Wave band selection method for dynamically regulating and controlling micro-nano structure calculation reconstruction spectrometer

Abstract

The invention provides a wave band selection method for dynamically regulating and controlling a micro-nano structure calculation reconstruction spectrometer, which comprises the following steps of: introducing a fractional order derivative into a multi-target sine / cosine algorithm comprising non-dominated sorting and crowding degree calculation, and solving a target function; the waveband selection problem of the dynamically-controllable micro-nano structure calculation reconstruction spectrometer is converted into the multi-target optimization problem without constraint conditions to be solved; that is to say, according to the band selection method based on the fractional order multi-target sine and cosine algorithm, the optimal result can be quickly obtained, and the balance is obtained between the average correlation coefficient and the band range, so that the multi-target sine and cosine algorithm can jump out of the local optimal solution to obtain the global optimal solution; finally, the reconstruction precision of the micro-nano structure calculation reconstruction spectrometer capable of being dynamically regulated and controlled is improved, the most appropriate working wave band is selected for the micro-nano structure calculation reconstruction spectrometer capable of being dynamically regulated and controlled, and meanwhile sufficient data guarantee is provided for follow-up spectrum reconstruction.

Description

Technical Field

[0001] The present invention belongs to the technical field of computational reconstruction spectrometers, and in particular relates to a band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer. Background Art

[0002] From basic scientific research to industrial applications, spectrometers are indispensable optical analysis tools in many fields such as astronomical photonics, earth science, precision agriculture, clinical and chemical analysis. However, due to factors such as their large size and precise optical path, traditional spectrometers are limited in their application in some scenarios that require miniaturized and micro-miniaturized spectrometers. Currently, computational reconstruction spectrometers are a common solution to this problem. Moreover, with the continuous advancement of basic research and engineering technology, dynamically controlled materials such as graphene, black phosphorus, liquid crystal, and micro-nano deformable materials have further reduced the size of spectrometers, making sub-wavelength and sub-pixel spectrometers possible.

[0003] However, the contradiction between the spectral resolution and the types of micro-nano materials in common computational reconstruction spectrometers is also prominent in computational reconstruction spectrometers for micro-nano dynamically controllable materials. On the one hand, the control ability of micro-nano dynamically controllable materials is limited, which makes the correlation coefficient of spectral responses between different control states large, which will cause the subsequent calculation results to be completely distorted or even have multiple solutions; on the other hand, although there will be spectral regions with small correlation coefficients locally, when this spectral region is small, that is, the detection range of the spectrometer is too small, this spectrometer will lose its application value. In addition, due to the complexity and changeability of spectral data and the large amount of data, calculating the average correlation coefficient of spectral responses under different control states is a serious non-convex problem, and it is difficult to assign weights in contradiction with the band width. Therefore, a non-weighted multi-objective non-convex optimization algorithm with simple principles, fast operation speed, and strong ability to jump out of local optimal values ​​is required. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a band selection method for a dynamically adjustable micro-nano structure computational reconstruction spectrometer, which can improve the reconstruction accuracy of the micro-nano dynamically adjustable structure computational reconstruction spectrometer.

[0005] A band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer comprises the following steps:

[0006] S1: Randomly generate N groups of different candidate bands;

[0007] S2: Perform non-dominated sorting and crowding calculation on the average correlation coefficient and band width corresponding to each candidate band, and obtain the non-dominated rank, average correlation coefficient crowding and band width crowding corresponding to each candidate band;

[0008] S3: Select the optimal candidate band for the current iteration according to the set screening rules;

[0009] S4: Substitute the optimal candidate band and each candidate band into the fractional-order sine-cosine algorithm respectively, so as to update each candidate band;

[0010] S5: re-execute steps S2 to S4 using the updated candidate bands until the set upper limit of the number of iterations is reached, and the optimal candidate band obtained in the last iteration is the band finally selected.

[0011] Furthermore, N groups of different candidate bands are randomly generated as follows:

[0012]

[0013] in, The starting wavelength points of each group of candidate bands are respectively The ending wavelength points of each group of candidate bands.

[0014] Furthermore, the method for obtaining the non-dominated rank corresponding to each candidate band is as follows:

[0015] S21: setting the initial values ​​of the non-dominated parameter nq and the non-dominated rank NDR corresponding to each candidate band to 0 respectively;

[0016] S22: All candidate bands are compared in pairs in sequence. In each pairwise comparison, one of the candidate bands is determined. Is the average correlation coefficient and band width greater than that of another candidate band? The average correlation coefficient and band width of The non-dominated parameter nq of the candidate band is incremented by 1 based on the current value. The non-dominated parameter nq remains unchanged at the current value. If it is not, the candidate band The non-dominated parameter nq of remains unchanged at its current value;

[0017] S23: After all candidate bands have completed the pairwise comparison, the non-dominated ranks NDR of all candidate bands with nq=0 are set to the number of rounds of the current pairwise comparison;

[0018] S24: Reset the non-dominated parameters nq of all candidate bands with nq≠0 to 0 and then re-execute steps S22 to S23 until the non-dominated ranks NDR of all candidate bands are set.

[0019] Furthermore, the method for obtaining the average correlation coefficient crowdedness and band width crowdedness corresponding to each candidate band is as follows:

[0020]

[0021] Where k = 1, 2, and is the average correlation coefficient crowding corresponding to the i-th candidate band, is the band width crowding degree corresponding to the i-th candidate band, is the average correlation coefficient corresponding to the i+1th candidate band, is the band width corresponding to the i+1th candidate band, is the average correlation coefficient corresponding to the i-1th candidate band, i = 1, 2, ..., N, is the band width corresponding to the i-1th candidate band, is the maximum average correlation coefficient among all candidate bands, is the maximum band width among all candidate bands, is the minimum average correlation coefficient among all candidate bands, It is the minimum band width among all candidate bands.

[0022] Furthermore, the setting screening rule in step S3 is:

[0023] When the minimum non-dominated rank corresponds to only one candidate band, this candidate band is the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands, the candidate band with the largest average correlation coefficient crowdedness and band width crowdedness is selected as the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands and there is no candidate band with both the maximum average correlation coefficient crowdedness and the maximum band width crowdedness, the candidate band with the largest average value of the average correlation coefficient crowdedness and the band width crowdedness is selected as the optimal candidate band.

[0024] Furthermore, in step S4, the method of updating each candidate band using the fractional-order sine-cosine algorithm is as follows:

[0025]

[0026] Among them, r1, r2, r3, and r4 are all set coefficients in the fractional-order sine-cosine algorithm, and r2∈[0,2π], r3∈[0,5], r4∈[0,1], a is a fixed constant, T is the upper limit of the number of iterations, Γ(ε+1), Γ(g+1), Γ(ε-g+1) are Gamma functions with values ​​of ε+1, g+1, ε-g+1 respectively, ε is the fractional order of the fractional sine cosine algorithm, the traversal number g=1,2,...,e, e is the cutoff order of the fractional sine cosine algorithm, B(t) is the optimal candidate band obtained at the tth iteration, x i (t) is the i-th candidate band at the t-th iteration before updating, x i(t+1-g) is the i-th candidate band at the t+1-g-th iteration, x i (t+1) is the i-th candidate band at the t-th iteration after update.

[0027] Furthermore, the calculation method of the average correlation coefficient corresponding to any candidate band is:

[0028] Changing the bias voltage of the dynamically adjustable micro-nanostructure, wherein the bias voltage is different and the spectral response corresponding to each dynamically adjustable micro-nanostructure is different;

[0029] The optical signal of the current candidate wavelength band is sequentially coupled with the spectral response of the dynamically adjustable micro-nanostructure under different bias voltages to obtain coupled optical signals under different bias voltages;

[0030] The average value of the coupled optical signal under different bias voltages is the average correlation coefficient corresponding to the current candidate band.

[0031] Furthermore, the dynamically adjustable micro-nano structure includes a deformation layer, a support layer and a base layer; wherein the deformation layer and the base layer serve as a positive electrode and a negative electrode connected to a bias voltage, respectively.

[0032] Beneficial effects:

[0033] The present invention provides a band selection method for a dynamically adjustable micro-nano structure computational reconstruction spectrometer, introduces fractional-order derivatives into a multi-objective sine / cosine algorithm including non-dominated sorting and crowding calculation, solves the objective function, and transforms the band selection problem of the dynamically adjustable micro-nano structure computational reconstruction spectrometer into an unconstrained multi-objective optimization problem for solution; that is, the band selection method based on the fractional-order multi-objective sine-cosine algorithm of the present invention can quickly obtain the optimal result, and achieve a trade-off between the average correlation coefficient and the band range, so that the multi-objective sine-cosine algorithm can jump out of the local optimal solution and obtain the global optimal solution, and finally improve the reconstruction accuracy of the micro-nano dynamically adjustable structure computational reconstruction spectrometer, select the most suitable working band for the dynamically adjustable micro-nano structure computational reconstruction spectrometer, and provide sufficient data guarantee for subsequent spectral reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 A flow chart of the method for selecting a band of a dynamically adjustable computationally reconstructed spectrometer provided by the present invention;

[0035] Figure 2 The micro-nano dynamically controllable structure provided by the present invention;

[0036] Figure 3 A schematic diagram of selecting the average correlation coefficient provided by the present invention;

[0037] Figure 4 This is the reconstruction result of the computational reconstruction spectrometer based on the micro-nano dynamically controllable structure provided by the present invention. DETAILED DESCRIPTION

[0038] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0039] Figure 1 The flowchart of the method for selecting the band of a dynamically adjustable computational reconstruction spectrometer based on a fractional-order multi-objective sine-cosine algorithm provided by the present invention is shown below:

[0040] S1: Randomly generate N groups of different candidate bands:

[0041]

[0042] in, The starting wavelength points of each group of candidate bands are respectively The ending wavelength points of each group of candidate bands.

[0043] S2: Perform non-dominated sorting and crowding calculation on the average correlation coefficient and band width corresponding to each candidate band, and obtain the non-dominated rank, average correlation coefficient crowding and band width crowding corresponding to each candidate band;

[0044] Furthermore, the calculation method of the average correlation coefficient corresponding to any candidate band is:

[0045] The bias voltage of the dynamically adjustable micro-nano structure is changed, wherein the spectral response corresponding to each dynamically adjustable micro-nano structure is different when the bias voltage is different; the optical signal of the current candidate band is sequentially coupled with the spectral response of the dynamically adjustable micro-nano structure under different bias voltages to obtain coupled optical signals under different bias voltages; the average value of the coupled optical signals under different bias voltages is the average correlation coefficient corresponding to the current candidate band F1=f[s1(X),s2(X),s3(X)…s M (X)], where M is the number of bias voltages, s1(X)~s M (X) is the spectral response.

[0046] It should be noted that if Figure 2 As shown, the dynamically adjustable micro-nano structure includes a deformation layer 201, a support layer 202 and a base layer 203; wherein the deformation layer and the base layer serve as a positive electrode and a negative electrode connected to a bias voltage, respectively.

[0047] At the same time, the band width corresponding to each candidate band n=1,2,…,N.

[0048] Furthermore, the method for obtaining the non-dominated rank corresponding to each candidate band is as follows:

[0049] S21: setting the initial values ​​of the non-dominated parameter nq and the non-dominated rank NDR corresponding to each candidate band to 0 respectively;

[0050] S22: All candidate bands are compared in pairs in sequence. In each pairwise comparison, one of the candidate bands is determined. Is the average correlation coefficient and band width greater than that of another candidate band? The average correlation coefficient and band width of The non-dominated parameter nq of the candidate band is incremented by 1 based on the current value. The non-dominated parameter nq remains unchanged at the current value. If it is not, the candidate band The non-dominated parameter nq of remains unchanged at its current value;

[0051] S23: After all candidate bands have completed the pairwise comparison, the non-dominated ranks NDR of all candidate bands with nq=0 are set to the number of rounds of the current pairwise comparison;

[0052] S24: Reset the non-dominated parameters nq of all candidate bands with nq≠0 to 0 and then re-execute steps S22 to S23 until the non-dominated ranks NDR of all candidate bands are set.

[0053] Furthermore, the method for obtaining the average correlation coefficient crowdedness and band width crowdedness corresponding to each candidate band is as follows:

[0054]

[0055] Where k = 1, 2, and is the average correlation coefficient crowding corresponding to the i-th candidate band, is the band width crowding degree corresponding to the i-th candidate band, is the average correlation coefficient corresponding to the i+1th candidate band, is the band width corresponding to the i+1th candidate band, is the average correlation coefficient corresponding to the i-1th candidate band, i = 1, 2, ..., N, is the band width corresponding to the i-1th candidate band, is the maximum average correlation coefficient among all candidate bands, is the maximum band width among all candidate bands, is the minimum average correlation coefficient among all candidate bands, It is the minimum band width among all candidate bands.

[0056] It should be noted that, in the boundary case of i=1orN, the corresponding congestion degree is directly set to the set maximum value.

[0057] S3: Select the optimal candidate band for the current iteration according to the set screening rules;

[0058] Set the filtering rules as:

[0059] When the minimum non-dominated rank corresponds to only one candidate band, this candidate band is the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands, the candidate band with the largest average correlation coefficient crowdedness and band width crowdedness is selected as the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands and there is no candidate band with both the maximum average correlation coefficient crowdedness and the maximum band width crowdedness, the candidate band with the largest average value of the average correlation coefficient crowdedness and the band width crowdedness is selected as the optimal candidate band.

[0060] That is to say, after the non-dominated sorting and crowding calculation are performed, each candidate band contains a non-dominated rank and two crowdings; when there is only one candidate band with the smallest non-dominated rank, the candidate band is the optimal band; when there are many smallest non-dominated ranks, the candidate band with the largest crowding is selected as the optimal band; if the above method cannot select the optimal band, then the average crowding is calculated, and then the band with the largest average crowding is selected as the candidate band.

[0061] S4: Substitute the optimal candidate band and each candidate band into the fractional-order sine-cosine algorithm respectively, so as to update each candidate band;

[0062] Specifically, the method of updating each candidate band using the fractional-order sine-cosine algorithm is as follows:

[0063]

[0064] Among them, r1, r2, r3, and r4 are all set coefficients in the fractional-order sine-cosine algorithm, and r2∈[0,2π], r3∈[0,5], r4∈[0,1], a is a fixed constant, T is the upper limit of the number of iterations, Γ(ε+1), Γ(g+1), Γ(ε-g+1) are Gamma functions with values ​​of ε+1, g+1, ε-g+1 respectively, ε is the fractional order of the fractional sine cosine algorithm, the traversal number g=1,2,...,e, e is the cutoff order of the fractional sine cosine algorithm, B(t) is the optimal candidate band obtained at the tth iteration, x i (t) is the i-th candidate band at the t-th iteration before updating, x i(t+1-g) is the i-th candidate band at the t+1-g-th iteration, x i (t+1) is the i-th candidate band at the t-th iteration after update.

[0065] S5: re-execute steps S2 to S4 using the updated candidate bands until the set upper limit of the number of iterations is reached, and the optimal candidate band obtained in the last iteration is the band finally selected.

[0066] Furthermore, in order to verify the performance of the band selection method of the micro-nano dynamic adjustable structure calculation reconstruction spectrometer based on the fractional-order multi-objective sine-cosine algorithm provided by the present invention, the following Figure 2 The micro-nanostructure shown is used as the basic unit of the computational reconstruction spectrometer. Figure 2 In the figure, 201 can be a two-dimensional material such as graphene, black phosphorus, halogen perovskite, etc., and can also be a metal such as gold, silver, aluminum, etc. that has been etched as a deformation layer. For the above two-dimensional material 201, it can be a multilayer film structure or a single-layer film structure. For the deformation layer 201, it is a support layer that reserves a deformation space for deformation after being etched. 203 is the base material of the micro-nano material. Specifically, in this embodiment, a deformable structure is selected as the basic unit of the computational reconstruction spectrometer.

[0067] Firstly, the performance of the fractional-order multi-objective sine-cosine algorithm is verified. Figure 3 The traversal method is given Figure 2 The relationship between the average correlation coefficient of the material and the band. The fractional-order multi-objective sine-cosine algorithm provided by the present invention can obtain the same results as the traversal. Regarding the running time, the running time of the traversal method on an Intel(R) Core(TM) i5-10300H computer is 2h37min, and the running time of the fractional-order multi-objective sine-cosine algorithm provided by the present invention on the same computing device is 56min.

[0068] Furthermore, Figure 4 The accuracy diagram of the fractional-order multi-objective sine-cosine algorithm provided by the present invention for spectrum reconstruction is given; the present invention uses a three-layer multi-layer perceptron as the spectrum reconstruction algorithm, and the number of neurons in each layer is 50, 250, and 600 respectively, which can be obtained from Figure 4 It can be seen that the spectrum reconstructed using the method provided by the present invention can achieve high-precision reconstruction of the spectrum.

[0069] In summary, the present invention provides a band selection method based on a fractional-order sine-cosine optimization algorithm, by judging the average correlation coefficient and bandwidth of the spectral response (reflection / transmission spectrum) of a dynamically adjustable micro-nanostructure under all control states in different bands as the objective function of the method; considering that the objective function is a non-convex function, the present invention introduces fractional-order derivatives into a multi-objective sine / cosine algorithm including non-dominated sorting and crowding calculation, and solves the objective function, thereby converting the band selection problem of a dynamically adjustable micro-nanostructure computational reconstruction spectrometer into an unconstrained multi-objective optimization problem for solution; the method provided by the present invention can not only efficiently select the most suitable working band for a dynamically adjustable micro-nanostructure computational reconstruction spectrometer, but also provide sufficient data guarantee for subsequent spectral reconstruction.

[0070] Of course, the present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art may certainly make various corresponding changes and modifications based on the present invention, but these corresponding changes and modifications should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. A band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer, characterized in that: The following steps are involved: S1: Randomly generate N groups of different candidate bands; S2: Perform non-dominated sorting and crowding calculation on the average correlation coefficient and band width corresponding to each candidate band, and obtain the non-dominated rank, average correlation coefficient crowding and band width crowding corresponding to each candidate band; S3: Select the optimal candidate band for the current iteration according to the set screening rules; S4: Substitute the optimal candidate band and each candidate band into the fractional-order sine-cosine algorithm respectively, so as to update each candidate band; S5: re-execute steps S2 to S4 using the updated candidate bands until the set upper limit of the number of iterations is reached, and the optimal candidate band obtained in the last iteration is the band finally selected.

2. A band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer as claimed in claim 1, characterized in that: Randomly generate N different groups of candidate bands as follows: in, The starting wavelength points of each group of candidate bands are respectively The ending wavelength points of each group of candidate bands.

3. A band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer as claimed in claim 2, characterized in that: The method for obtaining the non-dominated rank corresponding to each candidate band is: S21: setting the initial values ​​of the non-dominated parameter nq and the non-dominated rank NDR corresponding to each candidate band to 0 respectively; S22: All candidate bands are compared in pairs in sequence. In each pairwise comparison, one of the candidate bands is determined. Is the average correlation coefficient and band width greater than that of another candidate band? The average correlation coefficient and band width of The non-dominated parameter nq of the candidate band is incremented by 1 based on the current value. The non-dominated parameter nq remains unchanged at the current value. If it is not, the candidate band The non-dominated parameter nq of remains unchanged at its current value; S23: After all candidate bands have completed the pairwise comparison, the non-dominated ranks NDR of all candidate bands with nq=0 are set to the number of rounds of the current pairwise comparison; S24: Reset the non-dominated parameters nq of all candidate bands with nq≠0 to 0 and then re-execute steps S22 to S23 until the non-dominated ranks NDR of all candidate bands are set.

4. The band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer according to claim 1, characterized in that: The method for obtaining the average correlation coefficient crowdedness and band width crowdedness corresponding to each candidate band is as follows: Where k = 1, 2, and is the average correlation coefficient crowding corresponding to the i-th candidate band, is the band width crowding degree corresponding to the i-th candidate band, is the average correlation coefficient corresponding to the i+1th candidate band, is the band width corresponding to the i+1th candidate band, is the average correlation coefficient corresponding to the i-1th candidate band, i = 1, 2, ..., N, is the band width corresponding to the i-1th candidate band, is the maximum average correlation coefficient among all candidate bands, is the maximum band width among all candidate bands, is the minimum average correlation coefficient among all candidate bands, It is the minimum band width among all candidate bands.

5. The band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer according to claim 1, characterized in that: The setting screening rules in step S3 are: When the minimum non-dominated rank corresponds to only one candidate band, this candidate band is the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands, the candidate band with the largest average correlation coefficient crowdedness and band width crowdedness is selected as the optimal candidate band; when the minimum non-dominated rank corresponds to more than two candidate bands and there is no candidate band with both the maximum average correlation coefficient crowdedness and the maximum band width crowdedness, the candidate band with the largest average value of the average correlation coefficient crowdedness and the band width crowdedness is selected as the optimal candidate band.

6. The band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer according to claim 1, characterized in that: The method of updating each candidate band using the fractional-order sine-cosine algorithm in step S4 is: Among them, r1, r2, r3, and r4 are all set coefficients in the fractional-order sine-cosine algorithm, and r2∈[0,2π], r3∈[0,5], r4∈[0,1], a is a fixed constant, T is the upper limit of the number of iterations, Γ(ε+1), Γ(g+1), Γ(ε-g+1) are Gamma functions with values ​​of ε+1, g+1, ε-g+1 respectively, ε is the fractional order of the fractional sine cosine algorithm, the traversal number g=1,2,...,e, e is the cutoff order of the fractional sine cosine algorithm, B(t) is the optimal candidate band obtained at the tth iteration, x i (t) is the i-th candidate band at the t-th iteration before updating, x i (t+1-g) is the i-th candidate band at the t+1-g-th iteration, x i (t+1) is the i-th candidate band at the t-th iteration after update.

7. The band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer according to claim 1, characterized in that: The calculation method of the average correlation coefficient corresponding to any candidate band is: Changing the bias voltage of the dynamically adjustable micro-nanostructure, wherein the spectral response corresponding to each dynamically adjustable micro-nanostructure is different when the bias voltage is different; The optical signal of the current candidate wavelength band is sequentially coupled with the spectral response of the dynamically adjustable micro-nanostructure under different bias voltages to obtain coupled optical signals under different bias voltages; The average value of the coupled optical signal under different bias voltages is the average correlation coefficient corresponding to the current candidate band.

8. The band selection method for a dynamically controllable micro-nanostructure computational reconstruction spectrometer according to claim 7, characterized in that: The dynamically adjustable micro-nano structure comprises a deformation layer, a support layer and a base layer; wherein the deformation layer and the base layer serve as a positive electrode and a negative electrode connected to a bias voltage, respectively.

Images