Structural Design Method and System of Fourier Transform Spectrometer Based on Deep Learning

Abstract

The present disclosure provides a structural design method and system for a Fourier transform spectrometer based on deep learning, which relates to the technical field of spectrometer design. It includes decomposing and designing rules for the optical path structure of an interferometer to realize the parameterization of the optical structure of the interferometer; parameterizing actual design requirements and the optical structure parameters of multiple common interferometers to realize the configuration of training samples required for the initial structure derivation model; deriving a new interferometer structure according to the trained structure derivation model; evaluating all interferometer structures and screening out interferometers with better imaging quality; putting the interferometers with better imaging quality into the training samples, retraining and deriving the structure derivation model, and finally designing an interferometer optical structure that meets the requirements. The present disclosure can quickly give a solution according to the spectrometer design requirements and save the design cost.

Description

Technical Field

[0001] The present disclosure relates to the technical field of spectrometer design, and particularly to a structural design method and system of a Fourier transform spectrometer based on deep learning. Background Art

[0002] The statements in this part merely provide background technical information related to the present disclosure, and do not necessarily constitute prior art.

[0003] Spectrum is a fundamental property of matter and an important means for studying atomic energy levels, material structures, chemical compositions, and determining the elemental compositions of substances. Spectrometers have the advantages of fast analysis speed, high precision, non-destructive detection, etc., and are widely used in many fields such as chemistry, environment, materials, and life sciences. Spectrometers can be divided into filter-type spectrometers, diffraction grating spectrometers, Fourier transform spectrometers, etc. Compared with the other two types of spectrometers, Fourier transform spectrometers have the advantages of multi-channel, large radiation throughput, and low stray light. With the emergence of the fast Fourier transform algorithm and the rapid development of computers, Fourier transform spectrometers have gradually become the focus of current spectrometer research. Fourier transform spectrometers use interferometers to generate interference fringes and obtain the spectral information of the target by using the Fourier transform relationship between the interferogram and the spectrum.

[0004] Fourier transform spectrometers can be divided into time-modulated spectrometers and space-modulated spectrometers. At present, the research on time-modulated spectrometers is relatively mature and has a high spectral resolution, and many scholars have studied it. The time-modulated Fourier transform spectrometer takes the Michelson interferometer as the core component. The incident light beam is split into two beams by a beam splitter: a reflected beam and a transmitted beam; the reflected beam is reflected by a static mirror and transmitted by the beam splitter to reach the focusing mirror; the transmitted beam is reflected by a moving mirror and reflected by the beam splitter to reach the focusing mirror; the two beams are focused on the detector to form an image and present interference fringes, and then the spectral information is obtained by using the Fourier transform.

[0005] The existing solutions for improving the design of spectrometers under different requirements mainly involve changing their optical structures, such as rotating interferometers, double moving mirror interferometers, etc., but the above changes all require a certain amount of design time. In addition, or improving the data design of the interferometer, and realizing the improvement of the interferometer through data calculation, which are all by calculating the positions and tilting angles of each optical device. However, due to different requirements for the size, resolution, and measurement spectral range of the Fourier transform spectrometer in different test environments, it will be more troublesome to calculate the positions and tilting angles of each optical device respectively, and the practicability and real-time performance are not high, and it is impossible to quickly design a suitable spectrometer optical path. Summary of the Invention

[0006] To solve the above problems, the present disclosure proposes a structural design method and system for a Fourier transform spectrometer based on deep learning. By decomposing the optical path structure of the interferometer and designing rules, the parameterization of the optical structure of the interferometer is realized. Based on the structure derivation model, a new interferometer structure is derived, and all the derived interferometer structures are evaluated. The interferometer with better imaging quality is selected, and the structure derivation model is retrained and derived based on the interferometer with better imaging quality. Finally, the optimal interferometer design scheme that meets the requirements is obtained.

[0007] According to some embodiments, the present disclosure adopts the following technical solutions:

[0008] A structural design method for a Fourier transform spectrometer based on deep learning, comprising:

[0009] Obtain the target spectrometer model, and decompose the optical path structure of the interferometer in the target spectrometer;

[0010] According to the design rules, parameterize the decomposed optical path structure, and express the optical path structure of the interferometer in the form of a symbol sequence;

[0011] Input the symbol sequence expression of the optical path structure of the interferometer into the structure derivation model to obtain the optical path structure design scheme of the interferometer. Use the design evaluation index to screen the feasible optical path structure design scheme, and re-enter the feasible optical path structure design scheme into the structure derivation model for derivation to obtain the final optimal optical path structure design scheme;

[0012] Among them, the structure derivation model is a long short-term memory network model. During the training process, the interferometer size and shape, resolution, measurement spectral range, and interferometer parameterization results are used as the input of the long short-term memory network. The loss functions are constructed respectively with the differences between the derivation results and the actual results and the differences between the interferometer sizes and the required sizes as the constraint conditions, so that the long short-term memory network gradually derives the optical path structure of the interferometer.

[0013] Further, obtaining the target spectrometer and decomposing the optical path structure of the interferometer in the target spectrometer includes: decomposing the optical path structure of the interferometer, where the interferometer is composed of a beam splitter and a mirror. In the decomposition result, the beam splitter is represented by its tilt angle. Since the mirror has two attributes, tilt angle and installation distance, and the installation distance will change due to the movement of the mirror, the symbol arrangement order of the mirror is the installation distance, tilt angle, and the installation distance after movement.

[0014] Further, parameterize the structural parameters of the decomposed optical path structure according to the design rules, and express the optical path structure of the interferometer in the form of a symbol sequence, including: converting the optical path structure of the interferometer into a symbol sequence expression, converting the incident light beam into a transmitted light beam and a reflected light beam, parameterizing the optical path structure of the interferometer according to the mirrors passed by the two light beams. Based on the beam splitter, the mirrors passed by the transmitted light beam are arranged on the left side of the beam splitter, and the arrangement order is from right to left according to the order of the mirrors passed by the transmitted light beam. The mirrors passed by the reflected light beam are arranged on the right side of the beam splitter, and the arrangement order is from left to right according to the order of the mirrors passed by the transmitted light beam.

[0015] Further, the structure derivation model is a long short-term memory network model. Train the long short-term memory network model, initialize the long short-term memory network model, set the initial cell state to zero, the input of the training sample is the size and shape of the interferometer, the resolution, the measurement spectral range and the parameterization result of the interferometer. Take the comparison result between the number of mirrors of the interferometer and the input as the output of the long short-term memory network. Arrange the parameterization results of the interferometers in the sample from less to more, and put them into the input and output of the long short-term memory network for training in turn.

[0016] Further, construct the loss function of the long short-term memory network. The loss function consists of two parts. The first part is to calculate the difference between the derived optical structure result and the actual result. The second part is to calculate the difference between the derived interferometer size and the required size. The construction method of the loss function in the second part is as follows:

[0017] According to the required resolution, calculate the maximum optical path difference, which is expressed as the maximum optical path difference being the reciprocal of the required resolution. The maximum optical path difference determines the moving distance of the moving mirror. According to the parameterization result of the derived interferometer, first find the position of the tilt angle of the beam splitter. Arrange the symbol sequence to the left of the tilt angle of the beam splitter from right to left to form sequence 1, and arrange the symbol sequence to the right of the tilt angle of the beam splitter from left to right to form sequence 2, so as to reconstruct the two optical paths of the interferometer and obtain the coordinates of each mirror.

[0018] Further, calculate the position of each mirror in sequence 1 and sequence 2 according to the coordinates of each mirror, and set the total distance difference between the mirrors in the two sequences as the maximum optical path difference to calculate the derived interferometer size. Assume that the length and width of the calculated size are x max and y max respectively, and assume that the length and width of the required size are x need and y need， Therefore, the loss function of the second part is:

[0019] 。

[0020] According to some embodiments, the present disclosure adopts the following technical solutions:

[0021] A Fourier transform spectrometer structure design system based on deep learning, comprising:

[0022] A decomposition module, configured to obtain a target spectrometer model and decompose the interferometer optical path structure in the target spectrometer;

[0023] A parameterization module, configured to parameterize the decomposed optical path structure according to design rules and express the interferometer optical path structure in the form of a symbol sequence;

[0024] A derivation module, configured to input the symbol sequence expression of the interferometer optical path structure into a structure derivation model to obtain an interferometer optical path structure design scheme, screen a feasible optical path structure design scheme using design evaluation indicators, and re-input the feasible optical path structure design scheme into the structure derivation model for derivation to obtain a final optimal optical path structure design scheme;

[0025] Wherein, the structure derivation model is a long short-term memory network model. During the training process, the interferometer size and shape, resolution, measurement spectral range, and the parameterization result of the interferometer are used as the input of the long short-term memory network, and loss functions are respectively constructed with the differences between the derivation result and the actual result and the differences between the interferometer size and the required size as constraint conditions, so that the long short-term memory network gradually derives the interferometer optical path structure.

[0026] According to some embodiments, the present disclosure adopts the following technical solutions:

[0027] A computer program product, comprising a computer program, where when the computer program is executed by a processor, it implements the above-mentioned Fourier transform spectrometer structure design method based on deep learning.

[0028] According to some embodiments, the present disclosure adopts the following technical solutions:

[0029] A non-transitory computer-readable storage medium, which is used to store computer instructions, and when the computer instructions are executed by a processor, it implements the above-mentioned Fourier transform spectrometer structure design method based on deep learning.

[0030] According to some embodiments, the present disclosure adopts the following technical solutions:

[0031] An electronic device includes: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory so that the electronic device executes the method for designing the structure of a Fourier transform spectrometer based on deep learning as described above.

[0032] Compared with the prior art, the beneficial effects of the present disclosure are as follows:

[0033] In the method for designing the structure of a Fourier transform spectrometer based on deep learning of the present disclosure, the optical path structure parameters of the interferometer in the spectrometer are parameterized, and then a long short-term memory network is used to derive various interferometer structures. Then, the formed interferometers are evaluated, and a suitable interferometer is selected and further trained with a long short-term memory network and evaluated. Finally, the optimal design of the spectrometer is achieved. By inputting the size and shape, resolution, and measurement spectral range required by the spectrometer, a Fourier transform spectrometer with a suitable size and good imaging effect can be quickly designed by a computer, saving a large amount of time.

[0034] In the method for designing the structure of a Fourier transform spectrometer based on deep learning of the present disclosure, through the parameterization method, the interferometer structure can be expressed in the form of a symbol sequence, so that the long short-term memory network can effectively identify it; by modifying the loss function, the loss function consists of two parts. The first part is to calculate the difference between the optical structure result derived and the actual result, and the second part is to calculate the difference between the size of the derived interferometer and the required size. Considering the interferometer size into the long short-term memory network enables it to derive an interferometer structure with a better size; through cyclic training, a better interferometer structure can be further derived based on the interferometer structure with a better size. Description of the Drawings

[0035] The specification drawings forming a part of the present disclosure are used to provide a further understanding of the present disclosure. The schematic embodiments of the present disclosure and their descriptions are used to explain the present disclosure and do not constitute an improper limitation to the present disclosure.

[0036] Figure 1 It is the full-automatic design flow chart of the interferometer for the embodiment of the present disclosure;

[0037] Figure 2 It is the parameterization result of the Michelson interferometer for the embodiment of the present disclosure;

[0038] Figure 3 It is the interferometer structure diagram derived for the embodiment of the present disclosure. Detailed Embodiments

[0039] The present disclosure will be further described below in conjunction with the drawings and embodiments.

[0040] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present disclosure. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field to which the present disclosure belongs.

[0041] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present disclosure. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0042] Example 1

[0043] In an embodiment of the present disclosure, a structural design method of a Fourier transform spectrometer based on deep learning is provided. The method parameterizes the internal optical structure of the interferometer to enable the structure derivation model to identify. The structure derivation model is a long short-term memory network model. The feasible solutions are derived using the long short-term memory network. Finally, the optimal interferometer solution required specifically is screened out through evaluation indicators, and the long short-term memory network is retrained with this solution as a training sample to further derive a better interferometer, and finally the optimal spectrometer design solution is obtained. While saving human effort, this method designs a spectrometer with better performance. The method steps include:

[0044] Step 1: Obtain the target spectrometer model and decompose the optical path structure of the interferometer in the target spectrometer;

[0045] Step 2: Parameterize the decomposed optical path structure according to the design rules and express the optical path structure of the interferometer in the form of a symbol sequence;

[0046] Step 3: Input the symbol sequence expression of the optical path structure of the interferometer into the structure derivation model to obtain the optical path structure design solution of the interferometer. The feasible solutions of the optical path structure design are screened by the design evaluation indicators, and the feasible solutions of the optical path structure design are re-input into the structure derivation model for derivation to obtain the final optimal optical path structure design solution;

[0047] Among them, the structure derivation model is a long short-term memory network model. During the training process, the size and shape of the interferometer, the resolution, the measurement spectral range, and the parameterization result of the interferometer are used as the input of the long short-term memory network, and the loss functions are respectively constructed with the differences between the derivation results and the actual results and the differences between the interferometer size and the required size as the constraint conditions, so that the long short-term memory network gradually derives the optical path structure of the interferometer.

[0048] As an embodiment, the present disclosure discloses a structural design method of a Fourier transform spectrometer based on deep learning. Among them, the design requirements of the Fourier transform spectrometer are that the target size is less than 450 mm in both length and width, and the resolution is 0.045 cm -1 , and the wavenumber of the measured spectral range is 350~7500 cm -1 . The overall implementation process of the structural design method of the Fourier transform spectrometer based on deep learning is as Figure 1 shown, and the specific implementation process is as follows:

[0049] Step 1: Obtain the target spectrometer model and decompose the optical path structure in the target spectrometer;

[0050] Specifically, decompose the optical path structure of the interferometer. The interferometer consists of a beam splitter and a mirror. In the decomposition result, the beam splitter is represented by its tilt angle β. Since the mirror has two attributes, the tilt angle α and the installation distance l, and the installation distance l may change due to the movement of the mirror, the symbol arrangement order of the mirror is the installation distance l1, the tilt angle α, and the installation distance l2 after movement, and its mirror expression is l1αl2.

[0051] Step 2: Parametrize the decomposed optical path structure according to the design rules and express the optical path structure of the interferometer in the form of a symbol sequence;

[0052] Specifically, according to the design rules, parameterize the optical path structure of the interferometer. The result of parameterization is to convert the interferometer structure into a symbol sequence expression. The interferometer can convert the incident light beam into a transmitted light beam and a reflected light beam. Parameterize the optical path structure of the interferometer according to the mirrors passed by the two light beams. Based on the beam splitter β, the mirrors (l1αl2) passed by the transmitted light beam are arranged on the left side of β, and the arrangement order is from right to left according to the order of the mirrors passed by the transmitted light beam. The mirrors l1αl2 passed by the reflected light beam are arranged on the right side of β, and the arrangement order is from left to right according to the order of the mirrors passed by the transmitted light beam.

[0053] As an embodiment. Taking the Michelson interferometer as an example, this interferometer consists of a beam splitter, a fixed mirror and a moving mirror. Let the tilt angle of the beam splitter be β; the tilt angle of the fixed mirror be α, and the installation distance be l1; the tilt angle of the moving mirror be α, and the installation distance be l1, and the distance after movement be l2. Then the parameterization result of the Michelson interferometer is l1αl2βl1αl2. Since the fixed mirror does not move, one of the l2 symbols in l1αl2βl1αl2 in the parameterization result is empty. The parameterization result of the Michelson interferometer is as Figure 2 shown. This method can parameterize all translational interferometers.

[0054] Step 3: Input the symbolic sequence expression of the interferometer optical path structure into the structure derivation model to obtain the design scheme of the interferometer optical path structure. Screen the feasible optical path structure design schemes according to the design evaluation indicators, and use the feasible optical path structure design schemes to re-enter the structure derivation model for derivation to obtain the final optimal optical path structure design scheme;

[0055] Specifically, the structure derivation model is a long short-term memory network model. First, train the long short-term memory network model. During the training process, use the interferometer size and shape, resolution, measurement spectral range, and interferometer parameterization results as the input of the long short-term memory network. Construct loss functions respectively with the differences between the derivation results and the actual results and the differences between the interferometer size and the required size as the constraint conditions, so that the long short-term memory network gradually derives the interferometer optical path structure. Specifically, it includes:

[0056] 1) Design training samples, that is, use the interferometer size and shape of 450 mm, resolution of 0.045 cm -1 parameterization and the parameterization results of common interferometers as the input of the training samples; use the parameterization results of common interferometers as the sample output. Among them, the interferometer parameterization results are both used as the input and the output of the training samples, but in terms of specific performance, the two are interferometers of different optical structures. The parameterization results of common interferometers are specifically selected as: l1αl2βl1αl2, l1αl2l1αl2βl1αl2, l1αl2βl1αl2l1αl2, l1αl2l1αl2βl1αl2l1αl2.

[0057] 2) Construct a long short-term memory network model, and its specific output expression h i is:

[0058]

[0059]

[0060] Among them, W and b are the weight matrix and bias to be trained, [ ] represents concatenating two vectors, σ is the sigmoid function, h i-1 is the output value of the i -1th long short-term memory network, x i is the input parameter, C i is the i th unit state of the neural network.

[0061] Among them, the unit state ci The expression is:

[0062]

[0063] Among them, f i is the forget gate of the neural network. The cell state is obtained by element-wise multiplying the previous cell state by the forget gate, then element-wise multiplying the current input state by the input gate, and then adding the two products together.

[0064] 3) Initialize the long short-term memory network model, set the initial cell state c0 to zero, and use the interferometer size, shape, resolution, measurement spectral range, and interferometer parameterization results in the training samples as the input of the long short-term memory network. x i , and use the number of mirrors of the interferometer and the comparison result of the input (the number of mirrors of the interferometer is the same as or one more than the input x i ) as the output of the long short-term memory network. h i . Arrange the common interferometer parameterization results from less to more and put them into the input and output of the long short-term memory network in turn.

[0065] 4) Construct loss functions respectively with the differences between the derivation results and the actual results and the differences between the interferometer sizes and the required sizes as the constraint conditions. The loss function consists of two parts. The first part is to calculate the difference between the derivation result of the network model and the actual result. The second part is to calculate the difference between the derived interferometer size and the required size.

[0066] Furthermore, the construction method of the loss function in the first part is as follows:

[0067]

[0068] Among them, | V | is the total number of interferometers derived by the long short-term memory network, y i = 1 or 0. When the long short-term memory network predicts correctly, it is 1, otherwise it is 0. p i is the difference between the real structure and the derived structure.

[0069] Furthermore, the construction method of the loss function in the second part is as follows:

[0070] First, according to the required resolution, the maximum optical path difference L can be calculated. It is shown that the maximum optical path difference is the reciprocal of the required resolution, specifically 112 mm. The maximum optical path difference determines the moving distance of the moving mirror. According to the parameterized results of the interferometer predicted by the model, first use the computer to find the position of the beam splitter tilt angle β. For the symbol sequence on the left side of β, arrange it from right to left to form sequence 1, and for the symbol sequence on the right side of β, arrange it from left to right to form sequence 2, so as to reconstruct the two optical paths of the interferometer. Set up a coordinate axis at the position of the beam splitter, then the coordinates of each mirror can be expressed as:

[0071]

[0072]

[0073] Among them, l is l1 or l2, and specifically, they are the minimum installation distances of the mirrors. The minimum installation distance is selected as 100 mm, and the movable distance of the mirror is 212 mm; n represents the n th mirror that the light beam has passed through. Therefore, according to the above formula, the positions of each mirror in sequence 1 and sequence 2 can be calculated, and the total distance difference between the two sequences of mirrors is set as the maximum optical path difference L, so as to calculate the deduced interferometer size. Assume that the calculated length and width of the size are x max and y max , and the required length and width of the size are 450 mm and 450 mm respectively. Therefore, the loss function of the second part is:

[0074]

[0075] 5) Train the long short-term memory network model and update the parameters in the network through gradient descent. The update rules are as follows:

[0076]

[0077] Among them, is the training parameter of the long short-term memory network ( W and b ), and the subscripts new and old represent the parameters after network update and the parameters before update respectively. is the learning rate, is the gradient of the loss function with respect to the parameter.

[0078] As an example, a trained long short-term memory network is used to derive a new interferometer structure. The design requirements and the parameterized results of the training samples are input into the trained long short-term memory network, so that the long short-term memory network gradually derives a new interferometer structure. The derived interferometer structures are l1αl2l1αl2l1αl2βl1αl2l1αl2l1αl2, l1αl2l1αl2βl1αl2l1αl2l1αl2l1αl2, l1αl2l1αl2l1αl2βl1αl2l1αl2l1αl2l1αl2.

[0079] As an example, design evaluation indicators are used to screen feasible optical path structure design solutions, including:

[0080] The Zemax software is used to evaluate the derived interferometer structure and the interferometer structure of the training samples. The best interferometer is selected according to the root mean square radius of the interferometer imaging.

[0081] The root mean square radius results of the imaging of all interferometers are shown in the following table:

[0082]

[0083] Among them, the root mean square radius is an index used to describe the beam size. The smaller the root mean square radius, the better the imaging quality. It can be found that the interferometer structure l1αl2l1αl2l1αl2βl1αl2l1αl2l1αl2 has the smallest root mean square radius and the best imaging quality. The specific optical path diagram is as Figure 3 shown.

[0084] Furthermore, the long short-term memory network is retrained using the derived interferometer optical structure, so that the long short-term memory network re-derives a similar new interferometer structure based on this sample. The new interferometer structure is re-evaluated. If the imaging quality of the new interferometer structure is better, then this interferometer is used as a training sample and returned to the training process again until, due to size limitations, the long short-term memory network can no longer form an updated interferometer structure, and the obtained interferometer is used as the final interferometer.

[0085] If the derived interferometer structure l1αl2l1αl2l1αl2βl1αl2l1αl2l1αl2 is put into the training samples of the long short-term memory network model, the long short-term memory network is retrained. It is found during the training process that due to size limitations, the long short-term memory network can no longer form an updated interferometer structure. Therefore, the interferometer with the structure l1αl2l1αl2l1αl2βl1αl2l1αl2l1αl2 is used as the final interferometer. Finally, the corresponding shell and auxiliary equipment are matched to the interferometer, and a complete spectrometer structure is formed to complete the design of the spectrometer.

[0086] Example 2

[0087] In one embodiment of the present disclosure, a structural design system for a Fourier transform spectrometer based on deep learning is provided, including:

[0088] A decomposition module, configured to obtain a target spectrometer model and decompose the interferometer optical path structure in the target spectrometer;

[0089] A parameterization module, configured to parameterize the decomposed optical path structure according to design rules and express the interferometer optical path structure in the form of a symbol sequence;

[0090] A derivation module, configured to input the symbol sequence expression of the interferometer optical path structure into a structure derivation model to obtain an interferometer optical path structure design scheme, screen feasible design schemes of the optical path structure using design evaluation indicators, and re-input the feasible design schemes of the optical path structure into the structure derivation model for derivation to obtain a final optimal optical path structure design scheme;

[0091] Wherein, the structure derivation model is a long short-term memory network model. During the training process, the interferometer size and shape, resolution, measurement spectral range, and the parameterization result of the interferometer are used as the input of the long short-term memory network, and loss functions are respectively constructed with the differences between the derivation result and the actual result and the differences between the interferometer size and the required size as constraint conditions, so that the long short-term memory network gradually derives the interferometer optical path structure.

[0092] Embodiment 3

[0093] A computer program product, including a computer program, which when executed by a processor implements the above-mentioned structural design method for a Fourier transform spectrometer based on deep learning.

[0094] Embodiment 4

[0095] A non-transitory computer-readable storage medium, which is used to store computer instructions, and when the computer instructions are executed by a processor, the above-mentioned structural design method for a Fourier transform spectrometer based on deep learning is implemented.

[0096] Embodiment 5

[0097] An electronic device, including: a processor, a memory, and a computer program; wherein, the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device runs, the processor executes the computer program stored in the memory so that the electronic device executes and implements the above-mentioned structural design method for a Fourier transform spectrometer based on deep learning.

[0098] This disclosure is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to embodiments of the disclosure. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or means for implementing the functions specified in one block or multiple blocks.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, so that the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 or steps for implementing the functions specified in one block or multiple blocks.

[0100] Although the specific embodiments of the present disclosure have been described above in conjunction with the accompanying drawings, this is not a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that, based on the technical solutions of the present disclosure, various modifications or variations that can be made by those skilled in the art without creative efforts are still within the protection scope of the present disclosure.

Claims

1. A Fourier transform spectrometer structure design method based on deep learning, characterized in that: include: Obtain the target spectrometer model and decompose the interferometer optical path structure in the target spectrometer; According to the design rules, the decomposed optical path structure is parameterized and the interferometer optical path structure is expressed in the form of a symbol sequence; The symbolic sequence expression of the interferometer optical path structure is input into the structure deduction model to obtain the interferometer optical path structure design scheme, the design evaluation index is used to screen the feasible optical path structure design scheme, and the feasible optical path structure design scheme is re-input into the structure deduction model for deduction to obtain the final optimal optical path structure design scheme; Among them, the structure derivation model is a long short-term memory network model. During the training process, the interferometer size, shape, resolution, measurement spectrum range and interferometer parameterization results are used as the input of the long short-term memory network. The difference between the derivation result and the actual result and the difference between the interferometer size and the required size are used as constraints to construct loss functions respectively, so that the long short-term memory network gradually derives the interferometer optical path structure. The decomposed optical path structure is parameterized according to the design rules, and the interferometer optical path structure is expressed in the form of a symbol sequence, including: converting the interferometer optical path structure into a symbol sequence expression, converting the incident light beam into a transmitted light beam and a reflected light beam, and parameterizing the interferometer optical path structure according to the reflectors through which the two light beams pass. Based on the beam splitter, the reflectors through which the transmitted light beam passes are arranged on the left side of the beam splitter, and the arrangement order is arranged from right to left according to the order of the reflectors through which the transmitted light beam passes, and the reflectors through which the reflected light beam passes are arranged on the right side of the beam splitter, and the arrangement order is arranged from left to right according to the order of the reflectors through which the transmitted light beam passes.

2. The method for designing a Fourier transform spectrometer structure based on deep learning according to claim 1, characterized in that: Obtain a target spectrometer and decompose the interferometer optical path structure in the target spectrometer, including: decomposing the optical path structure of the interferometer, wherein the interferometer is composed of a beam splitter and a reflector. In the decomposition result, the beam splitter is represented by its tilt angle, and the reflector has two properties, namely, the tilt angle and the installation distance, and the installation distance will change due to the movement of the reflector, that is, the symbols of the reflector are arranged in the order of installation distance, tilt angle and installation distance after movement.

3. The method for designing a Fourier transform spectrometer structure based on deep learning according to claim 1, characterized in that: The structure derivation model is a long short-term memory network model. The long short-term memory network model is trained and initialized, and the initial unit state is set to zero. The input of the training sample is the interferometer size, resolution, measurement spectral range and interferometer parameterization result. The comparison result of the number of reflectors of the interferometer and the input is used as the output of the long short-term memory network. The interferometer parameterization results in the sample are arranged from few to many and are sequentially put into the input and output of the long short-term memory network for training.

4. The method for designing a Fourier transform spectrometer structure based on deep learning according to claim 1, characterized in that: Construct the loss function of the long short-term memory network. The loss function consists of two parts. The first part is the difference between the calculated optical structure result and the actual result. The second part is the difference between the calculated interferometer size and the required size. The construction method of the loss function of the second part is as follows: According to the required resolution, the maximum optical path difference is calculated, which is expressed as the reciprocal of the required resolution. The maximum optical path difference determines the moving distance of the moving mirror. According to the derived interferometer parameterization results, the position of the beam splitter tilt angle is first found, and the symbol sequence on the left side of the beam splitter tilt angle is arranged from right to left to form sequence 1, and the symbol sequence on the right side of the beam splitter tilt angle is arranged from left to right to form sequence 2, thereby reconstructing the two optical paths of the interferometer and obtaining the coordinates of each reflector.

5. The method for designing a Fourier transform spectrometer structure based on deep learning as claimed in claim 4, characterized in that: The position of each reflector in sequence 1 and sequence 2 is calculated according to the coordinates of each reflector, and the total distance difference between the reflectors in the two sequences is set as the maximum optical path difference. The derived interferometer size is calculated, assuming that the length and width of the calculated size are x max and y max , assuming the required dimensions are x need and y need， Therefore, the loss function of the second part is: 。 6. A Fourier transform spectrometer structure design system based on deep learning, based on the Fourier transform spectrometer structure design method based on deep learning as claimed in any one of claims 1 to 5, characterized in that: include: A decomposition module is used to obtain the target spectrometer model and decompose the interferometer optical path structure in the target spectrometer; A parameterization module is used to parameterize the decomposed optical path structure according to the design rules and express the interferometer optical path structure in the form of a symbol sequence; A derivation module is used to input the symbol sequence expression of the interferometer optical path structure into the structure derivation model to obtain the interferometer optical path structure design scheme, design evaluation indicators to screen the feasible schemes of the optical path structure design, and use the feasible schemes of the optical path structure design to re-input the structure derivation model for derivation to obtain the final optimal optical path structure design scheme; Among them, the structure derivation model is a long short-term memory network model. During the training process, the interferometer size and shape, resolution, measurement spectral range, and interferometer parameterization results are used as inputs of the long short-term memory network. The difference between the deduced results and the actual results and the difference between the interferometer size and the required size are used as constraints to construct loss functions, so that the long short-term memory network can gradually derive the interferometer optical path structure.

7. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for designing a Fourier transform spectrometer structure based on deep learning according to any one of claims 1 to 5 is implemented.

8. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the deep learning-based Fourier change spectrometer structure design method as described in any one of claims 1 to 5 is implemented.

9. An electronic device, characterized in that: include: A processor, a memory and a computer program; wherein the processor is connected to the memory, the computer program is stored in the memory, and when the electronic device is running, the processor executes the computer program stored in the memory so that the electronic device executes the Fourier change spectrometer structure design method based on deep learning as described in any one of claims 1-5.

Images