A design method of wide-spectrum coding filter for improving spectral reconstruction capability of computational reconfigurable spectrometer

Abstract

The application discloses a design method of a wide-spectrum coding filter for improving the spectrum reconstruction capability of a computational reconstruction spectrometer, thin-film transmission matrix is adopted to generate coded spectrum, and a particle swarm algorithm is used to optimize the structure parameters of the wide-spectrum coding filter, so that spectrum data with small correlation between spectrums and large difference within spectrums are obtained, and finally the film system structure of the wide-spectrum coding filter is designed. The application can automatically optimize and design the wide-spectrum coding filter with small correlation and high complexity, thereby reducing the difficulty of film system design, enhancing the coding capability of the coding filter, and improving the reconstruction precision of the computational reconstruction spectrometer.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of spectral analysis, in particular to a design method of a wide-spectrum coded filter for improving the spectral reconstruction capability of a computational reconstruction spectrometer, and the designed filter combined with the spectrometer is expected to be applied in many fields such as environmental monitoring, chemical analysis, industrial and agricultural production, safety inspection and anti-counterfeiting. BACKGROUND

[0002] The development of spectrometers has evolved from simple light splitting devices to complex analytical instruments. Modern spectrometers are not only light splitting and measuring devices, but also high-precision and multifunctional analytical platforms. They play an increasingly important role in scientific research, industrial detection, biomedicine and other fields. At the same time, in addition to the development of hardware, the progress of related computer algorithms has also promoted the development of a new type of spectrometer - computational reconstruction spectrometer. The unique feature of this type of spectrometer is that they combine computational optics and machine learning algorithms, and can reconstruct complete spectral information from a small amount of measurement data according to calibrated detection elements (such as detectors integrated with wide-spectrum encoders). These technologies greatly expand the application range of spectrometers and improve the efficiency and speed of spectral data acquisition and processing. In the present application, the wide-spectrum encoder selects a thin-film filter. Optical thin films are the most widely used filter materials at present. When light is incident on an optical thin film material, the reflected light between different film layers interferes with the incident light, thereby producing different transmittance and reflectance for different wavelengths of light. Therefore, its spectral response can be controlled by the thickness and refractive index of each thin film. In theory, optical thin films can produce any shape of continuous spectral response. The number of filters in the computational reconstruction spectrometer represents the number of encodings required for the measured light. Under the condition of determining the number of encodings, the encoding is realized by designing the same number of filters as the number of encodings, and the film system structures of these filters are different. Or only one kind of film system filter is used, and the polarization state and incident angle of the incident light are changed to realize the encoding. One polarization of light passing through the filter at one angle represents one encoding.

[0003] The regular film system filter designed by using conventional thin film design software may have problems such as high correlation coefficient, insufficient number of peaks and valleys, and insufficient complexity of the spectrum, which cannot meet the encoding accuracy of the encoding layer in the inversion network of the computational reconstruction spectrometer, greatly restricting the spectral reconstruction capability of the spectrometer. If the correlation degree of the designed transmittance spectrum data is smaller, the redundancy of the encoding of the incident light is smaller, and the detector can obtain more light information, which is beneficial to improve the reconstruction accuracy of the computational reconstruction spectrometer.

[0004] The Chinese patent document with publication number 2022115413251 discloses a spectral reconstruction method and a spectrometer based on a multi-angle polarized light broadband filter. However, it does not involve the design method of the key encoding element, the encoding filter. Since the angle rotation reconstruction spectrometer only uses a single filter for encoding operation, the encoding ability has an important influence on the reconstruction effect. The encoding spectrum of the randomly generated thickness wide spectrum encoding filter at different angles often has the characteristics of high correlation and low internal difference, so that better encoding and decoding effect cannot be obtained. In order to reduce the difficulty of designing the encoding filter film system of the angle reconstruction spectrometer and improve the encoding precision of the broadband filter based on optical thin film, the present patent proposes a wide spectrum encoding filter design method based on particle swarm algorithm. Finally, the correlation between the several spectral transmittances corresponding to the conventional random design film system structure and the correlation between the several spectral transmittances automatically optimized by the algorithm are compared to verify the feasibility of the present invention. SUMMARY

[0005] To overcome the shortcomings of the prior art, the present application provides a design method of a wide spectrum encoding filter for improving the spectral reconstruction ability of a computational reconstruction spectrometer. The film transmission matrix is used to generate encoding spectrum, and the particle swarm algorithm is used to optimize the structure parameters of the wide spectrum encoding filter to obtain spectral data with smaller spectral correlation and larger internal difference. Finally, the film system structure of the wide spectrum encoding filter is designed. The technical solution of the present application is as follows:

[0006] A design method of a wide spectrum encoding filter for improving the spectral reconstruction ability of a computational reconstruction spectrometer, which uses an encoding filter to output several groups of transmission spectrum for encoding the measured spectrum according to different incident light polarization states and different incident angles. The design method comprises the following steps:

[0007] Step 1: Determine the initial film system structure of the wide spectrum encoding filter to be designed, including the optimization design range of the substrate material, high refractive index material, low refractive index material, thin film layer number, material of each thin film, and thickness of each thin film, and the incident angle of polarized light under P and S polarization states.

[0008] Step 2: Calculate the spectral data values of the initial film system structure at different incident angles under P and S polarization states, i.e. the kth data value of the ith spectrum, where i=1, 2, 3…N, k=1, 2, 3…n, N is the total number of spectra under different polarization light and different incident angles, and n is the total number of data values included in each spectrum.

[0009] Step 3: Construct the evaluation function E of the spectrum, which is as follows:

[0010]

[0011] wherein A, B are weight coefficients, is the correlation coefficient r between the i-th and j-th spectrum ij is the average value of the absolute value of the difference between the i-th and j-th spectrum, X ik represents the k-th data value of the i-th spectrum, represents the average value of the n data values of the i-th spectrum, X jk represents the k-th data value of the j-th spectrum, represents the average value of the n data values of the j-th spectrum, is the combination number formula, i = 1, 2, 3 … N, j = 1, 2, 3 … N, and i≠j; is the average value of the inverse of the difference between the N spectra V i , wherein, represents the difference between the individual data values of the i-th spectrum;

[0012] Step four, set the number of populations, acceleration constant, weight value range, minimum error gradient tolerance, particle initial position, iteration number, and obtain the optimized spectral transmittance data and thickness of each layer of thin film by using the particle swarm algorithm.

[0013] Preferably, in the step one, the base material includes silicon, germanium, sulfide, selenide, fluoride or oxide; the high refractive index material includes zinc sulfide, zinc selenide, zirconium dioxide, hafnium dioxide, titanium dioxide, silicon or germanium, etc. single material or composite material; the low refractive index material includes silicon dioxide, aluminum oxide or metal fluoride; the film system structure includes Sub / (HL) n / Air, Sub / (LH) n / Air, Sub / (HLHL) n / Air, Sub / (H(LH) m L) n / Air, etc., wherein m and n are both integers, Sub is the substrate, H is the high refractive index film layer, L is the low refractive index film layer, and Air is air.

[0014] Preferably, in the step one, the number of thin film layers varies in the range of 2-60 layers, the thickness of each layer of film varies in the range of 5-3000 nanometers, and the incident angle of polarized light varies in the range of 0°-85°.

[0015] Preferably, in the step two, the thin film transmission matrix method is used.

[0016] Preferably, in the step two, the spectrum is the transmittance spectrum or the reflectance spectrum.

[0017] Preferably, in the step three, the value range of A and B is 0-1000.

[0018] Preferably, in the fourth step, the population size is in the range of 20-50, the acceleration constant is in the range of 2-4, the weight is in the range of 0.4-0.9, and the minimum error gradient tolerance is in the range of 1x10 -100 -1x10 -3 The initial position of the particle can be assigned a determined initial value or a random value, and the number of algorithm update iterations is 500-5000.

[0019] In addition to designing a single film system filter, the present application is also applicable to designing filters with multiple different film system structures by changing the polarization state and incident angle of the incident light to achieve coding.

[0020] Determine the initial film system structure of a plurality of wide-spectrum coded filters to be designed, including the base material, high refractive index material, low refractive index material, the variation range of the number of film layers and the thickness of each film layer. The plurality of filters can use the same base material, high refractive index material, and low refractive index material, only change the number of film layers and the thickness of each film layer to achieve the difference of the film system structure, or keep the number of film layers and the thickness of each film layer the same, use different base materials, high refractive index materials, and low refractive index materials to achieve the difference of the film system structure, or use different base materials, high refractive index materials, low refractive index materials, film layer numbers, and film thicknesses. The algorithm optimization method is the same as that in steps three and four.

[0021] Compared with the prior art, the present application has the following advantages:

[0022] (1) The present application can automatically optimize and design a plurality of transmission spectra with small correlation and high complexity, and at the same time give the film system structure of the coded filter, reducing the difficulty of film system design.

[0023] (2) The thin film filter designed by the present application as a coder of a spectrometer can improve the reconstruction accuracy of the computational reconstruction spectrometer. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 . The flowchart of the design method of the wide-spectrum coded filter for improving the spectral reconstruction capability of the computational reconstruction spectrometer proposed in the present application.

[0025] Figure 2 . The transmittance spectrum data of P-polarized light at different angles in the infrared waveband in the embodiment of the present application.

[0026] Figure 3 . The transmittance spectrum data of S-polarized light at different angles in the infrared waveband in the embodiment of the present application.

[0027] Figure 4The schematic diagram of the middle infrared waveband thin film filter structure designed by the embodiment of the application is shown in the figure. In the figure, 201 is sapphire, 202 is silicon dioxide (SiO2), and 203 is silicon (Si).

[0028] Figure 5. (a) P-polarized light transmittance spectral curve of the first set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles; (b) S-polarized light transmittance spectral curve of the first set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles.

[0029] Figure 6. (a) P-polarized light transmittance spectral curve of the second set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles; (b) S-polarized light transmittance spectral curve of the second set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles.

[0030] Figure 7. (a) P-polarized light transmittance spectral curve of the third set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles; (b) S-polarized light transmittance spectral curve of the third set of wide-spectrum filter structure parameters designed by the conventional random film system method at different angles. DETAILED DESCRIPTION

[0031] The application will be further described in detail below in combination with the drawings and examples, but should not limit the protection scope of the application.

[0032] Example 1

[0033] Step one, the initial film system structure is determined to be Sub / (Si / SiO2) 5 / Air, the base material is sapphire, and the thickness of each film varies in the range of 20-1500 nm. The P-polarized light is set at nine incident angles of 30°, 35°, 40°, 45°, 50°, 55°, 60°, 65°, and 70°, and the S-polarized light is set at nine incident angles of 30°, 35°, 40°, 45°, 50°, 55°, 60°, 65°, and 70°;

[0034] Step two, N=18 transmittance spectra at different polarized light and different incident angles are calculated by using the thin film transmission matrix method, each spectrum includes n=371 data values, the wavelength range is 3000 nm-4850 nm, and the data value interval is 5 nm;

[0035] Step three, the evaluation function E of the spectrum is constructed,

[0036]

[0037] In the formula, A and B are weight coefficients, and the value range is 0-1000, r is the correlation coefficient between the ith spectrum and the jth spectrumij the average value of the absolute value of X ik represents the kth data value of the ith spectrum, represents the average value of n data values of the ith spectrum, X jk represents the kth data value of the jth spectrum, represents the average value of n data values of the jth spectrum, is a combination number formula, i = 1, 2, 3…N, j = 1, 2, 3…N, and i≠j; is the average value of the inverse of the internal difference V i of the N spectra, wherein, represents the difference between each data value of the ith spectrum;

[0038] Step four, set the number of population to 20, the acceleration constant to 2.1, the weight value range to 0.6-0.9, the minimum error gradient tolerance to 1×10 -90 , and randomly assign the initial value of the particle;

[0039] Step five, set the number of algorithm update iterations to 1000, optimize the spectral data by using the particle swarm algorithm, and design the filter film system structure with lower correlation coefficient and greater internal difference of transmittance spectrum, as shown in Table 1. Figure 2 and Figure 3 respectively give the transmittance spectrum data at different incident angles under P and S two polarization states, and the correlation degree of the 18 spectral data is 0.218.

[0040] Table 1 Structure parameters of wide spectrum encoding filter designed in the embodiment

[0041] Number of film layers Film layer material Physical thickness / nm 1 SiO2 1215.70 2 Si 238.18 3 SiO2 1443.81 4 Si 1052.60 5 SiO2 1459.60 6 Si 1244.18 7 SiO2 1500.00 8 Si 1140.33 9 SiO2 1235.44 10 Si 226.63

[0042] The thin film filter can be designed by using the determined base material and the wide spectrum encoding filter structure parameters in Table 1, and the film system structure is shown in Figure 4 .

[0043] Table 2, Table 3, Table 4 are three groups of wide spectrum filter structure parameters designed by conventional random film system method, the thickness of each layer of film ranges from 200-1500nm randomly. Figure 5(a), Figure 5(b), Figure 6(a), Figure 6(b), Figure 7(a), Figure 7(b) are P, S polarized light different angle transmittance spectrum data under the above three groups of wide spectrum filter structure parameters, each group has 18 pieces of transmittance spectrum data, the correlation degree is 0.575, 0.49, 0.459 respectively. The correlation degree comparison results of the four groups of transmittance spectrum data are shown in Table 5, the correlation degree of the transmittance spectrum data optimized by the particle swarm algorithm is lower than that of the conventional random film system method. The results show that the application can automatically optimize and design a wide spectrum coding filter with low correlation degree and high complexity, thereby reducing the difficulty of film system design, enhancing the coding ability of the coding filter, and improving the reconstruction accuracy of the calculated reconstruction spectrometer.

[0044] Table 2 is a first group of wide spectrum coding filter structure parameters designed by a conventional random film system method

[0045] Number of film layers Film layer material Physical thickness / nm 1 SiO2 1259 2 Si 1378 3 SiO2 365 4 Si 1388 5 SiO2 1022 6 Si 326 7 SiO2 562 8 Si 911 9 SiO2 1445 10 Si 1455

[0046] Table 3 is a second group of wide spectrum coding filter structure parameters designed by a conventional random film system method

[0047] Number of film layers Film layer material Physical thickness / nm 1 SiO2 405 2 Si 1462 3 SiO2 1445 4 Si 831 5 SiO2 1241 6 Si 384 7 SiO2 748 8 Si 1391 9 SiO2 1230 10 Si 1448

[0048] Table 4 is a third group of wide spectrum coding filter structure parameters designed by a conventional random film system method

[0049] Group Relevance Random method first group 1 SiO2 1053 2 Random method second group 246 3 SiO2 1304 4 Random method third group 1415 5 SiO2 1083 6 Algorithm optimization group 1185 7 SiO2 1166 8 ​ 710 9 SiO2 1052 10 ​ 422

[0050] Table 5 is a comparison of the correlation degree of the transmittance spectrum data of the coding filter

[0051] ​ ​ ​ 0.575 ​ 0.49 ​ 0.459 ​ 0.218

[0052] Example 2

[0053] The base material is silicon (Si), the high refractive index material is germanium (Ge), and the low refractive index material is zinc sulfide (ZnS), and the filter film system structure is set as Sub / (ZnS / Ge) 4 / Air.

[0054] Example 3

[0055] The base material is sapphire, the high refractive index material is silicon (Si) and germanium (Ge), and the low refractive index material is silicon dioxide (SiO2) and zinc sulfide (ZnS), and the filter film system structure is set as Sub / (Ge / (Si / SiO2) 2 / ZnS) 2 / Air.

[0056] Other embodiments of the disclosure will be apparent to those skilled in the art from consideration of the specification and practice of the features disclosed herein. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the disclosure being indicated by the following claims.

Claims

1. A design method for a broadband coded filter to improve the spectral reconstruction capability of a computational reconstruction spectrometer, comprising using a single coded filter to output several sets of transmission spectra based on different incident light polarization states and different incident angles for encoding the spectrum to be measured, characterized in that, The design method includes the following steps: Step 1: Determine the initial film structure of the broadband coded filter to be designed, including the substrate material, high refractive index material, low refractive index material, number of thin film layers, material of each thin film, and the optimal design range of the thickness of each thin film, as well as the incident angle of polarized light under P and S polarization states. Step 2: Calculate the spectral data values ​​of the initial film structure under different incident angles in P and S polarization states, i.e., the k-th data value of the i-th spectrum, where i = 1, 2, 3...N, k = 1, 2, 3...n, N is the total number of spectra under different polarizations and different incident angles, and n is the total number of data values ​​included in each spectrum. Step 3: Construct the evaluation function E for the spectrum, as shown in the following formula: In the formula, A and B are weighting coefficients. The correlation coefficient r between the i-th and j-th spectra ij The average of the absolute values, X ik This represents the k-th data value of the i-th spectrum. X represents the average of the n data values ​​of the i-th spectrum. jk This represents the k-th data value of the j-th spectrum. This represents the average of the n data values ​​of the j-th spectrum. Here is the formula for combinations, where i = 1, 2, 3...N, j = 1, 2, 3...N, and i ≠ j; For the internal differences V of N spectra i The reciprocal of the average, in, This represents the differences between the data values ​​of the i-th spectrum; Step 4: Set the population size, acceleration constant, weight range, minimum error gradient tolerance, initial particle position, and number of iterations. Use the particle swarm optimization algorithm to obtain the optimized spectral data and the thickness of each film layer. In step four, the population size ranges from 20 to 50, the acceleration constant ranges from 2 to 4, the weight ranges from 0.4 to 0.9, and the minimum error gradient tolerance ranges from 1 × 10⁻⁶. -100 -1×10 -3 The initial position of the particle can be assigned a fixed initial value or a random value, and the number of algorithm update iterations is 500-5000.

2. The design method for a broadband coded filter to improve the spectral reconstruction capability of a computational reconstruction spectrometer according to claim 1, characterized in that, In step one, the substrate material includes silicon, germanium, sulfides, selenides, fluorides, or oxides; the high refractive index material includes zinc sulfide, zinc selenide, zirconium dioxide, hafnium dioxide, titanium dioxide, silicon, or germanium, either as a single material or a composite material; the low refractive index material includes silicon dioxide, aluminum oxide, or metallic fluorides; and the film structure includes Sub / (HL). n / Air、Sub / (LH) n / Air、Sub / (HLHL) n / Air or Sub / (H(LH)) m L) n / Air, where m and n are integers, Sub is the substrate, H is the high refractive index film, L is the low refractive index film, and Air is air.

3. The design method according to claim 1, characterized in that, In step one, the number of thin film layers varies from 2 to 60, the thickness of each layer varies from 5 to 3000 nanometers, and the incident angle of polarized light ranges from 0° to 85°.

4. The design method according to claim 1, characterized in that, In step two, the thin-film transfer matrix method is used.

5. The design method according to claim 1, characterized in that, In step two, the spectrum is either a transmittance spectrum or a reflectance spectrum.

6. The design method according to claim 1, characterized in that, In step three, the values ​​of A and B range from 0 to 1000.

Images