Air target height inversion method and system based on thermal infrared hyperspectrum

Abstract

The invention discloses an aerial target height inversion method and system based on a thermal infrared hyperspectrum, and belongs to the technical field of spectral data processing and passive ranging, and the method comprises the steps: screening and removing low-transmittance wavebands through atmospheric window transmittance analysis, removing wavebands with waveband redundancy and strong correlation through a method based on distance attenuation correlation analysis, and obtaining a high-resolution target height inversion result; dynamically selecting a wave band set by utilizing target energy distribution characteristics; and finally, constructing a nonlinear equation set based on the selected wave band, and solving key parameters such as the height, the emissivity and the speed of the target by adopting a multi-swarm collaborative particle swarm optimization (SOMC-PSO) inversion method based on sensitivity optimization. According to the target height inversion method based on the sensitive band selection and the dynamic optimization algorithm, high-precision inversion of the target height is achieved, and the method has high efficiency, robustness and real-time performance adapting to complex environments and can be widely applied to the field of thermal infrared hyperspectral target height inversion and parameter optimization.

Description

Technical Field

[0001] The present invention belongs to the technical field of spectral data processing and passive ranging, and more specifically, relates to a method and system for inverting the altitude of an airborne target based on thermal infrared hyperspectral data. Background Art

[0002] The thermal infrared hyperspectral technology provides rich spectral information for target detection and ranging by capturing the radiation characteristics of a target in the thermal infrared band (8 - 14 μm), and is widely used in fields such as military, remote sensing, and environmental monitoring. In target ranging, the thermal infrared passive ranging technology uses the relationship between the target radiation characteristics and the atmospheric transmission model to achieve the inversion and estimation of the target distance. Compared with active ranging technologies such as radar and laser, passive ranging has the advantages of good concealment and strong anti-interference ability, and shows significant application potential especially under complex background conditions.

[0003] Traditional passive ranging technologies usually use the infrared radiation intensity differences in single or multiple bands of multispectral data to achieve target distance inversion through radiation ratios, geometric derivations, or empirical formulas. Among them, the single-band method relies on the direct measurement of radiation intensity, is easily affected by atmospheric attenuation and background radiation, and has low ranging accuracy; the multi-band method reduces environmental interference by comparing the radiation intensity ratios of different bands, but most rely on preset fixed bands and do not consider the influence of atmospheric path radiation under long-distance conditions.

[0004] In recent years, the thermal infrared hyperspectral technology has become a new research hotspot for passive ranging because it can obtain target radiation information in multiple continuous bands. The spectral resolution of hyperspectral data is extremely high, and it can invert physical characteristics such as the temperature and emissivity of a target, providing important support for accurate ranging. After obtaining the thermal infrared hyperspectral data of an airborne target using an observation platform such as a satellite, the distance between the airborne target and the observation platform can be measured, and since the altitude of the observation platform is known, the altitude inversion of the airborne target can be achieved. Therefore, the altitude inversion of airborne targets based on thermal infrared hyperspectral data has also been widely applied. However, the high-dimensional characteristics of hyperspectral data also bring many challenges, such as redundant band information, complex non-linear relationships, and high computational costs. The target ranging process in the thermal infrared band is comprehensively affected by various factors such as atmospheric absorption, scattering, and background radiation. Traditional methods are prone to falling into local optimal solutions when dealing with complex non-linear atmospheric transmission models, resulting in a decrease in inversion accuracy. In addition, passive ranging technology has high requirements for real-time performance, especially in military applications. How to quickly process hyperspectral data while ensuring accuracy has become a key problem.

[0005] To solve the complex problems in target ranging, existing research has made some progress in multispectral ranging technology. For example, methods based on the ratio of radiation intensities in two or three bands invert the target distance by analyzing the ratio of radiation characteristics in fixed bands. However, such methods are usually only applicable to preset band combinations and it is difficult to dynamically adjust the bands to adapt to different targets or environmental conditions. In addition, many methods do not fully consider the non-linear influence of atmospheric radiation transmission characteristics on band intensities during long-distance ranging. In the latest research, a three-band multimodal fusion method using SDGSAT-1 thermal infrared data combined with a radiation transmission model is used for altitude inversion. Although certain results have been achieved, its bands are fixed and the number is limited, and it cannot make full use of continuous spectral information to optimize ranging accuracy. In addition, such methods have weak adaptability to the uncertainties of target radiation characteristics caused by changes in altitude, speed, and materials in complex environments and are difficult to meet the requirements of high-precision ranging.

[0006] Generally speaking, existing methods usually lack a dynamic band selection mechanism, cannot flexibly adjust the band combination according to target characteristics and environmental changes, are difficult to effectively cope with the significant fluctuations in target radiation characteristics caused by changes in multiple parameters, have insufficient robustness, and the inversion accuracy and calculation real-time performance need to be further improved. Summary of the Invention

[0007] In view of the defects and improvement requirements of the existing technology, the present invention provides a method and system for altitude inversion of airborne targets based on thermal infrared hyperspectrum, aiming to dynamically select sensitive bands of thermal infrared hyperspectral data according to target characteristics and environmental changes, and improve the accuracy, adaptability, and robustness of altitude inversion of airborne targets.

[0008] To achieve the above object, according to one aspect of the present invention, a method for altitude inversion of airborne targets based on thermal infrared hyperspectrum is provided, including:

[0009] Obtain the radiance L of the airborne target at each band obs,i (λ i ), where i represents the band number, and λ i ∈ ∧ selected represents the wavelength of the i-th band in ∧ selected , and ∧ selected represents the candidate band set in the current atmospheric scene;

[0010] For each non-empty subset of the candidate band set ∧ selected , for the corresponding band number set S, calculate the total radiance L sum (S) = ∑ i∈S L obs,i (λ i ) and the average correlation And calculate the comprehensive score accordingly |S| represents the size of the set S of band numbers, and R i,j represents the correlation between the i-th band and the j-th band in the current non-empty subset;

[0011] Select the non-empty subset with the highest comprehensive score as the inversion band set ∧ inversion ;

[0012] For each band λ i ∈∧ inversion , construct the radiance equation L sim,i (H, ε, v) that describes the relationship between radiance and height H, emissivity ε, and velocity v, and use (H, ε, v) as the parameter combination to be optimized, with the goal of minimizing the error function to solve for the optimal parameter combination (H * , ε * , v * ), and take the height H * as the height inversion result.

[0013] Furthermore, to solve for the optimal parameter combination (H * , ε * , v * ), it includes:

[0014] Use (H, ε, v) as the particles of the particle swarm optimization algorithm. After initializing the particle swarm, divide the particle swarm into a height subgroup, an emissivity subgroup, and a velocity subgroup, which are respectively used to iteratively search for the optimal height, optimal emissivity, and optimal velocity. And after every k iterations, each subgroup will share the current search results with other subgroups;

[0015] After the search ends, extract the height H * , emissivity ε * , and v * from the optimization results of the height subgroup, emissivity subgroup, and velocity subgroup respectively to obtain the optimal parameter combination (H * , ε * , v * ).

[0016] Furthermore, the weight for the height subgroup to update the particle velocity is:

[0017]

[0018] The weight for the emissivity subgroup to update the particle velocity is:

[0019]

[0020] The weight for the velocity subgroup to update the particle velocity is:

[0021]

[0022] Among them, p ∈ {H, ε, v}, L sim = ∑L sim,i (H, ε, v); α > 1 is the amplification factor.

[0023] Furthermore, the fitness value of each particle is F fitness = RMSE, and to solve the optimal parameter combination (H * , ε * , v * ), it further includes:

[0024] When the fitness value of the particle no longer changes, a random perturbation amount ΔH that follows the standard normal distribution is selected and applied to the height parameter being currently searched;

[0025] Among them, RMSE represents the sum of the root mean square errors of all bands.

[0026] Furthermore, the construction method of the candidate band set ∧ selected under the current atmospheric scenario includes:

[0027] According to the radiance equation under the current atmospheric scenario, calculate the radiance transmittance of the airborne target at different heights and different bands, and construct a radiance transmittance matrix; in the radiance transmittance matrix, each row corresponds to a height and each column corresponds to a band;

[0028] Calculate the correlation between every two columns in the radiance transmittance matrix as the correlation between the corresponding two bands; if the correlation between two bands is greater than a preset correlation threshold, then eliminate one of the bands;

[0029] The remaining bands form the candidate band set ∧ selected under the current atmospheric scenario.

[0030] Furthermore, if the correlation between two bands is greater than a preset correlation threshold, then eliminating one of the bands includes:

[0031] According to the radiance transmittance matrix, calculate the maximum amplitude of the radiance transmittance change with height for these two bands respectively, and eliminate the band with the smaller maximum amplitude of the radiance transmittance change with height.

[0032] Furthermore, the bands corresponding to the radiance transmittance matrix belong to the atmospheric window band set under the current atmospheric scenario; the construction method of the atmospheric window band set under the current atmospheric scenario includes:

[0033] For each band in the thermal infrared band, calculate the atmospheric transmittance distribution under the current atmospheric scenario respectively to calculate the atmospheric transmittance within a preset altitude range;

[0034] Exclude the bands whose atmospheric transmittance within the preset altitude range is lower than the preset transmittance threshold, and the remaining bands form the atmospheric window band set under the current atmospheric scenario.

[0035] Furthermore, the radiance equation is:

[0036]

[0037] where L represents radiance, L skin represents the self-radiation of the skin of the airborne target, represents the atmospheric radiation reflected by the skin of the airborne target, L path represents the path radiation from the airborne target to the observation platform, τ path represents the radiation transmittance from the airborne target to the observation platform.

[0038] According to another aspect of the present invention, there is provided a computer-readable storage medium, including a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the airborne target altitude inversion method based on thermal infrared hyperspectral provided by the present invention.

[0039] According to another aspect of the present invention, there is provided an airborne target altitude inversion system based on thermal infrared hyperspectral, including:

[0040] A computer-readable storage medium for storing a computer program;

[0041] And a processor for reading the computer program stored in the computer-readable storage medium and executing the above-mentioned airborne target altitude inversion method based on thermal infrared hyperspectral provided by the present invention.

[0042] Generally speaking, through the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0043] (1) By analyzing the distribution of the radiance observation values of airborne targets, the present invention calculates a comprehensive score for evaluating the average radiance and the inter-band correlation of each non-empty subset of the candidate band sets in the current atmospheric scene. The larger the average radiance and the lower the inter-band correlation, the higher the comprehensive score. Then, the band set with the largest comprehensive score is selected as the inversion band set. The inversion band set selected in this way has a relatively large overall radiance and a relatively low inter-band correlation, which contributes greatly to the inversion of the altitude of airborne targets. Based on this inversion band set screening method, the present invention can be adaptively adjusted according to the observation data, can avoid the limitations of artificially fixed bands, and improve the accuracy, adaptability, and robustness of altitude inversion.

[0044] (2) In a preferred embodiment of the present invention, a multi-group cooperative particle swarm optimization algorithm based on sensitivity optimization is proposed to solve the optimal combination of parameters after screening out the inversion band set that makes a greater contribution to altitude inversion. Specifically, the parameter combination (H, ε, v) to be optimized is used as the particle of the particle swarm optimization algorithm. Then, the particle swarm is divided into a height subgroup, an emissivity subgroup, and a velocity subgroup, which are respectively used to iteratively search for the optimal height, the optimal emissivity, and the optimal velocity. Since each subgroup independently optimizes the corresponding parameter, the computational efficiency of the inversion can be improved, and the convergence speed and global search ability of the algorithm can be enhanced. At the same time, after every k iterations, information sharing is performed among the subgroups, which can reduce the optimization error caused by the parameter compensation effect and quickly process hyperspectral data while ensuring the accuracy.

[0045] (3) In a preferred embodiment of the present invention, when using the proposed multi-group cooperative particle swarm optimization algorithm based on sensitivity optimization to solve the optimal parameter combination, the partial derivative of the radiance with respect to each parameter is calculated as the sensitivity of the corresponding parameter to the simulated radiance, and different sensitivity weights are set for different parameters on this basis. The sensitivity weight of the height parameter is obtained by multiplying its sensitivity value ratio by a magnification factor greater than 1. This can highlight the dominance of the height parameter in the particle swarm optimization search, make the search focus on height optimization, and further improve the accuracy of altitude inversion.

[0046] (4) In a preferred embodiment of the present invention, when using the proposed multi-group cooperative particle swarm optimization algorithm based on sensitivity optimization to solve the optimal parameter combination, when the particle fitness value no longer changes, a random perturbation is introduced into the height parameter, which can avoid the height parameter falling into the local optimal solution formed by the parameter compensation effect.

[0047] (5) In a preferred embodiment of the present invention, before screening the set of inversion bands contributing to height inversion, the correlation between bands is calculated through the radiance transmittance at different heights and different bands, so as to eliminate redundant and highly correlated bands based on the correlation between bands, and retain the set of bands with high physical information independence, thereby effectively reducing the dimension of the inversion equation system, reducing redundant calculations, and further improving the stability and accuracy of height inversion. In a further preferred embodiment, when eliminating the bands with high correlation, the bands with a small maximum amplitude of the change of radiance transmittance with height are eliminated, while the bands with a large maximum amplitude of the change of radiance transmittance with height are retained, thereby effectively screening out the bands that are more sensitive to height changes.

[0048] (6) In a preferred embodiment of the present invention, before eliminating redundant and highly correlated bands, the bands with high atmospheric transmittance are preliminarily screened out as the set of atmospheric window bands in the current atmospheric scenario, thereby eliminating the interference of low transmittance bands, effectively reducing the noise bands irrelevant to height inversion, simplifying the construction process of the candidate inversion equation system, and ensuring that the finally selected bands contribute the most to height inversion. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 Flowchart of the method for inverting the height of an aerial target based on thermal infrared hyperspectral provided by an embodiment of the present invention;

[0050] Figure 2 Flowchart of band selection provided by an embodiment of the present invention;

[0051] Figure 3 Flowchart of the sensitivity-based optimized multi-group cooperative particle swarm optimization algorithm (SOMC-PSO) provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0053] In the present invention, the terms "first", "second", etc. (if any) in the present invention and the accompanying drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence.

[0054] Aiming at the problems of insufficient robustness of existing air target altitude inversion methods in complex environments, and the need to further improve inversion accuracy and calculation real-time performance, the present invention provides an air target altitude inversion method and system based on thermal infrared hyperspectral. The overall concept is to utilize the high resolution and flexible band selection of hyperspectral data, and based on the analysis of the radiation characteristics of the environment and air targets, adaptively select sensitive bands, break through the fixed band limit, and improve the altitude inversion accuracy and robustness in complex scenarios.

[0055] The following are examples.

[0056] Example 1:

[0057] An air target altitude inversion method based on thermal infrared hyperspectral. This example is divided into two stages: in the first stage, bands that contribute more to air target altitude inversion are screened out to construct an inversion band set for the current atmospheric scene; in the second stage, based on the constructed inversion band set, the altitude of the space target is inverted.

[0058] As Figure 1 and Figure 2 shown, in this example, constructing an inversion band set for the current atmospheric scene includes: initially screening out bands with relatively high atmospheric filtering as the atmospheric window band set for the current atmospheric scene; removing redundant and highly correlated bands from the atmospheric window band set and retaining the band set with high physical information independence; by analyzing the distribution of radiance observation values, dynamically selecting bands with higher radiance values and sensitive to altitude changes to construct an inversion band set for the current atmospheric scene.

[0059] To achieve the selection of highly sensitive bands, it is first necessary to construct a target infrared radiation transfer model to describe the relationship between the radiance of the target and related parameters. Optionally, in this example, based on the physical theory of radiation transfer, a mathematical model is constructed in the thermal infrared band, as shown in the following radiance equation:

[0060]

[0061] where L represents radiance, L skin represents the self-radiation of the air target skin, represents the atmospheric radiation reflected by the air target skin, L path represents the path radiation from the air target to the observation platform, τ path represents the radiation transmittance from the air target to the observation platform, L skin 、 L path and τ pathAll are related to altitude, emissivity, and velocity. In the thermal infrared band, the aerial skin can be approximated as an opaque Lambertian gray body with a uniform emissivity. According to Planck's law, in the band with wavelength λ, the radiation of the aerial target skin can be expressed as: ε skin represents the skin emissivity, T skin represents the skin temperature, and M() represents Planck's formula.

[0062] This radiance equation comprehensively considers the physical characteristics of target spontaneous emission, reflected atmospheric radiation, and path radiation. Based on this radiance equation, the simulated radiance values under different parameters can be calculated by using an atmospheric transmission simulation tool (such as MODTRAN). Optionally, the atmospheric transmission simulation tool used in this embodiment is MODTRAN. When setting the MODTRAN simulation tool to simulate atmospheric radiation, reasonable configurations need to be made for the observation geometry, atmospheric conditions, and simulation range. The temperature profile is selected according to the target area to match the distribution model. Usually, the altitude of the aerial target is in the range of 0 - 20 km, and the wavelength range corresponding to the thermal infrared band is 8 - 14 μm. Correspondingly, in this embodiment, when setting the MODTRAN simulation tool to simulate atmospheric radiation, the altitude of the target is set in the range of 0 - 20 km according to actual needs, and considering the influence of the observation angle, it is set as the "slant path" oblique path; the band range is set at 8 - 14 μm, and a resolution of 1 cm -1 is selected for simulation. Through calculation in the Thermal Radiance mode, the atmospheric transmittance and path radiation characteristics on the path from the target to the observation point are obtained.

[0063] By using an atmospheric transmission simulation tool (such as MODTRAN), different standard atmospheric models (such as mid-latitude summer, mid-latitude winter, tropics, etc.) and aerosol types (such as rural, urban, marine, dust) are respectively set, and at the same time, the observation geometry paths in different altitude ranges (such as 0 - 20 km) are set, and the transmittance distribution in each wavenumber range within the 8 - 14 μm band is calculated. For different atmospheric scenarios and altitudes, a set of bands with higher transmittance is extracted, and a transmittance screening threshold is set to eliminate the bands with strong atmospheric absorption and low transmittance. By summarizing the band sets screened under different scenarios, a comprehensive atmospheric window band set is formed, laying a foundation for subsequent correlation analysis and altitude inversion, effectively eliminating the interference of unfavorable bands, and improving the accuracy and adaptability of altitude inversion.

[0064] Based on the above radiance equation, in this embodiment, the bands with relatively high atmospheric filtering are initially screened out as the atmospheric window band set under the current atmospheric scenario, specifically including:

[0065] For each band in the thermal infrared band, calculate the atmospheric transmittance distribution under the current atmospheric scenario respectively to calculate the atmospheric transmittance within the preset altitude range.

[0066] Eliminate the bands whose atmospheric transmittance within the preset altitude range is lower than the preset transmittance threshold, and the remaining bands form the atmospheric window band set under the current atmospheric scenario.

[0067] In this embodiment, by initially screening out the bands with higher atmospheric transmittance as the atmospheric window band set under the current atmospheric scenario, the interference of low transmittance bands is eliminated, the noise bands irrelevant to altitude inversion can be effectively reduced, the construction process of the candidate inversion equation set is simplified, and it is ensured that the finally selected bands have the greatest contribution to altitude inversion. It is easy to understand that in practical applications, the transmittance threshold for screening bands can be determined accordingly according to the transmittance distribution under the current scenario.

[0068] On the basis of screening out the atmospheric window band set under the current scenario, this embodiment eliminates the redundant and highly correlated bands from the atmospheric window band set and retains the band set with high physical information independence, specifically including:

[0069] According to the radiance equation under the current atmospheric scenario, calculate the radiance transmittance of the aerial target at different altitudes and different bands, and construct a radiance transmittance matrix; the radiance transmittance matrix can be expressed as follows:

[0070]

[0071] In the radiance transmittance matrix, each row corresponds to an altitude, and each column corresponds to a band; m and n respectively represent the number of altitudes and the number of bands.

[0072] Calculate the correlation between every two columns in the radiance transmittance matrix as the correlation between the corresponding two bands; optionally, in this embodiment, the Pearson coefficient is used as the correlation measure between every two columns in the radiance transmittance matrix, and the specific formula is:

[0073]

[0074] where, R i,j represents the Pearson coefficient between the i-th band and the j-th band, represents the mean value of the radiance transmittance of the i-th band at all altitudes, represents the mean value of the radiance transmittance of the j-th band at all altitudes, τ i,k represents the radiance transmittance of the i-th band at the k-th distance, τ j,k represents the radiance transmittance of the i-th band at the k-th distance;

[0075] If the correlation between two bands is greater than a preset correlation threshold, then one of the bands is removed;

[0076] The remaining bands form the candidate band set ∧ for the current atmospheric scene selected .

[0077] In this embodiment, based on the atmospheric window band set, the correlation between bands is first calculated through the radiance transmittance at different heights and different bands, so as to remove redundant and highly correlated bands based on the correlation between bands, and retain the band set with high physical information independence, thereby effectively reducing the dimension of the inversion equation system, reducing redundant calculations, and further improving the stability and accuracy of altitude inversion.

[0078] In order to further improve the accuracy of candidate altitude inversion, as a preferred implementation manner, in this embodiment, for two bands with relatively high correlation in the atmospheric window band set, the maximum amplitudes of the transmittance changing with height corresponding to the respective bands are calculated respectively, and the band with the smaller maximum amplitude changing with height is removed, while the band with the larger maximum amplitude changing with height is retained. For the i-th band, its maximum amplitude changing with height can be calculated by Δτ i = max(τ i (H k )) - min(τ i (H k )) where max(τ i (H k )) and max(τ i (H k )) respectively represent the maximum value and the minimum value in the i-th column of the radiance transmittance matrix.

[0079] By removing redundant and highly correlated bands in the above manner, this embodiment can effectively screen out bands that are more sensitive to altitude changes, further improving the accuracy, adaptability, and robustness of subsequent altitude inversion.

[0080] In practical applications, the atmospheric window band set under different atmospheric scenes can be established offline in advance, and the candidate band set under different atmospheric scenes can be further constructed. When performing altitude inversion of an airborne target, the current atmospheric scene can be judged according to time and longitude and latitude, and then the corresponding atmospheric window band set and candidate band set can be selected.

[0081] After constructing the candidate band set ∧ for the current atmospheric scene selected This embodiment dynamically selects bands with higher radiance values and sensitive to altitude changes by analyzing the distribution of radiance observation values to construct the inversion band set for the current atmospheric scene. The specific steps include:

[0082] Obtain the radiance L of the airborne target at each wavelength band obs,i (λ i ) observation results; where i represents the wavelength band serial number, λ i ∈∧ selected represents the wavelength of the i-th wavelength band in ∧ selected , and ∧ selected represents the candidate wavelength band set under the current atmospheric scenario;

[0083] For each non-empty subset of the candidate wavelength band set ∧ selected , for the corresponding wavelength band serial number set S, calculate the total radiance L sum (S) = ∑ i∈S L obs,i (λ i ) and the average correlation And thereby calculate the comprehensive score |S| represents the size of the wavelength band serial number set S, and R i,j represents the correlation between the i-th wavelength band and the j-th wavelength band in the current non-empty subset. Optionally, this correlation is also measured by the Pearson coefficient;

[0084] Select the non-empty subset with the highest comprehensive score, that is, the wavelength band set with the largest average radiance and the lowest correlation between wavelength bands, as the inversion wavelength band set ∧ inversion .

[0085] Through the above inversion wavelength band set construction method, this embodiment can perform adaptive adjustment according to the observation data. The finally constructed inversion wavelength band set has a relatively large overall radiance and a relatively low correlation between wavelength bands, which makes a great contribution to the airborne target altitude inversion. Therefore, it can avoid the limitations of artificially fixed wavelength bands and improve the accuracy, adaptability, and robustness of altitude inversion.

[0086] It should be noted that in practical applications, if the entire thermal infrared wavelength band set is used as the candidate wavelength band set ∧ selected , after constructing the inversion wavelength band set ∧ unversion , if the corresponding altitude inversion accuracy, adaptability, and robustness can meet the corresponding inversion requirements, the operations of initially screening the atmospheric window wavelength band set and removing redundant wavelength bands with relatively high correlations may not be performed.

[0087] Based on the constructed inversion wavelength band set, the process of performing spatial target altitude inversion in this embodiment specifically includes:

[0088] For each wavelength band λ i ∈∧ inversion , construct a radiance equation L sim,i(H, ε, v), taking (H, ε, v) as the parameter combination to be optimized, with the goal of minimizing the error function to solve for the optimal parameter combination (H * , ε * , v * ), and taking the height H among them * as the height inversion result.

[0089] To improve the inversion efficiency, this embodiment proposes a multi-group cooperative particle swarm optimization algorithm based on sensitivity optimization (SOMC-PSO) for solving the optimal parameter set. When using this algorithm to solve for the optimal parameter combination, this embodiment takes (H, ε, v) as the particles of the particle swarm optimization algorithm and initializes the particle swarm; when initializing the particle swarm, the search space range of the particle swarm will be set according to the common physical characteristics of airborne targets;

[0090] To improve the overall optimization efficiency, after initializing the particle swarm in this embodiment, the particle swarm will be divided into a height subgroup, an emissivity subgroup, and a velocity subgroup. Each subgroup independently optimizes the corresponding parameter, that is, the height subgroup, the emissivity subgroup, and the velocity subgroup are respectively used to iteratively search for the optimal height, the optimal emissivity, and the optimal velocity. And after every k iterations, each subgroup will share the current search result with other subgroups. Through this cooperative mechanism, the optimization error caused by the parameter compensation effect can be reduced.

[0091] To further improve the height inversion accuracy, during the process of particle swarm search optimization in this embodiment, the sensitivity of each parameter to the simulated radiance will be calculated first:

[0092]

[0093] where L sim = ∑L sim,i (H, ε, v); in this embodiment, the sensitivity value can quantify the importance of each parameter to the height inversion task;

[0094] To accurately calculate the sensitivity of each parameter, as a preferred implementation, this embodiment first uses uniform grid sampling to take N groups of particles within the entire search space range and calculates the sensitivity for each particle; for the k-th particle, its sensitivity is:

[0095]

[0096] L sim (k) represents the sum of the radiances of the k-th particle in all bands;

[0097] Then calculate the global equilibrium sensitivity:

[0098]

[0099] Meanwhile, to highlight the dominance of the height parameter in optimization and focus the search on height optimization, in this embodiment, different weights are set for different parameters based on the sensitivity calculation results, as follows:

[0100]

[0101] Then, after normalization, the sensitivity weights of each parameter are obtained; specifically, the weight for updating the particle velocity in the height particle swarm is:

[0102]

[0103] The weight for updating the particle velocity in the emissivity subgroup is:

[0104]

[0105] The weight for updating the particle velocity in the velocity subgroup is:

[0106]

[0107] Among them, α > 1 is the amplification factor, through which the dominant position of the height parameter can be strengthened.

[0108] Meanwhile, based on the error function of height inversion, the fitness value of each particle is calculated as F fitness = RMSE, where RMSE represents the sum of the root mean square errors of all bands, and the formula for each subgroup to update the particle velocity is:

[0109] v i,d (t + 1) = w(t)·v i,d (t) + c1·r1·γ d ·(p i,d - x i,d )·W d + c2·r2·γ d ·(g d - x i,d )·W d

[0110] Among them, v i,d (t + 1) and v i,d (t) represent the velocities of the particle at the (t + 1)-th moment and the t-th moment respectively; w(t) is the dynamic inertia weight, which gradually decreases with the number of iterations: w max and w min represent the maximum and minimum values of the dynamic inertia weight respectively; W d represents the weighted error sum of the sensitivity weights under this parameter; γ drepresents the balance factor, which restricts the influence of error on the particle velocity; c1 and c2 represent the acceleration factors, which control the moving speed of the particle towards the local optimal solution and the global optimal solution; r1 and r2 represent random numbers, which enable the particle to jump out of the local optimal solution; p i,d represents the personal best position of particle i in the d-th dimension; x i,d represents the current position of particle i; g d represents the global optimal solution;

[0111] And, when the fitness value of the particle no longer changes, execute x i,H ← x i,H + ΔH, ΔH ~ U(-δ, δ), so as to apply the random perturbation amount ΔH that obeys the standard normal distribution U(-δ, δ) to the height parameter x being currently searched i,H ; By introducing a random perturbation amount to the height parameter when the change of the particle fitness value stagnates, it is possible to avoid the height parameter falling into the local optimal solution formed by the compensation effect.

[0112] The specific process of the multi-group collaborative particle swarm optimization algorithm (SOMC-PSO) based on sensitivity optimization for solving the optimal parameter set in this embodiment is as Figure 3 shown, including:

[0113] S1: Using (H, ε, v) as particles, after initializing the particle swarm, divide the particle swarm into a height subgroup, an emissivity subgroup, and a velocity subgroup, which are respectively used to iteratively search for the optimal height, the optimal emissivity, and the optimal velocity, and initialize the weights of each subgroup for updating the particle velocity;

[0114] S2: Calculate the fitness value of the particles in the current iteration;

[0115] S3: Each subgroup updates the particle velocity and position;

[0116] S4: Determine whether the current iteration differs from the iteration when information sharing was performed among the subgroups last time by k. If so, after each subgroup shares the current search results, transfer to S3; otherwise, transfer to S5;

[0117] S5: Determine whether the algorithm has converged. If so, transfer to S7; otherwise, transfer to S6;

[0118] S6: Determine whether the change of the particle fitness value has stagnated. If so, introduce a random perturbation to the height parameter, and then transfer to S3; otherwise, transfer to S2;

[0119] S7: Extract the height H * , emissivity ε * and v * , respectively, from the optimization results of the height subgroup, the emissivity subgroup, and the velocity subgroup, to obtain the optimal parameter combination (H * , ε* , v * )。

[0120] Generally speaking, the full-process altitude inversion scheme provided in this embodiment forms a complete altitude inversion technical link through four steps: initial band screening, correlation elimination, adaptive sensitive band selection, and improved particle swarm optimization algorithm. This scheme can achieve high-precision solution of altitude inversion under different observation conditions and complex non-linear radiation transfer models, and has strong practicability.

[0121] Embodiment 2:

[0122] A computer-readable storage medium includes a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the method for aerial target altitude inversion based on thermal infrared hyperspectral provided in Embodiment 1 above.

[0123] Embodiment 3:

[0124] An aerial target altitude inversion system based on thermal infrared hyperspectral includes:

[0125] A computer-readable storage medium for storing a computer program;

[0126] And a processor for reading the computer program stored in the computer-readable storage medium and executing the method for aerial target altitude inversion based on thermal infrared hyperspectral provided in Embodiment 1 above.

[0127] This method breaks through the limitation of the fixed bands of traditional multispectral methods. Through dynamic sensitive band selection of hyperspectral data and the application of an improved non-linear optimization algorithm, it realizes high-precision inversion of the target distance, and significantly improves the adaptability and robustness of passive ranging technology in complex environments.

[0128] Those skilled in the art can easily understand that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for inverting the height of aerial targets based on thermal infrared hyperspectral, characterized in that: include: Get the radiance L of the aerial target in each band obs,i (λ i ) observation results; where i represents the band number, λ i ∈∧ selected Represents ∧ selected The wavelength of the i-th band in , ∧ selected Represents the candidate band set under the current atmospheric scene; For the candidate band set ∧ selected For each non-empty subset of , the corresponding band number set S is used to calculate the total radiance L sum (S) = ∑ i∈S L obs,i (λ i ) and the average correlation The comprehensive score is calculated from this |S| represents the size of the band number set S, R i,j Represents the correlation between the i-th band and the j-th band in the current non-empty subset; Select the non-empty subset with the highest comprehensive score as the inversion band set ∧ inversion ; For each band λ i ∈∧ inversion , construct the radiance equation L to describe the relationship between radiance and height H, emissivity ε and velocity v sim,i (H, ε, v), with (H, ε, v) as the parameter combination to be optimized to minimize the error function As the goal, find the optimal parameter combination (H * ,ε * ,v * ), where the height H * As a result of height inversion.

2. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 1, characterized in that: Find the optimal parameter combination (H * ,ε * ,v * ),include: With (H, ε, v) as the particle of the particle swarm optimization algorithm, after initializing the particle swarm, the particle swarm is divided into height subgroup, emissivity subgroup and speed subgroup, which are used to iteratively search for the optimal height, optimal emissivity and optimal speed respectively, and after every k iterations, each subgroup shares the current search results with other subgroups; After the search is completed, the height H is extracted from the optimization results of the height subgroup, the emissivity subgroup, and the velocity subgroup. * , emissivity ε * and v * , get the optimal parameter combination (H * ,ε * ,v * ); Wherein, k is a preset positive integer.

3. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 2, characterized in that: The weight of the height subgroup to update the particle velocity is: The weight of the emissivity subgroup to update the particle velocity is: The weight of the velocity subgroup to update the particle velocity is: in, L sim =∑L sim,i (H,ε,v); α>1 is the amplification factor.

4. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 3, characterized in that: The fitness value of each particle is F fitness =RMSE, and find the optimal parameter combination (H * ,ε * ,v * ), and also includes: When the fitness value of the particle no longer changes, a random perturbation ΔH that obeys the standard normal distribution is selected and applied to the height parameter of the current search; Among them, RMSE is the sum of the root mean square errors of all bands.

5. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to any one of claims 1 to 4, characterized in that: Candidate band set ∧ under the current atmospheric scene selected The construction methods include: According to the radiance equation in the current atmospheric scene, the radiance transmittance of the aerial target at different altitudes and different bands is calculated, and a radiance transmittance matrix is ​​constructed; in the radiance transmittance matrix, each row corresponds to a height, and each column corresponds to a band; Calculating the correlation between every two columns in the radiance transmittance matrix as the correlation between the corresponding two bands; if the correlation between the two bands is greater than a preset correlation threshold, eliminating one of the bands; The remaining bands constitute the candidate band set ∧ under the current atmospheric scene selected .

6. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 5, characterized in that: If the correlation between two bands is greater than the preset correlation threshold, one of the bands is removed, including: The maximum amplitude of the variation of radiance transmittance with height in the two bands is calculated according to the radiance transmittance matrix, and the band with the smaller maximum amplitude of the variation of radiance transmittance with height is eliminated.

7. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 5, characterized in that: The bands corresponding to the radiance transmittance matrix belong to the atmospheric window band set in the current atmospheric scene. The atmospheric window band set in the current atmospheric scene is constructed in the following ways: For each band in the thermal infrared band, the atmospheric transmittance distribution in the current atmospheric scene is calculated to calculate the atmospheric transmittance within the preset altitude range; The bands whose atmospheric transmittance within the preset altitude range is lower than the preset transmittance threshold are eliminated, and the remaining bands constitute the atmospheric window band set under the current atmospheric scene.

8. The method for inverting the height of an aerial target based on thermal infrared hyperspectroscopy according to claim 7, characterized in that: The radiance equation is: Where L represents the radiance, L skin represents the self-radiation of the aerial target skin, represents the atmospheric radiation reflected by the air target skin, L path represents the path radiation from the aerial target to the observation platform, τ path It represents the radiation transmittance from the aerial target to the observation platform.

9. A computer-readable storage medium, characterized in that: It comprises a stored computer program; when the computer program is executed by a processor, it controls the device where the computer-readable storage medium is located to execute the aerial target height inversion method based on thermal infrared hyperspectroscopy as described in any one of claims 1 to 8.

10. An aerial target altitude inversion system based on thermal infrared hyperspectral, characterized in that: include: A computer-readable storage medium for storing a computer program; And a processor, used for reading the computer program stored in the computer-readable storage medium, and executing the aerial target height inversion method based on thermal infrared hyperspectroscopy as described in any one of claims 1 to 8.

Images