A non-uniform infrared correction method and system for dynamically adjusting correction parameters

Abstract

The present application relates to the technical field of infrared temperature measurement, and particularly relates to a non-uniformity infrared correction method and system for dynamically adjusting correction parameters, the method comprising: acquiring a visible light image and a short-wave infrared image, and performing space-time registration with a target infrared image; extracting guide features representing scene spatial structures from the registered visible light image and short-wave infrared image, which are used to guide infrared correction; constructing a non-linear response model to quantify the non-linear effect of a detector; constructing a joint optimization objective function, and obtaining a non-uniformity parameter sequence and a real infrared image sequence by minimizing the objective function; and correcting the input target infrared image frame by frame using the non-uniformity parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of infrared temperature measurement, in particular to a non-uniformity infrared correction method and system for dynamically adjusting correction parameters. BACKGROUND

[0002] Non-uniformity correction in infrared thermal imaging is a key step to improve image quality and temperature measurement accuracy, and its core purpose is to compensate for the inconsistent response characteristics (mainly represented by gain and offset) between each pixel on the focal plane array (FPA) of the detector. Long-wave infrared (LWIR) is widely used because of its strong atmospheric penetration, which can effectively detect the thermal radiation of objects at normal temperature. Existing infrared thermal imaging mainly uses long-wave infrared. Short-wave infrared (SWIR) mainly relies on environmental reflection imaging, and its characteristics are clear image target details and high contrast. It plays an irreplaceable role in high-temperature monitoring, active covert imaging, and complex environment penetration detection.

[0003] Existing technologies have methods for non-uniformity correction based on integration time dynamic adjustment and image entropy. Integration time adjustment is mainly hardware parameter optimization, aiming to make the detector work in the linear region to create better conditions for subsequent correction (usually traditional calibration). Image entropy adjustment mainly uses information as an indicator to evaluate correction effect or drive parameter optimization, usually embedded in an optimization framework (as a cost function), but in complex scenes, there are strong non-uniformities or targets, and the optimal solution may not exist (for example, maximizing entropy may cause over-sharpening or introduce noise).

[0004] However, existing technologies still cannot effectively solve the three core problems in non-uniformity correction: frequent calibration, limitations of two-point linear models, and inability to efficiently implement multi-modal cooperation (fusion of short-wave infrared imaging) to improve the effectiveness of non-uniformity correction. SUMMARY

[0005] The present application establishes a non-linear response model to simulate complex non-linear effects such as detector saturation, crosstalk, and temperature drift, and uses fusion of visible light and short-wave infrared (SWIR) image structural information to constrain real radiation detection values to achieve non-uniformity correction in complex scenes.

[0006] The technical solution proposed by the present application is a non-uniformity infrared correction method for dynamically adjusting correction parameters, which includes:

[0007] Obtain visible light images and short-wave infrared images, and perform spatio-temporal registration with target infrared images;

[0008] Extract guide features representing scene spatial structure from the registered visible light images and short-wave infrared images to guide infrared correction;

[0009] constructing a non-linear response model to quantify the non-linear effect of the detector;

[0010] constructing a joint optimization objective function, and obtaining the non-uniformity parameter sequence and the real infrared image sequence by minimizing the objective function;

[0011] correcting the input target infrared image frame by frame using the non-uniformity parameters.

[0012] Preferably, the visible light image and the short-wave infrared image are acquired and spatio-temporally registered with the target infrared image, comprising:

[0013] acquiring a plurality of visible light image data within the acquisition period to form a visible light image set: I vis = {I vis (t) | t = t0, t1, …, t n}; wherein t n represents the nth acquisition time point within the acquisition period; I vis (t) represents the visible light image data;

[0014] acquiring a plurality of short-wave infrared image data within the acquisition period to form a short-wave infrared image set: I swir = {I swir (t) | t = t0, t1, …, t n}; I swir (t) represents the short-wave infrared image data;

[0015] affine transforming the visible light image data and the short-wave infrared image data to realize registration with the target infrared image, i.e.

[0016] wherein, and represent the image data after affine transformation; φ t and ψ t represent the registration parameters of the visible light image and the registration parameters of the short-wave infrared image; represents the affine transformation.

[0017] Preferably, the guide features representing the spatial structure of the scene are extracted from the registered visible light image and short-wave infrared image to guide infrared correction, comprising:

[0018] acquiring the visible light image after matching extracting the edge features of the visible light image represent the x-axis direction gradient and the y-axis direction gradient; σ represents the Sigmoid activation function;

[0019] Acquiring a registered short-wave infrared image Extracting texture features of the short-wave infrared image wherein LBP represents a local binary pattern operator;

[0020] Constructing adaptive fusion features based on edge features of the visible light image and texture features of the short-wave infrared image wherein K represents an attention weighting function.

[0021] Preferably, the constructing a nonlinear response model and quantifying the nonlinear effect of the detector comprises:

[0022] Establishing a relationship between the observation value and the true radiation value of each pixel of the target infrared image, namely: raw (x,y,t) = f(θ(x,y))I true (x,y,t) + ∈(x,y,t); wherein I raw (x,y,t) represents the observation value of the pixel point at (x,y) of the target infrared image at time t; I true (x,y,t) represents the true observation value of the pixel point at (x,y) at time t; ∈(x,y,t) represents the environmental noise of the pixel point at (x,y) at time t; f θ (x,y) represents a pixel nonlinear response function; θ represents a parameter of the pixel nonlinear response function;

[0023] wherein θ = [a0, a1, a2, a3]; a0, a1, a2, and a3 represent a detector dark current offset, a linear gain amount, a quadratic nonlinear term, and an exponential term, respectively.

[0024] Preferably, the constructing a joint optimization objective function and obtaining a non-uniformity parameter sequence and a true infrared image sequence by minimizing the objective function comprises:

[0025] Establishing a structure guiding term, taking the structure guiding term as a supervision signal, and guiding the non-uniformity correction process of the target infrared image comprises:

[0026] Let the structure guiding term be wherein DoG represents a difference of Gaussian edge detection operator; α t represents a dynamic weight of the visible light feature; β t represents a dynamic weight of the short-wave infrared feature; γ t represents a fusion feature weight;

[0027] wherein SNR swir represents a signal-to-noise ratio of the short-wave infrared image; SNR vis represents a signal-to-noise ratio of the visible light image; ∈ represents environmental noise;

[0028] Constructing joint optimization objective function:

[0029]

[0030] where I raw (x,y,t) represents the original observation value of the target infrared image (x,y) point pixel, I true (x,y,t) represents the corrected observation value of the target infrared image; represents the spatial gradient of the corrected infrared image (x,y) point pixel; represents the gradient operator; θ prior represents the calibrated initial value of the parameter;

[0031] Solving f = min G by alternatingly solving sub-problems ADMM algorithm to obtain the real infrared image sequence {I true (x,y,t)|t∈[1,T]} and non-uniformity parameter sequence {λ s , λ p , θ(x,y)}.

[0032] Preferably, the establishment structure guiding term, taking the structure guiding term as a supervision signal, guides the non-uniform correction process of the target infrared image, further comprising:

[0033] Reconstructing the structure guiding term, constructing a new structure guiding term by a nonlinear differential constraint algorithm to solve the adaptation problem between the linear expression of the structure guiding term and the nonlinear nature of the nonlinear correction;

[0034] Let the new structure guiding term be:

[0035]

[0036] where Ψ represents the amplitude-direction joint loss function; Ψ = a·exp(-k|b|), a, b represent two variables of the amplitude-direction joint loss function; k represents the direction sensitive coefficient; W vis , W swir respectively represent the learning convolution kernel of the visible light gradient and the short wave infrared gradient of the ReLU network; W v represents the fusion gradient convolution kernel of the ReLU network; represents the texture feature gradient of the short wave infrared image; ⊙ represents the multiplication of corresponding position elements; represents the normalized value of the direction angle between ;

[0037] Constructing a new adaptive fusion feature:

[0038]

[0039] establishing a new dynamic weight α t , β t update strategy, namely:

[0040] wherein, TV represents total variation calculation; η, λ' represent update weight coefficient and weighting coefficient.

[0041] Preferably, the input target infrared image is corrected frame by frame using the non-uniformity parameters, comprising:

[0042] The input target infrared image is corrected using the pre-calibrated non-uniformity parameter sequence, and the corrected infrared image is:

[0043] wherein, O t (x,y) represents the observation value of the tth frame of the target infrared image at position (x,y); t = 1, 2, …, T; O t (x,y) represents the offset field parameter of the tth frame at position (x,y); G t (x,y) represents the gain field parameter of the tth frame at position (x,y); represents the corrected pixel observation value;

[0044] O t (x,y) = ||(a 0, a 2) t ||, wherein (a 0, a 2) t represents the offset field vector composed of the detector dark current offset and the quadratic nonlinearity term at the tth frame; G t (x,y) = ||(a 1, a 3) t ||, (a 1, a 3) t represents the linear gain amount and the exponential term of the detector at the tth frame.

[0045] Preferably, the input target infrared image is corrected frame by frame using the non-uniformity parameters, further comprising:

[0046] An adaptive adjustment mechanism for non-uniformity parameters is constructed, specifically:

[0047] adjusted gain field parameter wherein η 0 represents the adaptive intensity;

[0048] adjusted offset field parameter

[0049] wherein, represents the neighborhood of position (x,y);

[0050] I represents all the sampling points geometry of the target infrared image; I ref (x,y) represents the infrared reference radiation value at position (x,y); med represents the absolute median deviation.

[0051] A non-uniformity infrared correction system for dynamically adjusting correction parameters, the system being used to perform the non-uniformity infrared correction method for dynamically adjusting correction parameters.

[0052] A computer-readable storage medium storing a computer program, the computer program being executed by a processor to implement the non-uniformity infrared correction method for dynamically adjusting correction parameters.

[0053] Advantages of the present application:

[0054] The present application, in combination with the non-linear nature of non-uniformity correction, optimizes the construction of the structure guide term, so that it can better match the non-linear characteristics of the non-uniformity correction process. At the same time, through the guidance of the structure guide term, high-precision non-linear correction is realized without calibration, to adapt to the correction needs of some complex scenes. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A flowchart of the non-uniformity infrared correction method for dynamically adjusting correction parameters of the present application. DETAILED DESCRIPTION

[0056] The following description is provided to disclose the present application so that those skilled in the art can implement the present application. The preferred embodiments in the following description are only examples, and other obvious modifications can be made by those skilled in the art. The basic principles of the present application defined in the following description can be applied to other embodiments, modifications, improvements, equivalents and other technical solutions without departing from the spirit and scope of the present application.

[0057] It can be understood that the term "one" should be understood as "at least one" or "one or more", that is, in one embodiment, the number of one element can be one, and in another embodiment, the number of the element can be multiple, and the term "one" cannot be understood as a limitation on the number.

[0058] Embodiment one:

[0059] Reference Figure 1 The technical solution provided by the present application is: a non-uniformity infrared correction method for dynamically adjusting correction parameters, comprising the following steps:

[0060] Step 1, obtaining a visible light image and a short-wave infrared image, and performing space-time registration with a target infrared image; specifically comprising the following steps:

[0061] Step 1.1, obtain multiple visible light image data in the acquisition period to form a visible light image set: I vis = {I vis (t) | t = t0, t1, …, t n}; where t n represents the nth acquisition time point in the acquisition period; I vis (t) represents visible light image data;

[0062] Step 1.2, obtain multiple short-wave infrared image data in the acquisition period to form a short-wave infrared image set: I swir = {I swir (t) | t = t0, t1, …, t n}; I swir (t) represents short-wave infrared image data;

[0063] Step 1.3, affine transform the visible light image data and short-wave infrared image data to realize registration with the target infrared image, i.e.

[0064] wherein, and represent image data after affine transformation; φ t and ψ t represent registration parameters of the visible light image and registration parameters of the short-wave infrared image; represents affine transformation.

[0065] Step 2, extract guide features representing the spatial structure of the scene from the registered visible light image and short-wave infrared image for guiding infrared correction; specifically including the following steps:

[0066] Step 2.1, obtain the visible light image after matching value extract edge features of the visible light image represent x-axis direction gradient and y-axis direction gradient; σ represents Sigmoid activation function; for example, use the 3rd layer of VGG16 network to extract edge features;

[0067] Step 2.2, obtain the registered short-wave infrared image extract texture features of the short-wave infrared image wherein LBP represents local binary pattern operator; for example, use ResNet network to extract texture features (texture gradient);

[0068] Step 2.3, construct adaptive fusion features based on edge features of the visible light image and texture features of the short-wave infrared image wherein K represents attention weighting function.

[0069] Step 3, constructing a nonlinear response model to quantify the nonlinear effect of the detector; specifically including the following steps:

[0070] Establishing the relationship between the observation value and the true radiation value of each pixel of the target infrared image, that is: raw (x,y,t) = f(θ(x,y))I true (x,y,t) + ∈(x,y,t) ;

[0071] Wherein, I raw (x,y,t) represents the observation value of the pixel point at (x,y) of the target infrared image at time t; I true (x,y,t) represents the true observation value of the pixel point at (x,y) at time t; ∈(x,y,t) represents the environmental noise of the pixel point at (x,y) at time t; f θ (x,y) represents the pixel nonlinear response function; θ represents the parameter of the pixel nonlinear response function;

[0072] Wherein, θ = [a0, a1, a2, a3]; a0, a1, a2, a3 represent the dark current offset of the detector, the linear gain, the quadratic nonlinear term (used to reflect the crosstalk effect of the detector), and the exponential term (used to fit the saturation effect), respectively.

[0073] Step 4, constructing a joint optimization objective function to obtain the non-uniformity parameter sequence and the true infrared image sequence by minimizing the objective function; specifically including the following steps:

[0074] Establishing a structure guide term, taking the structure guide term as a supervision signal to guide the non-uniformity correction process of the target infrared image, including:

[0075] Let the structure guide term be Wherein, DoG represents a Gaussian difference edge detection operator; α t represents the dynamic weight of the visible light feature; β t represents the dynamic weight of the short-wave infrared feature; γ t represents the fusion feature weight;

[0076] Wherein, SNR swir represents the signal-to-noise ratio of the short-wave infrared image; SNR vis represents the signal-to-noise ratio of the visible light image; ∈ represents the environmental noise;

[0077] Constructing a joint optimization objective function:

[0078]

[0079]

[0080] where I raw (x,y,t) represents the original observation value of the target infrared image (x,y) point pixel, I true (x,y,t) represents the corrected observation value of the target infrared image. represents the spatial gradient of the corrected infrared image (x,y) point pixel. represents the gradient operator; θ prior represents the initial value of the parameter calibration.

[0081] The ADMM algorithm is solved by alternately solving the sub-problems f = min G, to obtain the real infrared image sequence {I true (x,y,t) | t ∈ [1,T] and the non-uniformity parameter sequence {λ s , λ p , θ (x,y)}.

[0082] Step 5, using the non-uniformity parameter, the input target infrared image is corrected frame by frame; Specifically, the following steps are included:

[0083] The input target infrared image is corrected using the pre-calibrated non-uniformity parameter sequence, and the corrected infrared image is:

[0084] wherein, represents the observation value of the target infrared image at position (x,y) in the t frame; t = 1,2…,T; O t (x,y) represents the offset field parameter at position (x,y) in the t frame; G t (x,y) represents the gain field parameter at position (x,y) in the t frame. represents the corrected pixel observation value.

[0085] O t (x,y) = || (a0,a2) t ||, wherein (a0,a2) t represents the offset field vector composed of the detector dark current offset and the quadratic nonlinearity term at the t frame; G t (x,y) = || (a1,a3) t ||, (a1,a3) t represents the linear gain amount and the exponential term of the detector at the t frame.

[0086] Example two:

[0087] In example one, the structure guiding term is linear expression, but the essence of non-uniformity correction is nonlinear, in order to better match the nonlinear characteristics of correction, we will reconstruct the structure guiding term, the specific scheme is as follows:

[0088] The reconstruction structure guide item is constructed by a nonlinear differential constraint algorithm to solve the fitting problem between the linear expression of the structure guide item and the nonlinear nature of the nonlinear correction.

[0089] The new structure guide item is set as:

[0090]

[0091] Wherein, ψ represents the amplitude-direction joint loss function; ψ=a·exp(-k|b|), a and b represent two variables of the amplitude-direction joint loss function; k represents a direction sensitive coefficient; W vis , W swir respectively represent the learning convolution kernel of the visible light gradient and the short wave infrared gradient of the ReLU network; W v represents the fusion gradient convolution kernel of the ReLU network; represents the texture feature gradient of the short wave infrared image; ⊙ represents multiplication of corresponding position elements; represents the normalized value of the direction angle between .

[0092] The new adaptive fusion feature is constructed as:

[0093]

[0094] The update strategy of the new dynamic weight α t , β t is established, that is:

[0095] Wherein, TV represents the total variation calculation; η and λ' represent the update weight coefficient and the weighting coefficient;

[0096] The non-uniformity parameter adaptive adjustment mechanism is constructed, specifically as:

[0097] The adjusted gain field parameter Wherein, η0 represents the adaptive strength;

[0098] The adjusted offset field parameter

[0099] Wherein, represents the neighborhood of position (x, y);

[0100] represents the geometry of all sampling points of the target infrared image; I ref (x, y) represents the infrared reference radiation value of position (x, y); med represents the absolute median difference.

[0101] After obtaining the new structure guide item, the non-uniformity correction is guided by using the new structure guide item;

[0102] A new target optimization function is constructed:

[0103] Wherein, λ g , λ o Indicate gain weight and offset weight;

[0104] By using an alternating solution sub-problem algorithm, the infrared image true radiation I' true (x,y,t) and

[0105] The application further provides a non-uniformity infrared correction system for dynamically adjusting correction parameters, which is used for executing the non-uniformity infrared correction method for dynamically adjusting correction parameters.

[0106] A computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the non-uniformity infrared correction method for dynamically adjusting correction parameters.

[0107] The processes described above with reference to the flowcharts can be implemented as computer software programs in accordance with embodiments of the present disclosure. Embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication section, and / or installed from a detachable medium. When the computer program is executed by a central processing unit (CPU), the above-described functions defined in the methods of the present application are performed. It should be noted that the computer readable medium of the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device. In the present application, the computer readable signal medium can include a data signal carried in a baseband or as part of a carrier wave, in which the computer readable program code is carried. Such a propagated data signal can take a variety of forms, including but not limited to, an electromagnetic signal, an optical signal or any suitable combination of the above. The computer readable signal medium can also be any computer readable medium that can send, propagate or transfer the program for use by or in connection with an instruction execution system, apparatus or device. The program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to, wireless, wire, optical cable, RF or the like, or any suitable combination of the above.

[0108] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0109] Those skilled in the art should understand that the embodiments of the present invention described above and shown in the accompanying drawings are merely examples and do not limit the present invention. The purpose of the present invention has been fully and effectively achieved. The functions and structural principles of the present invention have been shown and explained in the embodiments. Without departing from the principles described, the implementation of the present invention may have any changes or modifications.

Claims

1. A non-uniform infrared correction method for dynamically adjusting correction parameters, characterized in that, The method includes: Acquire visible light and shortwave infrared images, and perform spatiotemporal registration with the target infrared image; Guiding features characterizing the spatial structure of the scene are extracted from the registered visible light and short-wave infrared images for use in guiding infrared correction. Construct a nonlinear response model to quantify the nonlinear effects of the detector; A joint optimization objective function is constructed, and by minimizing the objective function, a non-uniform parameter sequence and a real infrared image sequence are obtained. The non-uniformity parameter is used to perform frame-by-frame correction on the input target infrared image.

2. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 1, characterized in that, The acquisition of visible light images and short-wave infrared images, and the spatiotemporal registration with the target infrared image, includes: Multiple visible light image data acquired within the acquisition period constitute a visible light image set: I vis ={I vis (t)|t=t0、t1、…、t n }; where t n Indicates the nth acquisition time point within the acquisition period; I vis (t) represents visible light image data; Multiple shortwave infrared image data acquired within the acquisition period constitute a shortwave infrared image set: I swir ={I swir (t)|t=t0、t1、…、t n };I swir (t) represents shortwave infrared image data; Affine transformations are performed on visible light image data and shortwave infrared image data to achieve registration with the target infrared image, i.e.: in, and Represents the image data after affine transformation; φ t and ψ t These represent the registration parameters for visible light images and the registration parameters for shortwave infrared images. This represents an affine transformation.

3. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 2, characterized in that, The step of extracting guiding features characterizing the spatial structure of the scene from the registered visible light image and short-wave infrared image for guiding infrared correction includes: Visible light image after obtaining matching values Extracting edge features from visible light images The gradients along the x-axis and y-axis represent the gradients; σ represents the Sigmoid activation function. Acquire registered shortwave infrared images Extracting texture features from shortwave infrared images Wherein, LBP represents the Local Binary Mode Operator; An adaptive fusion feature is constructed based on edge features from visible light images and texture features from short-wave infrared images. Where K represents the attention weighting function.

4. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 3, characterized in that, The construction of the nonlinear response model and the quantification of the detector's nonlinear effects include: Establish the relationship between the observed value and the true radiance value of each pixel in the target infrared image, i.e.: I raw (x,y,t)=f(θ(x,y))I true (x,y,t)+∈(x,y,t); where, I raw (x,y,t) represents the observed value of the pixel at (x,y) in the infrared image of the target at time t; I true (x,y,t) represents the actual observed value of the pixel at (x,y) at time t; ∈(x,y,t) represents the environmental noise of the pixel at (x,y) at time t; f θ (x,y) represents the pixel nonlinear response function; θ represents the parameters of the pixel nonlinear response function; Where θ = [a0, a1, a2, a3]; a0, a1, a2, a3 represent the detector dark current offset, linear gain, quadratic nonlinear term, and exponential term, respectively.

5. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 4, characterized in that, The construction of the joint optimization objective function, by minimizing the objective function, yields a non-uniform parameter sequence and a true infrared image sequence, including: A structure-guided term is established and used as a monitoring signal to guide the non-uniformity correction process of the target infrared image, including: Set up a structural guiding item Where DoG represents the Gaussian difference edge detection operator; α t The dynamic weights representing visible light characteristics; β t Dynamic weights representing shortwave infrared characteristics; γ t Indicates the weights of the fused features; Among them, SNR swir The signal-to-noise ratio (SNR) of a shortwave infrared image. vis The signal-to-noise ratio of a visible light image is represented by ∈; environmental noise is represented by ∈. Construct a joint optimization objective function: Among them I raw (x,y,t) represents the original observation value of the pixel at point (x,y) in the infrared image of the target. true (x,y,t) represents the observed value after correction of the infrared image of the target; This represents the spatial gradient of the pixel at point (x,y) in the corrected infrared image; Represents the gradient operator; θ prior Indicates the initial calibration value of the parameter; The ADMM algorithm, which alternately solves subproblems, is used to solve f = minG to obtain the real infrared image sequence {I}. true (x,y,t)|t∈[1,T]} and the non-uniform parameter sequence {λ s , λ p ,θ(x,y)}.

6. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 5, characterized in that, The establishment of the structure guidance term, using the structure guidance term as a monitoring signal to guide the non-uniformity correction process of the target infrared image, further includes: The structural guiding term is reconstructed by constructing a new structural guiding term through a nonlinear differential constraint algorithm to solve the adaptation problem between the linear expression of the structural guiding term and the nonlinear nature of nonlinear correction. Let a new structure guide be defined: Where ψ represents the magnitude-direction joint loss function; ψ = a·exp(-k|b|), a and b represent two variables of the magnitude-direction joint loss function; k represents the direction sensitivity coefficient; W vis W swir W represents the learned convolutional kernels for the visible light gradient and short-wave infrared gradient of the ReLU network, respectively; v This represents the fused gradient convolution kernel of the ReLU network; The gradient represents the texture features of a shortwave infrared image; ⊙ represents element-wise multiplication at corresponding positions. express The normalized value of the angle between the directions; Constructing new adaptive fusion features: Establish new dynamic weights α t β t The update strategy is as follows: in, TV represents total variation calculation; η and λ′ represent the updated weight coefficients and weighting coefficients.

7. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 6, characterized in that, The step of performing frame-by-frame correction on the input target infrared image using non-uniform parameters includes: The input target infrared image is corrected using a pre-calibrated non-uniform parameter sequence. The corrected infrared image is as follows: in, Let represent the observation value of the target in the t-th frame of the infrared image at position (x, y); t = 1, 2, ..., T; O t (x,y) represents the offset field parameter at position (x,y) in frame t; G t (x,y) represents the gain field parameters at position (x,y) in frame t; This represents the corrected pixel observation value; O t (x,y)=||(a0,a2) t ||, where (a0, a2) t G represents the offset field vector consisting of the detector dark current offset and the second-order nonlinear term at frame t; t (x,y)=||(a1,a3) t ||,(a1,a3) t This represents the linear gain and exponential term of the detector at frame t.

8. The non-uniform infrared correction method for dynamically adjusting correction parameters according to claim 7, characterized in that, The step of performing frame-by-frame correction on the input target infrared image using non-uniform parameters further includes: An adaptive adjustment mechanism for non-uniform parameters is constructed, specifically as follows: Adjusted gain field parameters Where η0 represents the adaptive strength; Adjusted offset field parameters in, Represents the neighborhood of position (x, y); Represents the geometry of all sampling points in the infrared image of the target; I ref (x,y) represents the infrared reference radiation value at position (x,y); med represents the calculation of the absolute median difference.

9. A non-uniform infrared correction system with dynamically adjustable correction parameters, characterized in that, The system is used to perform a non-uniform infrared correction method for dynamically adjusting correction parameters as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that is executed by a processor to implement a non-uniform infrared correction method for dynamically adjusting correction parameters as described in any one of claims 1-8.

Images