WO2023221588A1 - 光谱重构方法和装置、光谱仪、存储介质和电子设备 - Google Patents

光谱重构方法和装置、光谱仪、存储介质和电子设备 Download PDF

Info

Publication number
WO2023221588A1
WO2023221588A1 PCT/CN2023/078205 CN2023078205W WO2023221588A1 WO 2023221588 A1 WO2023221588 A1 WO 2023221588A1 CN 2023078205 W CN2023078205 W CN 2023078205W WO 2023221588 A1 WO2023221588 A1 WO 2023221588A1
Authority
WO
WIPO (PCT)
Prior art keywords
spectrum
data
current
processed
positive
Prior art date
Application number
PCT/CN2023/078205
Other languages
English (en)
French (fr)
Inventor
王少伟
刘清权
尹知沁
陆卫
Original Assignee
中国科学院上海技术物理研究所
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 中国科学院上海技术物理研究所 filed Critical 中国科学院上海技术物理研究所
Priority to US18/354,510 priority Critical patent/US20230375405A1/en
Publication of WO2023221588A1 publication Critical patent/WO2023221588A1/zh

Links

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
    • G01J3/28—Investigating the spectrum
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/25—Colour; Spectral properties, i.e. comparison of effect of material on the light at two or more different wavelengths or wavelength bands
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/15—Correlation function computation including computation of convolution operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
    • G06F17/10—Complex mathematical operations
    • G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01J—MEASUREMENT OF INTENSITY, VELOCITY, SPECTRAL CONTENT, POLARISATION, PHASE OR PULSE CHARACTERISTICS OF INFRARED, VISIBLE OR ULTRAVIOLET LIGHT; COLORIMETRY; RADIATION PYROMETRY
    • G01J3/00—Spectrometry; Spectrophotometry; Monochromators; Measuring colours
    • G01J3/28—Investigating the spectrum
    • G01J2003/283—Investigating the spectrum computer-interfaced

Definitions

  • the present application relates to the technical field of spectrum analysis, and specifically relates to a spectrum reconstruction method and spectrum reconstruction device, a spectrometer, a computer-readable storage medium and an electronic device.
  • the reconstruction spectrometer includes a hardware part and a software part.
  • the hardware part mainly includes the detector, and the software part mainly includes the spectrum reconstruction algorithm. After the detector detects the response value, the response value is inverted and calculated using the spectral reconstruction method to obtain the reconstructed spectrum.
  • the accuracy of reconstructed spectra obtained by current spectral reconstruction methods is low and cannot meet user requirements.
  • embodiments of the present application provide a spectrum reconstruction method, a spectrum reconstruction device, a spectrometer, a computer-readable storage medium and an electronic device, which solve the problem of poor noise immunity of the spectrum reconstruction method.
  • a spectrum reconstruction method provided by an embodiment of the present application is applied to a spectrometer, including: determining the spectrum data to be processed; using an objective function to convert negative spectrum values in the spectrum data to be processed into positive spectrum values, Obtain positive spectrum data, and use the objective function to perform loss calculation based on the positive spectrum data and response value data to determine the loss value. If the loss value meets the first preset condition or the number of loss calculations meets the second preset condition, the reconstruction is obtained Spectral data, wherein the response value data is obtained by measuring the spectrum to be reconstructed using a spectrometer.
  • an objective function is used to convert negative spectral values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and the objective function is used based on the positive spectrum data and the response value. Loss calculation is performed on the data to determine the loss value.
  • reconstructed spectral data is obtained, including: using the objective function to convert the current spectral data to be processed
  • the neutral and negative spectral values are converted into positive spectral values to obtain the current positive spectral data;
  • the objective function is used to determine the current loss value and the current reconstructed spectral data based on the current positive spectral data and response value data; based on the current The loss value and the current reconstructed spectrum data, adjust the spectrum data to be processed, and iterate the spectrum data to be processed until it is determined that the loss value meets the first preset condition or the number of loss calculations meets the second preset condition, and the reconstructed spectrum is obtained data.
  • an objective function is used to convert negative spectral values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data, including: the objective function is used to The current spectrum data to be processed takes the absolute value, and the negative spectrum values in the current spectrum data to be processed are converted into positive spectrum values to obtain the current positive spectrum data.
  • the objective function before using the objective function to convert negative spectral values in the current spectral data to be processed into positive spectral values to obtain the current positive spectral data, it also includes: determining The signed diagonal matrix of the current spectrum data to be processed; use the objective function to convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data, including: the objective function passes the current The negative spectral values in the current spectral data to be processed are converted into positive spectral values by multiplying the spectral data to be processed and the signed diagonal matrix to obtain the current positive spectral data.
  • determining the spectral data to be processed includes: based on the current reconstructed spectral data obtained in the k-th cycle and the current reconstruction obtained in the (k+1)-th cycle.
  • Spectral data determine the current spectrum data to be processed corresponding to the (k+2)th cycle, where k is a positive integer.
  • based on the current reconstructed spectral data obtained in the kth cycle and The current reconstructed spectrum data obtained in the (k+1)th cycle and the current to-be-processed spectral data corresponding to the (k+2)th cycle are determined, including: calculating the current reconstructed spectrum data obtained in the (k+1)th cycle.
  • the difference between the reconstructed spectrum data and the current reconstructed spectrum data obtained in the kth cycle is used to obtain the difference data; based on the current reconstructed spectrum data, difference data and preset obtained in the (k+1)th cycle Return function to determine the current spectrum data to be processed corresponding to the (k+2)th cycle.
  • an embodiment of the present application provides a spectrum reconstruction device, applied to a spectrometer, including: a determination module configured to determine the spectrum data to be processed; a reconstruction module configured to use an objective function to convert the spectrum data to be processed into Negative spectral values are converted into positive spectral values to obtain positive spectral data, and the objective function is used to perform loss calculation based on the positive spectral data and response value data to determine the loss value. If the loss value meets the first preset condition or the loss calculation The second preset condition is met several times to obtain reconstructed spectrum data, wherein the response value data is obtained by measuring the spectrum to be reconstructed using a spectrometer.
  • an embodiment of the present application provides a spectrometer, including: a detector, used to measure the spectrum to be reconstructed to obtain response value data, and send the response value data to a processor; and the processor, electrically connected to the detector. Connection, used for receiving response value data and executing the spectral reconstruction method of the first aspect to obtain reconstructed spectral data.
  • an embodiment of the present application provides a computer-readable storage medium.
  • the storage medium stores instructions.
  • the instructions When the instructions are executed by a processor of an electronic device, the electronic device can perform the spectral reconstruction mentioned in the first aspect. construction method.
  • an embodiment of the present application provides an electronic device.
  • the electronic device includes: a processor; and a memory for storing computer-executable instructions.
  • a processor configured to execute computer-executable instructions to implement the spectrum reconstruction method mentioned in the first aspect.
  • the spectrum reconstruction method uses an objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and uses the objective function to perform losses based on the positive spectrum data and response value data. Calculate to determine the loss value. If the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset condition, reconstructed spectral data is obtained. Based on the characteristic that the spectrum must be positive, this application uses the objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values, thereby improving the anti-noise ability of the iterative solution process of the objective function, and thereby improving the obtained heavy weights. accuracy of structural spectral data.
  • this application uses an objective function to convert negative spectral values in the spectrum data to be processed into positive spectrum values without adding constraints to the objective function, avoiding turning an unconstrained problem into a constrained problem, and further improving the obtained reconstructed spectrum.
  • the accuracy of the data is improved, and the convergence efficiency of the objective function is improved.
  • Figure 1A shows a schematic diagram of an application scenario of a spectrum reconstruction method provided by an embodiment of the present application.
  • Figure 1B shows a schematic diagram of an application scenario of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 2 shows a schematic flow chart of a spectrum reconstruction method provided by an embodiment of the present application.
  • Figure 3 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 4 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 5 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 6 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 7 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • Figure 8 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • FIG. 9 is a schematic diagram of the measurement principle of a spectrometer provided by an embodiment of the present application.
  • Figure 10 shows a schematic structural diagram of a detector provided by an embodiment of the present application.
  • Figure 11 shows a response curve provided by an embodiment of the present application.
  • Figure 12A shows the reconstructed spectrum and incident spectrum obtained using the integrated filter and Tikhonov regularization method shown in Figure 10.
  • Figure 12B shows the reconstructed spectrum and incident spectrum obtained using the integrated filter shown in Figure 10 and the spectrum reconstruction method of the present application.
  • Figure 13 shows a schematic structural diagram of a detector provided by an embodiment of the present application.
  • Figure 14 shows a response curve provided by another embodiment of the present application.
  • Figure 15A shows the reconstructed spectrum and incident spectrum obtained using the detector shown in Figure 13 and the Tikhonov regularization method.
  • Figure 15B shows the reconstructed spectrum and incident spectrum obtained using the detector shown in Figure 13 and the spectrum reconstruction method of the present application.
  • Figure 16 shows a schematic structural diagram of a spectrum reconstruction device provided by an embodiment of the present application.
  • Figure 17 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • Figure 18 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • Figure 19 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • Figure 20 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • Figure 21 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • Figure 22 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • FIG. 23 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • Figure 1A shows a schematic diagram of an application scenario of a spectrum reconstruction method provided by an embodiment of the present application.
  • the scene shown in FIG. 1A includes the processor 110 in the spectrometer 100 and the spectrum measurement module 120 in the spectrometer 100 .
  • the spectrum measurement module 120 is communicatively connected with the processor 110 .
  • the processor 110 is used to determine the spectrum data to be processed; use an objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data; and use the objective function to calculate the spectrum based on the positive spectrum data and the response. Loss calculation is performed on the value data to determine the loss value.
  • the reconstructed spectrum data is obtained.
  • the spectrum measurement module 120 is used to measure the spectrum to be reconstructed to obtain response value data, and send the response value data to the processor 110 so that the processor 110 can perform the above operations.
  • Figure 1B shows a schematic diagram of an application scenario of a spectrum reconstruction method provided by another embodiment of the present application.
  • the scene shown in FIG. 1B includes spectrometer 100 and server 200.
  • the spectrometer 100 measures the spectrum to be reconstructed to obtain response value data, and sends the response value data to the server 200 .
  • the server 200 receives the response value data and determines the spectrum data to be processed; uses the objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and uses the objective function based on the positive spectrum data and the response value. Loss calculation is performed on the data to determine the loss value. If the loss value meets the first preset condition or the number of loss calculations meets the second preset condition, reconstructed spectrum data is obtained.
  • the server 200 after the server 200 obtains the reconstructed spectrum data, it can also send it to the spectrometer 100, so that the reconstructed spectrum data is displayed on the spectrometer 100.
  • FIG. 2 shows a schematic flow chart of a spectrum reconstruction method provided by an embodiment of the present application. As shown in Figure 2, the spectrum reconstruction method includes the following steps.
  • Step 210 Determine the spectral data to be processed.
  • the spectral data to be processed may be initialized spectral data.
  • the spectrum data to be processed may be spectrum data obtained by initialization methods such as random value initialization, random positive number initialization, Tikhonov initialization, etc.
  • the spectrum data to be processed can be a spectrum matrix, and the elements in the spectrum matrix represent the wavelength of a light wave.
  • Step 220 Use the objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and use the objective function to perform loss calculations based on the positive spectrum data and response value data to determine the loss value. If If the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset condition, the reconstructed spectrum data is obtained.
  • the response value data is obtained by measuring the spectrum to be reconstructed using a spectrometer.
  • the response value data may be a response value matrix.
  • the spectrum to be reconstructed is the spectrum that needs to be measured.
  • the reconstructed spectral data may be spectral data obtained by optimizing the spectral data to be processed based on the response value data using an objective function.
  • the objective function can characterize the functional relationship between the spectral data to be processed and the response value data.
  • the objective function can also convert negative spectral values in the spectral data to be processed into positive spectral values. Negative spectral values in the spectral data to be processed may be noise in the spectral data to be processed.
  • the absolute value of the spectrum value in the spectrum data to be processed can be calculated, thereby converting the negative spectrum value in the spectrum data to be processed into a positive spectrum value to obtain positive spectrum data.
  • the signed diagonal matrix of the spectral data to be processed can be determined first, and then the negative spectral values in the spectral data to be processed are converted into positive spectral values using the signed diagonal matrix to obtain positive spectral data.
  • the objective function may include an absolute value operator, which is used to convert negative spectral values in the spectral data to be processed into positive spectral values. That is, by introducing an absolute value operator in the objective function, the negative spectral values in the spectral data to be processed are converted into positive spectral values.
  • the objective function is used to perform loss calculation based on the positive spectrum data and the response value data to determine the loss value, which may be performed by taking the difference between the response value data and the positive spectrum data to determine the loss value.
  • the first preset condition may be a preset threshold. If the loss value meets the preset threshold, the reconstructed spectral data is obtained.
  • the first preset condition may also be a preset convergence condition of the loss value. If the loss value meets the preset convergence condition of the loss value, the reconstructed spectrum data is obtained.
  • the preset convergence condition of the loss value can be that the loss values calculated for several consecutive times are very close, or even the loss values calculated for the last few times are greater than the loss values calculated for the previous few times.
  • the second preset condition may be a preset number of loss calculations. If the number of loss calculations meets the preset number of loss calculations, the reconstructed spectrum data is obtained.
  • the user can determine the first preset condition and the second preset condition according to actual needs, which are not specifically limited in this application.
  • the spectrum reconstruction method uses an objective function to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and uses the objective function to perform losses based on the positive spectrum data and response value data. Calculate to determine the loss value. If the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset condition, reconstructed spectral data is obtained. Since the negative spectral values in the spectral data to be processed are interference data, the iterative solution process for the objective function has poor noise immunity.
  • this application uses an objective function to convert negative spectral values in the spectral data to be processed into positive spectral values, thereby improving the anti-noise capability of the iterative solution process of the objective function, thereby improving the accuracy of the obtained reconstructed spectral data.
  • this application uses an objective function to convert negative spectral values in the spectrum data to be processed into positive spectrum values without adding constraints to the objective function, avoiding turning an unconstrained problem into a constrained problem, and further improving the obtained reconstructed spectrum. The accuracy of the data is improved, and the convergence efficiency of the objective function is improved.
  • Figure 3 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • the embodiment shown in FIG. 3 is extended based on the embodiment shown in FIG. 2.
  • the following focuses on the differences between the embodiment shown in FIG. 3 and the embodiment shown in FIG. 2, and the similarities will not be described again.
  • an objective function is used to convert the negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and the objective function is used based on the positive spectrum data and response value data.
  • Loss calculation is performed to determine the loss value. If the loss value meets the first preset condition or the number of loss calculations meets the second preset condition, steps for reconstructing the spectral data are obtained, including the following steps.
  • Step 310 Use the objective function to convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data.
  • the current spectrum data to be processed refers to the spectrum data currently being processed.
  • the current positive spectrum data refers to the currently obtained positive spectrum data.
  • the objective function S can be expressed by the following formula (1).
  • I represents the response value data
  • T represents the reconstruction matrix
  • Q represents the preprocessing matrix
  • L represents the regularization matrix
  • the subscript k represents the kth cycle
  • X k represents the current spectral data to be processed corresponding to the kth cycle.
  • ⁇ represents the regularization coefficient
  • m represents the power number
  • n represents the type of norm.
  • T, Q, L, m, and n can all be obtained through pre-calibration or pre-setting, and the values are not specifically limited in this application.
  • the above formula (1) indicates that X k can be adjusted iteratively to make S obtain a smaller value, or even the minimum value.
  • Step 320 Use the objective function to determine the current loss value and the current reconstructed spectrum data based on the current positive spectrum data and response value data.
  • the current loss value can be calculated through the above formula (1).
  • the spectrum data to be processed used to calculate the current loss value is the current reconstructed spectrum data.
  • Step 330 Adjust the spectrum data to be processed based on the current loss value and the current reconstructed spectrum data, and iterate the spectrum data to be processed until it is determined that the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset Conditions to obtain reconstructed spectral data.
  • the spectrum data to be processed can be continuously adjusted, and the spectrum data to be processed can be iterated in a loop until it is determined that the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset condition, and the value used to calculate the current loss value will be used.
  • the current reconstructed spectrum data is determined as the reconstructed spectrum data.
  • the current spectrum data to be processed can be determined as the current reconstructed spectrum data. That is, for each cycle, after the loss value is calculated for this cycle, the spectrum data to be processed for this cycle is determined as the reconstructed spectrum data for this cycle.
  • Figure 4 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • the embodiment shown in FIG. 4 is extended based on the embodiment shown in FIG. 3.
  • the following focuses on the differences between the embodiment shown in FIG. 4 and the embodiment shown in FIG. 3, and the similarities will not be described again.
  • the step of using the objective function to convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data includes the following steps.
  • Step 410 The objective function converts the negative spectrum values in the current spectrum data to be processed into positive spectrum values by taking the absolute value of the current spectrum data to be processed, and obtains the current positive spectrum data.
  • the objective function can convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values by taking the absolute value of the current spectrum data to be processed, Get the current positive spectral data.
  • the method of obtaining the absolute value is simple and accurate, which improves the efficiency of solving the objective function.
  • Figure 5 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application. Based on the embodiment shown in FIG. 3 , the embodiment shown in FIG. 5 is extended. The following focuses on the differences between the embodiment shown in FIG. 5 and the embodiment shown in FIG. 3 , and the similarities will not be described again.
  • the step of using the objective function to convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data includes the following steps.
  • Step 510 Determine the signed diagonal matrix of the current spectral data to be processed.
  • the signed diagonal matrix can be represented by A k .
  • the signed diagonal matrix A k of the spectral data to be processed can be obtained through the following formula (2).
  • the subscript k represents the kth cycle.
  • Step 520 The objective function converts the negative spectrum values in the current spectrum data to be processed into positive spectrum values by multiplying the current spectrum data to be processed and the sign diagonal matrix to obtain the current positive spectrum data.
  • the objective function converts the negative spectral values in the current spectral data to be processed into positive spectral values by multiplying the current spectral data to be processed Spectral data, the method is simple and efficient.
  • Figure 6 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • the embodiment shown in FIG. 6 is extended based on the embodiment shown in FIG. 3.
  • the following focuses on the differences between the embodiment shown in FIG. 6 and the embodiment shown in FIG. 3, and the similarities will not be described again.
  • the step of determining the spectral data to be processed includes the following steps.
  • Step 610 Based on the current reconstructed spectrum data obtained in the k-th cycle and the current reconstructed spectrum data obtained in the (k+1)-th cycle, determine the current spectrum to be processed corresponding to the (k+2)-th cycle. data.
  • the current reconstructed spectral data obtained by the loop and the current reconstructed spectral data obtained by the second loop are used to determine the current to-be-processed spectral data corresponding to the third loop.
  • the current to-be-resolved spectrum data corresponding to the (k+2)-th cycle is determined. Processing the spectral data can make the current spectrum data to be processed in subsequent cycles refer to the current reconstructed spectrum data obtained in the first two cycles, thereby improving the accuracy of the current spectrum data to be processed in subsequent cycles.
  • Figure 7 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • the embodiment shown in Fig. 7 is extended based on the embodiment shown in Fig. 6.
  • the following focuses on the differences between the embodiment shown in Fig. 7 and the embodiment shown in Fig. 6, and the similarities will not be described again.
  • the (k+2)-th ) cycles corresponding to the steps of the current spectral data to be processed, including the following steps.
  • Step 710 Calculate the difference between the current reconstructed spectrum data obtained in the (k+1)th cycle and the current reconstructed spectrum data obtained in the k-th cycle to obtain difference data.
  • Y k+1 can be used to represent the current reconstructed spectrum data obtained in the (k+1)th cycle
  • Y k can be used to represent the current reconstructed spectrum data obtained in the k-th cycle.
  • Step 720 Based on the current reconstructed spectrum data, difference data and preset return function obtained in the (k+1)th cycle, determine the current spectrum data to be processed corresponding to the (k+2)th cycle.
  • Y k+2 can be used to represent the current reconstructed spectrum data obtained in the (k+2)th cycle.
  • the following formula (3) ie, preset return function
  • X k+2 corresponding to the (k+2)th cycle.
  • P k+2 is the intermediate value, which can be calculated by the following formula (4).
  • a k+2 represents the signed diagonal matrix of the current spectral data to be processed corresponding to the (k+2)th cycle, therefore, it can be obtained using the above formula (2).
  • l represents the iteration step size, and grad represents the gradient function.
  • P k+2 ⁇ Y k + ⁇ k+2 (Y k+1 -Y k ) (4)
  • ⁇ k+2 represents the preset return function corresponding to the (k+2)th cycle.
  • the difference data is obtained, and based on the (k+1)th cycle
  • the obtained current reconstructed spectrum data, difference data and preset return function determine the current spectrum data to be processed corresponding to the (k+2)th cycle, optimize the spectrum data to be processed, and then iterate through multiple cycles Improve the accuracy of reconstructed spectral data.
  • Tikhonov regularization assignment can be used to initialize X 1 .
  • the following formula (5) can be used to initialize X 1 .
  • X 1 ( TH T+ ⁇ J) -1 T H I (5)
  • T is the same as T in formula (1), representing the reconstruction matrix, and the superscript H is used to indicate that T H is the conjugate transpose moment of T
  • J represents the identity matrix with the same dimensions as T H T
  • ⁇ is the Tikhonov regularization coefficient. ⁇ can be preset.
  • Figure 8 shows a schematic flow chart of a spectrum reconstruction method provided by another embodiment of the present application.
  • the embodiment shown in Fig. 8 is extended based on the embodiment shown in Fig. 3.
  • the following focuses on the differences between the embodiment shown in Fig. 8 and the embodiment shown in Fig. 3, and the similarities will not be described again.
  • the step of determining the current loss value and the current reconstructed spectrum data based on the current positive spectrum data and response value data using the objective function includes the following steps.
  • Step 810 Use the objective function to preprocess the current positive spectrum data to obtain the current preprocessed spectrum data.
  • the preprocessing matrix Q in the above formula (1) can be used to preprocess the current positive spectrum data to obtain the current preprocessed spectrum data.
  • the preprocessing matrix Q may be a Fourier transform matrix or other matrices capable of discretization, which is not specifically limited in this application.
  • Step 820 Use the objective function to perform loss calculation based on the current preprocessed spectral data and response value data to determine the current loss value and the current reconstructed spectral data.
  • the continuous spectrum can be discretized, thereby reducing the amount of data and improving calculation efficiency.
  • FIG. 9 is a schematic diagram of the measurement principle of a spectrometer provided by an embodiment of the present application.
  • the spectrometer 1000 includes: a detector 1100 and a processor 1200.
  • the detector 1100 is used to measure the spectrum F( ⁇ ) to be reconstructed to obtain response value data (in this embodiment, the response value data may be a response value matrix In ), and send the response value data to the processor 1200.
  • the processor 1200 is electrically connected to the detector 1100 and is used to receive response value data and execute the spectrum reconstruction method in the above embodiment to obtain reconstructed spectrum data.
  • Detector 1100 may include integrated optical filters and photosensitive chips.
  • the photosensitive chip can be a complementary metal oxide semiconductor (Complementary Metal Oxide Semiconductor, CMOS) photosensitive chip or a charge-coupled device (Charge-coupled Device, CCD).
  • CMOS Complementary Metal Oxide Semiconductor
  • CCD Charge-coupled Device
  • FIG. 10 shows a schematic structural diagram of an integrated optical filter provided by an embodiment of the present application.
  • the integrated optical filter includes a substrate 1110 , a titanium dioxide film layer 1120 and a silicon dioxide film layer 1130 .
  • the titanium dioxide film layer 1120 and the silicon dioxide film layer 1130 may be arranged crosswise.
  • the specific layer thickness and specific number of layers of the titanium dioxide film layer 1120 and the silicon dioxide film layer 1130 can be set according to actual needs, and are not specifically limited in this application.
  • the integrated filter can be represented by the film system Sub
  • Sub represents the substrate 1110.
  • H represents a titanium dioxide film layer (TiO 2 ) with a thickness of 78 nm
  • L represents a first silicon dioxide film layer (SiO 2 ) 1121 with a thickness of 115 nm.
  • M represents the second silicon dioxide film layer 1122 except L.
  • the substrate 1110 may be a glass substrate, a quartz substrate, a gemstone substrate, etc., more specifically, it may be a k9 glass substrate in the international standard classification.
  • the appropriate thickness of the second silicon dioxide film layer 1122 can be determined through experiments so that the transmission peak position of the integrated filter increases from 600 nm to 750 nm at intervals of 0.5 nm, that is, the integrated filter has a total of 301 filters.
  • a response curve as shown in Figure 11 can be obtained.
  • the reconstruction matrix T in the above embodiment can be obtained.
  • the condition number t of T is 2.7 ⁇ 10 6 .
  • the spectrum F( ⁇ ) to be reconstructed (that is, the incident spectrum in this embodiment) is a curve with a double-peak spacing of 3 nm, and it is assumed that the detector has 1% Gaussian white noise.
  • the solid line in Figure 12A is the incident spectrum, and the dotted line is the reconstructed spectrum obtained using the integrated filter and Tikhonov regularization method shown in Figure 10.
  • the solid line in Figure 12B is the incident spectrum, and the dotted line is the reconstructed spectrum obtained using the integrated filter shown in Figure 10 and the spectrum reconstruction method of the present application. It can be seen that the reconstructed spectrum obtained by using the spectrum reconstruction method of the present application is closer to the incident spectrum, that is, the reconstructed spectrum obtained by using the spectrum reconstruction method of the present application is more accurate.
  • the mean square error between the reconstructed spectrum and the incident spectrum obtained using the spectrum reconstruction method of the present application is only 1/4 of the mean square error between the reconstructed spectrum and the incident spectrum obtained using the Tikhonov method. Therefore, the spectrum reconstruction method of the present application has better anti-noise capability.
  • Figure 13 shows a schematic structural diagram of a detector provided by an embodiment of the present application.
  • the detector 1100 includes a quantum dot layer 1140 and a photosensitive element 1150 .
  • the material of the quantum dot layer 1140 may be Cs 2 SnX 6 .
  • X represents the halogen elements Cl, Br, and I.
  • the band gap of the quantum dot material can be adjusted to obtain different transmission spectra.
  • the detector 1100 a response curve as shown in Figure 14 can be obtained.
  • the reconstruction matrix T in the above embodiment can be obtained.
  • the condition number t of T is 1.9 ⁇ 10 21 .
  • the condition number t can be calculated by the above formula (6).
  • the response curve shown in Figure 14 is an S-shaped light response curve with a gradually increasing central peak position.
  • the spectrum F( ⁇ ) to be reconstructed (that is, the incident spectrum in this embodiment) is a curve with a double-peak distance of 30 nm. And assume that the detector has 0.1% Gaussian white noise.
  • the solid line in Figure 15A is the incident spectrum, and the dotted line is the reconstructed spectrum obtained using the detector shown in Figure 13 and the Tikhonov regularization method.
  • the solid line in Figure 15B is the incident spectrum, and the dotted line is the reconstructed spectrum obtained using the detector shown in Figure 13 and the spectrum reconstruction method of the present application. It can be seen that the reconstructed spectrum obtained by using the spectrum reconstruction method of the present application is closer to the incident spectrum, that is, the reconstructed spectrum obtained by using the spectrum reconstruction method of the present application is more accurate.
  • the mean square error between the reconstructed spectrum and the incident spectrum obtained using the spectral reconstruction method of the present application is only the weight difference obtained using the Tikhonov method. 1/4 of the mean square error between the constitutive spectrum and the incident spectrum. Therefore, the spectrum reconstruction method of the present application has better anti-noise ability.
  • Figure 16 shows a schematic structural diagram of a spectrum reconstruction device provided by an embodiment of the present application.
  • the spectrum reconstruction device 900 in the embodiment of the present application includes: a determination module 910 and a reconstruction module 920.
  • the determining module 910 is configured to determine the spectrum data to be processed.
  • the reconstruction module 920 is configured to use an objective function to convert negative spectrum values in the spectrum data to be processed into positive spectrum values to obtain positive spectrum data, and use the objective function to perform loss calculations based on the positive spectrum data and response value data to determine Loss value, if the loss value satisfies the first preset condition or the number of loss calculations satisfies the second preset condition, reconstructed spectrum data is obtained, wherein the response value data is obtained by measuring the spectrum to be reconstructed using a spectrometer.
  • Figure 17 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in Fig. 17 is extended based on the embodiment shown in Fig. 16.
  • the following focuses on the differences between the embodiment shown in Fig. 17 and the embodiment shown in Fig. 16, and the similarities will not be described again.
  • the reconstruction module 920 in the embodiment of the present application includes: a conversion unit 921, a loss calculation unit 922, and an adjustment unit 923.
  • the conversion unit 921 is configured to use an objective function to convert the negative spectrum values in the current spectrum data to be processed into positive spectrum values to obtain the current positive spectrum data.
  • the loss calculation unit 922 is configured to determine the current loss value and the current reconstructed spectrum data based on the current positive spectrum data and the response value data using the objective function.
  • the adjustment unit 923 is configured to adjust the spectrum data to be processed based on the current loss value and the current reconstructed spectrum data, and iterate the spectrum data to be processed until it is determined that the loss value satisfies the first preset condition or the number of loss calculations satisfies the third 2. Preset conditions to obtain reconstructed spectral data.
  • Figure 18 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in Fig. 18 is extended based on the embodiment shown in Fig. 17.
  • the following focuses on the differences between the embodiment shown in Fig. 18 and the embodiment shown in Fig. 17, and the similarities will not be described again.
  • the conversion unit 921 of the embodiment of the present application includes: an absolute value conversion sub-unit 1110.
  • the absolute value conversion subunit 1110 is configured such that the objective function converts the negative spectrum values in the current spectrum data to be processed into positive spectrum values by taking the absolute value of the current spectrum data to be processed, and obtains the current Positive spectral data.
  • Figure 19 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in Fig. 19 is extended based on the embodiment shown in Fig. 17.
  • the following focuses on the differences between the embodiment shown in Fig. 19 and the embodiment shown in Fig. 17, and the similarities will not be described again.
  • the conversion unit 921 of the embodiment of the present application includes: a diagonal matrix determination sub-unit 1210 and a matrix conversion sub-unit 1220.
  • the diagonal matrix determination subunit 1210 is configured to determine the signed diagonal matrix of the current spectral data to be processed.
  • the matrix conversion subunit 1220 is configured such that the objective function converts the negative spectrum values in the current spectrum data to be processed into positive spectrum values by multiplying the current spectrum data to be processed and the symbolic diagonal matrix to obtain the current Positive spectral data.
  • Figure 20 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in Fig. 20 is extended based on the embodiment shown in Fig. 17.
  • the following focuses on the differences between the embodiment shown in Fig. 20 and the embodiment shown in Fig. 17, and the similarities will not be described again.
  • the determination module 910 in the embodiment of the present application includes: a current data determination unit 911.
  • the current data determining unit 911 is configured to determine the (k+2)-th based on the current reconstructed spectrum data obtained in the k-th cycle and the current reconstructed spectrum data obtained in the (k+1)-th cycle.
  • Figure 21 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in Fig. 21 is extended based on the embodiment shown in Fig. 20.
  • the following focuses on the differences between the embodiment shown in Fig. 21 and the embodiment shown in Fig. 20, and the similarities will not be described again.
  • the current data determination unit 911 in the embodiment of the present application includes: a current difference calculation sub-unit 1410 and a current spectrum data to be processed determination sub-unit 1420.
  • the current difference calculation subunit 1410 is configured to calculate the difference between the current reconstructed spectrum data obtained in the (k+1)th cycle and the current reconstructed spectrum data obtained in the kth cycle, and obtain the difference. data.
  • the current spectrum data to be processed determination subunit 1420 is configured to determine the corresponding value of the (k+2)th cycle based on the current reconstructed spectrum data, difference data and preset return function obtained in the (k+1)th cycle. of the current spectral data to be processed.
  • Figure 22 shows a schematic structural diagram of a spectrum reconstruction device provided by another embodiment of the present application.
  • the embodiment shown in FIG. 22 is extended based on the embodiment shown in FIG. 17.
  • the following focuses on the differences between the embodiment shown in FIG. 22 and the embodiment shown in FIG. 17, and the similarities will not be described again.
  • the loss calculation unit 922 in the embodiment of the present application includes: a preprocessing subunit 1510 and a loss value determination subunit 1520.
  • the preprocessing subunit 1510 is configured to use the objective function to preprocess the current positive spectrum data to obtain the current preprocessed spectrum data.
  • the loss value determination subunit 1520 is configured to use an objective function to perform loss calculation based on the current preprocessed spectral data and response value data to determine the current loss value and the current reconstructed spectral data.
  • the operations and functions of the processing spectral data determination sub-unit 1420, as well as the pre-processing sub-unit 1510 and the loss value determination sub-unit 1520 included in the loss calculation unit 922 can refer to the spectral reconstruction method provided in the above-mentioned Figures 2 to 8. In order to avoid duplication, I won’t go into details here.
  • FIG. 23 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • the electronic device 1600 includes: one or more processors 1601 and a memory 1602; and computer program instructions stored in the memory 1602. When executed by the processor 1601, the computer program instructions cause the processor 1601 to perform the following: The spectral reconstruction method of any of the above embodiments.
  • the processor 1601 may be a central processing unit (Central Processing Unit, CPU) or other forms of processing units with data transmission capabilities and/or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.
  • CPU Central Processing Unit
  • Memory 1602 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and/or non-volatile memory.
  • Volatile memory may include, for example, random access memory (Random Access Memory, RAM) and/or cache memory (Cache), etc.
  • Non-volatile memory may include, for example, read-only memory (Read Only Memory, ROM), hard disk, flash memory, etc.
  • One or more computer program instructions may be stored on a computer-readable storage medium, and the processor 1601 may execute the program instructions to implement the above steps in the spectral reconstruction method of various embodiments of the present application and/or other desired Function.
  • the electronic device 1600 may also include an input device 1603 and an output device 1604, and these components are interconnected through a bus system and/or other forms of connection mechanisms (not shown in Figure 23).
  • the input device 1603 may also include, for example, a keyboard, a mouse, a microphone, etc.
  • the output device 1604 can output various information to the outside.
  • the output device 1604 may include, for example, a display, a speaker, a printer, a communication network and remote output devices connected thereto, and the like.
  • the electronic device 1600 may also include any other appropriate components depending on the specific application.
  • embodiments of the present application may also be computer program products, including computer program instructions.
  • the computer program instructions When run by a processor, the computer program instructions cause the processor to perform the spectral reconstruction method as in any of the above embodiments. step.
  • the computer program product can be written in any combination of one or more programming languages to write program codes for executing the operations of the embodiments of the present application.
  • Programming languages include object-oriented programming languages, such as Java, C++, etc., and also include conventional programming languages.
  • the program code may execute entirely on the user's computing device, partly on the user's computing device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device, or entirely on the remote computing device or process executed on the server.
  • embodiments of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When executed by a processor, the computer program instructions cause the processor to perform the method described in the “Exemplary Method” section above of this specification. Steps in the spectral reconstruction methods of various embodiments of the present application.
  • Computer-readable storage media can take the form of any combination of one or more computer-readable media.
  • the readable medium may be a readable signal medium or a readable storage medium.
  • the readable storage medium may include, for example, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: electrical connection with one or more wires, portable disk, hard disk, RAM, ROM, Erasable Programmable Read Only Memory Memory, EPROM) or flash memory, optical fiber, portable compact disk read-only memory (Compact Dksk Read Only Memory, CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above.
  • each component or each step can be decomposed and/or recombined. These decompositions and/or recombinations shall be considered equivalent versions of this application.

Landscapes

  • Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Pure & Applied Mathematics (AREA)
  • Mathematical Optimization (AREA)
  • Mathematical Analysis (AREA)
  • Computational Mathematics (AREA)
  • Spectroscopy & Molecular Physics (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Chemical & Material Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biochemistry (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Health & Medical Sciences (AREA)
  • Investigating Or Analysing Materials By Optical Means (AREA)
  • Spectrometry And Color Measurement (AREA)

Abstract

一种光谱重构方法，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，从而提高对目标函数的迭代求解过程的抗噪能力，进而提高所得到的重构光谱数据的准确性。另外，这种光谱重构方法无需对目标函数增加约束，避免了将非约束问题变为约束问题，进一步提高了得到的重构光谱数据的准确性，且提高了目标函数的收敛效率，同时解决了光谱重构方法抗噪能力差的问题。还公开了一种光谱重构装置,光谱仪(1000)，计算机可读存储介质和电子设备（1600）。

Description

光谱重构方法和装置、光谱仪、存储介质和电子设备 技术领域
本申请涉及光谱分析技术领域，具体涉及一种光谱重构方法和光谱重构装置，光谱仪，以及计算机可读存储介质和电子设备。
发明背景
每种原子都对应有特征光谱，因此可以根据特征光谱来鉴别物质或确定化学成分。目前，主要使用重构型光谱仪来测量和分析光谱。重构型光谱仪包括硬件部分和软件部分，硬件部分主要包括探测器，软件部分主要包括光谱重构算法。探测器探测到响应值后，由光谱重构方法对响应值进行反演计算，从而得到重构光谱。然而，目前的光谱重构方法得到的重构光谱准确性低，不能满足用户要求。
发明内容
有鉴于此，本申请实施例提供了一种光谱重构方法和光谱重构装置，光谱仪，以及计算机可读存储介质和电子设备，解决了光谱重构方法抗噪能力差的问题。
第一方面，本申请一实施例提供的一种光谱重构方法，应用于光谱仪，包括：确定待处理光谱数据；利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据，其中，响应值数据为利用光谱仪对待重构光谱进行测量得到的。
结合本申请的第一方面，在一些实施例中，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据，包括：利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据；利用目标函数，基于当前的正光谱数据和响应值数据，确定当前的损失值和当前的重构光谱数据；基于当前的损失值和当前的重构光谱数据，调整待处理光谱数据，并循环迭代待处理光谱数据，直至确定损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。
结合本申请的第一方面，在一些实施例中，利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据，包括：目标函数通过对当前的待处理光谱数据取绝对值的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
结合本申请的第一方面，在一些实施例中，在利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据之前，还包括：确定当前的待处理光谱数据的符号对角矩阵；利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据，包括：目标函数通过对当前的待处理光谱数据和符号对角矩阵取乘积的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
结合本申请的第一方面，在一些实施例中，确定待处理光谱数据，包括：基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，其中，k为正整数。
结合本申请的第一方面，在一些实施例中，基于第k个循环得到的当前的重构光谱数据和 第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，包括：计算第(k+1)个循环得到的当前的重构光谱数据和第k个循环得到的当前的重构光谱数据的差值，得到差值数据；基于第(k+1)个循环得到的当前的重构光谱数据、差值数据和预设回传函数，确定第(k+2)个循环对应的当前的待处理光谱数据。
第二方面，本申请一实施例提供了一种光谱重构装置，应用于光谱仪，包括：确定模块，配置为确定待处理光谱数据；重构模块，配置为利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据，其中，响应值数据为利用光谱仪对待重构光谱进行测量得到的。
第三方面，本申请一实施例提供了一种光谱仪，包括：探测器，用于对待重构光谱进行测量得到响应值数据，并将响应值数据发送给处理器；处理器，与探测器电连接，用于接收响应值数据，并执行第一方面的光谱重构方法，以得到重构光谱数据。
第四方面，本申请一实施例提供了一种计算机可读存储介质，存储介质存储有指令，当指令由电子设备的处理器执行时，使得电子设备能够执行上述第一方面提及的光谱重构方法。
第五方面，本申请一实施例提供了一种电子设备，电子设备包括：处理器；以及用于存储计算机可执行指令的存储器。处理器，用于执行计算机可执行指令，以实现上述第一方面提及的光谱重构方法。
本申请实施例提供的光谱重构方法，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。本申请根据光谱一定为正值的特征，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，从而提高对目标函数的迭代求解过程的抗噪能力，进而提高得到的重构光谱数据的准确性。另外，本申请利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，无需对目标函数增加约束，避免了将非约束问题变为约束问题，进一步提高了得到的重构光谱数据的准确性，且提高了目标函数的收敛效率。
附图简要说明
图1A所示为本申请一实施例提供的光谱重构方法的应用场景示意图。
图1B所示为本申请另一实施例提供的光谱重构方法的应用场景示意图。
图2所示为本申请一实施例提供的光谱重构方法的流程示意图。
图3所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图4所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图5所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图6所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图7所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图8所示为本申请另一实施例提供的光谱重构方法的流程示意图。
图9所示为本申请一实施例提供的光谱仪的测量原理的示意图。
图10所示为本申请一实施例提供的探测器的结构示意图。
图11所示为本申请一实施例提供的响应曲线。
图12A所示为利用图10所示的集成滤光片和吉洪诺夫正则化方法得到的重构光谱和入射光谱。
图12B所示为利用图10所示的集成滤光片和本申请的光谱重构方法得到的重构光谱和入射光谱。
图13所示为本申请一实施例提供的探测器的结构示意图。
图14所示为本申请另一实施例提供的响应曲线。
图15A所示为利用图13所示的探测器和吉洪诺夫正则化方法得到的重构光谱和入射光谱。
图15B所示为利用图13所示的探测器和本申请的光谱重构方法得到的重构光谱和入射光谱。
图16所示为本申请一实施例提供的光谱重构装置的结构示意图。
图17所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图18所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图19所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图20所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图21所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图22所示为本申请另一实施例提供的光谱重构装置的结构示意图。
图23所示为本申请一实施例提供的电子设备的结构示意图。
实施本申请的方式
下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
图1A所示为本申请一实施例提供的光谱重构方法的应用场景示意图。图1A所示的场景包括光谱仪100中的处理器110和光谱仪100中的光谱测量模块120。光谱测量模块120与处理器110通信连接。具体而言，处理器110用于确定待处理光谱数据；利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。光谱测量模块120用于对待重构光谱进行测量得到响应值数据，并将响应值数据发送给处理器110，以便处理器110执行上述操作。
图1B所示为本申请另一实施例提供的光谱重构方法的应用场景示意图。图1B所示的场景包括光谱仪100和服务器200。光谱仪100对待重构光谱进行测量得到响应值数据，并将响应值数据发送给服务器200。服务器200接收响应值数据，并确定待处理光谱数据；利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。
在本申请一实施例中，服务器200得到重构光谱数据以后，也可以发送给光谱仪100，从而在光谱仪100上显示重构光谱数据。
图2所示为本申请一实施例提供的光谱重构方法的流程示意图。如图2所示，该光谱重构方法包括如下步骤。
步骤210，确定待处理光谱数据。
具体地，待处理光谱数据可以是初始化得到的光谱数据。例如，待处理光谱数据可以是通过随机值初始化、随机正数初始化、吉洪诺夫(Tikhonov)初始化等初始化的方法得到的光谱数据。待处理光谱数据可以是光谱矩阵，光谱矩阵中的元素代表一个光波的波长。
步骤220，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。
具体地，响应值数据为利用光谱仪对待重构光谱进行测量得到的。响应值数据可以是响应值矩阵。待重构光谱是需要被测量的光谱。重构光谱数据可以是利用目标函数，基于响应值数据对待处理光谱数据进行优化得到的光谱数据。目标函数可以表征待处理光谱数据与响应值数据之间的函数关系。目标函数还可以将待处理光谱数据中负的光谱数值转化为正的光谱数值。待处理光谱数据中负的光谱数值可以是待处理光谱数据中的噪声。在本 申请一实施例中，可以对待处理光谱数据中的光谱数值做绝对值，从而将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据。在本申请一实施例中，可以先确定待处理光谱数据的符号对角矩阵，然后利用符号对角矩阵，将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据。
具体地，目标函数可以包括绝对值算符，绝对值算符用于将待处理光谱数据中负的光谱数值转化为正的光谱数值。即，通过在目标函数中引入绝对值算符，将待处理光谱数据中负的光谱数值转化为正的光谱数值。
具体地，利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，可以是通过取响应值数据和正光谱数据的差值的方式进行损失计算，从而确定损失值。第一预设条件可以是预设的阈值。如果损失值满足预设的阈值，得到重构光谱数据。第一预设条件还可以是预设的损失值的收敛条件。如果损失值满足预设的损失值的收敛条件，得到重构光谱数据。预设的损失值的收敛条件可以是连续几次计算得到的损失值很接近，甚至是后几次计算得到的损失值大于前几次计算得到的损失值。第二预设条件可以是预设的损失计算的次数。如果损失计算的次数满足预设的损失计算的次数，得到重构光谱数据。用户可以根据实际需求确定第一预设条件和第二预设条件，本申请不做具体限定。
本申请实施例提供的光谱重构方法，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。由于待处理光谱数据中负的光谱数值是干扰数据，导致对目标函数的迭代求解过程的抗噪能力差。因此，本申请利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，从而提高对目标函数的迭代求解过程的抗噪能力，进而提高得到的重构光谱数据的准确性。另外，本申请利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，无需对目标函数增加约束，避免了将非约束问题变为约束问题，进一步提高了得到的重构光谱数据的准确性，且提高了目标函数的收敛效率。
图3所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图2所示实施例基础上延伸出图3所示实施例，下面着重叙述图3所示实施例与图2所示实施例的不同之处，相同之处不再赘述。
如图3所示，在本申请实施例中，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据的步骤，包括如下步骤。
步骤310，利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
具体地，当前的待处理光谱数据是指当前正在被处理的光谱数据。当前的正光谱数据是指当前得到的正光谱数据。
示例性地，目标函数S可以用以下的公式(1)表示。
其中，I表示响应值数据，T表示重构矩阵，Q表示预处理矩阵，L表示正则化矩阵，下标k表示第k次循环，Xk表示第k次循环对应的当前的待处理光谱数据，表示正则化项，α表示正则化系数，m表示幂数，n表示范数的类型。T、Q、L和m、n均可以通过预先标定或预先设置获得，本申请对的取值均不做具体限定。
以上公式(1)表示可以通过不断迭代调整Xk，使S得到一个较小的值，甚至是最小的值。|Xk|表示对Xk做绝对值，从而将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据。上述公式(1)也可以没有预处理矩阵Q和正则化矩阵L。甚至，上述公式(1)也可以没有正则化项
步骤320，利用目标函数，基于当前的正光谱数据和响应值数据，确定当前的损失值和当前的重构光谱数据。
示例性地，可以通过上述公式(1)计算当前的损失值。计算当前的损失值所使用的待处理光谱数据即为当前的重构光谱数据。
步骤330，基于当前的损失值和当前的重构光谱数据，调整待处理光谱数据，并循环迭代待处理光谱数据，直至确定损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。
具体地，可以不断调整待处理光谱数据，并循环迭代待处理光谱数据，直至确定损失值满足第一预设条件或损失计算的次数满足第二预设条件，将计算当前的损失值所使用的当前的重构光谱数据确定为重构光谱数据。
具体地，利用目标函数，基于当前的正光谱数据和响应值数据，确定当前的损失值后，可以将当前的待处理光谱数据确定为当前的重构光谱数据。即针对每循环，在本次循环计算完损失值以后，将本次循环的待处理光谱数据确定为本次循环的重构光谱数据。
通过基于当前的损失值和当前的重构光谱数据，调整待处理光谱数据，并循环迭代待处理光谱数据，直至确定损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据，可以从多次循环迭代各自对应的重构光谱数据中确定出较准确的重构光谱数据，从而进一步提高最终得到的重构光谱数据的准确性。
图4所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图3所示实施例基础上延伸出图4所示实施例，下面着重叙述图4所示实施例与图3所示实施例的不同之处，相同之处不再赘述。
如图4所示，在本申请实施例中，利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据的步骤，包括如下步骤。
步骤410，目标函数通过对当前的待处理光谱数据取绝对值的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
在实际应用中，如上述公式(1)所示，目标函数可以通过对当前的待处理光谱数据取绝对值的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。取绝对值的方式简单准确，提高了对目标函数的求解效率。
图5所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图3所示实施例基础上延伸出图5所示实施例，下面着重叙述图5所示实施例与图3所示实施例的不同之处，相同之处不再赘述。
如图5所示，在本申请实施例中，利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据的步骤，包括如下步骤。
步骤510，确定当前的待处理光谱数据的符号对角矩阵。
具体地，符号对角矩阵可以用Ak表示。可以通过以下公式(2)求待处理光谱数据的符号对角矩阵Ak。
Ak←diag(sign(Xk))      (2)
其中，下标k表示第k次循环。
步骤520，目标函数通过对当前的待处理光谱数据和符号对角矩阵取乘积的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
在实际应用中，目标函数通过对当前的待处理光谱数据X和符号对角矩阵A取乘积的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据，方法简单，效率高。
图6所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图3所示实施例基础上延伸出图6所示实施例，下面着重叙述图6所示实施例与图3所示实施例的不同之处，相同之处不再赘述。
如图6所示，在本申请实施例中，确定待处理光谱数据的步骤，包括如下步骤。
步骤610，基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据。
具体地，k为正整数。例如，当k＝1时，(k+1)＝2，(k+2)＝3，即，基于第1个循 环得到的当前的重构光谱数据和第2个循环得到的当前的重构光谱数据，确定第3个循环对应的当前的待处理光谱数据。
在实际应用中，基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，可以使后续循环中的当前的待处理光谱数据参考到前两次循环中的得到的当前的重构光谱数据，从而提高后续循环中的当前的待处理光谱数据的准确性。
图7所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图6所示实施例基础上延伸出图7所示实施例，下面着重叙述图7所示实施例与图6所示实施例的不同之处，相同之处不再赘述。
如图7所示，在本申请实施例中，基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据的步骤，包括如下步骤。
步骤710，计算第(k+1)个循环得到的当前的重构光谱数据和第k个循环得到的当前的重构光谱数据的差值，得到差值数据。
具体地，可以用Yk+1表示第(k+1)个循环得到的当前的重构光谱数据，用Yk表示第k个循环得到的当前的重构光谱数据。
步骤720，基于第(k+1)个循环得到的当前的重构光谱数据、差值数据和预设回传函数，确定第(k+2)个循环对应的当前的待处理光谱数据。
具体地，可以用Yk+2表示第(k+2)个循环得到的当前的重构光谱数据。示例性地，可以使用以下公式(3)(即，预设回传函数)确定第(k+2)个循环对应的当前的待处理光谱数据Xk+2。
其中，Pk+2为中间值，可以通过以下的公式(4)计算得到。Ak+2表示第(k+2)个循环对应的当前的待处理光谱数据的符号对角矩阵，因此，可以利用上述公式(2)求得。l表示迭代步长，grad表示梯度函数。
Pk+2←Yk+βk+2(Yk+1-Yk)      (4)
其中，βk+2表示第(k+2)个循环对应的预设回传函数。
第(k+2)个循环结束后，得到的Yk+2等于Xk+2。
通过计算第(k+1)个循环得到的当前的重构光谱数据和第k个循环得到的当前的重构光谱数据的差值，得到差值数据，并基于第(k+1)个循环得到的当前的重构光谱数据、差值数据和预设回传函数，确定第(k+2)个循环对应的当前的待处理光谱数据，优化了待处理光谱数据，进而通过多次循环迭代提高重构光谱数据的准确性。
在本申请一实施例中，当k＝1时，即第一次循环时，可以利用Tikhonov正则化赋值对X1进行初始化赋值。具体地，可以使用以下公式(5)对X1进行初始化赋值。
X1＝(THT+γJ)-1THI      (5)
其中，T和公式(1)中的T一样，代表重构矩阵，上标H用于表示TH为T的共轭转置矩
阵，J代表与THT的维度相同的单位矩阵，γ为吉洪诺夫正则化系数。γ可以预先设置。
图8所示为本申请另一实施例提供的光谱重构方法的流程示意图。在图3所示实施例基础上延伸出图8所示实施例，下面着重叙述图8所示实施例与图3所示实施例的不同之处，相同之处不再赘述。
如图8所示，在本申请实施例中，利用目标函数，基于当前的正光谱数据和响应值数据，确定当前的损失值和当前的重构光谱数据的步骤，包括如下步骤。
步骤810，利用目标函数，对当前的正光谱数据进行预处理，得到当前的预处理后的光谱数据。
具体地，可以使用上述公式(1)中的预处理矩阵Q来对当前的正光谱数据进行预处理，得到当前的预处理后的光谱数据。预处理矩阵Q可以是傅立叶变换矩阵，也可以是其他的能够实现离散化的矩阵，本申请不做具体限定。
步骤820，利用目标函数，基于当前的预处理后的光谱数据和响应值数据进行损失计算，以确定当前的损失值和当前的重构光谱数据。
利用目标函数，对当前的正光谱数据进行预处理，可以将连续的光谱进行离散化，从而减少数据量，提高计算效率。
图9所示为本申请一实施例提供的光谱仪的测量原理的示意图。如图9，光谱仪1000包括：探测器1100和处理器1200。探测器1100用于对待重构光谱F(λ)进行测量得到响应值数据(在本实施例中，响应值数据可以是响应值矩阵In)，并将响应值数据发送给处理器1200。处理器1200，与探测器1100电连接，用于接收响应值数据，并执行上述实施例中的光谱重构方法，以得到重构光谱数据。探测器1100可以包括集成滤光片和光敏芯片。光敏芯片可以是互补金属氧化物半导体(Complementary Metal Oxide Semiconductor，CMOS)光敏芯片，电荷耦合元件(Charge-coupled Device，CCD)。
图10所示为本申请一实施例提供的集成滤光片的结构示意图。如图9所示，集成滤光片包括衬底1110、二氧化钛膜层1120和二氧化硅膜层1130。其中，二氧化钛膜层1120和二氧化硅膜层1130可以交叉设置。二氧化钛膜层1120和二氧化硅膜层1130的具体层厚和具体层数可以根据实际需求进行设置，本申请不做具体限定。
具体地，集成滤光片可以用膜系Sub|(HL)5M(HL)5表示。其中，Sub表示衬底1110。H表示厚度为78nm的二氧化钛膜层(TiO2)，L表示厚度为115nm的第一二氧化硅膜层(SiO2)1121。M表示除L之外的第二二氧化硅膜层1122。衬底1110可以是玻璃衬底、石英衬底、宝石衬底等，更具体地，可以是国际标准分类中的k9玻璃衬底。
通过试验可以确定合适的第二二氧化硅膜层1122的厚度，使集成滤光片的透射峰位从600nm以0.5nm的间距增加至750nm，即，集成滤光片共301个滤光片。
通过探测器1100中的集成滤光片和光敏芯片，可以得到如图11所示的响应曲线，响应曲线经处理即可得到上述实施例中的重构矩阵T。经计算，T的条件数t为2.7×106。条件数t可以通过以下公式(6)计算。
t＝||T||·||T-1||      (6)
待重构光谱F(λ)(即本实施例的入射光谱)为双峰间距为3nm的曲线，并假定探测器有1％的高斯白噪声。图12A中实线为入射光谱，虚线为利用图10所示的集成滤光片和吉洪诺夫正则化方法得到的重构光谱。图12B实线为入射光谱，虚线为利用图10所示的集成滤光片和本申请的光谱重构方法得到的重构光谱。可以看出，利用本申请的光谱重构方法得到的重构光谱更接近入射光谱，即，利用本申请的光谱重构方法得到的重构光谱更加准确。更具体地，利用本申请的光谱重构方法得到的重构光谱与入射光谱的均方差仅是利用Tikhonov方法得到的重构光谱的与入射光谱的均方差的1/4。因此，本申请的光谱重构方法具有更好的抗噪声能力。
图13所示为本申请一实施例提供的探测器的结构示意图。如图13所示，探测器1100包括量子点层1140和光敏元1150。
具体地，量子点层1140的材料可以是Cs2SnX6。X代表卤素元素Cl、Br、I，通过改变卤素元素的组分，可以调节量子点材料的带隙，从而获得不同的透射光谱。通过探测器1100，可以得到如图14所示的响应曲线。对图14所示的响应曲线中波长500nm至800nm对应的响应曲线进行处理，即可得到上述实施例中的重构矩阵T。经计算，T的条件数t为1.9×1021。条件数t可以通过以上的公式(6)计算。图14所示的响应曲线为中心峰位逐渐递增的S型光响应曲线。
待重构光谱F(λ)(即本实施例的入射光谱)为双峰间距为30nm的曲线。并假定探测器有0.1％高斯白噪声的。图15A中实线为入射光谱，虚线为利用图13所示的探测器和吉洪诺夫正则化方法得到的重构光谱。图15B中实线为入射光谱，虚线为利用图13所示的探测器和本申请的光谱重构方法得到的重构光谱。可以看出，利用本申请的光谱重构方法得到的重构光谱更接近入射光谱，即，利用本申请的光谱重构方法得到的重构光谱更加准确。更具体地，利用本申请的光谱重构方法得到的重构光谱与入射光谱的均方差仅是利用Tikhonov方法得到的重 构光谱的与入射光谱的均方差的1/4。因此，本申请的光谱重构方法具有更好的抗噪声能力。
上文结合图2至图8，详细描述了本申请的方法实施例，下面结合图16至图22，详细描述本申请的装置实施例。应理解，方法实施例的描述与装置实施例的描述相互对应，因此，未详细描述的部分可以参见前面方法实施例。
图16所示为本申请一实施例提供的光谱重构装置的结构示意图。如图16所示，本申请实施例的光谱重构装置900包括：确定模块910和重构模块920。
具体地，确定模块910配置为，确定待处理光谱数据。重构模块920配置为，利用目标函数将待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用目标函数基于正光谱数据和响应值数据进行损失计算，以确定损失值，如果损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据，其中，响应值数据为利用光谱仪对待重构光谱进行测量得到的。
图17所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图16所示实施例基础上延伸出图17所示实施例，下面着重叙述图17所示实施例与图16所示实施例的不同之处，相同之处不再赘述。
如图17所示，本申请实施例的重构模块920包括：转化单元921、损失计算单元922和调整单元923。
具体地，转化单元921配置为，利用目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。损失计算单元922配置为，利用目标函数，基于当前的正光谱数据和响应值数据，确定当前的损失值和当前的重构光谱数据。调整单元923配置为，基于当前的损失值和当前的重构光谱数据，调整待处理光谱数据，并循环迭代待处理光谱数据，直至确定损失值满足第一预设条件或损失计算的次数满足第二预设条件，得到重构光谱数据。
图18所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图17所示实施例基础上延伸出图18所示实施例，下面着重叙述图18所示实施例与图17所示实施例的不同之处，相同之处不再赘述。
如图18所示，本申请实施例的转化单元921包括：绝对值转化子单元1110。
具体地，绝对值转化子单元1110配置为，目标函数通过对当前的待处理光谱数据取绝对值的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
图19所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图17所示实施例基础上延伸出图19所示实施例，下面着重叙述图19所示实施例与图17所示实施例的不同之处，相同之处不再赘述。
如图19所示，本申请实施例的转化单元921包括：对角矩阵确定子单元1210和矩阵转换子单元1220。
具体地，对角矩阵确定子单元1210，配置为确定当前的待处理光谱数据的符号对角矩阵。矩阵转换子单元1220，配置为目标函数通过对当前的待处理光谱数据和符号对角矩阵取乘积的方式，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据。
图20所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图17所示实施例基础上延伸出图20所示实施例，下面着重叙述图20所示实施例与图17所示实施例的不同之处，相同之处不再赘述。
如图20所示，本申请实施例的确定模块910包括：当前数据确定单元911。
具体地，当前数据确定单元911配置为，基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，其中，k为正整数。
图21所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图20所示实施例基础上延伸出图21所示实施例，下面着重叙述图21所示实施例与图20所示实施例的不同之处，相同之处不再赘述。
如图21所示，本申请实施例的当前数据确定单元911包括：当前差值计算子单元1410和当前待处理光谱数据确定子单元1420。
具体地，当前差值计算子单元1410配置为，计算第(k+1)个循环得到的当前的重构光谱数据和第k个循环得到的当前的重构光谱数据的差值，得到差值数据。当前待处理光谱数据确定子单元1420配置为，基于第(k+1)个循环得到的当前的重构光谱数据、差值数据和预设回传函数，确定第(k+2)个循环对应的当前的待处理光谱数据。
图22所示为本申请另一实施例提供的光谱重构装置的结构示意图。在图17所示实施例基础上延伸出图22所示实施例，下面着重叙述图22所示实施例与图17所示实施例的不同之处，相同之处不再赘述。
如图22所示，本申请实施例的损失计算单元922包括：预处理子单元1510和损失值确定子单元1520。
具体地，预处理子单元1510配置为，利用目标函数，对当前的正光谱数据进行预处理，得到当前的预处理后的光谱数据。损失值确定子单元1520配置为，利用目标函数，基于当前的预处理后的光谱数据和响应值数据进行损失计算，以确定当前的损失值和当前的重构光谱数据。
图16至图22提供的光谱重构装置中的确定模块910和重构模块920，以及重构模块920中包括的转化单元921、损失计算单元922和调整单元923，以及转化单元921中包括的绝对值转化子单元1110、对角矩阵确定子单元1210和矩阵转换单元1220，以及确定模块910包括的当前数据确定单元911，以及当前数据确定单元911包括的当前差值计算子单元1410和当前待处理光谱数据确定子单元1420，以及损失计算单元922包括的预处理子单元1510和损失值确定子单元1520的操作和功能可以参考上述图2至图8提供的光谱重构方法，为了避免重复，在此不再赘述。
图23所示为本申请一实施例提供的电子设备的结构示意图。如图23所示，该电子设备1600包括：一个或多个处理器1601和存储器1602；以及存储在存储器1602中的计算机程序指令，计算机程序指令在被处理器1601运行时使得处理器1601执行如上述任一实施例的光谱重构方法。
处理器1601可以是中央处理单元(Central Processkng Unkt,CPU)或者具有数据传输能力和/或指令执行能力的其他形式的处理单元，并且可以控制电子设备中的其他组件以执行期望的功能。
存储器1602可以包括一个或多个计算机程序产品，计算机程序产品可以包括各种形式的计算机可读存储介质，例如易失性存储器和/或非易失性存储器。易失性存储器例如可以包括随机存取存储器(Random Access Memory,RAM)和/或高速缓冲存储器(Cache)等。非易失性存储器例如可以包括只读存储器(Read Only Memory,ROM)、硬盘、闪存等。在计算机可读存储介质上可以存储一个或多个计算机程序指令，处理器1601可以运行程序指令，以实现上文的本申请的各个实施例的光谱重构方法中的步骤以及/或者其他期望的功能。
在一个示例中，电子设备1600还可以包括：输入装置1603和输出装置1604，这些组件通过总线系统和/或其他形式的连接机构(图23中未示出)互连。
此外，该输入装置1603还可以包括例如键盘、鼠标、麦克风等等。
该输出装置1604可以向外部输出各种信息。该输出装置1604可以包括例如显示器、扬声器、打印机、以及通信网络及其所连接的远程输出设备等等。
当然，为了简化，图23中仅示出了该电子设备1600中与本申请有关的组件中的一些，省略了诸如总线、输入装置/输出接口等组件。除此之外，根据具体应用情况，电子设备1600还可以包括任何其他适当的组件。
除了上述方法和设备以外，本申请的实施例还可以是计算机程序产品，包括计算机程序指令，计算机程序指令在被处理器运行时使得处理器执行如上述任一实施例的光谱重构方法中的步骤。
计算机程序产品可以以一种或多种程序设计语言的任意组合来编写用于执行本申请实施例操作的程序代码，程序设计语言包括面向对象的程序设计语言，诸如Java、C++等，还包括常 规的过程式程序设计语言，诸如“C”语言或类似的程序设计语言。程序代码可以完全地在用户计算设备上执行、部分地在用户设备上执行、作为一个独立的软件包执行、部分在用户计算设备上部分在远程计算设备上执行、或者完全在远程计算设备或处理器上执行。
此外，本申请的实施例还可以是计算机可读存储介质，其上存储有计算机程序指令，计算机程序指令在被处理器运行时使得处理器执行本说明书上述“示例性方法”部分中描述的根据本申请各种实施例的光谱重构方法中的步骤。
计算机可读存储介质可以采用一个或多个可读介质的任意组合。可读介质可以是可读信号介质或者可读存储介质。可读存储介质例如可以包括但不限于电、磁、光、电磁、红外线、或半导体的系统、装置或器件，或者任意以上的组合。可读存储介质的更具体的例子(非穷举的列表)包括：具有一个或多个导线的电连接、便携式盘、硬盘、RAM、ROM、可擦式可编程只读存储器(Erasable Programmable Read Only Memory,EPROM)或闪存、光纤、便携式紧凑盘只读存储器(Compact Dksk Read Only Memory,CD-ROM)、光存储器件、磁存储器件、或者上述的任意合适的组合。
以上结合具体实施例描述了本申请的基本原理，但是，需要指出的是，在本申请中提及的优点、优势、效果等仅是示例而非限制，不能认为这些优点、优势、效果等是本申请的各个实施例必须具备的。另外，上述公开的具体细节仅是为了示例的作用和便于理解的作用，而非限制，上述细节并不限制本申请为必须采用上述具体的细节来实现。
本申请中涉及的器件、装置、设备、系统的方框图仅作为例示性的例子并且不意图要求或暗示必须按照方框图示出的方式进行连接、布置、配置。如本领域技术人员将认识到的，可以按任意方式连接、布置、配置这些器件、装置、设备、系统。诸如“包括”、“包含”、“具有”等等的词语是开放性词汇，指“包括但不限于”，且可与其互换使用。这里所使用的词汇“或”和“和”指词汇“和/或”，且可与其互换使用，除非上下文明确指示不是如此。这里所使用的词汇“诸如”指词组“诸如但不限于”，且可与其互换使用。
还需要指出的是，在本申请的装置、设备和方法中，各部件或各步骤是可以分解和/或重新组合的。这些分解和/或重新组合应视为本申请的等效方案。
提供所公开的方面的以上描述以使本领域的任何技术人员能够做出或者使用本申请。对这些方面的各种修改对于本领域技术人员而言是非常显而易见的，并且在此定义的一般原理可以应用于其他方面而不脱离本申请的范围。因此，本申请不意图被限制到在此示出的方面，而是按照与在此公开的原理和新颖的特征一致的最宽范围。
为了例示和描述的目的已经给出了以上描述。此外，此描述不意图将本申请的实施例限制到在此公开的形式。尽管以上已经讨论了多个示例方面和实施例，但是本领域技术人员将认识到其某些变型、修改、改变、添加和子组合。
以上仅为本申请的较佳实施例而已，并不用以限制本申请，凡在本申请的精神和原则之内，所作的任何修改、等同替换等，均应包含在本申请的保护范围之内。

Claims (14)

  1. 一种光谱重构方法，应用于光谱仪，包括：
    确定待处理光谱数据；
    利用目标函数将所述待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用所述目标函数基于所述正光谱数据和响应值数据进行损失计算，以确定损失值，如果所述损失值满足第一预设条件或所述损失计算的次数满足第二预设条件，得到重构光谱数据，其中，所述响应值数据为利用所述光谱仪对待重构光谱进行测量得到的。
  2. 根据权利要求1所述的光谱重构方法，其中，所述利用目标函数将所述待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用所述目标函数基于所述正光谱数据和响应值数据进行损失计算，以确定损失值，如果所述损失值满足第一预设条件或所述损失计算的次数满足第二预设条件，得到重构光谱数据，包括：
    利用所述目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据；
    利用所述目标函数，基于所述当前的正光谱数据和所述响应值数据，确定当前的损失值和当前的重构光谱数据；
    基于所述当前的损失值和所述当前的重构光谱数据，调整所述待处理光谱数据，并循环迭代所述待处理光谱数据，直至确定所述损失值满足所述第一预设条件或所述损失计算的次数满足所述第二预设条件，得到所述重构光谱数据。
  3. 根据权利要求2所述的光谱重构方法，其中，所述利用所述目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据，包括：
    所述目标函数通过对所述当前的待处理光谱数据取绝对值的方式，将所述当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到所述当前的正光谱数据。
  4. 根据权利要求2所述的光谱重构方法，其中，所述利用所述目标函数，将当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到当前的正光谱数据，包括：
    确定所述当前的待处理光谱数据的符号对角矩阵；
    所述目标函数通过对所述当前的待处理光谱数据和所述符号对角矩阵取乘积的方式，将所述当前的待处理光谱数据中负的光谱数值转化为正的光谱数值，得到所述当前的正光谱数据。
  5. 根据权利要求2至4任一项所述的光谱重构方法，其中，所述确定待处理光谱数据，包括：
    基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，其中，k为正整数。
  6. 根据权利要求5所述的光谱重构方法，其中，所述基于第k个循环得到的当前的重构光谱数据和第(k+1)个循环得到的当前的重构光谱数据，确定第(k+2)个循环对应的当前的待处理光谱数据，包括：
    计算所述第(k+1)个循环得到的当前的重构光谱数据和所述第k个循环得到的当前的重构光谱数据的差值，得到差值数据；
    基于所述第(k+1)个循环得到的当前的重构光谱数据、所述差值数据和预设回传函数，确定所述第(k+2)个循环对应的当前的待处理光谱数据。
  7. 根据权利要求2至6任一项所述的光谱重构方法，其中，所述利用所述目标函数，基于所述当前的正光谱数据和所述响应值数据，确定当前的损失值和当前的重构光谱数据，包括：
    利用所述目标函数，对所述当前的正光谱数据进行预处理，得到当前的预处理后的光谱数据；
    利用所述目标函数，基于所述当前的预处理后的光谱数据和所述响应值数据进行损失计算，以确定所述当前的损失值和所述当前的重构光谱数据。
  8. 根据权利要求1至7任一项所述的光谱重构方法，其中，所述目标函数包括以下函数：
    其中，I表示响应值数据，T表示重构矩阵，Q表示预处理矩阵，L表示正则化矩阵，Xk表示第k次循环对应的当前的待处理光谱数据，表示正则化项，α表示正则化系数，m表示幂数，n表示范数的类型。
  9. 根据权利要求1至8任一项所述的光谱重构方法，其中，所述第一预设条件是预设的阈值。
  10. 根据权利要求1至8任一项所述的光谱重构方法，其中，所述第二预设条件是预设的损失计算的次数。
  11. 一种光谱重构装置，应用于光谱仪，包括：
    确定模块，配置为确定待处理光谱数据；
    重构模块，配置为利用目标函数将所述待处理光谱数据中负的光谱数值转化为正的光谱数值，得到正光谱数据，并利用所述目标函数基于所述正光谱数据和响应值数据进行损失计算，以确定损失值，如果所述损失值满足第一预设条件或所述损失计算的次数满足第二预设条件，得到重构光谱数据，其中，所述响应值数据为利用所述光谱仪对待重构光谱进行测量得到的。
  12. 一种光谱仪，包括：
    探测器，用于对待重构光谱进行测量得到响应值数据，并将所述响应值数据发送给处理器；
    所述处理器，与所述探测器电连接，用于接收所述响应值数据，并执行权利要求1至10任一项所述的光谱重构方法，以得到重构光谱数据。
  13. 一种计算机可读存储介质，所述存储介质存储有指令，当所述指令由电子设备的处理器执行时，使得所述电子设备能够执行上述权利要求1至10任一项所述的光谱重构方法。
  14. 一种电子设备，所述电子设备包括：
    处理器；
    用于存储计算机可执行指令的存储器；
    所述处理器，用于执行所述计算机可执行指令，以实现上述权利要求1至10任一项所述的光谱重构方法。
PCT/CN2023/078205 2022-05-17 2023-02-24 光谱重构方法和装置、光谱仪、存储介质和电子设备 WO2023221588A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/354,510 US20230375405A1 (en) 2022-05-17 2023-07-18 Method and device for reconstructing spectrum, spectrometer, non-transitory computer-readable storage medium, and electronic device

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210536422.5A CN115235628B (zh) 2022-05-17 2022-05-17 光谱重构方法和装置、光谱仪、存储介质和电子设备
CN202210536422.5 2022-05-17

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/354,510 Continuation US20230375405A1 (en) 2022-05-17 2023-07-18 Method and device for reconstructing spectrum, spectrometer, non-transitory computer-readable storage medium, and electronic device

Publications (1)

Publication Number Publication Date
WO2023221588A1 true WO2023221588A1 (zh) 2023-11-23

Family

ID=83668331

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/078205 WO2023221588A1 (zh) 2022-05-17 2023-02-24 光谱重构方法和装置、光谱仪、存储介质和电子设备

Country Status (2)

Country Link
CN (1) CN115235628B (zh)
WO (1) WO2023221588A1 (zh)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115235628B (zh) * 2022-05-17 2023-12-01 中国科学院上海技术物理研究所 光谱重构方法和装置、光谱仪、存储介质和电子设备

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150181131A1 (en) * 2012-07-20 2015-06-25 Carl Zeiss Ag Method and apparatus for image reconstruction
CN104849220A (zh) * 2015-06-09 2015-08-19 武汉大学 一种平面式文物光谱图像获取方法
CN108520488A (zh) * 2018-04-10 2018-09-11 深圳劲嘉集团股份有限公司 一种重构光谱并进行复制的方法以及电子设备
CN108830796A (zh) * 2018-06-20 2018-11-16 重庆大学 基于谱空结合和梯度域损失的高光谱图像超分辨重构方法
US20190096049A1 (en) * 2017-09-27 2019-03-28 Korea Advanced Institute Of Science And Technology Method and Apparatus for Reconstructing Hyperspectral Image Using Artificial Intelligence
CN113506235A (zh) * 2021-09-08 2021-10-15 武汉纺织大学 一种对抗曝光变化的自适应加权光谱重建方法
CN115235628A (zh) * 2022-05-17 2022-10-25 中国科学院上海技术物理研究所 光谱重构方法和装置、光谱仪、存储介质和电子设备

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150181131A1 (en) * 2012-07-20 2015-06-25 Carl Zeiss Ag Method and apparatus for image reconstruction
CN104849220A (zh) * 2015-06-09 2015-08-19 武汉大学 一种平面式文物光谱图像获取方法
US20190096049A1 (en) * 2017-09-27 2019-03-28 Korea Advanced Institute Of Science And Technology Method and Apparatus for Reconstructing Hyperspectral Image Using Artificial Intelligence
CN108520488A (zh) * 2018-04-10 2018-09-11 深圳劲嘉集团股份有限公司 一种重构光谱并进行复制的方法以及电子设备
CN108830796A (zh) * 2018-06-20 2018-11-16 重庆大学 基于谱空结合和梯度域损失的高光谱图像超分辨重构方法
CN113506235A (zh) * 2021-09-08 2021-10-15 武汉纺织大学 一种对抗曝光变化的自适应加权光谱重建方法
CN115235628A (zh) * 2022-05-17 2022-10-25 中国科学院上海技术物理研究所 光谱重构方法和装置、光谱仪、存储介质和电子设备

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
ZHANG JIRUI, ZHANG CHUNMIN, YAN TINGYU, CHEN ZEYU: "Spectrum reconstruction for five-channel autocorrelation function of spectropolarimeter", OPTIK., WISSENSCHAFTLICHE VERLAG GMBH, DE, vol. 157, 1 March 2018 (2018-03-01), DE , pages 1259 - 1266, XP093109540, ISSN: 0030-4026, DOI: 10.1016/j.ijleo.2017.12.067 *
潘之玮 (PAN, ZHIWEI): "基于图像融合的多光谱图像超分辨率重建算法 (Non-official translation: Multispectral Image Super-resolution Reconstruction Algorithm Based on Image Fusion)", 浙江大学博士论文 (NON-OFFICIAL TRANSLATION: DOCTORAL DISSERTATIONS OF ZHEJIANG UNIVERSITY), 15 April 2019 (2019-04-15) *

Also Published As

Publication number Publication date
CN115235628A (zh) 2022-10-25
CN115235628B (zh) 2023-12-01

Similar Documents

Publication Publication Date Title
Kitaev et al. The soft mode in the Sachdev-Ye-Kitaev model and its gravity dual
WO2023221588A1 (zh) 光谱重构方法和装置、光谱仪、存储介质和电子设备
Kasiviswanathan et al. Online l1-dictionary learning with application to novel document detection
US20160127405A1 (en) Information processing system, information processing apparatus, and information processing method
TWI804793B (zh) 用於光學關鍵尺寸計量的方法及系統
CN111209661B (zh) 一种光谱椭偏测量带宽和数值孔径退偏效应修正建模方法及装置
EP3816829B1 (en) Detection device and detection method
US9147162B2 (en) Method for classification of newly arrived multidimensional data points in dynamic big data sets
US20230375405A1 (en) Method and device for reconstructing spectrum, spectrometer, non-transitory computer-readable storage medium, and electronic device
Li et al. Poisson phase retrieval with Wirtinger flow
Ezzell et al. The quantum low-rank approximation problem
JP2021134115A5 (zh)
CN114693107A (zh) 企业财务风险预测方法、系统和电子设备
CN113450287A (zh) 高分辨率光谱恢复方法、装置和电子设备
Avramidi Heat Kernel on Lie Groups and Maximally Symmetric Spaces
WO2022194040A1 (zh) 基于光谱芯片的图像传感方法、光谱恢复方法、装置和电子设备
CN113421208A (zh) 高分辨率光谱恢复方法
Helstrom Detection of incoherent objects by a quantum-limited optical system
Pla et al. Bona fide riesz projections for density estimation
Yang et al. Large data throughput optimization model with full C order model parallel flow number prediction optical domain
WO2023279338A1 (en) Neural spectral field reconstruction for spectrometer
Theiler et al. Iterative R and R (rotation and remarginalization) for detecting targets in spectral imagery
Jackson et al. The zeros of the QCD partition function
Yamazaki et al. Adaptive projection subspace dimension for the thick-restart Lanczos method
Holzmann The nonrelativistic limit of Dirac operators with Lorentz scalar δ‐shell interactions

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 23806549

Country of ref document: EP

Kind code of ref document: A1