An Infrared Signal Denoising and Feature Extraction Method for Dim Space Targets

Abstract

The present invention discloses a method for denoising and feature extraction of infrared signals of space dim targets. Step 1: Use a space target infrared simulation system to obtain the noisy infrared radiation intensity signals of targets with regular shapes in two bands. Step 2: Use the average weight fuzzy distance entropy as the fitness function. Step 3: Use the dragonfly optimization algorithm to optimize the fitness function and solve the optimal decomposition layer number and quadratic penalty factor of variational mode decomposition. Step 4: Perform variational mode decomposition on the infrared radiation intensity signal according to the optimal parameter combination to obtain the decomposition signals of K modes. The method for denoising and feature extraction of infrared signals of space dim targets disclosed by the present invention has the effect of being able to effectively remove the noise components of infrared noisy signals, suppressing the influence of noise on the extraction of the product feature of temperature feature and effective radiation area, and significantly improving the feature extraction accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of infrared signal, and in particular to a method for denoising and feature extraction of infrared signals of dim targets in space. Background Art

[0002] The research on the infrared radiation characteristics of space targets has attracted wide attention in many application fields, such as space kill assessment, space debris monitoring and removal, and weak target detection. Infrared radiation characteristics are an important information source for target detection, tracking, and recognition. Due to the long imaging distance between the space target and the infrared detector, the target usually appears as only one or a few pixels in the infrared image of the detector, and the infrared signal of the target received by the detector is weak. The infrared radiation of the target includes the surface temperature of the target, the product of emissivity and area, micro-motion, and other information, all of which are contained in a few pixels. This brings the possibility of feature extraction, classification, and recognition of space weak targets based on these pixels.

[0003] In the process of target detection, the measurement and analysis of infrared radiation are particularly important. However, factors such as the temperature change effect, sensor position calibration error, and optoelectronic noise in the infrared radiation measurement system seriously interfere with the extraction of the infrared radiation signal, resulting in a low signal-to-noise ratio of the infrared signal. This will seriously restrict the classification and recognition effect of space infrared targets using infrared sensors. Therefore, it is of great significance to study a denoising and feature extraction model with better robustness for target classification and recognition.

[0004] Currently, the commonly used signal denoising methods include wavelet transform, singular value decomposition, empirical mode decomposition, etc. However, these methods all have limitations to varying degrees. The effect of wavelet transform depends on the wavelet basis function, decomposition level, and threshold; the singular value decomposition method needs to overcome how to determine the break point of the singular value vector. Empirical mode decomposition is an effective automatic decomposition algorithm that can decompose the noise signal into multiple orthogonal functions without any prior matrix. However, EMD has limitations in terms of mode mixing, noise sensitivity, and end effect. To overcome mode mixing, several improved EMD methods have been proposed by scholars, such as the ensemble empirical mode decomposition that uses noise-assisted EMD, which can suppress mode mixing to a certain extent. However, these improved empirical mode decomposition methods will inevitably produce error accumulation and lack strict mathematical theory support.

[0005] The commonly used methods for temperature feature extraction are single-band temperature measurement method, multi-band temperature measurement method, and dual-band temperature measurement method. Single-band temperature measurement is the most commonly used method in the infrared radiation characteristic measurement system. Through radiometric calibration, the functional relationship between the incident radiation and the detector output is established. However, the measured radiation includes the self-radiation of the target and the reflected external radiation. The self-radiation cannot be separated from the total radiation received by the detector, and the emissivity information of the target is unknown. Therefore, this method can only obtain the brightness temperature of the target, which limits the judgment of the target attributes. Multi-band temperature measurement uses the prior knowledge of the target emissivity to obtain the radiation information of the target in multiple bands, represents the target emissivity as a function of wavelength, and then combines the target radiation and emissivity model to obtain the surface temperature and emissivity of the target. However, if there is a large difference between the set emissivity model and the actual model of the target, the temperature measurement accuracy will drop sharply. For some unknown materials or lack of prior knowledge of tar emissivity, it is difficult to meet the requirements of temperature measurement accuracy. Summary of the Invention

[0006] The present invention discloses a method for denoising and feature extraction of infrared signals of space dim targets, aiming to solve the technical problems in the background technology that the measured radiation includes the self-radiation of the target and the reflected external radiation, the self-radiation cannot be separated from the total radiation received by the detector, and the emissivity information of the target is unknown.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A method for denoising and feature extraction of infrared signals of space dim targets specifically includes the following steps:

[0009] Step 1: Use the infrared simulation system of space targets to obtain the noisy infrared radiation intensity signals of targets with regular shapes in two bands, and respectively execute Steps 2 to 5 on the infrared signals in each infrared band;

[0010] Step 2: Use the average weight fuzzy distance entropy as the fitness function;

[0011] Step 3: Use the dragonfly optimization algorithm to optimize the fitness function, and solve the optimal decomposition layer number and quadratic penalty factor of variational mode decomposition;

[0012] Step 4: Perform variational mode decomposition on the infrared radiation intensity signal according to the optimal parameter combination to obtain the decomposition signals of K modes;

[0013] Step 5: Calculate the Pearson correlation coefficient between each modal component and the original radiation intensity signal, select the effective modal components for signal reconstruction, and achieve signal denoising;

[0014] Step 6: Obtain the denoised target infrared radiation intensity signals in two spectral bands, and extract the temperature characteristics of the target by using the dual-band temperature measurement method;

[0015] Step 7: With the help of the distance information between the target and the infrared detector, extract the product characteristic of the infrared emissivity area of the target.

[0016] In a preferred solution, in the above-mentioned Step 1, the infrared noisy radiation intensity signal of the target is simulated by a space target infrared simulation system, and the conventional shapes are cone, cone with spherical bottom, cone with cylindrical bottom, cylinder and sphere. In the above-mentioned Step 2, the solution model of the average weighted fuzzy distance entropy is as follows:

[0017] Estimate the empirical probability density function of the similarity matrix of the signal according to the histogram, and calculate the normalized fuzzy distance entropy as:

[0018]

[0019] where M is the number of units of the histogram;

[0020] Calculate the mutual information between the modal component and the original infrared radiation intensity signal Y as:

[0021]

[0022] where p(u), p(Y), and p(u, Y) represent the marginal probability distribution and the joint probability distribution;

[0023] Then the average weighted fuzzy distance entropy is:

[0024]

[0025] where (u(i) represents the i-th modal component, and K is the number of modal components.

[0026] In a preferred solution, the dragonfly algorithm in the above-mentioned Step 3 is a swarm intelligence optimization algorithm, and its basic idea is as follows:

[0027] Step S31: Determine the scale of the dragonfly population, randomly generate X dragonflies, set [K, α] as the position vector of the dragonflies, and initialize the velocity vector;

[0028] Step S32: Calculate the fitness function of all dragonflies, and update the food position and the natural enemy position;

[0029] Step S33: Calculate the separation factor, the companion flying factor, the collision factor, the food attraction factor, and the factor of staying away from the natural enemy of the dragonflies;

[0030] Step S34: Update the position and velocity vectors according to the surrounding radius;

[0031] Step S35: If the maximum number of iterations is reached, terminate the search and output the global optimal solution; otherwise, continue the iteration.

[0032] In a preferred embodiment, the separation factor S i , the co-flight factor A i , the collision factor C i , the food attraction factor F i and the predator avoidance factor E i of the dragonfly algorithm in step S33 are updated as follows:

[0033]

[0034]

[0035]

[0036] F i = X + - X

[0037] E i = X - + X

[0038] where X is the position of the current individual, N is the number of dragonflies within the vicinity of the current individual, X j is the position of the j-th dragonfly within the vicinity of the current individual, X j is the velocity vector of the j-th dragonfly within the vicinity of the current individual, X 十 is the food position, and X 一 is the predator position;

[0039] The method for updating the position and velocity vector of the dragonfly algorithm in step S34 is as follows:

[0040] If there are no other individuals around the current individual, the velocity vector and position vector are:

[0041] X t+1 = X t + Lévy(d) × X t

[0042] ΔX t+1 = 0

[0043] Otherwise

[0044] X t+1 = X t + ΔX t+1

[0045] ΔX t+1 = (sS i + aA i + cC i + fFi +eE i )+wΔX t ;

[0046] The global optimal solution in step S35 is the position where the optimal individual is located, corresponding to the optimal combination [K, α] of variational mode decomposition parameters.

[0047] In a preferred embodiment, the variational mode decomposition process in step S4 is specifically as follows:

[0048] Let u k (t) be the BLIMF component, and its time-domain expression is as follows:

[0049] u k (t) = A k (t)cos(φ k (t))

[0050] where A k (t) and φ k (t) represent the envelope and phase respectively. This algorithm mainly consists of two steps: constructing a variational problem and solving the variational problem. First, a variational problem is constructed to estimate the center frequency and bandwidth of the BLIMFs. The Hilbert transform is used to obtain the single-sided spectrum of the relevant analysis signal of each. The spectrum of the IMF is shifted to the baseband of each u k . The bandwidth of the demodulated signal is estimated by Gaussian smoothness (i.e., the square norm). Based on the above, this problem can be formulated as the following constrained variational problem:

[0051]

[0052] where u k represents the k-th BLIMF, w k is the center frequency of u k , represents the derivative operator, δ(t) is the Dirac function, * represents the convolution operation, and f is the original signal;

[0053] Secondly, to solve the above variational problem, the Lagrange multiplier λ and the quadratic penalty term α are introduced to transform the above problem into an unconstrained variational problem and find the optimal solution. The augmented Lagrangian function is expressed as follows:

[0054]

[0055] Then, the alternating direction multiplier method (ADMM) is used to find the saddle point of the function as the solution of the above augmented Lagrangian equation, and update The iterative equations are as follows:

[0056]

[0057]

[0058] Repeat the iterative process according to the above equation until the following convergence condition is met and the iteration stops;

[0059]

[0060] The calculation method of the Pearson correlation coefficient in step 5 is as follows:

[0061]

[0062] where R(T) is the ratio of the radiation intensities of the blackbody at temperature T in two infrared bands, S 1 ’ is the denoised radiation intensity of the target in band 1, S 2 ’ is the denoised radiation intensity of the target in band 2;

[0063] The method for extracting the infrared emissivity-area product feature in step 7 is as follows:

[0064]

[0065] where S’ is the denoised radiation intensity in any band and T is the extracted temperature feature.

[0066] As can be seen from the above, a method for denoising and feature extraction of infrared signals of space dim targets specifically includes the following steps: Step 1: Use a space target infrared simulation system to obtain the noisy infrared radiation intensity signals of a target with a regular shape in two bands, and separately perform steps 2 to 5 on the infrared signals in each infrared band; Step 2: Use the average weight fuzzy distance entropy as the fitness function; Step 3: Use the dragonfly optimization algorithm to optimize the fitness function and solve the optimal decomposition layer number and quadratic penalty factor of variational mode decomposition; Step 4: Perform variational mode decomposition on the infrared radiation intensity signal according to the optimal parameter combination to obtain the decomposition signals of K modes; Step 5: Calculate the Pearson correlation coefficient between each mode component and the original radiation intensity signal, select the effective mode components for signal reconstruction to achieve signal denoising; Step 6: Obtain the denoised target infrared radiation intensity signals in the two spectral bands, and use the two-band temperature measurement method to extract the temperature feature of the target; Step 7: With the help of the distance information between the target and the infrared detector, extract the infrared emissivity-area product feature of the target. The method for denoising and feature extraction of infrared signals of space dim targets provided by the present invention has the following technical effects:

[0067] 1: Variational mode decomposition is an adaptive signal decomposition method that decomposes the input signal into a series of signals with limited frequency bandwidths, which can reflect the sparse characteristics of the signal, has good signal decomposition accuracy and denoising performance, and can effectively avoid the mode mixing problem during signal decomposition;

[0068] 2: The present invention can avoid the influence of human subjective factors on the parameter combination of variational mode decomposition, can achieve the selection of adaptive parameter combinations, and effectively find the optimal parameter combination;

[0069] 3: The present invention can effectively suppress the adverse effects of noise on the extraction of temperature characteristics and emissivity-area product characteristics, improve the accuracy of feature extraction, and is conducive to the separation and recognition of space infrared dim targets. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 It is a schematic flow chart of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0071] Figure 2 It is a schematic diagram of the infrared radiation intensity with noise of the target of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0072] Figure 3 It is a schematic diagram of the modal decomposition result of the infrared radiation intensity signal with noise of the target of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0073] Figure 4 It is the Pearson correlation coefficient between each modal component and the infrared radiation intensity signal with noise of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0074] Figure 5 It is a schematic diagram of the infrared radiation intensity after denoising of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0075] Figure 6 It is a schematic diagram for comparing the extracted temperature characteristics of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention.

[0076] Figure 7 It is a schematic diagram for comparing the extracted emissivity-area product characteristics of a method for denoising and feature extraction of infrared signals of space dim targets proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0077] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0078] What kind of scenarios is the infrared signal denoising and feature extraction method for space dim targets disclosed in the present invention mainly applied to?

[0079] Referring to Figure 1-7 , an infrared signal denoising and feature extraction method for space dim targets specifically includes the following steps:

[0080] Step 1: Use the infrared simulation system of space targets to obtain the noisy infrared radiation intensity signals of targets with regular shapes in two bands, and respectively execute Steps 2 to 5 on the infrared signals in each infrared band;

[0081] Step 2: Use the average weight fuzzy distance entropy as the fitness function;

[0082] Step 3: Use the dragonfly optimization algorithm to optimize the fitness function, and solve for the optimal decomposition layer number and quadratic penalty factor of variational mode decomposition;

[0083] Step 4: Perform variational mode decomposition on the infrared radiation intensity signal according to the optimal parameter combination to obtain the decomposition signals of K modes;

[0084] Step 5: Calculate the Pearson correlation coefficient between each modal component and the original radiation intensity signal, select the effective modal components for signal reconstruction, and achieve signal denoising;

[0085] Step 6: Obtain the denoised target infrared radiation intensity signals in the two spectral bands, and use the dual-band temperature measurement method to achieve the extraction of the temperature characteristics of the target;

[0086] Step 7: With the help of the distance information between the target and the infrared detector, achieve the extraction of the product feature of the infrared emissivity area of the target.

[0087] In a preferred embodiment, in Step 1, the infrared noisy radiation intensity signal of the target is obtained by simulation through the infrared simulation system of space targets, and the regular shapes are cone, cone with spherical bottom, cone with cylindrical bottom, cylinder, and sphere.

[0088] In a preferred embodiment, the dragonfly algorithm in Step 3 is a swarm intelligence optimization algorithm, and its basic concept is:

[0089] Step S31: Determine the scale of the dragonfly population, randomly generate X dragonflies, set [K, α] as the position vector of the dragonflies, and initialize the velocity vector;

[0090] Step S32: Calculate the fitness function of all dragonflies, and update the food position and the natural enemy position;

[0091] Step S33: Calculate the separation factor, flocking factor, collision factor, food attraction factor, and natural enemy avoidance factor of the dragonflies;

[0092] Step S34: Update the position and velocity vectors according to the surrounding radius;

[0093] Step S35: If the maximum number of iterations is reached, terminate the search and output the global optimal solution; otherwise, continue the iteration.

[0094] In a preferred embodiment, the separation factor S i of the dragonfly algorithm in step S33, the companion factor A i the collision factor C i the food attraction factor F i and the factor E for staying away from natural enemies i are updated according to the following formulas:

[0095]

[0096]

[0097]

[0098] F i = X + - X

[0099] E i = X - + X

[0100] where X is the position of the current individual, N is the number of dragonflies within the surrounding of the current individual, X j is the position of the j-th dragonfly within the surrounding of the current individual, X j is the velocity vector of the j-th dragonfly within the surrounding of the current individual, X 十 is the food position, and X 一 is the position of the natural enemy.

[0101] In a preferred embodiment, the method for updating the position and velocity vectors of the dragonfly algorithm in step S34 is as follows:

[0102] If there are no other individuals around the current individual, the velocity vector and position vector are:

[0103] X t+1 = X t + Lévy(d) × X t

[0104] ΔX t+1 = 0

[0105] Otherwise

[0106] X t+1 = X t + ΔX t+1

[0107] ΔX t+1 =(sS i +aA i +cC i +fF i +eE i )+wΔX t 。

[0108] In a preferred embodiment, the global optimal solution in step S35 is the position where the optimal individual is located, corresponding to the optimal combination [K, α] of variational mode decomposition parameters.

[0109] In a preferred embodiment, the variational mode decomposition process in step S4 is specifically as follows:

[0110] Let u k (t) be the BLIMF component, and its time-domain expression is as follows:

[0111] u k (t)=A k (t)cos(φ k (t))

[0112] Wherein, A k (t) and φ k (t) respectively represent the envelope and the phase. This algorithm mainly consists of two steps: constructing a variational problem and solving the variational problem. First, a variational problem is constructed to estimate the center frequency and bandwidth of the BLIMFs. The Hilbert transform is used to obtain the single-sided spectrum of the relevant analysis signal of each one. The spectrum of the IMF is shifted to the baseband of each u k . The bandwidth of the demodulated signal is estimated by Gaussian smoothness (i.e., the square norm). Based on the above, this problem can be expressed as the following constrained variational problem:

[0113]

[0114] Where u k represents the k-th BLIMF, w k is the center frequency of u k , represents the derivative operator, δ(t) is the Dirac function, * represents the convolution operation, and f is the original signal;

[0115] Secondly, to solve the above variational problem, the Lagrange multiplier λ and the quadratic penalty term α are introduced to transform the above problem into an unconstrained variational problem and find the optimal solution. The augmented Lagrangian function is expressed as follows:

[0116]

[0117] Then, the alternating direction method of multipliers (ADMM) is used to find the saddle point of the function as the solution of the above augmented Lagrangian equation, and updates are made separately in the frequency domain. The iterative equations are shown as follows:

[0118]

[0119]

[0120] The iterative process is repeated according to the above equations until the following convergence conditions are met and the iteration stops;

[0121]

[0122] In a preferred embodiment, the Pearson correlation coefficient calculation method in step 5 is as follows:

[0123]

[0124] where R(T) is the ratio of the radiation intensities of the blackbody at temperature T in two infrared bands, S 1 ’ is the denoised radiation intensity of the target in band 1, and S 2 ’ is the denoised radiation intensity of the target in band 2.

[0125] In a preferred embodiment, the infrared emissivity-area product feature extraction method in step 7 is as follows:

[0126]

[0127] where S’ is the denoised radiation intensity in any band, and T is the extracted temperature feature.

[0128] Example: In this part, a space target with a cylindrical bottom and a conical top is taken as an example. Its radius is 1 m, the height of the cone is 1 m, the height of the cylinder is 1 m, the thickness of the surface material is 1 mm, the emissivity is 0.8, the visible light absorptivity is 0.45, the conical rotation rate is, the initial temperature is 300 K, the observation duration is 30 s, and the two infrared bands used are 8–12 and 6–7. The noisy infrared radiation intensity signals of the target with a conventional shape in the two bands are obtained by using a space target infrared simulation system. To analyze the effect of the denoising part of the algorithm, taking the infrared radiation intensity signal in the 8–12 band as an example, Figure 2 shows the noisy radiation intensity in this band. First, the dragonfly algorithm randomly generates a population of 20 individuals, the maximum number of iterations is 20, and the ranges of the parameters K and are [5, 10] and [1500, 8000] respectively. When the fitness function curve converges during the optimization process, the corresponding parameter combination is the optimal parameter combination of variational mode decomposition, which is [10, 2016] respectively. After inputting this group into variational mode decomposition, the obtained mode decomposition results are as Figure 3As shown. Calculate the Pearson correlation coefficient between each modal component and the original signal, as Figure 4 shown. According to the effective mode selection criterion, select the first two components as effective components for signal reconstruction. The reconstructed and denoised signal is shown in Figure 5. Perform the same operation on the signal in the 6–7 band to obtain the denoised radiation intensity signals in the two bands. Perform dual-band temperature measurement on the signals in the two bands to obtain the temperature characteristic curve as Figure 6 shown. For comparison, the temperature characteristic curve before denoising is also shown. With the help of the distance information between the target and the detector and the temperature characteristics, further extract the emissivity-area product characteristics of the target to obtain the emissivity-area product characteristic curve as Figure 7 shown. For comparison, the emissivity-area product characteristic curve before denoising is also shown. It can be seen that under this method, using this method can effectively improve the accuracy of target feature extraction and reduce the influence of noise on feature extraction.

[0129] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An infrared signal denoising and feature extraction method for space dim targets, characterized in that, it specifically includes the following steps: Step 1: Use a space target infrared simulation system to obtain the noisy infrared radiation intensity signals of targets with regular shapes in two bands, and perform Steps 2 to 5 on the infrared signals in each infrared band respectively; Step 2: Use the average weighted fuzzy distance entropy as the fitness function; Step 3: Use the dragonfly optimization algorithm to optimize the fitness function, and solve the optimal decomposition layer number and quadratic penalty factor of variational mode decomposition; Step 4: Perform variational mode decomposition on the infrared radiation intensity signal according to the optimal parameter combination to obtain the decomposition signals of K modes; Step 5: Calculate the Pearson correlation coefficient between each mode component and the original radiation intensity signal, and select the effective mode components for signal reconstruction to achieve signal denoising; Step 6: Obtain the denoised target infrared radiation intensity signals in the two spectral bands, and use the dual-band temperature measurement method to achieve the extraction of the temperature characteristics of the target; Step 7: With the help of the distance information between the target and the infrared detector, achieve the extraction of the product characteristic of the infrared emissivity area of the target; In the said Step 2, the solution model of the average weighted fuzzy distance entropy is as follows: Estimate the empirical probability density function of the similarity matrix of the signal according to the histogram, and calculate the normalized fuzzy distance entropy as: where M is the number of cells of the histogram; Calculate the mutual information between the mode component and the original infrared radiation intensity signal Y as: where p(u), p(Y), and p(u, Y) represent the marginal probability distribution and the joint probability distribution; Then the average weighted fuzzy distance entropy is: where u(i) represents the i-th mode component, and K is the number of mode components; The dragonfly algorithm in the said Step 3 is a swarm intelligence optimization algorithm, and its basic concept is: Step S31: Determine the scale of the dragonfly population, randomly generate X dragonflies, set [K, α] as the position vector of the dragonflies, and initialize the velocity vector; Step S32: Calculate the fitness function of all dragonflies, and update the food position and the natural enemy position; Step S33: Calculate the separation factor, the accompanying factor, the collision factor, the food attraction factor, and the factor of staying away from the natural enemy of the dragonflies; Step S34: Update the position and velocity vectors according to the surrounding radius; Step S35: When the maximum number of iterations is met, terminate the search and output the global optimal solution; otherwise, continue the iteration.

2. According to the method for infrared signal denoising and feature extraction of a space dim target described in Claim 1, characterized in that, in the said Step 1, the infrared noisy radiation intensity signal of the target is obtained by simulation through a space target infrared simulation system, and the regular shapes are cone, spherical bottom cone, cylindrical bottom cone, cylinder, and sphere.

3. According to the method for infrared signal denoising and feature extraction of a space dim target described in Claim 1, characterized in that, The separation factor S of the dragonfly algorithm in step S33 i , the accompanying flight factor A i , the collision factor C i , the food attraction factor F i and the factor E of staying away from natural enemies i are updated as follows: F i = X + - X E i = X - + X Among them, X is the position of the current individual, N is the number of dragonflies within the current individual's vicinity, X j is the position of the j-th dragonfly within the current individual's vicinity, X j is the velocity vector of the j-th dragonfly within the current individual's vicinity, X 十 is the food position, X 一 is the position of the natural enemy.

4. According to the method for infrared signal denoising and feature extraction of a space dim target described in Claim 1, characterized in that, The method for updating the position and velocity vectors of the dragonfly algorithm in the said Step S34 is: If there are no other individuals around the current individual, the velocity vector and the position vector are: X t+1 = X t + Lévy(d) × X t ΔX t+1 = 0 Otherwise X t+1 = X t + ΔX t+1 ΔX t+1 =(sS i +aA i +cC i +fF i +eE i )+wΔX t 。 5. A method for denoising and feature extraction of infrared signals of dim space targets according to claim 1, characterized in that, the global optimal solution in step S35 is the position where the optimal individual is located, corresponding to the optimal combination [K, α] of variational mode decomposition parameters.

6. A method for denoising and feature extraction of infrared signals of dim space targets according to claim 1, characterized in that, the variational mode decomposition process in step S4 is specifically as follows: Let u k (t) be the BLIMF component, and its time-domain expression is as follows: u k u(t) = A k u(t)cos(φ k (t)) Among them, where A k (t) and φ k (y) represent the envelope and phase respectively. This algorithm mainly consists of two steps: constructing a variational problem and solving the variational problem. First, a variational problem is constructed to estimate the center frequency and bandwidth of BLIMFs. The Hilbert transform is used to obtain the single-sided spectrum of the correlation analysis signal of each one. The spectrum of the IMF is shifted to the baseband of each u k . The bandwidth of the demodulated signal is estimated by Gaussian smoothness. Based on the above, this problem can be formulated as the following constrained variational problem: where u k represents the k-th BLIMF, w k is the center frequency of u k , represents the derivative operator, δ(t) is the Dirac function, * represents the convolution operation, and f is the original signal; The augmented Lagrangian function adds a quadratic penalty term to the Lagrangian function. Secondly, to solve the above variational problem, the Lagrange multiplier λ and the quadratic penalty term α are introduced to transform the above problem into an unconstrained variational problem and find the optimal solution. The augmented Lagrangian function corresponding to this unconstrained variational problem is expressed as follows: Then, the alternating direction multiplier method is used to find the saddle point of the function as the solution of the above augmented Lagrangian equation, and updates are made separately in the frequency domain The iterative equation is shown as follows: Repeat the iterative process according to the above equation until the following convergence condition is met and the iteration stops; 7. A method for denoising and feature extraction of infrared signals of dim space targets according to claim 1, characterized in that, the calculation method of the Pearson correlation coefficient in step 5 is: where R(T) is the ratio of the radiation intensities of a blackbody at temperature T within two infrared bands, S 1 ’ is the denoised radiation intensity of the target within band 1, S 2 ’ is the denoised radiation intensity of the target within band 2.

8. A method for denoising and feature extraction of infrared signals of dim space targets according to claim 1, characterized in that, the method for extracting the infrared emissivity area product feature in step 7 is: Among them, S ’ is the radiation intensity after denoising in any wavelength band, and T is the extracted temperature feature.

Images