm, 450nm, 600nm, and 680nm, a bandwidth of 5nm, and intensity values ​​of I1-I4, we calculated R1=I2 / I1, K1=I3 / I2, R2=I3 / I4, and K2=I4 / I1. R1, R2, K1, and K2 were used to evaluate the spectral flatness and stability of the 420-680nm band, where R1 characterizes the spectral jitter of the blue light excitation source. Algorithm error verification: N×M sets of spectral data were collected by piezoelectric ceramic step displacement, corresponding to the step size STEP. Adjacent displacement straight lines were constructed using the odd-even grouping method. The expected root mean square (RMS) (μ0) and standard deviation (RMS) (Δ) of the displacement deviation of the wavelength extraction algorithm were calculated. The accuracy of the algorithm was evaluated by RMS (μ0), the precision was evaluated by RMS (Δ), and the minimum step size STEP (min) was used as the system's limiting resolution. Displacement-wavelength calibration: Polynomial fitting is performed on the odd and even grouped data to obtain S1(λ) and S2(λ). By iteratively correcting the weights k1, k2 and the linear term C=a*λ+b, the residual δ=∑∆(j) approaches zero, and the final calibration relationship S(λ)=k1*S1(λ)+k2*S2(λ)+C is output. Emission angle-wavelength calibration: The reference plate with known thickness H is continuously moved perpendicular to the optical axis of the probe, and the wavelengths λ1 and λ2 of the upper / lower surface and the displacement difference S(λ2)-S(λ1) at different positions are collected. The emission angle θ(λ2) is inverted by combining the Sellmeier equation n(λ2), and the θ-λ relationship is obtained by alternating fitting correction.

2. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 1, characterized in that, In the spectral data preprocessing: dark0 is obtained by blocking the blackboard with the light source off, and includes dark current and system noise; dark1 is obtained by turning on the light source and blocking the blackboard, and includes dark current and reflection noise between lenses; Normalized signals eliminate light source intensity modulation and backscattering noise.

3. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 1, characterized in that, In the evaluation of spectral stability: When K1=K2 and R1=R2, the spectrum in the 420-680nm band is determined to have a trapezoidal flat top parallel to the horizontal axis. The change in R1 reflects the power drift in the 450nm / 420nm band caused by fluctuations in the blue light source current.

4. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 1, characterized in that, The algorithm error verification includes: Calculate the displacement deviation μ1 of even-numbered points by constructing a straight line using odd-numbered index points, and calculate the displacement deviation μ2 of odd-numbered points by constructing a straight line using even-numbered index points; The mean deviation μ0 = (μ1 + μ2) / N, where N is the total number of displacement points; Standard deviation Δ= ; RMS(μ0)= ,RMS(Δ)= 。 5. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 4, characterized in that, The limiting resolution STEP(min) is defined as the minimum step size STEP when RMS(μ0) and RMS(Δ) reach a set threshold.

6. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 1, characterized in that, The polynomial fitting uses five polynomials, with odd and even groups of data, to perform polynomial fitting to obtain S1(λ) and S2(λ). The iterative correction in the displacement-wavelength calibration includes: (a) Initially set the correction term C=0, and calculate the fitting error: ； Among them, Si(λ)=k1⋅S1(λ)+k2⋅S2(λ); (b) Calculate the residuals according to the Malikoff principle: Divide all N data points into two equal segments in index order. First half: Residuals Δ(j) from index j=1 to ⌊N / 2⌋; Second half: The residual Δ(j) from index j = ⌊N / 2⌋ + 1 to N; Calculate the difference between the sums of the residuals: ； (c) If |δ|>ε, where ε is the threshold for approaching zero, update the correction term C=a⋅λ+b, recalculate Si(λ)=k1⋅S1(λ)+k2⋅S2(λ)+C and iteratively execute steps (a)-(b) until |δ|≤ε.

7. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 1, characterized in that, In the emission angle-wavelength calibration: The thickness H of the reference plate needs to be greater than the thickness corresponding to the minimum identifiable spectral interval of the sensor, and the Sellmeier equation of the reference plate is known. Verticality is determined by the maximum intensity of the spectral signal on the upper surface; The emission angle θ(λ2) is obtained by inversion from the formula: 。 8. The novel method for precise thickness measurement calibration and evaluation using a single spectral confocal sensor according to claim 7, characterized in that, The θ-λ relationship is optimized using the alternating fitting correction method of claim 6.

9. A spectral confocal thickness measurement system, used to perform the method as described in claim 1, characterized in that, Stored in the controller: The displacement-wavelength relationship S(λ) calibrated by the method; The emission angle-wavelength relationship θ(λ) calibrated by the method; When used for real-time thickness calculation, the refractive index n(λ) of the measured material and the formula are combined: .

Images