Joint recovery method for low-extinction long-wave infrared focal plane polarimetric camera calibration

Abstract

The application discloses a combined recovery method for low-extinction ratio long-wave infrared focal plane polarization camera calibration, and relates to the field of optical detection, which comprises the following steps: for a long-wave infrared focal plane polarization camera to be calibrated, calibrating the extinction ratio and estimating a system equivalent analysis matrix, and then determining an initial Stokes parameter recovery result and an initial recovered polarization degree image; taking the initial recovered polarization degree image as an initial value, guiding DOLP by using an S0 total intensity component in the initial Stokes parameter recovery result, and constructing a polarization physical constraint combined recovery model; the recovery requirement is to directly recover in a DoLP domain or not directly recover in the DoLP domain; according to the recovery requirement, the polarization physical consistency constraint is used to solve the polarization physical constraint combined recovery model, and a combined recovery result is obtained. The application can recover the polarization precision while effectively suppressing a large amount of noise caused by matrix inversion.

Description

Technical Field

[0001] This application relates to the field of optical inspection, and in particular to a joint recovery method for calibration of a low extinction ratio long-wave infrared split-plane polarization camera. Background Technology

[0002] A focal plane polarization camera achieves single-exposure acquisition of scene polarization information by integrating micro-polarization arrays in different directions at the pixel level of the focal plane. Long-wave infrared focal plane polarization cameras have significant application value in target detection, camouflage identification, scene segmentation, material classification, and remote sensing due to their advantages such as passive detection, strong low-light adaptability, and sensitivity to target polarization characteristics.

[0003] Existing split-plane polarization cameras typically acquire the scene's radiative response through four polarization direction sub-channels, and then calculate the Stokes parameters S0, S1, S2, as well as the polarization degree parameter DoLP and polarization angle AoP. However, due to manufacturing errors, inconsistent transmittance, polarization axis deviation, and low extinction ratio in actual devices, micro-polarizers deviate significantly from the ideal analysis model in actual measurements. Therefore, it is usually necessary to construct a system analysis matrix or an equivalent Mueller matrix, and then perform polarization calibration recovery through matrix inversion. However, in the long-wave infrared band, the system analysis matrix often deviates from the ideal state due to limitations in device fabrication, thermal noise, response non-uniformity, and the low extinction ratio of micro-polarizers. When this matrix is ​​inverted, measurement noise, non-uniform noise, and modeling errors are significantly amplified, especially in the DoLP parameter, which is composed of the nonlinearity of S1, S2, and S0, manifesting as large-area noise, pseudo-textures, edge fragmentation, and local distortion. This not only reduces the quality of the polarization image but also weakens the actual calibration effect.

[0004] In existing technologies, one type of method focuses on improving the accuracy of the calibration matrix, but struggles to suppress noise amplification after inverse kinematics (IDK). Another type of method performs conventional filtering or smoothing on DoLP images, but often ignores the physical correlation between polarization parameters, easily leading to blurred edges, distorted polarization features, and a lack of physical consistency in the restored results. Especially under low extinction ratio conditions, relying solely on traditional Mueller matrix inverse kinematics is insufficient to balance polarization restoration accuracy and image robustness. Summary of the Invention

[0005] The purpose of this application is to provide a joint restoration method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera, which can effectively suppress a large amount of noise caused by matrix inversion while restoring polarization accuracy.

[0006] To achieve the above objectives, this application provides the following solution: This application provides a joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera, including: For a long-wave infrared focal plane polarization camera to be calibrated, the extinction ratio is calibrated and the system equivalent analysis matrix is ​​estimated. Based on the system equivalent analysis matrix, the initial Stokes parameter recovery result and the initial recovered polarization degree image are determined. Using the initial restored polarization degree image as the initial value, the total intensity component of S0 in the initial Stokes parametric restoration result is used to guide DOLP, and a polarization physical constraint joint restoration model is constructed. Construct polarization physical consistency constraints and select recovery requirements; the recovery requirements are either direct recovery in the DoLP domain or recovery not directly in the DoLP domain. Based on the recovery requirements, the polarization physical consistency constraint is used to solve the joint recovery model of the polarization physical constraint, and the joint recovery result is obtained.

[0007] According to the specific embodiments provided in this application, the following technical effects are disclosed: For a long-wave infrared focal plane polarization camera to be calibrated, this application performs extinction ratio calibration and estimates the system equivalent analysis matrix, thereby determining the initial Stokes parametric recovery result and the initial recovered polarization degree image. Then, based on this, a guided filtering joint correction mechanism based on complementary intensity-polarization features is introduced. Specifically, using the initial recovered polarization degree image as the initial value, the S0 total intensity component in the initial Stokes parametric recovery result is used to guide the DOLP, constructing a polarization physical constraint joint recovery model. This process does not rely on ideal high extinction ratio device conditions, nor does it require significant modifications to the existing hardware system; it can be achieved simply by jointly optimizing the calibration recovery process. Furthermore, this process can effectively suppress a large amount of noise caused by matrix inversion while restoring polarization accuracy.

[0008] A polarization physical consistency constraint is constructed, and a recovery requirement is selected. Based on the recovery requirement, the polarization physical consistency constraint is used to solve the joint recovery model, obtaining the joint recovery result. Corresponding to different recovery requirements, this application includes a direct recovery implementation in the DoLP domain and a joint recovery implementation in the Stokes component domain. A polarization parameter physical consistency constraint is introduced during the recovery process to ensure that the recovery result conforms to the physical laws of polarization images, thereby avoiding non-physical results that may occur during conventional numerical optimization. Furthermore, direct recovery in the DoLP domain can be understood as directly generating the final DoLP image and constraining its pixels to keep the DoLP value within a physically reasonable range. Recovery not directly in the DoLP domain involves calculating the DoLP from the Stokes component after the S1 and S2 recovery steps are completed. This allows the recovered DoLP to more accurately reflect the polarization characteristics of the target region while maintaining physical reasonableness, making it more reliable and convincing.

[0009] In summary, this application, while retaining the system's polarization calibration capability, effectively suppresses matrix pseudo-inverse recovery noise under low extinction ratio conditions. In particular, it significantly improves the random noise, pseudo-texture, and edge fragmentation problems in DoLP images, thereby obtaining recovery results that combine polarization accuracy, physical consistency, and image structure preservation. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the joint recovery method for calibration of a low extinction ratio long-wave infrared split-plane polarization camera in one embodiment of this application.

[0012] Figure 2 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0013] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0014] This application proposes a novel calibration processing method that, while preserving the polarization recovery capability of the inverse Mueller matrix, suppresses noise propagation caused by low extinction ratio, and maintains the physical rationality of polarization parameters and the authenticity of image structure.

[0015] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0016] Before implementing the joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera protected in this application, this application first establishes a non-ideal polarization measurement model for the long-wave infrared focal plane polarization camera.

[0017] Each superpixel of a long-wave infrared focal plane polarization camera consists of multiple polarization sub-channels with different analysis directions. For any i-th polarization analysis sub-channel, its response under actual device conditions depends not only on the total intensity component of the incident light field but also on the transmittance of the micro-polarizer, the analysis direction deviation, and the non-ideal extinction ratio of that polarization analysis sub-channel. We assume that the micro-polarizer transmittance and analysis direction deviation are ideal, with only the extinction ratio deviation existing.

[0018] Let the incident Stokes vector of the scene under test at a certain pixel position be... .

[0019] Wherein, S0 is the total intensity component, S1 and S2 are the linear polarization components, S1 represents the light intensity difference between 0° polarization azimuth angle and 90° polarization azimuth angle, and S2 represents the light intensity difference between 45° polarization azimuth angle and 135° polarization azimuth angle.

[0020] Assume the transmittance along the transmission axis of the polarizer is q (0 < 1), and the transmittance perpendicular to the transmission axis is r (0 < 1). e = q / r represents the extinction ratio of the polarizer. Due to the non-ideal nature of the micro-polarizer, the total transmitted light intensity should be the sum of the light intensities parallel to and perpendicular to the principal axis, i.e., q + r. For m in the Mueller matrix... 01 m 02 In the terms, S1 and S2 represent the intensity difference between the horizontal and vertical polarization azimuth angles, respectively, and the intensity difference between the 45° and 135° polarization azimuth angles. q represents the parameter of the parallel polarization azimuth angle, and r represents the parameter of the vertical polarization azimuth angle, which are (qr)cos²θ and (qr)sin²θ, respectively. Therefore, the modulation formula for the total light intensity is: .

[0021] Correspondingly, the actual output response of each polarization analysis sub-channel in a set of experimental measurement data is expressed as: .

[0022] Where, q i Let r be the equivalent transmittance of the i-th polarization analysis sub-channel in the direction parallel to the principal axis. i Let θ be the equivalent transmittance of the i-th polarization analysis sub-channel in the direction perpendicular to the principal axis. i Let n be the polarization azimuth angle of the i-th polarization analysis sub-channel. i Let be the combined noise term for the i-th polarization analysis sub-channel.

[0023] Writing all polarization analysis sub-channels in matrix form yields the following representation: .

[0024] , .

[0025] .

[0026] Where m is the number of polarization analysis sub-channels, and I is the actual output response of each polarization analysis sub-channel. Here, q and r are used to more accurately describe the mapping relationship between the incident Stokes vector and the response of each polarization analysis sub-channel, aiming to make the analysis matrix in the subsequent matrix inversion more accurate. Matrix A represents the system analysis relationship from the incident Stokes vector to the sub-channel response of the actual focal plane polarization camera, which is the basis for subsequent calibration recovery and joint optimization.

[0027] Next, we introduce the extinction ratio parameter and establish an analysis matrix degradation model caused by low extinction ratio for analysis.

[0028] The extinction ratio of the i-th polarization analysis sub-channel is defined as: , .

[0029] Further define the polarization modulation of the polarization analysis sub-channels for: .

[0030] q i +r i The total energy transmission level q of this polarization analysis subchannel is determined by... i -r i The ratio δ of the polarization analysis sub-channel's ability to resolve differences in polarization direction determines the polarization analysis sub-channel's ability to do so. i This indicates the weight of polarization information in the total output, which is essentially the normalized modulation contrast of the polarization analysis sub-channel in response to different polarization states.

[0031] Therefore, the actual output response of the i-th polarization analysis sub-channel can be rewritten as: .

[0032] Ideally, the analysis directions of the four polarization analysis sub-channels correspond to 0, 45, 90, and 135, respectively. Furthermore, when analyzing the dominant influence of low extinction ratio on system stability, the following approximation can be made: the total transmittance levels of the four polarization analysis sub-channels are assumed to be approximately the same, and their corresponding q... i +r i The parameters can be uniformly approximated by the same parameter g; it is assumed that the polarization modulation capabilities of the four sub-channels are approximately the same, and their corresponding q i -r i It can be uniformly approximated by the same parameter d. Under this approximation, the system analysis matrix can be written as: .

[0033] To analyze the invertible stability of the system's analytical matrix, further research is needed on matrix A. T A. For non-square matrix analysis matrices, A T A reflects the system's ability to distinguish between Stokes components and the degree of coupling between them. Substituting the above approximate analysis matrix, we get: .

[0034] Its diagonal elements reflect the effective response intensity of the system to the total intensity component and the two linearly polarized components, respectively. Among them, the response intensity corresponding to S0 is g. 2 The response intensities corresponding to S1 and S2 are both d. 2 / 2. Further, matrix A T The eigenvalues ​​of A are equal to the squares of the singular values ​​of the system analysis matrix. Therefore, the singular values ​​of the analysis matrix A are as follows: , .

[0035] Its condition number is: .

[0036] As can be clearly seen from the above expressions, when the extinction ratio e is high, the polarization analysis sub-channel has a strong ability to distinguish different polarization directions, the condition number of the analysis matrix is ​​small, and the system inverse solution is relatively stable. However, when the extinction ratio decreases, q and r gradually approach each other, causing d=qr to decrease, corresponding to a decrease in the effective response intensity of S1 and S2, which in turn leads to a rapid increase in the condition number of the analysis matrix. An increase in the condition number means that the matrix pseudo-inverse operation is more sensitive to noise, that is, small noise and errors in the measurement process will be significantly amplified in the Stokes parameter recovery.

[0037] As can be seen from the condition number, a decrease in the extinction ratio will lead to an increase in the condition number of the system analysis matrix, which in turn makes the pseudo-inverse recovery of the matrix more sensitive to noise.

[0038] Therefore, from a physical perspective, a low extinction ratio does not significantly weaken the system's ability to measure the total intensity component S0, but it does significantly reduce the system's ability to modulate and distinguish the linearly polarized components S1 and S2. This leads to a situation where, during matrix inversion or pseudo-inverse recovery, in order to separate the weaker polarization information from the measurement signal, the subtle differences related to polarization must be amplified more strongly, inevitably amplifying noise synchronously.

[0039] Thus far, this application has analyzed why matrix inversion calculation increases noise under low extinction ratio conditions. Based on the above analysis, this application provides a joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera. This method is executed by computer equipment, specifically by a terminal or server alone, or by both a terminal and a server.

[0040] In one exemplary embodiment, such as Figure 1 As shown, it includes the following steps 101 to 105.

[0041] Step 101: For the long-wave infrared split-plane polarization camera to be calibrated, perform extinction ratio calibration and estimate the system equivalent analysis matrix.

[0042] In a specific application, to obtain q i With r i The actual value is used to calibrate the extinction ratio using the standard linear polarization radiation input method. Step 101 includes steps (11) to (13).

[0043] (11) Input multiple sets of known linearly polarized Stokes vectors into the long-wave infrared focal plane polarization camera to be calibrated through the long-wave infrared polarization calibration optical path.

[0044] Specifically, a long-wave infrared polarization calibration optical path is constructed, which includes a blackbody, a collimator, a long-wave infrared high extinction ratio polarizer, and the camera under test. A uniformly polarized radiation field with a known polarization direction is input to the long-wave infrared focal plane polarization camera to be calibrated through this long-wave infrared polarization calibration optical path.

[0045] (12) For any given set of linearly polarized Stokes vectors, calculate the extinction ratio of each polarization analysis sub-channel using the parallel output response and the vertical output response; a set of known linearly polarized Stokes vectors and the corresponding parallel and vertical output responses of each polarization analysis sub-channel constitute a set of experimental measurement data. Here, the principle for measuring the extinction ratio is as follows: to measure the extinction ratio of the 0-degree channel, rotate the polarizer to 0 degrees, and then read the images of the 0-degree channel and the 90-degree channel to calculate the extinction ratio of each pixel.

[0046] Specifically, extinction ratios were measured for four types of polarization analysis sub-channels at 0°, 45°, 90°, and 135°. During measurement, the polarizer was rotated to a position approximately parallel to the principal axis of the corresponding polarization analysis sub-channel, and one or more frames of images were acquired. The average response value of the sub-channel under this input state was recorded as the parallel output response I. 平行 Next, rotate the polarizer to a position approximately perpendicular to the principal axis of the polarization analysis sub-channel, acquire the corresponding image, and take the average response value, which is denoted as the vertical output response I.垂直 .

[0047] To reduce the impact of random noise during the above measurement process, multiple frames of images can be continuously acquired and averaged for each polarization direction. For example, at least 10-20 frames of images should be acquired for each input direction, and the average value of pixels in the same sub-channel should be statistically analyzed to obtain a relatively stable output value under that input condition. Based on this, the number of intermediate angle groups can be further increased as auxiliary measurement data to improve the reliability of parameter estimation.

[0048] In short, to obtain the equivalent analysis matrix of the actual camera system, this embodiment uses a standard polarization state input device to input multiple sets of known linearly polarized Stokes vectors to the long-wave infrared focal plane polarization camera to be calibrated: .

[0049] Correspondingly, the system measurement output is acquired, that is, the actual output response of each polarization analysis sub-channel in the k-th set of experimental measurement data: I （k） .

[0050] (13) Based on multiple sets of experimental measurement data, the equivalent analysis matrix of the system is estimated by fitting the least squares method, as shown in the following formula: .

[0051] in, The system equivalent analysis matrix, This represents the weight of the measurement data in the k-th experimental group. This represents the known linearly polarized Stokes vector corresponding to the k-th experimental measurement data. The response is predicted using matrix A; matrix A characterizes the systematic analysis relationship from the incident linearly polarized Stokes vector to the polarization analysis subchannel response.

[0052] Step 102: Based on the system equivalent analysis matrix, determine the initial Stokes parameter recovery result and the initial recovered polarization degree image.

[0053] In a specific application, based on the system equivalent analysis matrix Through the corresponding pseudo-inverse matrix The formula for obtaining the initial Stokes parameter recovery result is as follows: .

[0054] The initial recovered polarization image is calculated using the following formula. : .

[0055] in, The Stokes vector is recovered after the matrix is ​​inverted, which is the result of recovering the initial Stokes parameters; The total intensity component S0 of the initial Stokes parameter recovery result. The linear polarization component S1 is the initial Stokes parametric recovery result. S2 represents the linear polarization component of the initial Stokes parametric recovery result, ε is a preset small positive number introduced to prevent the denominator from being zero. I represents the actual output response corresponding to each polarization analysis sub-channel in a set of experimental measurement data.

[0056] This step achieves polarization calibration recovery based on the system analysis matrix. However, due to the ill-conditioned nature of the analysis matrix caused by the low extinction ratio, the initial recovery results usually still contain obvious noise amplification effects, which are more prominent in DoLP images.

[0057] Step 103: Using the initial restored polarization degree image as the initial value, the total intensity component of S0 in the initial Stokes parameter restoration result is used to guide DOLP, and a polarization physical constraint joint restoration model is constructed.

[0058] Considering the total intensity component in the initial recovery result Typically, it has a relatively high signal-to-noise ratio and can stably reflect the geometric edges, thermal radiation distribution, and structural contours of a scene. This application utilizes... As guiding information, structural constraints are imposed on the DoLP restoration process to suppress the propagation of inverse demodulation noise in DoLP images under low extinction ratio conditions.

[0059] The polarization physical constraint joint recovery model is as follows: .

[0060] Among them, D The restored DoLP image represents the joint restoration result; D represents the DoLP image to be restored. The initial polarization degree image is restored; p and q are the pixel positions in the DoLP image; p is the current pixel, q is a pixel in the neighborhood of p, and N(p) is the neighborhood set of pixel p; Based on Constructed guiding weights, The total intensity component S0 of the initial Stokes parametric recovery result, (D p -D q ) 2 For the DoLP difference between pixel p and pixel q, D p D is the DoLP value at pixel position p. qλ is the DoLP value at pixel q; λ1 and λ2 are weight parameters; Ψ(D) is the polarization physical feasible region constraint function.

[0061] To be more specific, This is a data fidelity item, the purpose of which is to make the optimization result approximate the original recovery result.

[0062] The purpose of the structural domain constraint term in S0 is to determine whether pixel p and pixel q belong to the same structural region in S0.

[0063] based on The constructed guiding weights are expressed as: .

[0064] Where, σ s σ is the intensity adjustment parameter. r This is a parameter for adjusting spatial distance; The total intensity component S0 of pixel p in the initial Stokes parametric recovery result. S0 is the total intensity component of pixel q in the initial Stokes parametric recovery result.

[0065] The intensity difference term represents the difference between two pixels on S0, used to determine whether two pixels belong to the same region.

[0066] The spatial distance term represents the distance between two pixels in space; both factors determine this distance. The magnitude of the weights.

[0067] λ1 controls the intensity of S0-guided smoothing; a large λ1 emphasizes denoising and smoothing, while a small λ1 emphasizes preserving the original recovery result. For polarization physical constraint terms: λ2 controls the strength of the physical constraint. A large λ2 strictly limits the result to a reasonable range; a small λ2 has a weaker effect on the physical constraint.

[0068] Based on the above weights, when When the change is gradual within the neighborhood, it indicates that the region is homogeneous and has a larger weight, which can enhance DoLP's smoothing and noise reduction; when When there are obvious edge transitions, the weights are reduced to suppress cross-edge smoothing and preserve the target boundary and structural details.

[0069] Step 104: Construct polarization physical consistency constraints and select recovery requirements; the recovery requirements are either direct recovery in the DoLP domain or recovery not directly in the DoLP domain.

[0070] Since the polarization parameters in linear polarization imaging must satisfy the basic physical constraints, the polarization physical consistency constraint is set as: S0≥0, S1 2 +S2 2 ≤S0 2 0 ≤ DOLP ≤ 1. The above physical feasible region constraint is introduced into the joint recovery model to avoid non-physical values ​​in the recovery results, thereby improving the reliability and stability of the recovery results.

[0071] This application can perform recovery directly in the DoLP domain, or jointly optimize S1 and S2 in the Stokes domain. Through the combined effect of measurement consistency constraints, structural regularization terms, and physical consistency constraints, it obtains the optimal recovery result that better matches the original measurement data, better matches the prior knowledge of the scene structure, and satisfies the laws of polarization physics.

[0072] Since S1 and S2 represent the polarization discrimination capability, direct recovery in the DoLP domain can be understood as directly generating the final DoLP image and constraining its pixels to keep the DoLP values ​​within a physically reasonable range. However, this method only applies to the final DoLP and may not fully utilize the Stokes component information of each pixel, potentially losing intermediate detail information in the S1 and S2 components.

[0073] The approach of not directly restoring in the DoLP domain involves first restoring the S1 and S2 values ​​of each pixel in the Stokes component domain, and then applying physical constraints and regularization terms to S1 and S2. After restoration, the DoLP is calculated from the Stokes components. This method allows the restored DoLP to more accurately reflect the polarization characteristics of the target region while maintaining physical plausibility, making it more reliable and convincing.

[0074] Step 105: Based on the recovery requirements, the polarization physical consistency constraint is used to solve the joint recovery model of the polarization physical constraint to obtain the joint recovery result.

[0075] When the recovery requirement is to be directly recovered in the DoLP domain, the polarization physical feasible region constraint function is defined as: .

[0076] To ensure that each pixel is within a reasonable range, the result is brought back to a reasonable range. This formula is expressed as checking each pixel individually.

[0077] To further improve measurement consistency and physical rigor, specifically when the recovery requirement is not directly in the DoLP domain, the recovery is performed in the Stokes component domain. and Joint optimization is performed, where the variables to be optimized are S1 and S2, and these variables are fixed. As a guiding component, the joint recovery model of polarization physical constraints is rewritten as follows: .

[0078] This is used to ensure that the S1 and S2 substitution models are as consistent as possible with I.

[0079] , indicating by The guided regularization functions S1 and S2 sum the squared differences between adjacent pixels in the image according to their weights, serving as a quantitative indicator of the degree of unsmoothness. 1,p S 1,q Let S be the values ​​of the linear polarization component S1 to be optimized at pixels p and q. 2,p S 2,q Let S be the values ​​of the linear polarization component S2 to be optimized at pixels p and q.

[0080] This represents the difference between S1 and S2 between the p-th pixel and the q-th pixel. Essentially, it measures the difference between two adjacent pixels in the polarization component space.

[0081] , represents the Stokes physical consistency constraint, checking whether the p-th pixel violates S1. 2 +S2 2 ≤S0 2 This physical law. S0 is the total intensity component of pixel p in the initial Stokes parametric recovery result.

[0082] and S1 and S2 are the optimized recovery results that minimize the objective function under the combined effect of measurement consistency term, structural regularization term, and physical constraint term.

[0083] In obtaining and Then, the recovered DoLP image is calculated using the following formula: .

[0084] Compared to direct DoLP domain recovery, this approach better unifies measurement consistency, structure guidance, and polarization physics constraints into a single recovery framework.

[0085] Through the joint optimization of the above steps, the final recovered polarization parameter image is obtained. Specifically, multiple polarization parameter images can be generated according to the code, or only the final DOLP image can be generated directly. If the DOLP domain is selected, the S0 image, the inverted image, and the final recovered image can be obtained, or only the final DOLP image can be generated; if the S1 and S2 domains are selected, the corrected S1 and S2 images, as well as the final DOLP image, can be generated.

[0086] In summary, this application utilizes the high signal-to-noise ratio information of S0 to constrain and suppress random noise generated in DoLP due to the large inverse extinction caused by low extinction ratio, without destroying the true structural edges. While preserving the system's polarization calibration capability, this application achieves effective suppression of matrix pseudo-inverse recovery noise under low extinction ratio conditions, significantly improving random noise, pseudo-textures, and edge fragmentation problems in DoLP images, thereby obtaining recovery results that combine polarization accuracy, physical consistency, and image structure preservation.

[0087] In practical applications, the average gradient at the edges can also be used to evaluate the clarity of the outline, the degree of boundary enhancement, and the ability to preserve the structure of the restored polarized image in the edge region.

[0088] (1) To evaluate the edge enhancement effect of the polarization images before and after restoration, the average gradient value of the polarization image is calculated within the edge region extracted from the intensity map. Let the edge region be Ωe, then the average gradient G of the edge region is defined as: .

[0089] Where, N e Let be the total number of pixels in the edge region, and Dp be the total number of pixels in the edge region. Let be the gradient at the p-th pixel.

[0090] This index is used to characterize the intensity of structural changes in the target edge region of a polarized image. The larger the value of G, the clearer the restored edge contour and the more obvious the structure preservation and boundary enhancement effects.

[0091] (2) To evaluate the distinguishability of the target region relative to the background region in the restored polarized image, its contrast C is calculated using the following formula: .

[0092] in, The average value of the target area. This is the average value of the background area. The standard deviation is the background area.

[0093] This metric is used to evaluate the distinguishability and detectability of the target region relative to the background region in the restored polarization image. The restored polarization image should be superior to the unrestored polarization image.

[0094] (3) To evaluate its physical rationality, the percentage of pixels that do not satisfy 0≤DoLP≤1 is statistically analyzed and defined as the DoLP out-of-bounds rate R: .

[0095] Where N (DoLP<0, DoLP>1) represents the number of out-of-bounds pixels, and N represents the total number of pixels included in the statistics. The closer this indicator is to 0, the more the restoration result conforms to the polarization physical range constraint, and the better the physical rationality.

[0096] After evaluating the above three parameters, if the recovery result does not meet the expected requirements, the parameters of the joint recovery stage can be adjusted according to different evaluation indicators. Only the corresponding parameters in the polarization physical constraint joint recovery model need to be adjusted, without having to repeat the preceding system modeling and calibration steps.

[0097] In practical applications, when the average gradient index of the edge region is low, indicating that the restored polarization image has blurred edges, excessively smoothed structure, or insufficient contour enhancement, it is preferable to reduce the weight parameter of the structural smoothing constraint term or appropriately increase the number of iterations to reduce the suppression of target edge details by excessive smoothing and enhance the ability of the restoration process to preserve structural changes.

[0098] When the contrast index of the target area relative to the background area is low, it indicates that the polarization difference between the target and the background is not obvious after restoration and the target is not distinguishable enough. In this case, it is preferable to appropriately reduce the smoothing constraint intensity and appropriately increase the weight of the physical constraint term or the guiding constraint term to enhance the restoration ability of the polarization characteristics of the target area and improve the target-background difference.

[0099] When the DoLP out-of-bounds rate or physical constraint violation rate is too high, it indicates that there are non-physical polarization values ​​or Stokes components in the recovery result that do not meet the physical feasible region constraints. In this case, it is preferable to increase the weight parameter of the physical constraint term and appropriately increase the number of iterations to enhance the convergence of the recovery result to the physical reasonable range.

[0100] Based on the feedback results of different evaluation indicators, this application can adaptively adjust the parameters of the joint recovery stage and re-execute the recovery iteration until a polarization recovery result that takes into account edge preservation, target distinguishability and physical consistency is obtained.

[0101] Based on the same inventive concept, this application also provides a system. The solution provided by this system is similar to the solution described in the above method. Therefore, the specific limitations of one or more system embodiments provided below can be found in the limitations of the method above, and will not be repeated here.

[0102] In one exemplary embodiment, a joint recovery system for calibration of a long-wave infrared focal plane polarization camera is provided, comprising the following multiple modules.

[0103] The calibration module is used to calibrate the extinction ratio and estimate the system equivalent analysis matrix for a long-wave infrared focal plane polarization camera to be calibrated.

[0104] The initial recovery module is used to determine the initial Stokes parameter recovery result and the initial recovered polarization degree image based on the system equivalent analysis matrix.

[0105] The polarization physical constraint joint restoration model construction module is used to construct a polarization physical constraint joint restoration model by using the initially restored polarization degree image as the initial value and guiding the DOLP with the S0 total intensity component in the initial Stokes parametric restoration result. In this module, the polarization physical constraint joint restoration model guided by S0 is used to process the initially inversely restored polarization image.

[0106] The constraint and requirement determination module is used to construct polarization physical consistency constraints and select recovery requirements; the recovery requirements are either directly recovered in the DoLP domain or not directly recovered in the DoLP domain.

[0107] The joint recovery module is used to solve the joint recovery model of polarization physical constraints according to the recovery requirements and the polarization physical consistency constraints to obtain the joint recovery result.

[0108] In summary, this application sequentially performed the following processes: establishing a non-ideal polarization measurement model for a long-wave infrared focal plane polarization camera; analyzing extinction ratio degradation; calibrating and initial restoration; constructing a joint restoration model based on S0-guided polarization physical constraints; outputting result images; and comparing DOLP images before and after correction. Compared with existing technologies, this application has the following advantages: (1) To address the common problem of noise amplification caused by low extinction ratio inversion in long-wave infrared focal plane polarization cameras, a joint correction mechanism based on complementary intensity and polarization features is proposed: using the correction result of the extinction ratio inversion matrix as the initial value, a high signal-to-noise ratio intensity image S0 is introduced to guide the DOLP, which can effectively suppress a large amount of noise caused by matrix inversion while restoring polarization accuracy. This process does not depend on ideal high extinction ratio device conditions, nor does it require significant modification to the existing hardware system; it can be achieved simply by jointly optimizing the calibration and recovery process.

[0109] (2) Physical consistency constraints of polarization parameters were introduced during the restoration process to ensure that the restoration results meet the physical laws of polarization images, thereby avoiding non-physical results that may occur during conventional numerical optimization.

[0110] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 2 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and databases. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera.

[0111] Those skilled in the art will understand that Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0112] In one exemplary embodiment, a computer device is also provided, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the above-described method embodiments.

[0113] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0114] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0115] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0116] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0117] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0118] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0119] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera, characterized in that, The method includes: For a long-wave infrared focal plane polarization camera to be calibrated, the extinction ratio is calibrated and the system equivalent analysis matrix is ​​estimated. Based on the system equivalent analysis matrix, the initial Stokes parameter recovery result and the initial recovered polarization degree image are determined. Using the initial restored polarization degree image as the initial value, the total intensity component S0 in the initial Stokes parametric restoration result is used to guide DOLP, and a polarization physical constraint joint restoration model is constructed. Construct polarization physical consistency constraints and select recovery requirements; the recovery requirements are either direct recovery in the DoLP domain or recovery not directly in the DoLP domain. Based on the recovery requirements, the polarization physical consistency constraint is used to solve the joint recovery model of the polarization physical constraint, and the joint recovery result is obtained.

2. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 1, characterized in that, For the long-wave infrared focal plane polarization camera to be calibrated, the extinction ratio is calibrated and the system equivalent analysis matrix is ​​estimated, including: By using the long-wave infrared polarization calibration optical path, multiple sets of known linear polarization Stokes vectors are input to the long-wave infrared focal plane polarization camera to be calibrated. For any given set of linearly polarized Stokes vectors, the extinction ratio of each polarization analysis sub-channel is calculated using the parallel output response and the vertical output response. A set of known linearly polarized Stokes vectors and the corresponding parallel and vertical output responses of each polarization analysis sub-channel constitute a set of experimental measurement data. Based on multiple sets of experimental measurement data, the equivalent analysis matrix of the system is estimated using the least squares method, as shown in the following formula: ； in, The system equivalent analysis matrix, This represents the weight of the measurement data in the k-th experimental group. This represents the actual output response of each polarization analysis sub-channel in the k-th set of experimental measurement data. This represents the known linearly polarized Stokes vector corresponding to the k-th experimental measurement data. The response is predicted using matrix A; matrix A represents the systematic analysis relationship from the incident linearly polarized Stokes vector to the polarization analysis subchannel response, and K is the number of sets of experimental measurement data.

3. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 1, characterized in that, Based on the system equivalent analysis matrix, the initial Stokes parameter recovery result and the initial recovered polarization degree image are determined, including: Based on the system equivalent analysis matrix Through the corresponding pseudo-inverse matrix The formula for obtaining the initial Stokes parameter recovery result is as follows: ； The initial recovered polarization image is calculated using the following formula. : ； in, The Stokes vector is recovered after the matrix is ​​inverted, which is the result of recovering the initial Stokes parameters; The total intensity component S0 of the initial Stokes parameter recovery result. The linear polarization component S1 is the result of the initial Stokes parametric recovery. S2 represents the linear polarization component of the initial Stokes parameter recovery result, ε is a preset small positive number, and I is the actual output response corresponding to each polarization analysis sub-channel in a set of experimental measurement data.

4. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 1, characterized in that, The polarization physical constraint joint recovery model is as follows: ； Among them, D The restored DoLP image represents the joint restoration result; D represents the DoLP image to be restored. The initial polarization degree image is restored; p and q are the pixel positions in the DoLP image; p is the current pixel, q is a pixel in the neighborhood of p, and N(p) is the neighborhood set of pixel p; Based on Constructed guiding weights, The total intensity component S0 of the initial Stokes parametric recovery result, (D p -D q ) 2 For the DoLP difference between pixel p and pixel q, D p D is the DoLP value at pixel position p. q λ is the DoLP value at pixel q; λ1 and λ2 are weight parameters; Ψ(D) is the polarization physical feasible region constraint function.

5. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 4, characterized in that, based on The constructed guiding weights are expressed as: ； Where, σ s σ is the intensity adjustment parameter. r This is a parameter for adjusting spatial distance; The total intensity component S0 of pixel p in the initial Stokes parametric recovery result. S0 is the total intensity component of pixel q in the initial Stokes parametric recovery result.

6. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 1, characterized in that, The polarization physical consistency constraint is: S0≥0, S1 2 +S2 2 ≤S0 2 , 0≤DOLP≤1; Where S0 is the total intensity component, S1 and S2 are the linear polarization components, and DOLP is the degree of polarization.

7. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 4, characterized in that, When the recovery requirement is to be directly recovered in the DoLP domain, the polarization physical feasible region constraint function is defined as: 。 8. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 4, characterized in that, The recovery requirement refers to the recovery process within the Stokes component domain when recovery is not performed directly in the DoLP domain. and Joint optimization is performed, and the joint recovery model of polarization physical constraints is rewritten as follows: ； Where S1 and S2 are linear polarization components, and For the optimized S1 and S2; Let I be the system equivalent analysis matrix, and let I be the actual output response of each polarization analysis subchannel in a set of experimental measurement data. , indicating by Guided regularization functions for S1 and S2; S 1,p S 1,q Let S be the values ​​of the linear polarization component S1 to be optimized at pixels p and q. 2,p S 2,q Let S be the values ​​of the linear polarization component S2 to be optimized at pixels p and q. , represents the Stokes physical consistency constraint; The total intensity component S0 of pixel p in the initial Stokes parametric recovery result; In obtaining and Then, the recovered DoLP image is calculated using the following formula: ； Where ε is a preset small positive number.

9. The joint recovery method for calibration of a low extinction ratio long-wave infrared focal plane polarization camera according to claim 2, characterized in that, The actual output response of each polarization analysis sub-channel in a set of experimental measurement data is expressed as follows: ； Where, q i Let r be the equivalent transmittance of the i-th polarization analysis sub-channel in the direction parallel to the principal axis. i Let θ be the equivalent transmittance of the i-th polarization analysis sub-channel in the direction perpendicular to the principal axis. i Let n be the polarization azimuth angle of the i-th polarization analysis sub-channel. i S0 represents the total noise term of the i-th polarization analysis sub-channel; S1 and S2 represent the total intensity component and the linear polarization components. Writing all polarization analysis sub-channels in matrix form yields the following representation: ； ， ； ； Where m is the number of polarization analysis sub-channels.

Images