Disclosure of Invention
The application aims to provide a space target pBRDF model construction method based on an improved SBA algorithm, which can improve the accuracy of polarization characteristic measurement of the space target.
In order to achieve the above object, the present application provides the following.
In a first aspect, the present application provides a method for constructing a spatial target pBRDF model based on an improved SBA algorithm, where the method for constructing a spatial target pBRDF model based on the improved SBA algorithm includes the following steps.
Initially constructing pBRDF a model, and obtaining a polarization characteristic measurement result of a sample to be measured, wherein the polarization characteristic measurement result comprises the sample to be measured under various anglesValues.
And initializing the position of the parent plant by using a Tent chaotic mapping algorithm in the improved SBA algorithm, and updating the position of the creeping stems of the parent plant by using a simulated annealing algorithm.
And carrying the parameter inversion result into the pBRDF model to calculate errors of simulation values and actual measurement values of the model, and selecting whether to output the pBRDF model according to the error precision to obtain a final pBRDF model, wherein the final pBRDF model is used for analyzing the infrared polarization characteristics of the sample to be tested under multiple angles.
Optionally, a model pBRDF is initially constructed, specifically including the following steps.
According to pBRDF theory, light elastic scattering theory and blackbody radiation theory, initially establishing a pBRDF model of the target material, and determiningThe target material refers to a material corresponding to the sample to be detected.
Optionally, according to the polarization characteristic measurement result, performing parameter inversion by adopting an improved SBA algorithm to obtain a parameter inversion result, and specifically comprising the following steps.
Defining initial parameters of an SBA algorithm, wherein the initial parameters comprise the number of inversion parameters, the initial parent plant population number, the maximum iteration times of a strawberry optimization algorithm and a Tent chaotic mapping algorithm and the maximum iteration times of a simulated annealing algorithm.
And initializing the position of the parent plant by adopting a Tent chaotic mapping algorithm according to the initial parameters to obtain the initialized position of the parent plant.
Generating the son root and the creeping stem of the parent plant randomly according to the initialized parent plant position to obtain a plant propagation matrix, wherein each parent plant in the plant propagation matrix generates one son root and one creeping stem randomly.
And according to the plant propagation matrix, updating the position of the stolons by adopting a simulated annealing algorithm to obtain the updated position of the stolons.
And based on the updated position of the stolons, carrying out fitness calculation by taking the minimum objective function solution as a target, and selecting a mother plant population of the next iteration by utilizing the calculated fitness value and a roulette algorithm.
And determining whether to output a parameter inversion result according to whether any one of the termination conditions is met, wherein the parameter inversion result comprises optimal inversion values of a plurality of parameters to be inverted, and the plurality of parameters to be inverted comprise sigma, n, ks、km and kv, wherein sigma is roughness, n is the real part of complex refractive index, k is the imaginary part of complex refractive index, ks is specular reflectivity, km is diffuse reflectivity and kv is bulk scattering rate.
And representing inversion precision according to the average value of the relative errors of the inversion values and the measured values of the parameters sigma, n and k to be inverted.
Optionally, the parameter inversion result is brought into the pBRDF model to calculate an error between a simulation value and an actual measurement value of the model, and whether to output the pBRDF model is selected according to the error precision, so as to obtain a final pBRDF model, which specifically comprises the following steps.
Substituting the optimal inversion value corresponding to each parameter to be inverted in the parameter inversion result into the parameter inversion resultIn the formula, the simulation of the pBRDF model is calculatedValues.
The simulation of the pBRDF model using an evaluation functionAnd evaluating the values, and determining a final pBRDF model according to the evaluation result.
Optionally, said simulation of said pBRDF model with an evaluation functionAnd evaluating the values, and determining a final pBRDF model according to an evaluation result, wherein the method specifically comprises the following steps.
Calculating the simulation of the pBRDF model using MAE functionValues and experimental measurementsMAE values between values, measured by the experimentThe value refers to the measured polarization characteristic measurement resultValues.
And determining a final pBRDF model according to the MAE value and a preset threshold value.
Optionally, determining a final pBRDF model according to the MAE value and a preset threshold, which specifically includes the following steps.
Determining the simulation of the pBRDF model when the MAE value is less than 5%The values are within reasonable threshold ranges, with the current pBRDF model being the final pBRDF model.
Determining the simulation of the pBRDF model when the MAE value is greater than or equal to 5%And (3) when the value is not in a reasonable threshold range, improving the pBRDF model and returning to the step of initializing the parent plant position by adopting a Tent chaotic mapping algorithm according to the initial parameters to obtain the initialized parent plant position until the MAE value is less than 5%.
Optionally, obtaining a measurement result of polarization characteristics of the sample to be measured specifically includes the following steps.
And measuring the sample to be measured by using a space target infrared polarization characteristic measuring device to obtain a polarization characteristic measuring result of the sample to be measured.
The space target infrared polarization characteristic measuring device comprises a gas generation system, a closed moving system, an electric object stage, a light source system, a polarization acquisition system, a control system and a computer.
The gas generation system is communicated with the inside of the closed moving system through a pipeline, the electric object stage is arranged in the closed moving system, the light source system and the polarization acquisition system are arranged in the closed moving system, the gas generation system, the electric object stage, the light source system and the polarization acquisition system are all connected with the control system, and the control system is also connected with the computer.
The gas generation system is used for generating mixed gas and filling the mixed gas into the closed mobile system through the pipeline, wherein the mixed gas comprises aerosol and water vapor.
The electric object stage is used for placing a sample to be tested and enabling the sample to be tested to rotate.
The light source system is used for emitting infrared light beams to the sample to be detected.
The polarization acquisition system is used for acquiring polarization images of the sample to be detected.
The closed moving system is used for providing a closed environment for the mixed gas, enabling the light source system to move in the closed environment to emit the infrared light beams at multiple angles, and enabling the polarization acquisition system to move in the closed environment to acquire the polarization images at multiple angles.
The control system is used for receiving the instruction sent by the computer, respectively controlling the working states of the gas generating system, the electric object stage, the light source system and/or the polarization acquisition system according to the instruction, and acquiring the polarization image and transmitting the polarization image to the computer.
And the computer is used for analyzing and processing the polarized image to obtain a polarization characteristic measurement result of the sample to be measured.
Optionally, the closed moving system comprises a semicircular closed box and a semicircular guide rail.
The center of the top of the semicircular closed box is communicated with the gas generation system through the pipeline, and the electric object stage is arranged right below the center of the top of the semicircular closed box.
The semicircular guide rail is laid on the inner surface of the semicircular closed box and is connected with the control system.
The semicircular guide rail is slidably provided with the light source system and the polarization acquisition system.
Optionally, the semicircular guide rail comprises a first guide rail and a second guide rail.
The first guide rail is slidably provided with the light source system, and the second guide rail is slidably provided with the polarization acquisition system.
The first guide rail and the second guide rail are both provided with first angle marks, the first angle marks are 0-90 degrees, and the indexing value of the first angle marks is 1 degree.
The first angle marks of 0 degree of the first guide rail and the second guide rail are positioned at one end close to the top of the semicircular closed box.
Optionally, the electric object stage is provided with a second angle mark, the second angle mark is 0-360 degrees, the indexing value of the second angle mark is 1 degree, and the second angle mark of 0 degree of the electric object stage is opposite to the first guide rail.
According to the specific embodiments provided by the application, the following technical effects are disclosed.
The application provides a space target pBRDF model construction method based on an improved SBA algorithm, which comprises the steps of firstly initially constructing pBRDF model and obtaining the polarization characteristic measurement result of a sample to be measured, wherein the measurement result comprises the sample to be measured under various anglesThe method comprises the steps of obtaining a model, obtaining a parameter inversion result, carrying out parameter inversion by adopting an improved SBA algorithm according to a polarization characteristic measurement result, initializing a parent plant position by using a Tent chaotic mapping algorithm in the improved SBA algorithm, updating the position of a creeping stem of the parent plant by using a simulated annealing algorithm, carrying the parameter inversion result into a initially constructed pBRDF model to calculate errors of a model simulation value and an actual measurement value, and selecting whether to output a pBRDF model according to the error precision to obtain a final pBRDF model. The final pBRDF model is used for analyzing the long-wave infrared polarization characteristics of the sample to be detected under multiple angles, and can improve the accuracy of polarization detection of a space target.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present application, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
At present, the existing polarization measuring device does not consider the influence of the atmospheric environment on the measurement precision, but the measured multi-angleThe values are the basis data for the parametric inversion and construction pBRDF of the model. In addition, the traditional parameter inversion method mainly comprises a nonlinear least square method and a particle swarm optimization algorithm. The nonlinear least square method is sensitive to an initial value, requires a great amount of experimental data, has large calculation amount, can be inaccurate when a particle swarm optimization algorithm is used for processing some specific problems, and can be influenced by random factors, so that the robustness is low.
Based on this, the embodiment aims to provide a space target pBRDF model construction method based on an improved SBA (Strawberry optimization ) algorithm, wherein the pBRDF model is established according to pBRDF theory, elastic scattering theory of light and blackbody radiation theory, and the pBRDF model is used as a space target infrared polarization characteristic model, can be used for further analyzing infrared polarization characteristics of the space target under multiple angles, and infrared refers to a wave band of 8-14 μm. Firstly, the infrared polarization characteristic measuring device of the space target is utilized to accurately measure and simulate the infrared polarization characteristic of the space target in the atmospheric environment, a more accurate, reliable and multi-angle polarization characteristic measuring result is obtained, and meanwhile, a preliminary pBRDF model is constructed. And then, applying the polarization characteristic measurement result to the process of perfecting the pBRDF model, and obtaining a final pBRDF model through parameter inversion and evaluation verification. The method can effectively solve the problems of dependence on initial values and poor robustness of the traditional parameter inversion method, overcomes the defects of low convergence speed and easy sinking into a local optimal solution of the traditional SBA algorithm, and can improve the accuracy of parameter inversion of the metal object.
In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to the appended drawings and appended detailed description.
As shown in fig. 1, the present embodiment provides a spatial target infrared polarization characteristic measurement device, which includes a gas generation system 1, a closed moving system, an electric stage 4, a light source system 5, a polarization acquisition system 6, a control system 7, and a computer 8.
The gas generating system 1 is communicated with the inside of the closed mobile system through a pipeline, the electric object stage 4 is arranged in the inside of the closed mobile system, the light source system 5 and the polarization acquisition system 6 are arranged in the inside of the closed mobile system, the gas generating system 1, the electric object stage 4, the light source system 5 and the polarization acquisition system 6 are all connected with the control system 7, and the control system 7 is also connected with the computer 8.
In this embodiment, the gas generating system 1 is configured to generate a mixed gas, and fill the mixed gas into the closed mobile system through the pipe. The mixed gas comprises aerosol and water vapor. The gas generating system 1 adopts the HRF-4B series gas generating system 1 of Suzhou Honey purification technology Co., ltd, and is used for generating mixed gas of aerosol and water vapor with different proportions.
In this embodiment, the light source system 5 is configured to emit an infrared beam toward the sample to be measured. The light source system 5 is a blackbody light source. The blackbody light source adopts SLS303 series light source of Thorlabs, which can generate 550nm-15 μm beam with output beam power of 4.5w and output beam diameter of 50mm.
In this embodiment, the polarization acquisition system 6 is configured to acquire a polarization image of the sample to be measured. The polarization acquisition system is a long-wave infrared polarization camera, the long-wave infrared polarization camera adopts a polarization camera of GWPL X0318A model with wide north, the polarization acquisition system is a focal plane polarization camera, the response wave band is 8-14 mu m, and the polarization acquisition system can be used for acquiring long-wave infrared polarization information of an object.
In this embodiment, the closed moving system is configured to provide a closed environment for the mixed gas, and enable the light source system 5 to move in the closed environment to emit the infrared light beam at multiple angles, and enable the polarization acquisition system 6 to move in the closed environment to acquire the polarized image at multiple angles.
In this embodiment, the control system 7 is configured to receive an instruction sent by the computer 8, and control the working states of the gas generating system 1, the electric stage 4, the light source system 5 and/or the polarization acquisition system 6 according to the instruction, and acquire the polarization image and transmit the polarization image to the computer 8.
In this embodiment, the computer 8 is configured to analyze the polarization image to obtain a measurement result of the polarization characteristic of the sample to be measured. Wherein the polarization characteristic measurement result comprises the sample to be measured under various anglesValues. In this embodiment, the polarization characteristic measurement result obtained by the measurement of the spatial target infrared polarization characteristic measurement device may be used for parameter inversion and construction and output pBRDF models to further analyze the infrared polarization characteristics of the spatial target under multiple angles in detail.
In this embodiment, the computer 8 has built therein a confast software for extracting the degree of polarization of the polarized image.
In this embodiment, the closed moving system includes a semicircular closed box 2 and a semicircular guide rail 3. The top center of the semicircular closed box 2 is communicated with the gas generation system 1 through the pipeline, and the electric object stage 4 is arranged right below the top center of the semicircular closed box 2. The semicircular guide rail 3 is laid on the inner surface of the semicircular closed box 2, and the semicircular guide rail 3 is connected with the control system 7. The semicircular guide rail 3 is slidably provided with the light source system 5 and the polarization acquisition system 6.
In this embodiment, the semicircular guide rail 3 is a customized model, and is divided into a left side and a right side, wherein the left side semicircular guide rail 3 is a first guide rail and is used for dynamically measuring an angle between 0 ° and 90 °. The right semicircular guide rail 3 is a second guide rail and is used for dynamically measuring an angle between 0 degrees and 90 degrees.
The first guide rail is slidably provided with the light source system 5, that is, the light source system 5 can slide along the first guide rail, and the second guide rail is slidably provided with the polarization acquisition system 6, that is, the polarization acquisition system 6 can slide along the second guide rail.
In this embodiment, first angle marks are provided on the first guide rail and the second guide rail, the first angle marks are 0 ° to 90 °, and the index value of the first angle marks is 1 °. Wherein, the first angle sign of 0 degree of first guide rail and the second guide rail is located near the one end at semi-circular seal box 2 top.
In this embodiment, the electric stage 4 is used for placing a sample to be tested and rotating the sample to be tested. The electric object stage 4 adopts DDR100 series electric object stages of Thorlabs company, dynamically measures the angle between 0 degree and 360 degrees, and is provided with a second angle mark, the second angle mark is 0 degree to 360 degrees, the indexing value of the second angle mark is 1 degree, wherein the second angle mark of 0 degree is opposite to the first guide rail.
In the present embodiment, the control system 7 includes a gas generation control system, a light source control system, an imaging device control system, and an electric turntable control system. The gas generation control system is used for controlling the working state, the working power and the like of the gas generation system 1. The light source control system is used for controlling the working state, working power and the like of the light source system 5. The imaging device control system is used for controlling the working state, working power and the like of the long-wave infrared polarization camera 6. The electric turntable control system is used for controlling the working state, the rotation angle and the like of the electric object stage 4.
In this embodiment, the sample to be measured is a wafer made of three common metal materials of aluminum alloy, titanium alloy and stainless steel, and the diameter and the beam aperture of the wafer are equal and are both 50mm.
In an exemplary embodiment, a method for constructing a spatial target pBRDF model based on an improved SBA algorithm is provided, and the method for constructing the spatial target pBRDF model based on the improved SBA algorithm can be applied to the measurement of the polarization characteristic by the spatial target infrared polarization characteristic measurement device, and performs parameter inversion and verification on the basis of the measurement of the polarization characteristic, so as to determine a final pBRDF model. As shown in FIG. 2, the method for constructing the space object pBRDF model based on the improved SBA algorithm mainly comprises the following steps.
And S1, initially constructing pBRDF a model, and acquiring a polarization characteristic measurement result of a sample to be measured.
In this embodiment, the spatial target infrared polarization characteristic measurement device may be used to measure the polarization characteristic measurement result of the sample to be measured, to obtain the polarization characteristic measurement result of the sample to be measured at various anglesValues.
And S2, performing parameter inversion by adopting an improved SBA algorithm according to the polarization characteristic measurement result to obtain a parameter inversion result.
In this embodiment, compared with the conventional SBA algorithm, the improved SBA algorithm is mainly improved in that the Tent chaotic mapping algorithm is used to initialize the position of the parent plant, and the simulated annealing algorithm is used to update the position of the stolons of the parent plant.
And S3, bringing the parameter inversion result into the pBRDF model to calculate errors of the simulation value and the actual measurement value of the model, and selecting whether to output the pBRDF model according to the error precision to obtain a final pBRDF model. The final pBRDF model can be used for further comprehensively and in detail analyzing the infrared polarization characteristics of the sample to be tested under multiple angles.
In this embodiment, step S1 initially builds pBRDF a model, which specifically includes the following steps.
According to pBRDF theory, light elastic scattering theory and blackbody radiation theory, initially establishing a pBRDF model of the target material, and determiningThe target material refers to a material corresponding to the sample to be detected.
In this embodiment, step S2 performs parameter inversion by using an improved SBA algorithm according to the polarization characteristic measurement result, to obtain a parameter inversion result, and specifically includes the following steps.
And S21, defining initial parameters of an SBA algorithm, wherein the initial parameters comprise the number of inversion parameters, the initial parent plant population number, the maximum iteration times of a strawberry optimization algorithm and a Tent chaotic mapping algorithm and the maximum iteration times of a simulated annealing algorithm.
And S22, initializing the position of the parent plant by adopting a Tent chaotic mapping algorithm according to the initial parameters to obtain the initialized position of the parent plant.
Step S23, generating the son root and the creeping stem of the mother plant randomly according to the initialized mother plant position to obtain a plant propagation matrix, wherein each mother plant in the plant propagation matrix generates one son root and one creeping stem randomly.
And step S24, according to the plant propagation matrix, updating the position of the stolons by adopting a simulated annealing algorithm to obtain the updated position of the stolons.
And S25, calculating fitness by taking the minimum objective function solution as a target based on the updated position of the stolon, and selecting a new parent population for the next iteration by utilizing the fitness value and a roulette algorithm.
And step S26, judging whether to output a parameter inversion result according to whether any termination condition is met. The parameter inversion result comprises optimal inversion values of a plurality of parameters to be inverted. The plurality of parameters to be inverted comprise sigma, n, ks、km and kv, wherein sigma is roughness, n is the real part of complex refractive index, k is the imaginary part of complex refractive index, ks is specular reflectivity, km is diffuse reflectivity, and kv is bulk scattering rate. Wherein the termination condition is as follows.
(1) The three algorithms of the Tent chaotic map, the strawberry optimization algorithm and the simulated annealing algorithm all reach the maximum iteration times.
(2) A set of optimal solutions occurs such that the fitness value of the objective function is greater than 20.
And S27, representing inversion precision according to the average value of the relative errors of inversion values and measured values of parameters sigma, n and k to be inverted, wherein the relative errors are the ratio of absolute error values to true values. The absolute error value is the absolute value of the difference between the measured value and the inversion value of the parameter to be inverted, and the true value is the measured value of the parameter to be inverted.
In this embodiment, step S3 brings the parameter inversion result into the pBRDF model to calculate an error between the simulation value and the actual measurement value of the model, and selects whether to output the pBRDF model according to the error accuracy, so as to obtain a final pBRDF model.
S31, substituting the optimal inversion value corresponding to each parameter to be inverted in the parameter inversion result into the parameter inversion resultIn the formula, the simulation of pBRDF model is calculatedValues.
Step S32, adopting an evaluation function to simulate the pBRDF modelAnd evaluating the values, and determining a final pBRDF model according to the evaluation result.
In this embodiment, step S32 uses an evaluation function to simulate the pBRDF modelsAnd evaluating the values, and determining a final pBRDF model according to an evaluation result, wherein the method specifically comprises the following steps.
Step S321, calculating the simulation of the pBRDF model by using MAE (mean absolute error) functionValues and experimental measurementsMAE values between values. Wherein the experiment is measuredThe value refers to the measured polarization characteristic measurement resultValues, e.g. in the measurement of polarization properties of a sample to be measured using the above-mentioned spatial target infrared polarization property measuring deviceValues.
Step S322, determining a final pBRDF model according to the MAE value and a preset threshold.
In this embodiment, step S322 determines a final pBRDF model according to the MAE value and the preset threshold, which specifically includes the following two cases.
(1) Determining the simulation of the pBRDF model when the MAE value is less than 5%The values are within reasonable threshold ranges, with the current pBRDF model being the final pBRDF model.
(2) Determining the simulation of the pBRDF model when the MAE value is greater than or equal to 5%And returning to the step S22 'initializing the parent plant position by adopting a Tent chaotic mapping algorithm according to the initial parameters after the model is required to be further improved to obtain an initialized parent plant position', re-initializing the parent plant position and obtaining a corresponding parameter inversion result, and carrying the parameter inversion result into the pBRDF model again according to the re-obtained parameter inversion result so as to calculate errors of the model simulation value and the actual measurement value, and selecting whether to output the pBRDF model according to the error precision until the corresponding MAE value is less than 5%.
In order to make the technical solution of the present embodiment clearer, the specific structure of the apparatus and the implementation steps of the method of the present embodiment will be described in detail below in an illustrative manner.
The method for constructing the space target pBRDF model based on the improved SBA algorithm is divided into a polarization information acquisition stage, a parameter inversion stage and a verification stage as shown in fig. 3, wherein the polarization information acquisition stage uses a model creation module and an image signal acquisition module, and the model creation module is mainly used for determining pBRDF theory based on specular reflection components, diffuse reflection components and volume scattering components as shown in fig. 4, and preliminarily constructing pBRDF a model by comprehensively considering pBRDF theory, elastic scattering theory of light and blackbody radiation theory and solvingA formula. As shown in fig. 5, the image signal acquisition module is mainly used for starting to inflate by using the gas generating system 1 and generating a light beam by using a blackbody light source, setting a long-wave infrared light source, a long-wave infrared polarization camera and a moving range and a stepping angle of the electric object table 4 by using the control system 7, wherein the light source on the first guide rail slides within a range of 20-60 degrees according to a stepping angle of 10 degrees, the electric object table 4 rotates once every step of the light source by 120 degrees, the light source and the electric object table 4 slide from 20-60 degrees to 60-60 degrees according to a stepping angle of 2 degrees, and meanwhile, a photo of a sample to be measured is taken, so that the acquisition of polarization signals is realized, then the polarization degree of the photo is extracted by using a contast software, and polarization degree data is obtained by arrangement, namely experimental measurement is obtainedValues.
In this embodiment, the polarization information obtaining stage specifically includes the following steps.
Step 1, the gas generating system 1 starts to charge. The ratio of aerosol to water vapor in the mixed gas generated by the gas generating system 1 is set by the control system 7, and then the gas is blown into the semicircular closed box 2.
The main component of the injected gas is water vapor, and a small amount of aerosol is also adopted, because the two gases in the atmosphere have the greatest influence on the long-wave infrared polarization detection of the space target.
And 2, generating a light beam by using the blackbody light source. The control system 7 is used for setting a long-wave infrared light source, a long-wave infrared polarization camera and the moving range and the stepping angle of the electric object stage 4, wherein the long-wave infrared light source is positioned on the first guide rail, and the long-wave infrared polarization camera is positioned on the second guide rail. The light source of the first guide rail, namely the long-wave infrared light source and the long-wave infrared polarization camera of the second guide rail, slide in the range of 20-60 degrees in zenith angle, the electric object stage 4 rotates in 0-360 degrees, wherein 0 degree of the electric object stage 4 is a position opposite to the first guide rail, the stepping angle of the long-wave infrared light source is 10 degrees, the stepping angle of the long-wave infrared polarization camera 6 is 2 degrees, and the stepping angle of the electric object stage 4 is 120 degrees.
And 3, controlling the long-wave infrared polarization camera 6 to start shooting, wherein the electric object stage 4 rotates one circle every time the long-wave infrared light source steps, and the long-wave infrared polarization camera 6 steps from 20 degrees to 60 degrees every time the long-wave infrared light source steps and the electric object stage 4 steps, and 5 multiplied by 4 multiplied by 21=420 polarized images can be obtained by accumulating one round of acquisition, and the number is 001-420.
And 4, extracting the polarization degree of the photo by using a comparison software. Inputting all polarized images in the data set into a computer 8, and obtaining the polarization degree of the sample to be tested in each polarized image through a const software, so as to finally obtain the polarization degree data of different zenith angles and different azimuth angles of the sample to be tested under different simulated atmospheric environments.
Compared with the traditional nonlinear least square method and particle swarm optimization algorithm, under the experimental data of the same set number, the improved SBA algorithm has higher precision and better robustness for parameter inversion of metal objects, is very suitable for parameter inversion of objects which are mainly made of metal materials and take the influence of atmospheric environment on the polarization characteristics of the objects into consideration, wherein the space targets are mainly made of metal materials. The SBA algorithm is a heuristic algorithm that describes how strawberry plants migrate from place to place and produce multiple sub-roots and stolons. The strawberry plant utilizes the root and the stolons to carry out local and global search to find out life resources, whether the position is rich in resources means whether the objective function solution is good or bad, the root and the stolons are moved to new positions with rich resources to produce offspring plants, the offspring plants are regarded as finding out an optimized solution, and the optimal solution of the objective function can be obtained through continuous iteration. The basic idea of the SBA algorithm in this embodiment is that the process of generating new sub-roots and stolons from each parent root represents inversion of a set of parameters, and the positions of each new root and new stolons represent a set of inverted parameter values, and the parent strain required for the next iteration is selected by combining the fitness value of the objective function and the roulette algorithm, so that the iteration is continued until a predetermined termination condition is met, and a final optimal parameter solution is obtained. Meanwhile, the Tent chaotic mapping algorithm and the simulated annealing algorithm are introduced into the method, so that the traditional SBA algorithm is improved, and the problem that the traditional SBA algorithm is low in convergence speed and easy to fall into a local optimal solution is solved.
In this embodiment, a parameter inversion module is used in the parameter inversion stage, and the workflow of the parameter inversion module specifically includes the following steps.
The parameters to be inverted in this embodiment include roughness σ, real part n of complex refractive index, imaginary part k of complex refractive index, specular reflectivity ks, diffuse reflectivity km, and bulk scattering rate kv.
The position solutions of the neutron root and the stolon in the parameter inversion algorithm represent a group of inverted unknown parameters, each parent plant inverts 6 parameters to be inverted simultaneously, and the obtained optimal solution is the optimal inversion value of the parameters to be inverted in the final inversion.
Step 1, firstly, initially constructing a pBRDF model according to pBRDF theory, light elastic scattering theory and blackbody radiation theory, and further obtaining the polarization degree,The formula is as follows.
。
The known parameters are as follows: Representing the muller matrix of the atmospheric aerosol and water vapor mixture,For the irradiance of the surface of the object,For the incident zenith angle of the light,In order to detect the zenith angle,In order to detect the azimuth angle,Is a mueller matrix of specular reflection components,、、Are allIn the presence of an element of the group,AndRepresenting the diffuse and bulk scatter components respectively,Is a blackbody radiation formula of an object,As a function of the shadow mask,Is the probability distribution function of the surface normal, σ is the roughness, n is the real part of the complex refractive index, and k is the imaginary part of the complex refractive index. The unknown parameters are specular reflectance ks, diffuse reflectance km, and bulk scattering rate kv. Wherein sigma, n, ks、km、kv are parameters to be inverted.
The pBRDF theory describes the reflection characteristics of the target material at various observation angles, the elastic scattering theory of light comprises Rayleigh scattering and Mie scattering, the interaction between light and atmospheric particles is mainly described, and the blackbody radiation theory mainly describes spontaneous radiation of an object.
Firstly, setting initial parameters of an SBA algorithm, wherein the initial parameters comprise the number m of inversion parameters, the initial parent plant population N, tent chaotic mapping algorithm and the maximum iteration number tmax1 of a simulated annealing algorithm, the maximum iteration number tmax2 of a strawberry optimization algorithm, the number m=6 of inversion parameters is generally 30-50, the number N=50 is generally 50-500 due to the fact that the inversion parameters are more, the iteration number is generally 50-500, and the iteration number can be properly reduced due to the fact that the N is larger in value, more searches can be performed for each iteration, and the iteration number tmax1=50,tmax2 =200 can be properly reduced.
And 3, initializing the parent plant position by using a Tent chaotic mapping algorithm.
Because the conventional SBA algorithm uses a random operator to initialize the position of the parent plant, the problems of uneven distribution of the parent plant position, weak global searching capability and low population diversity are easily encountered, so that the problem of local optimum is caused, and the overall optimization efficiency is influenced. Therefore, the parent population is initialized by introducing a Tent chaotic mapping algorithm into the SBA algorithm, and the Tent chaotic mapping algorithm is added with a random number rand (0, 1)/N, so that the randomness, the ergodic property and the regularity of the Tent chaotic mapping algorithm are maintained, and the iteration can be effectively prevented from falling into a small period point and an unstable period point. The Tent chaotic map sequence is defined as follows.
。
。
Wherein, theRepresents the t+1st iteration value of the chaotic map,The t iteration value of the chaotic map is represented by 0< tmax1, rand (0, 1) is a random number uniformly distributed between 0 and 1, N is the initial parent plant population number, Xlb is the lower bound of the position variable, Xub is the upper bound of the position variable,The position of the ith parent strain after the t-th iteration is shown.
In this embodiment, when the parent plant position is initialized by using the Tent chaotic mapping algorithm, as shown in fig. 6, the position of the initial population is calculated first, then a random number sequence rand (0, 1) in the [0,1] interval is generated, then the chaotic distribution value of each individual in the [0,1] interval is generated by using the Tent mapping formula, and then the chaotic distribution value of each individual is converted into an actual position parameter, and finally the initial population is generated.
And 4, randomly generating a son root and a creeping stem for each mother plant.
The initial parent plant population is represented by an mxn matrix. Each stock plant then randomly generates a closer root and a farther stolon in each iteration based on the initial population, thereby generating new roots and stolon. To obtain an optimal solution, the stolons are searched around. When the stolons just reach the local minima, the algorithm will get faster speed and better global searching capability. Through successive iterations, each variable can search for a possible optimal solution after the t+1st iteration. This process can be expressed as follows:
。
Wherein, theRepresenting a plant propagation matrix at iteration t+1 times, 0< tmax2, the dimension of which is m×2N, m being the number of stolons, N being the initial parent plant population; AndThe position of the root and the position of the stolon at iteration t+1 times are represented respectively, and the dimensions of the positions are m×n; Is the best solution for the root and stolon positions at the t-th iteration, its dimension is also mxN, droot and drunner are two scalar quantities representing the root-to-stock and stolon-to-stock distances, respectively, and typically drunner>droot;r1 and r2 are random matrices with members uniformly distributed over the range [ -0.5,0.5 ].
And 5, updating the position of the stolons by using a simulated annealing algorithm.
In this embodiment, the simulated annealing algorithm gradually changes from global searching to local searching as the number of iterations increases by combining gaussian and cauchy distributions. In the early stage, the algorithm can explore a larger solution space because global searches predominate, and in the later stage, local searches predominate, helping to find more accurate solutions in a smaller range. In this way, the algorithm can more effectively avoid trapping in the locally optimal solution, and at the same time ensure that a better solution is found in the convergence stage. The position of the updated stolons is calculated by the following formula.
。
。
。
Wherein, theRepresents the position of the stolon obtained in the t+1st iteration, 0< tmax1,AndThe t-th iteration is the new position generated from the gaussian and cauchy distributions,As a scale parameter of the gaussian distribution,Is the scale parameter of the cauchy distribution, N (0, 1) and C (0, 1) are respectively the standard normal distribution and the standard cauchy distribution,The position of the stolons is indicated for the t-th iteration. Where a gaussian distribution is used for local search, smaller step sizes are required, so the scale parameters of the gaussian distributionTypically between 0.01 and 0.1 is taken, whereas the cauchy distribution is used for global searching, requiring a large step size, so the scale parameters of the cauchy distributionTypically between 0.1 and 1.0, in this embodiment=0.05,=0.5. The creeping stem position updating formula can avoid sinking into a local optimal solution as much as possible, and can ensure that a better solution is found in a convergence stage.
And 6, calculating the fitness value.
In this embodiment, the fitness value of all optimization solutions based on the objective functionCan be expressed as follows.
。
Where a is an adjustable parameter, which is used to adjust the calculation mode of the fitness value, typically 0,Is an objective function, the expression of which is as follows.
。
Wherein sigma, n, ks、km and kv are parameters to be inverted, sigma is roughness, n is the real part of complex refractive index, k is the imaginary part of complex refractive index, ks is specular reflectivity, km is diffuse reflectivity, kv is bulk scattering rate,Obtained by experimental measurementThe value of the sum of the values,For substitution into optimal solutionThe value of the sum of the values,For the incident zenith angle of the light,In order to detect the zenith angle,The difference between the incident zenith angle and the detected zenith angle.
And 7, selecting a new parent population for the next iteration.
First, the fitness value of the calculated objective function is calculated in ascending orderSorting is carried out, an optimized solution with the first N/2 smaller fitness values is selected, then the rest N/2 optimized solutions are selected by using a roulette algorithm, and finally the N optimized solutions after combination of the two choices are used as new parent populations of the next iteration.
In this embodiment, the expression of the roulette algorithm is as follows.
。
。
Wherein, theRepresenting the individual probability that individual i is selected,Indicating the fitness of the individual i,The sum, i.e. the cumulative probability, for all individual fitness.
Then randomly generating an array m, wherein the value range of elements in the array is between 0 and 1, and sequencing the elements from small to large. If the cumulative probability Ftotal is greater than the element m [ i ] in the array, then the individual x (i) is selected, if it is less than m [ i ], then the next individual x (i+1) is compared until one is selected, and this step is repeated N/2 times to obtain the remaining N/2 optimal solutions.
In this embodiment, when selecting a new parent population for the next iteration using fitness values and roulette algorithm, as shown in fig. 7, fitness values are first calculated, then the calculated fitness values are arranged in ascending order, and the first N/2 smaller fitness values are selected. Meanwhile, calculating the sum of fitness values of all individuals in the population by using a roulette algorithm, calculating the selection probability of each individual by using the roulette algorithm, generating a random number, selecting the individual, repeating N/2 times to obtain N/2 fitness values, obtaining N fitness values by combining the selected first N/2 smaller fitness values, and finally taking the obtained N fitness values as parent plant positions of the next iteration.
And 8, circulating or outputting a final solution.
In this embodiment, the cycle is terminated and the optimal inversion values of the 6 parameters to be inverted are output when either of the following two termination conditions occurs.
(1) The three algorithms of the Tent chaotic mapping algorithm, the strawberry optimization algorithm and the simulated annealing algorithm all reach the maximum iteration times.
(2) A set of optimal solutions occurs such that fitness values of the objective functionGreater than the fitness threshold, the fitness threshold is set to 20 in this embodiment.
And 9, calculating the accuracy of inversion parameters.
In this embodiment, the relative errors of the parameters σ, n, and k are calculated first, and then the inversion result is represented by using the average value of the relative errors of the three parameters, where the relative error expression is as follows.
。
Wherein delta is a relative error value, delta is an absolute error value, namely an absolute value of a difference between an actual measured value of a parameter to be inverted and an inversion value, and L is a true value, namely an actual measured value of the parameter to be inverted.
In this embodiment, the verification stage uses a verification module, and the workflow of the verification module specifically includes the following steps.
Firstly, after obtaining the optimal inversion values of 6 parameters to be inverted, namely sigma, n, ks、km and kv, substituting the optimal inversion values of the 6 parameters to be inverted intoSimulation of pBRDF model obtained from the formulaAnd (5) characterizing the polarization degree data simulated by the pBRDF model. Then judging the simulation by an evaluation functionWhether the value is within a reasonable threshold value or not, the evaluation function in this embodiment adopts MAE, and the expression is as follows.
。
Wherein, theSimulation for pBRDF modelThe value of the sum of the values,Obtained by experimental measurementThe value, n, is the total number of iterations of the inversion algorithm, n=300.
Then, the polarization degree data simulated by the model is judged (pBRDF simulation of the model)Value) and experimentally measured polarization degree data (experimentally measuredValues) are less than 5%, when the MAE value is less than 5%, the MAE value represents that pBRDF model simulatesThe values are within reasonable threshold ranges, resulting in a final pBRDF model. When the MAE value is greater than or equal to 5%, the model is further corrected pBRDF%, and then the step S22 'the step of initializing the parent plant position by adopting the Tent chaotic mapping algorithm according to the initial parameters to obtain the initialized parent plant position' is performed, the parent plant position is re-initialized, the corresponding parameter inversion result is obtained, and the pBRDF model and the simulation are re-constructed according to the re-obtained parameter inversion resultCalculation and evaluation of the values until their corresponding MAE values are less than 5%.
Fig. 8 is a three-dimensional graph of DOLP (linear polarization degree) generated by pBRDF model, a detection azimuth angle and a detection zenith angle, wherein the polarization degree is increased and then reduced along with the change of the zenith angle and reaches the maximum at the 45-degree observation zenith angle at the 180-degree azimuth angle, and fig. 9 is a polar graph generated by pBRDF model, wherein the radial direction is the observation zenith angle, the circumference is the observation azimuth angle, and the polarization characteristic distribution of the aluminum plate material in the simulated atmospheric environment is symmetrical about the 0-180-degree azimuth angle. Wherein the darker the color, the lower the degree of polarization, the whiter the color, and the higher the degree of polarization in fig. 8 and 9, the azimuth angle of the light source is set to 0 ° and the zenith angle is set to 45 °.
Aiming at the problems that the prior pBRDF-based parameter inversion method has low dependence on initial values and poor robustness and the traditional SBA algorithm has low convergence speed and is easy to fall into a local optimal solution, the embodiment initially builds a pBRDF model of a target material based on pBRDF theory, light elastic scattering theory and blackbody radiation theory and obtainsThe formula is that the space target infrared polarization characteristic measuring device is used to collect the polarization image and obtain the multi-angle polarization characteristic measuring result of the sample to be measured, the SBA algorithm with improved coupling is used to invert the roughness and complex refractive index waiting inversion parameters, and finally the optimal inversion value of the parameter to be inverted is substitutedAfter the formula, the simulation of the target material can be obtainedAnd determining whether the established pBRDF model precision meets the requirement or not by using the average absolute error, and determining a final pBRDF model. The method can effectively solve the defects of the traditional parameter inversion method, and compared with the traditional parameter inversion method such as a particle swarm optimization algorithm, the inversion accuracy of a metal object can be improved by about 20%, so that a more accurate pBRDF model can be obtained.
The pBRDF model is used for analyzing the long-wave infrared polarization characteristics of the sample material to be detected under multiple angles, and for the detection mode of low resolution and large background noise interference of long-wave infrared polarization detection, the accuracy of polarization detection of a space target can be improved by comparing the polarization data result obtained by the pBRDF model. In addition, a reference can be provided for constructing polarization characteristic models in different environments in other fields.
The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.
The principles and embodiments of the present application have been described in detail in this application, the above examples are provided to facilitate understanding of the method of the present application and its core ideas, and modifications may be made by those skilled in the art in light of the present teachings, both in the detailed description and the application scope. In view of the foregoing, this description should not be construed as limiting the application.