A method, system, device and storage medium for improving spectral image resolution based on nonlinear regression data processing
Through nonlinear regression data processing and fitting algorithm, the problem of limited resolution of the spectral image is solved, and a 4-fold increase in the resolution of the spectral image is achieved, achieving higher accuracy.
Patent Information
- Application Number
- CN202110042951.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-01-13
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2041-01-13
AI Technical Summary
The existing spectrometers have limited resolution when measuring spectral images, which cannot meet the needs of higher accuracy, and hardware breakthroughs are difficult to achieve.
Using a nonlinear regression data processing method, the discrete data of the optical measurement device is processed by fitting the algorithm to solve the transmission function to improve the resolution of the spectral image. The specific steps include calculating the sum of the squares of the distances from the measurement scatter point to the transmission function and finding the deviation. The derivative is connected in parallel to the system of equations, and the solution is to obtain the transmission function parameters, thereby improving the resolution.
Without increasing hardware costs, the resolution of the spectrometer was improved by 4 orders of magnitude to reach 0.0012pm, far exceeding the accuracy of existing equipment on the market.
Smart Images

Figure CN112700373B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method, system, device and storage medium for improving the resolution of spectral images based on nonlinear regression data processing, and belongs to the technical field of optical measurement. Background Art
[0002] Optical measurement is widely used in scientific research and production: for example, in the field of fiber Bragg grating sensing, which is widely used in bridge stress, power cabinet temperature, etc., the target parameters are mainly determined by detecting the central wavelength of the spectrum returned by the fiber Bragg grating; in the medical field, the concentration of target components in liquids is also detected by measuring the central wavelength of the transmission spectrum of structures such as Mach Zehnder interferometers.
[0003] At present, the mainstream method of obtaining a spectrum is to measure it with a spectrometer. The spectrometer measures from a certain wavelength to a certain wavelength, with a step size of a tiny wavelength. The corresponding light power is given at each wavelength, and all the data are connected to form an image of the spectrum. Therefore, in fact, the data obtained by the spectrometer is discrete. Taking the ANRITSU (Anritsu) brand MS9710C spectrometer as an example, its measurement step size (the professional term is wavelength resolution) is 0.05nm; another example is the YOGOKAWA (Yokogawa) brand AQ6374, its wavelength resolution is also 0.05nm. 0.05nm is the mainstream resolution of high-end spectrometers. Therefore, the data image obtained by the spectrometer is like Figure 1 As shown, from Figure 1 It can be clearly seen that with the current hardware level, the central wavelength value that can be obtained is 1551.45nm, which cannot be finer. This resolution is currently achievable with a 200,000 yuan device, but it is far from meeting current needs. If we seek a breakthrough in hardware, it is very difficult with the current hardware technology level. Table 1 gives the readings of a spectrometer with an actual resolution of 0.05nm.
[0004] Table 1
[0005]
[0006]
[0007] Optical measurement equipment such as spectrometers measure the optical power value at a certain wavelength. Many power values at many wavelengths constitute a spectral image. The measurement of optical power has its limits. The minimum resolution that most spectrometers can detect is 1nW. This means that the optical power value at a certain wavelength given by the spectrometer is 3.586μW. If it is measured in microwatts, it can only be given to the third decimal place. If it is measured in nanowatts, it can only be given to the unit place, which is not accurate enough. Summary of the invention
[0008] In view of the shortcomings of the prior art, the present invention provides a method for improving the resolution of spectral images based on nonlinear regression data processing;
[0009] The present invention also provides a system, a computer device and a storage medium for improving the resolution of spectral images based on nonlinear regression data processing;
[0010] The method of the present invention can improve the resolution of optical measuring equipment such as a spectrometer when searching for a central wavelength by 4 orders of magnitude without changing any hardware.
[0011] The technical solution of the present invention is:
[0012] A method for improving the resolution of spectral images based on nonlinear regression data processing comprises the following steps:
[0013] (1) Obtaining discrete data through optical measurement equipment;
[0014] (2) Use fitting algorithms to process discrete data and restore the transfer function, which is the high-resolution spectral image obtained; the transfer function refers to the output spectrum of the optical device divided by the input spectrum. If the input is flat white light (the power of various wavelengths is equal), the output spectrum of the optical device can be regarded as the transfer function. The transfer function represents all the working characteristics and all the information of the optical device. If the input is not flat white light, then the output spectrum can be divided by the input spectrum (power at the same wavelength is divided), and the result is the transfer function.
[0015] According to the preferred embodiment of the present invention, step (2) specifically refers to: setting m parameters k1, k2, ..., k m The transfer function is The n groups of data points (x1, y1), (x2, y2), …, (x n ,y n ), find the m parameters k1, k2, ..., k of the transfer function m , and we get the transfer function.
[0016] Further preferably, the specific method includes:
[0017] A. Calculate the measurement scatter points, i.e. n sets of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0018] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0019] B. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is where the partial derivative is zero.
[0020] For m parameters k1, k2, ..., k m Calculate the partial derivatives respectively and establish them into a system of equations, as shown in formula (I):
[0021]
[0022] By solving equation (I), we can get k1, k2, …, k m The value of is used to solve for the transfer function.
[0023] In actual measurement, when the real transfer function changes by a very small amount (much less than 0.05nm), the traditional measurement method that only uses the maximum value cannot obtain this change. However, in the method provided by the present invention, as long as any data point in the n sets of data changes, the fitting result will be different. The n data have a synergistic effect, and the resolution that can be achieved is far better than 0.05nm.
[0024] Further preferably, the optical measuring device is a spectrometer.
[0025] Preferably, according to the present invention, when the optical device is a Mach Zehnder interferometer, the transmission function is set as shown in formula (II):
[0026]
[0027] In formula (II), M is the diffraction order, a, b, λ0 are three unknown parameters;
[0028] The spectrum is not symmetrical about the central wavelength of a certain order. There are errors in using parabola fitting, so a special formula is needed for fitting.
[0029] a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0030] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0031] b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero.
[0032] Expand the square term and take partial derivatives with respect to a, b, and λ0 to obtain formula (III):
[0033]
[0034] By solving equation (III), we can obtain the values of a, b, and λ0 and solve the transfer function.
[0035] Preferably, according to the present invention, when the optical device is a waveguide array grating, the transmission function is set as shown in formula (IV):
[0036] f(x)=a0+a1x+a2x 2 (IV)
[0037] In formula (IV), a0, a1, a2 are three parameters to be determined;
[0038] a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0039] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0040] b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero.
[0041] Expand the square term and take partial derivatives with respect to a0, a1, a2 to obtain formula (V):
[0042]
[0043] By solving equation (V), we can obtain the values of a0, a1, and a2, and solve the transfer function.
[0044] A system for improving the resolution of spectral images based on nonlinear regression data processing, comprising a discrete data acquisition module and a transfer function obtaining module.
[0045] The discrete data acquisition module is used to: acquire discrete data by measuring with an optical measuring device;
[0046] The transfer function obtaining module is used to process discrete data using a fitting algorithm to restore the transfer function, which is the obtained high-resolution spectral image.
[0047] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for improving the resolution of a spectral image based on nonlinear regression data processing are implemented.
[0048] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for improving the resolution of a spectral image based on nonlinear regression data processing.
[0049] The beneficial effects of the present invention are:
[0050] When the present invention performs optical measurement of the central wavelength, the resolution of the device can be increased by 4 orders of magnitude to 0.0012pm by simply relying on scientific data processing without increasing the hardware cost of the device; if the method proposed by the present invention is not adopted, the existing commercial devices on the market cannot achieve such a resolution. The present invention is applicable to all optical detections that require detection of the central wavelength, and its application scope is of great significance. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] Figure 1 Schematic diagram of discrete data obtained for an existing optical spectrum analyzer;
[0052] Figure 2 It is a schematic diagram of the change of the fitted spectrum center wavelength λ0 when the wavelength of the spectrum changes slightly;
[0053] Figure 3 is a schematic diagram of discrete data obtained and the obtained transfer function when the optical device is a waveguide array grating;
[0054] Figure 4 is a schematic diagram of discrete data obtained and the obtained transfer function when the optical device is a Mach Zehnder interferometer;
[0055] Figure 5 The structure block diagram of the system for improving the resolution of spectral images based on nonlinear regression data processing of the present invention. DETAILED DESCRIPTION
[0056] The present invention will be further defined below in conjunction with the accompanying drawings and embodiments, but is not limited thereto.
[0057] Example 1
[0058] A method for improving the resolution of spectral images based on nonlinear regression data processing comprises the following steps:
[0059] (1) Obtaining discrete data through optical measurement equipment;
[0060] (2) Use fitting algorithms to process discrete data and restore the transfer function, which is the high-resolution spectral image obtained; the transfer function refers to the output spectrum of the optical device divided by the input spectrum. If the input is flat white light (the power of various wavelengths is equal), the output spectrum of the optical device can be regarded as the transfer function. The transfer function represents all the working characteristics and all the information of the optical device. If the input is not flat white light, then the output spectrum can be divided by the input spectrum (power at the same wavelength is divided), and the result is the transfer function.
[0061] Step (2) specifically refers to: setting m parameters k1, k2, ..., k m The transfer function is The n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ), find the m parameters k1, k2, ..., k of the transfer function m , and the transfer function is obtained. The specific methods include:
[0062] A. Calculate the measurement scatter points, i.e. n sets of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0063] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0064] B. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is where the partial derivative is zero.
[0065] For m parameters k1, k2, ..., k m Calculate the partial derivatives respectively and establish them into a system of equations, as shown in formula (I):
[0066]
[0067] By solving equation (I), we can get k1, k2, …, k m The value of is used to solve for the transfer function.
[0068] In actual measurement, when the real transfer function changes by a very small amount (much less than 0.05nm), the traditional measurement method that only uses the maximum value cannot obtain this change. However, in the method provided by the present invention, as long as any data point in the n sets of data changes, the fitting result will be different. The n data have a synergistic effect, and the resolution that can be achieved is far better than 0.05nm.
[0069] The optical measurement device is a spectrum analyzer.
[0070] Example 2
[0071] The method for improving the resolution of spectral images based on nonlinear regression data processing according to Example 1 is different in that:
[0072] Taking the most common Mach Zehnder interferometer in optical devices as an example, its transmission function is shown in formula (II):
[0073]
[0074] Among them, a, b, and λ0 are three unknown parameters.
[0075] It is necessary to use nonlinear regression theory and rely on formula (I) to find these three parameters from a set of measurement data so that the sum of the distances from the transmission function to each data is minimized.
[0076] Assume that the device power is 100 microwatts (this is a very typical power for integrated optical devices), the minimum power resolution of the optical measurement equipment is set to 1nW, and the minimum wavelength resolution is 0.02nm. That is to say, in the calculation program, when sampling the function, the interval of x should be 0.02, and y should be taken to the fifth decimal place. In the simulation calculation, the wavelength range we sampled is 1520nm-1580nm. There are 1200 sets of data points in this range, that is, n=1200 in formula (Ⅰ).
[0077] Table 2 shows the data of actual wavelength change and fitted wavelength;
[0078] Table 2
[0079]
[0080]
[0081] Figure 2 The figure shows the change of the fitted spectrum center wavelength λ0 when there is a slight change in the wavelength of the spectrum. Figure 2 It can be seen that when the wavelength changes to 1.94×10 -5 nm to 2.05×10 -5 nm, the fitting result does not change and is always 1550.3233451018nm until the wavelength becomes 2.06×10 -5 nm, the fitting wavelength becomes 1550.3233453856 nm. Therefore, it can be considered that the resolution can be improved to 2.06×10 -5 -1.94×10 -5=1.2×10 -6 nm, which means that in the process of finding the central wavelength, although the hardware has not changed, the actual resolution that can be achieved has been improved by 0.05÷(1.2×10 -6 ) = 41666.7 times, more than 40,000 times, which is more than 4 orders of magnitude.
[0082] The spectrum is not symmetrical about the central wavelength of a certain order. There are errors in using parabola fitting, so a special formula is needed for fitting.
[0083] a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0084] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0085] b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero.
[0086] Expand the square term and take partial derivatives with respect to a, b, and λ0 to obtain formula (III):
[0087]
[0088] By solving equation (III), we can obtain the values of a, b, and λ0, and use the numerical method to solve the transfer function.
[0089] To prove the validity of the formula, here we provide data with only three data points: (1510, 0.453256546), (1541, 0.966711020), (1544, 0.985169733). From equation group (III), we can calculate a=0.5, b=0.5, λ0=1550.00000. In fact, these three data points come from This proves the correctness of the method and equation group (III) proposed in the present invention.
[0090] Figure 4 What is given is the application for a set of data points, the circles are the data points (discrete data), and the curves are the results after fitting (transfer function).
[0091] Example 3
[0092] The method for improving the resolution of spectral images based on nonlinear regression data processing according to Example 1 is different in that:
[0093] When the optical device is a waveguide array grating, the central wavelength is measured using the waveguide array grating. The output spectrum of the waveguide array grating is a Gaussian-like shape that is symmetrical about the central wavelength. For such optical devices, parabola fitting can meet the actual requirements for finding the central wavelength. The transmission function is set as shown in formula (IV):
[0094] f(x)=a0+a1x+a2x 2 (IV)
[0095] In formula (IV), a0, a1, a2 are three parameters to be determined;
[0096] a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function;
[0097] The square of the distance from the measured scatter point to the transfer function is The sum of the squares of the distances from each scattered point to the transfer function is
[0098] b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero.
[0099] Expand the square term and take partial derivatives with respect to a0, a1, a2 to obtain formula (V):
[0100]
[0101] By solving equation (V), we can obtain the values of a0, a1, and a2, and solve the transfer function.
[0102] Table 3 shows 36 data points:
[0103] Table 3
[0104]
[0105]
[0106] According to equation group (V), the fitting results are a0 = -10.43583, a1, = 0.0135, a2 = -4.36687x10 -6 . From this, we can calculate that the central wavelength value is 1545.72955nm. The fitting results are as follows Figure 3 As shown, the dots are 36 groups of data points (as shown in Table 3, the dot test curve is distorted to a certain extent due to the test error), and the curve is the quadratic function after fitting.
[0107] Example 4
[0108] A system for improving the resolution of spectral images based on nonlinear regression data processing, such as Figure 5 As shown, it includes a discrete data acquisition module and a transfer function calculation module. The discrete data acquisition module is used to: obtain discrete data by measuring with an optical measuring device; the transfer function calculation module is used to: process discrete data using a fitting algorithm to restore the transfer function. The transfer function refers to the shape law of the real output spectrum of the optical device when the input light is flat white light, so as to obtain a high-resolution spectral image.
[0109] Example 5
[0110] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a method for improving the resolution of a spectral image based on nonlinear regression data processing are implemented.
[0111] Example 6
[0112] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for improving the resolution of a spectral image based on nonlinear regression data processing.
Claims
1. A method for improving the resolution of spectral images based on nonlinear regression data processing, characterized in that: The steps include: (1) Obtaining discrete data through optical measurement equipment; (2) Using a fitting algorithm to process discrete data and restore the transfer function, which is the high-resolution spectral image obtained; the transfer function refers to the output spectrum of the optical device divided by the input spectrum; Step (2) specifically refers to: setting m parameters k1, k2, ..., k m The transfer function is The n groups of data points (x1, y1), (x2, y2), …, (x n ,y n ), find the m parameters k1, k2, ..., k of the transfer function m , that is, the transfer function; When the optical device is a Mach Zehnder interferometer, the transmission function is set as shown in formula (II): In formula (II), M is the diffraction order, a, b, λ0 are three unknown parameters; a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function; The square of the distance from the measured scatter point to the transfer function is i=1,2,…,n, the sum of the squares of the distances from each scattered point to the transmission function is b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero. Expand the square term and take partial derivatives with respect to a, b, and λ0 to obtain formula (III): By solving equation (III), we can obtain the values of a, b, and λ0 and solve the transfer function.
2. The method for improving spectral image resolution based on nonlinear regression data processing according to claim 1, characterized in that: Specific methods include: A. Calculate the measurement scatter points, i.e. n sets of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function; The square of the distance from the measured scatter point to the transfer function is i=1,2,…,n, the sum of the squares of the distances from each scattered point to the transmission function is B. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is where the partial derivative is zero. For m parameters k1, k2, ..., k m Calculate the partial derivatives respectively and establish them into a system of equations, as shown in formula (I): By solving equation (I), we can get k1, k2, …, k m The value of is used to solve for the transfer function.
3. The method for improving spectral image resolution based on nonlinear regression data processing according to claim 1, characterized in that: The optical measuring device is a spectrum analyzer.
4. The method for improving spectral image resolution based on nonlinear regression data processing according to claim 1, characterized in that: When the optical device is a waveguide array grating, the transmission function is set as shown in formula (IV): f(x)=a0+a1x+a2x 2 (Ⅳ) In formula (IV), a0, a1, a2 are three parameters to be determined; a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function; The square of the distance from the measured scatter point to the transfer function is i=1,2,…,n, the sum of the squares of the distances from each scattered point to the transmission function is b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero. Expand the square term and take partial derivatives with respect to a0, a1, a2 to obtain formula (V): By solving equation (V), we can obtain the values of a0, a1, and a2, and solve the transfer function.
5. A system for improving spectral image resolution based on nonlinear regression data processing, characterized in that: The invention comprises a discrete data acquisition module and a transfer function acquisition module. The discrete data acquisition module is used to obtain discrete data by measuring with an optical measuring device. The transfer function acquisition module is used to process the discrete data using a fitting algorithm to restore the transfer function, which is the obtained high-resolution spectral image. The transfer function refers to the output spectrum of the optical device divided by the input spectrum. Specifically, m parameters k1, k2, ..., k are set. m The transfer function is The n groups of data points (x1, y1), (x2, y2), …, (x n ,y n ), find the m parameters k1, k2, ..., k of the transfer function m , that is, the transfer function; When the optical device is a Mach Zehnder interferometer, the transmission function is set as shown in formula (II): In formula (II), M is the diffraction order, a, b, λ0 are three unknown parameters; a. Calculate the measurement scatter points, i.e. n groups of data points (x1, y1), (x2, y2), ..., (x n ,y n ) is the sum of the squares of the distances to the transfer function; The square of the distance from the measured scatter point to the transfer function is i=1,2,…,n, the sum of the squares of the distances from each scattered point to the transmission function is b. Calculate the partial derivative for each parameter. The function with the minimum distance to each measurement point is the point where the partial derivative is zero. Expand the square term and take partial derivatives with respect to a, b, and λ0 to obtain formula (III): By solving equation (III), we can obtain the values of a, b, and λ0 and solve the transfer function.
6. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of the method for improving the resolution of a spectral image based on nonlinear regression data processing according to any one of claims 1 to 4 when executing the computer program.
7. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed by a processor, the steps of the method for improving the resolution of a spectral image based on nonlinear regression data processing according to any one of claims 1 to 4 are implemented.
Citation Information
Patent Citations
Method for comprehensively evaluating optical remote sensing image quality
CN103268602A
Image super-resolution and image quality enhancement method
CN103310427A
Phase recovery method for extended target imaging in Poisson noise model
CN110146258A
Spectral polarization morphology five-dimensional information measuring device and method based on interference illumination
CN111750799A
Hyperspectral image super-resolution method based on depth prior information
CN112184553A
