Infrared focal plane array non-uniformity correction method based on integral time adjustment

Abstract

The invention relates to optical plane correction, and discloses an infrared focal plane array non-uniformity correction method based on integral time adjustment, which comprises the following steps: for each pixel of an infrared focal plane array, respectively establishing a nonlinear response model representing an incidence relation among integral time, radiation flux and detector output response; according to the radiation distribution characteristics of the current scene and the independent response characteristics of each pixel, the optimal integral time is determined for each pixel through an adaptive optimization algorithm; dynamically updating the parameters of the nonlinear response model of each pixel based on the deviation between the actual output of the detector and the output of the nonlinear response model; and compensating the response nonlinear difference between different pixels based on the optimal integral time, and performing further linear compensation on the compensated signal based on a two-point correction formula. With the adoption of the method, the defects of a traditional correction method in the aspects of nonlinear correction, scene adaptation, drift adaptation and the like are overcome, and the imaging quality of the infrared focal plane array is effectively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to optical plane correction, in particular to an infrared focal plane array non-uniformity correction method based on integral time adjustment. BACKGROUND

[0002] As the core component of infrared imaging system, the detection performance of infrared focal plane array (IRFPA) directly determines the imaging quality. However, due to factors such as semiconductor material preparation process, detector unit response difference and environmental temperature drift, the same radiation input will produce different outputs for each pixel of the IRFPA, i.e. non-uniformity. This non-uniformity will cause problems such as fixed pattern noise and stripe interference in imaging, seriously affecting the accuracy of target recognition and scene analysis, especially in high-precision applications such as military reconnaissance, industrial detection and medical diagnosis.

[0003] Existing non-uniformity correction techniques mainly fall into three categories:

[0004] Traditional calibration correction methods (such as two-point correction and multi-point correction): fixed parameters (gain, bias) are obtained through laboratory calibration to linearly compensate the pixel output. However, this method can only correct linear response differences and cannot handle non-linear responses (such as signal distortion in the saturation region) of the detector under high radiation flux or long integration time. Moreover, the fixed calibration parameters cannot adapt to the response drift of the detector (such as dark current change and material aging) over a long period of operation.

[0005] Scene adaptive correction methods (such as time-domain filtering based on statistics and neural network correction): without pre-calibration, the correction parameters are adjusted in real time by analyzing the statistical characteristics of the scene image. However, this method relies on scene changes (such as target motion) and is prone to failure in static scenes. Moreover, it has high computational complexity and cannot meet the real-time requirements of high-frame-rate imaging systems.

[0006] Integral time adjustment method: by adjusting the integration time globally or in segments (such as using the same integration time for the entire array or dividing the fixed integration time by regions), the non-linear response interval of the detector is attempted to be avoided. However, existing methods do not consider the individual response differences of the pixels, and the same integration time cannot adapt to the non-linear characteristics of all pixels (such as some pixels being saturated while others being underexposed), resulting in limited correction accuracy.

[0007] Therefore, how to simultaneously correct the linear and non-linear response differences of the pixels while considering real-time performance, and adapt to the response drift of the detector, has become a key technical problem for improving the imaging quality of IRFPA. SUMMARY

[0008] The present application aims to provide an infrared focal plane array non-uniformity correction method based on integral time adjustment to solve the technical problems raised in the background.

[0009] To achieve the above object, the application discloses the following technical solutions: An infrared focal plane array non-uniformity correction method based on integration time adjustment, the method comprising the following steps:

[0010] For each pixel of the infrared focal plane array, based on multiple sets of calibration data, a nonlinear response model is established for each pixel to represent the correlation between integration time, radiant flux and detector output response;

[0011] According to the radiation distribution characteristics of the current scene and the independent response characteristics of each pixel, an optimal integration time is determined for each pixel by an adaptive optimization algorithm, wherein the optimization goal is to minimize the non-uniformity of the corrected image;

[0012] Based on the deviation between the actual output of the detector and the output of the nonlinear response model, the parameters of the nonlinear response model of each pixel are dynamically updated;

[0013] Based on the optimal integration time, the response nonlinearity difference between different pixels is compensated, and the compensated signal is further linearly compensated based on a two-point correction formula.

[0014] As a preferred, the adaptive optimization algorithm for determining the optimal integration time for each pixel comprises the following steps:

[0015] For each pixel, based on the effective response interval of its nonlinear response model, an initial value range of integration time is set, wherein the lower limit of the initial value range is not lower than the minimum response time of the infrared focal plane array detector, and the upper limit is not more than the maximum tolerance time of the detector to avoid signal saturation;

[0016] Select a particle swarm optimization algorithm or a genetic algorithm as the adaptive optimization algorithm, and independently configure algorithm parameters for each pixel, the algorithm parameters including initial population size and iteration termination condition;

[0017] For each pixel, based on the radiation distribution characteristics of the current scene, the integration time candidate value is iteratively screened within the initial value range by the adaptive optimization algorithm: and after each iteration, the pixel output non-uniformity corresponding to the candidate value is calculated, the candidate value that reduces the non-uniformity is retained, and the candidate value that increases the non-uniformity is eliminated;

[0018] When the algorithm meets the iteration termination condition, the value corresponding to the minimum non-uniformity in the currently retained integration time candidate value is determined as the optimal integration time of the pixel.

[0019] As a preferred, the effective response interval based on the nonlinear response model of each pixel comprises:

[0020] extracting an integral time range in which the output signal in the nonlinear response model of the pixel increases approximately linearly with the integral time, taking the lower limit of the range as the lower limit of the initial integral time value range;

[0021] extracting a maximum integral time in which the output signal in the nonlinear response model of the pixel does not reach a saturation threshold, taking the time value as the upper limit of the initial integral time value range.

[0022] Preferably, the nonlinear response model comprises a first-order correlation term of integral time and radiant flux, a second-order cross-correlation term of integral time and radiant flux, and a basic offset term; the first-order correlation term is used to represent the influence of the linear relationship between integral time and radiant flux on the pixel output, the second-order cross-correlation term is used to represent the influence of the nonlinear coupling relationship between integral time and radiant flux on the pixel output, and the basic offset term is used to represent the inherent output offset of the pixel without radiation input.

[0023] Preferably, the expression of the nonlinear response model is:

[0024]

[0025] wherein Y p is the actual output gray value of the pth pixel in the infrared focal plane array; t int is the integral time configured for the pth pixel in the infrared focal plane array; F p is the radiant flux actually received by the pth pixel in the infrared focal plane array; a p is the first-order cross-term coefficient of the pth pixel in the infrared focal plane array; b p is the inherent offset parameter of the pth pixel in the infrared focal plane array; c p is the second-order cross-term coefficient of the pth pixel in the infrared focal plane array.

[0026] Preferably, the nonlinear response model is obtained by fitting a plurality of sets of calibration data by the least square method, and in the parameter fitting process, based on the least square method, the deviation between the pixel output gray value predicted by the model and the pixel output gray value actually measured in the calibration experiment is minimized, the nonlinear response model parameters corresponding to each pixel are finally determined, and the construction of the single-pixel nonlinear response model is completed.

[0027] Preferably, the calibration data comprises a calibration radiation source, an integral time point, and a radiant flux level; wherein:

[0028] The calibration radiation source comprises a high-precision black body.

[0029] The integral time point covers the entire integral time range of the conventional work of the infrared focal plane array detector.

[0030] The radiation flux level covers low, medium and high radiation intervals of an effective dynamic range of the infrared focal plane array detector.

[0031] As preferred, the deviation of the actual output of the detector from the output of the nonlinear response model is used to dynamically update the parameters of the nonlinear response model for each pixel, including the following steps:

[0032] The actual output value of the detector for each pixel in the corrected image is extracted pixel by pixel as a collection period of fixed image frames, and the predicted output value of the pixel under the current scene radiation condition is calculated by calling the nonlinear response model; the difference between the actual output value of the detector and the predicted output value of the nonlinear response model is taken as the single-pixel residual value, and the single-pixel residual values of continuous multiple frames are smoothed to obtain the effective residual value reflecting the real response deviation of the pixel;

[0033] Based on the inherent noise level of the infrared focal plane array detector, a residual judgment standard is set; for each pixel, its effective residual value is continuously monitored, and if the effective residual value of the pixel continuously exceeds the residual judgment standard, the parameter updating process is triggered;

[0034] For the pixel triggering the update, the parameters of the nonlinear response model are adjusted according to the positive and negative of the effective residual value and the deviation size, wherein if the effective residual value is positive, the parameters representing the linear correlation between integration time and radiation flux, the parameters representing the nonlinear correlation and the basic offset parameter in the model are respectively adjusted by a small amount, and the adjustment amplitude is adapted according to the residual deviation size; if the effective residual value is negative, the parameters representing the linear correlation between integration time and radiation flux, the parameters representing the nonlinear correlation and the basic offset parameter in the model are respectively adjusted by a small amount.

[0035] As preferred, the deviation of the actual output of the detector from the output of the nonlinear response model is used to dynamically update the parameters of the nonlinear response model for each pixel, including the following steps:

[0036] After the parameter adjustment is completed, the predicted output value of the pixel is calculated by the nonlinear response model, and the new residual value of the new predicted output value and the actual output value of the detector is compared, if the new residual value meets the residual judgment standard and the output value of the pixel is in the effective response interval of the detector, it is confirmed that the parameter update is effective; if the new residual value does not meet the residual judgment standard or the output exceeds the effective interval, the model parameters before adjustment are returned, the adjustment amplitude is reduced and the parameters of the nonlinear response model are adjusted again until the new residual value meets the residual judgment standard and the output value of the pixel is in the effective response interval of the detector.

[0037] As preferred, the optimal integration time based compensation compensates for the response nonlinearity difference between different pixels, and the compensated signal is further linearly compensated based on a two-point correction formula, including the following steps:

[0038] For each pixel of the infrared focal plane array, the corresponding optimal integration time is called respectively, and the pixel output signal is adjusted through the adaptation of the optimal integration time and the pixel response characteristic;

[0039] The correction gain and bias are calculated by setting high and low reference radiation points, and the output signal of all the adjusted pixels is linearly adjusted.

[0040] Beneficial effects: the infrared focal plane array non-uniformity correction method based on integration time adjustment of the present application constructs a non-linear response model for each pixel, accurately characterizes the linear and non-linear coupling relationship between integration time and radiation flux, effectively solves the non-linear distortion under high radiation / long integration time, improves the non-linear response correction accuracy and reduces the image non-uniformity; through the adaptive optimization algorithm, the optimal integration time is determined for each pixel independently, so that the pixel output avoids the non-linear response interval of itself, realizes the pixel-level dynamic adaptation, and covers the radiation scene of the full dynamic range of the detector; through the dynamic updating of the model parameters based on the residual error, the parameters are adjusted in real time to adapt to the response drift of the detector, without frequent laboratory calibration, and the long-term stability of the system is improved; the optimal integration time non-linear compensation and two-point correction linear compensation are adopted, the non-linear difference is eliminated first, and then the linear residual error is refined, and the algorithm design complexity is low, and the correction accuracy and high frame frequency imaging demand are considered. Therefore, the present application solves the defects of the prior art in non-linear correction, scene adaptation, drift adaptation and the like, and improves the imaging quality of the IRFPA. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 The flow chart of the infrared focal plane array non-uniformity correction method based on integration time adjustment provided by the embodiments of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0044] In this document, the term "comprising" is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not explicitly listed, or further includes elements inherent to such a process, method, article or device. Without more limitations, the elements defined by the statement "comprising" do not exclude the presence of additional identical elements in the process, method, article or device including the elements.

[0045] The embodiments of the present application disclose an infrared focal plane array non-uniformity correction method based on integral time adjustment as shown in the following. Figure 1 The method comprises the following steps:

[0046] S1, for each pixel of the infrared focal plane array, a nonlinear response model representing the correlation between the integral time, the radiant flux and the detector output response is respectively established based on a plurality of sets of calibration data;

[0047] S2, the optimal integral time is determined for each pixel by an adaptive optimization algorithm according to the radiation distribution characteristics of the current scene and the independent response characteristics of each pixel, wherein the optimization target is to minimize the non-uniformity of the corrected image;

[0048] S3, the parameters of the nonlinear response model of each pixel are dynamically updated based on the deviation between the actual output of the detector and the output of the nonlinear response model;

[0049] S4, the response nonlinear difference between different pixels is compensated based on the optimal integral time, and the compensated signal is further linearly compensated based on a two-point correction formula.

[0050] According to the above, the infrared focal plane array non-uniformity correction method based on integral time adjustment of the embodiments of the present application solves the problem that the traditional linear correction cannot handle nonlinear distortion by establishing a nonlinear response model representing the correlation between the integral time, the radiant flux and the detector output response for each pixel; the optimal integral time is determined for each pixel by an adaptive optimization algorithm to adapt to the independent response characteristics of the pixel, avoiding the defect that the global integral time cannot take into account the pixel difference; the model parameters are dynamically updated to adapt to the change of the detector response deviation; the nonlinear difference is compensated by the optimal integral time, and the linear compensation is compensated by the two-point correction, taking into account the nonlinear and linear correction requirements, reducing the image non-uniformity as a whole, and improving the overall correction comprehensiveness and adaptability.

[0051] In the embodiment, the nonlinear response model comprises a first-order correlation term of the integration time and the radiant flux, a second-order cross-correlation term of the integration time and the radiant flux, and a basic offset term; the first-order correlation term is used to represent the influence of the linear relationship between the integration time and the radiant flux on the pixel output, the second-order cross-correlation term is used to represent the influence of the nonlinear coupling relationship between the integration time and the radiant flux on the pixel output, and the basic offset term is used to represent the inherent output offset of the pixel without radiation input.

[0052] Through the above, the linear relationship, the nonlinear coupling relationship and the inherent offset of the pixel are represented by the first-order correlation term, the second-order cross-correlation term and the basic offset term respectively, and the different response characteristics of the pixel are processed in a targeted manner; compared with the traditional single model, the linear response, the nonlinear coupling response under high radiation / long integration time and the inherent offset without radiation can be comprehensively captured, and the representation integrity of the model for the pixel response is improved.

[0053] Specifically, the expression of the nonlinear response model is:

[0054]

[0055] wherein Y p is the actual output gray value of the pth pixel in the infrared focal plane array, the numerical range of which matches the analog-to-digital conversion accuracy of the detector (such as 8-bit detector corresponding to 0-255); t int is the integration time configured for the pth pixel in the infrared focal plane array, i.e. the duration of the light signal collection of the detector for the pixel; F p is the radiant flux actually received by the pth pixel in the infrared focal plane array, i.e. the radiant energy per unit time through the photosensitive surface of the pixel; a p is the first-order cross-term coefficient of the pth pixel in the infrared focal plane array, used to quantify the influence degree of the linear correlation between the integration time and the radiant flux on the output gray value of the pixel; b p is the inherent offset parameter of the pth pixel in the infrared focal plane array, the value of which is equal to the output gray value of the pixel when there is no radiant flux input F p =0 and the integration time is the reference value, and is used to compensate the dark current offset of the pixel; c p is the second-order cross-term coefficient of the pth pixel in the infrared focal plane array, used to quantify the nonlinear influence intensity of the nonlinear coupling between the integration time and the radiant flux (such as high radiant flux or long integration time scene) on the output gray value of the pixel.

[0056] By the above, the expression of the nonlinear response model and the definition of each parameter (such as matching the output gray value with the analog-to-digital conversion accuracy, the integral time with the acquisition time length, and the radiation flux with the radiation energy per unit time) are clear, so that the model is quantifiable and the parameters have clear physical meanings; the values of the parameters are adapted to the hardware characteristics of the detector (such as the analog-to-digital conversion accuracy), and the model is reproducible and has high engineering practicability.

[0057] Secondly, the nonlinear response model is obtained by fitting a plurality of sets of calibration data by the least square method, and in the parameter fitting process, based on the least square method, the deviation between the pixel output gray value predicted by the model and the pixel output gray value actually measured in the calibration experiment is minimized, and the nonlinear response model parameters corresponding to each pixel are finally determined, and the construction of the single-pixel nonlinear response model is completed.

[0058] By the above, the calibration data is fitted by the least square method, the deviation between the predicted output of the model and the actually measured output is minimized, and the fitting accuracy of the nonlinear response model is ensured; compared with other fitting methods, the least square method has high stability and reliable calculation, and can make the model closer to the real response characteristics of the pixel and reduce the correction deviation caused by the fitting error.

[0059] Feasibly, the calibration data includes a calibration radiation source, integral time points, and radiation flux levels; wherein:

[0060] The calibration radiation source includes a high-precision black body, which needs to have the ability to stably output different radiation intensities, and the control accuracy of the radiation temperature thereof needs to reach the high-precision level required by the infrared detector calibration experiment, so as to ensure the stability and consistency of the radiation flux output during the calibration process and avoid the influence of the radiation source fluctuation on the accuracy of the calibration data;

[0061] The integral time points cover the entire integral time interval of the conventional work of the infrared focal plane array detector, and the distribution density of the integral time points needs to meet the complete sampling requirement of the response characteristics of the detector under different integral time lengths, that is, the law of the response characteristics of the detector with the integral time in the range from the shortest conventional integral time to the longest conventional integral time can be completely captured, so as to ensure the adaptability of the model to the full integral time working scene of the detector;

[0062] The radiation flux levels cover the low, medium and high radiation intervals of the effective dynamic range of the infrared focal plane array detector, and the number of the divided radiation flux levels needs to meet the overall representation requirement of the response characteristics of the detector under different radiation receiving intensities; the number of the image frames collected under each radiation flux level needs to effectively suppress the interference of the image noise on the calibration data through the statistical processing of the multiple frames of data, so as to ensure that the signal-to-noise ratio of the calibration data meets the model fitting accuracy requirement.

[0063] By the above, the high-precision black body is used as the calibration radiation source (ensuring stable and consistent radiation), full conventional integration time points are covered (fully capturing the response law), and full radiation levels are covered (fully characterizing the response under different radiation), so as to ensure accurate and comprehensive calibration data, avoid calibration data errors caused by radiation source fluctuations, incomplete integration time / radiation level sampling, and ensure that the model constructed based on the data can adapt to the full working scene of the detector and improve the scene adaptability of the model.

[0064] In the embodiment, the adaptive optimization algorithm is used to determine the optimal integration time for each pixel, including the following steps:

[0065] For each pixel, an initial value range of the integration time is set based on the effective response interval of the nonlinear response model of the pixel, wherein the lower limit of the initial value range is not lower than the minimum response time of the infrared focal plane array detector, and the upper limit is not higher than the maximum tolerance time of the detector to avoid signal saturation;

[0066] The particle swarm optimization algorithm or the genetic algorithm is selected as the adaptive optimization algorithm, and algorithm parameters including the initial population size and the iteration termination condition are independently configured for each pixel;

[0067] For each pixel, based on the radiation distribution characteristics of the current scene, the adaptive optimization algorithm is used to iteratively screen the integration time candidate values in the initial value range: and after each iteration, the non-uniformity of the pixel output corresponding to the candidate value is calculated, the candidate value that reduces the non-uniformity is retained, and the candidate value that increases the non-uniformity is eliminated;

[0068] When the algorithm meets the iteration termination condition, the value corresponding to the minimum non-uniformity in the currently retained integration time candidate values is determined as the optimal integration time of the pixel.

[0069] By the above, the determination process of the optimal integration time is clear, the acquisition of the optimal integration time is more standardized and operable, the algorithm parameters are independently configured to adapt to different pixel characteristics, the candidate values that reduce the non-uniformity are iteratively screened and retained, the integration time finally determined can effectively minimize the corrected non-uniformity, the optimization efficiency and accuracy are improved, and blind calculation is avoided.

[0070] Further, the effective response interval based on the nonlinear response model of the pixel includes:

[0071] The integration time range in which the output signal of the nonlinear response model of the pixel changes approximately linearly with the increase of the integration time is extracted, and the lower limit of the interval is taken as the lower limit of the initial value range of the integration time;

[0072] The maximum integral time of the output signal not reaching the saturation threshold in the nonlinear response model of the pixel is extracted, and the time value is taken as the upper limit of the initial integral time range.

[0073] Through the above, the initial integral time range is set based on the effective response interval of the pixel nonlinear model, the invalid interval (such as the nonlinear interval and the saturation interval) is excluded, the optimization calculation amount is reduced, the insufficient response caused by the too short integral time or the signal saturation caused by the too long integral time is avoided, the initial range is adapted to the real response of the pixel, and an accurate foundation is laid for subsequent optimization.

[0074] In the embodiment, the parameters of the nonlinear response model of each pixel are dynamically updated based on the deviation of the actual output of the detector from the output of the nonlinear response model, including the following steps:

[0075] The actual output value of the detector of the pixel in the corrected image is extracted pixel by pixel as a collection period of a fixed image frame group, and the nonlinear response model is called to calculate the predicted output value of the pixel under the current scene radiation condition. The difference between the actual output value of the detector and the predicted output value of the nonlinear response model is taken as a single-pixel residual value, and the single-pixel residual values of continuous multiple frames are smoothed to filter out transient residual fluctuations caused by random noise of the detector, so as to obtain effective residual values reflecting the real response deviation of the pixel.

[0076] Based on the inherent noise level of the infrared focal plane array detector, a residual judgment standard is set. For each pixel, the effective residual value thereof is continuously monitored. If the effective residual value of the pixel continuously exceeds the residual judgment standard, it is determined that the model parameters of the pixel cannot match the current response characteristics, and a parameter updating process is triggered.

[0077] For the pixel triggering the update, the parameters of the nonlinear response model thereof are adjusted according to the positive and negative of the effective residual value and the deviation size. If the effective residual value is positive, the parameters (corresponding to a p ) representing the linear correlation between the integral time and the radiation flux, the parameters (corresponding to c p ) representing the nonlinear correlation, and the basic offset parameter (corresponding to b p ) in the model are respectively adjusted by a small amount. The adjustment amount is adapted according to the residual deviation size (the larger the residual deviation, the larger the adjustment amount, so as to effectively reduce the residual). If the effective residual value is negative, the parameters representing the linear correlation between the integral time and the radiation flux, the parameters representing the nonlinear correlation, and the basic offset parameter in the model are respectively adjusted by a small amount.

[0078] By the above, the true response deviation is obtained by collecting the residual error through fixed frame groups and smoothing processing to filter out the instantaneous residual error fluctuation caused by random noise of the detector; the residual error determination standard is set based on the inherent noise to avoid false triggering of parameter updating; the model parameters are adjusted according to the positive and negative and size of the residual error to accurately reduce the residual error and ensure that the parameter updating is consistent with the current response state of the pixel and avoids model instability caused by blind adjustment.

[0079] Further, the parameter of the nonlinear response model of each pixel is dynamically updated based on the deviation of the actual output of the detector from the output of the nonlinear response model, and the method further comprises the following steps:

[0080] After the parameter adjustment is completed, the predicted output value of the pixel is calculated through the nonlinear response model, and the new residual error value of the new predicted output value and the actual output value of the detector is compared. If the new residual error value meets the residual error determination standard and the output value of the pixel is in the effective response interval of the detector, it is confirmed that the parameter updating is effective. If the new residual error value does not meet the residual error determination standard or the output exceeds the effective interval, the model parameters before adjustment are returned to, the adjustment amplitude is reduced, and the parameters of the nonlinear response model are adjusted again until the new residual error value meets the residual error determination standard and the output value of the pixel is in the effective response interval of the detector.

[0081] By the above, the updated parameters are verified and returned to adjust the mechanism to ensure that the updated parameters can make the residual error meet the standard and the pixel output be in the effective interval; the problem of residual error exceeding the standard or output exceeding the interval (over-saturation / over-weakness) after parameter adjustment is avoided, and the adaptive parameters are gradually found through re-adjustment to improve the reliability and stability of the model parameter updating.

[0082] In the embodiment, the response nonlinear difference between different pixels is compensated based on the optimal integration time, and the compensated signal is further linearly compensated based on a two-point correction formula, comprising the following steps:

[0083] For each pixel of the infrared focal plane array, the corresponding optimal integration time is called respectively, and the pixel output signal is adjusted through the adaptation of the optimal integration time and the pixel response characteristics;

[0084] The correction gain and bias are calculated by setting two reference radiation points, and the output signal of all the adjusted pixels is linearly adjusted.

[0085] The core of the first correction is to avoid the nonlinear response interval of the pixel output through the pixel-specific optimal integration time. The essence is to realize nonlinear→approximate linear signal conversion based on the pixel-level nonlinear response model, and the formula is as follows: Y lin,p = f p (t opt,p , F p), where p is the pixel index, and its value range is the total number of pixels of the infrared focal plane array (IRFPA), such as p = 1, 2, …, N (N is the total number of pixels); t opt,p is the optimal integration time of pixel p; F p is the actual radiation flux received by pixel p, which is determined by the current scene radiation distribution (such as the infrared radiation intensity of the target / background); f p (·) is the nonlinear response model function of pixel p, which represents the mapping relationship between the radiation flux and the optimal integration time and the pixel output.

[0086] The secondary correction is to further eliminate the residual linear difference of the approximately linear signal after the primary correction, and the formula is modified to the standard mathematical format (the original formula has format deviation): where Y lin,p is the output signal of pixel p after the primary correction, which has been adjusted by t opt,p and has eliminated most of the nonlinear differences, and the signal shows an approximately linear characteristic; Y corr,p is the final corrected signal of pixel p after the secondary correction, which is the output after eliminating the linear residual difference, and is directly used for image display or subsequent processing; G p is the two-point correction gain coefficient of pixel p, which is used to compensate for the response sensitivity difference of pixel p, and is calculated from the fluctuation range of the signal after the primary correction and the expected output range of the detector, and the specific calculation formula is Y exp,max is the upper limit of the expected output of the detector, which is determined by the hardware performance or application requirements of the detector (such as 255 for an 8-bit detector and 4095 for a 12-bit detector), and is a fixed value based on the detector model preset, Y exp,min is the lower limit of the expected output of the detector, which is usually set to 0 (the ideal output value when there is no signal) or a small positive value (such as 10) according to the noise suppression requirements to avoid zero drift interference, Y lin,p,max is the maximum value of the signal of pixel p after the primary correction, Y lin,p,min is the minimum value of the signal of pixel p after the primary correction; B p is the two-point correction bias coefficient of pixel p, which is used to compensate for the output reference offset of pixel p, and is calculated from the deviation of the mean value of the signal after the primary correction and the output reference value (such as the midpoint gray value) of the detector, and the specific calculation formula is: is the mean value of the signal after the primary correction, Y exp,ref is the output reference value of the detector, which is usually set to the midpoint of the expected output range to ensure that the signal is symmetrically distributed within the expected range, and is a fixed value.

[0087] By the above, the specific implementation manner of multi-stage correction is determined, and the logic of non-linear compensation and linear compensation is landed; wherein the optimal integration time can be adapted to the pixel characteristics to compensate for the non-linear difference, and the gain bias calculated by the high and low reference points can accurately correct the linear residual difference, ensuring the execution effect of multi-stage correction and further reducing the non-uniformity.

[0088] In summary, the infrared focal plane array non-uniformity correction method based on integration time adjustment of the embodiment, by establishing a non-linear response model representing the correlation between integration time, radiant flux and detector output response for each pixel of the infrared focal plane array based on multiple sets of calibration data, accurately captures the linear and nonlinear characteristics of pixel response, breaks through the limitations of traditional linear correction methods that cannot handle nonlinear distortion under high radiation or long integration time, and realizes efficient correction of the nonlinear response of the detector, reducing image non-uniformity; secondly, by determining the optimal integration time for each pixel according to the radiation distribution characteristics of the current scene and the independent response characteristics of each pixel, using an adaptive optimization algorithm, the output signal of each pixel can avoid its own nonlinear response interval, solving the problem that existing global or segmented integration time adjustment cannot adapt to individual differences of pixels, realizing pixel-level dynamic adaptation, effectively covering the full dynamic range of the radiation scene of the detector; at the same time, by updating the parameters of the nonlinear response model of each pixel based on the deviation between the actual output of the detector and the output of the nonlinear response model, the response drift of the detector caused by environmental changes or aging is adapted in real time, without the need for frequent laboratory calibration, realizing adaptive adjustment of the response drift of the detector, improving the stability of the system for long-term work; in addition, by compensating for the response nonlinear difference between different pixels based on the optimal integration time, and further linearly compensating for the compensated signal based on the two-point correction formula, a cooperative correction logic of non-linear compensation and linear compensation is formed, which eliminates nonlinear differences and then refines linear residual errors, taking into account the correction accuracy and real-time of engineering implementation, and without the need for large-scale modification of the existing infrared focal plane array hardware, it is easy to integrate and apply in existing imaging systems. Compared with the prior art, the present application solves the defects of traditional correction methods in non-linear correction, scene adaptation, drift adaptation, etc., effectively improving the imaging quality of the infrared focal plane array.

[0089] In the embodiments provided by the present application, it should be understood that the embodiments described herein can be realized by hardware, software, firmware, middleware, codes or any proper combination thereof. For hardware implementation, the processor can be realized in one or more of the following components: an application specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a processor, a controller, a microcontroller, a microprocessor, other electronic units designed to perform the functions described herein, or a combination thereof. For software implementation, the procedures described herein can be implemented with a computer program that is written in any suitable programming language. The program can be stored in a computer readable storage medium or transmitted as one or more instructions or codes on the computer readable storage medium. The computer readable storage medium includes any storage medium that can be accessed by a computer. The computer readable storage medium can include but is not limited to the following media: a RAM, a ROM, an EEPROM, a CD-ROM or other optical disc storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program codes in the form of instructions or data structures and can be accessed by a computer.

[0090] Finally, it should be noted that the above only describes the preferred embodiments of the present application and is not intended to limit the present application. Although the present application has been described in detail with reference to the foregoing embodiments, modifications or equivalent replacements of some technical features described in the foregoing embodiments can be made by those skilled in the art, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. An integration time adjustment based infrared focal plane array non-uniformity correction method, comprising: The method comprises the following steps: For each pixel of the infrared focal plane array, a nonlinear response model representing the correlation between the integration time, the radiant flux and the detector output response is respectively established based on multiple sets of calibration data; An optimal integration time is determined for each pixel by an adaptive optimization algorithm according to the radiation distribution characteristics of the current scene and the independent response characteristics of each pixel, wherein the optimization objective is to minimize the non-uniformity of the corrected image; The parameters of the nonlinear response model of each pixel are dynamically updated based on the deviation between the actual output of the detector and the output of the nonlinear response model; The response nonlinearity difference between different pixels is compensated based on the optimal integration time, and the compensated signal is further linearly compensated based on a two-point correction formula.

2. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 1, wherein, The adaptive optimization algorithm for determining the optimal integration time for each pixel comprises the following steps: For each pixel, an initial value range of the integration time is set based on the effective response interval of the nonlinear response model of the pixel, wherein the lower limit of the initial value range is not lower than the minimum response time of the infrared focal plane array detector, and the upper limit is not higher than the maximum tolerance time of the detector to avoid signal saturation; A particle swarm optimization algorithm or a genetic algorithm is selected as the adaptive optimization algorithm, and algorithm parameters including the initial population size and the iteration termination condition are independently configured for each pixel; For each pixel, an integration time candidate value is iteratively screened within the initial value range by the adaptive optimization algorithm based on the radiation distribution characteristics of the current scene, and after each iteration, the non-uniformity of the pixel output corresponding to the candidate value is calculated, the candidate value that reduces the non-uniformity is retained, and the candidate value that increases the non-uniformity is eliminated; When the algorithm meets the iteration termination condition, the value corresponding to the minimum non-uniformity among the currently retained integration time candidate values is determined as the optimal integration time of the pixel.

3. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 2, wherein, The effective response interval of the nonlinear response model comprises: The lower limit of the integration time range in which the output signal of the nonlinear response model of the pixel changes approximately linearly with the increase of the integration time is extracted as the lower limit of the initial value range of the integration time; The maximum integration time in which the output signal of the nonlinear response model of the pixel does not reach the saturation threshold is extracted as the upper limit of the initial value range of the integration time.

4. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 1, wherein, The nonlinear response model comprises a first-order correlation term of the integration time and the radiant flux, a second-order cross-correlation term of the integration time and the radiant flux and a basic offset term; wherein the first-order correlation term is used to represent the influence of the linear relationship between the integration time and the radiant flux on the pixel output, the second-order cross-correlation term is used to represent the influence of the nonlinear coupling relationship between the integration time and the radiant flux on the pixel output, and the basic offset term is used to represent the inherent output offset of the pixel without radiation input.

5. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 4, wherein, The expression of the nonlinear response model is: wherein Y p is the actual output gray value of the pth pixel in the infrared focal plane array; t int is the integration time configured for the pth pixel in the infrared focal plane array; F p is the radiation flux actually received by the pth pixel in the infrared focal plane array; a p is the first-order cross term coefficient of the pth pixel in the infrared focal plane array; b p is the inherent offset parameter of the pth pixel in the infrared focal plane array; c p is the second-order cross term coefficient of the pth pixel in the infrared focal plane array.

6. The integration time adjustment based infrared focal plane array non-uniformity correction method according to any one of claims 1-5, wherein, The nonlinear response model is obtained by fitting a plurality of sets of calibration data by least squares method, and in the parameter fitting process, based on least squares method, the deviation between the pixel output gray value predicted by the model and the pixel output gray value actually measured in the calibration experiment is minimized, and finally the nonlinear response model parameters corresponding to each pixel are determined, and the construction of the single pixel nonlinear response model is completed.

7. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 6, wherein, The calibration data includes a calibration radiation source, an integration time point, and a radiation flux level. The calibration radiation source includes a high-precision black body. The integration time point covers the entire integration time interval of the conventional operation of the infrared focal plane array detector. The radiation flux level covers the low, medium and high radiation intervals of the effective dynamic range of the infrared focal plane array detector.

8. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 1, wherein, The deviation between the actual output of the detector and the output of the nonlinear response model is used to dynamically update the parameters of the nonlinear response model of each pixel, including the following steps: A fixed image frame group is used as the acquisition period, and the actual output value of the detector of the pixel in the corrected image is extracted pixel by pixel, and the nonlinear response model is called to calculate the predicted output value of the pixel under the current scene radiation condition; the difference between the actual output value of the detector and the predicted output value of the nonlinear response model is taken as the single-pixel residual value, and the single-pixel residual values of continuous multiple frames are smoothed to obtain effective residual values that can reflect the real response deviation of the pixel; Based on the inherent noise level of the infrared focal plane array detector, a residual judgment standard is set; for each pixel, the effective residual value is continuously monitored, and if the effective residual value of the pixel continuously exceeds the residual judgment standard, the parameter updating process is triggered; For the pixels triggered for updating, the parameters of the nonlinear response model are adjusted according to the positive and negative of the effective residual value and the deviation size, wherein if the effective residual value is positive, the parameters in the model representing the linear correlation between integration time and radiation flux, the parameters representing the nonlinear correlation, and the basic offset parameters are respectively adjusted by a small amount, and the adjustment amplitude is adapted according to the residual deviation size; if the effective residual value is negative, the parameters in the model representing the linear correlation between integration time and radiation flux, the parameters representing the nonlinear correlation, and the basic offset parameters are respectively adjusted by a small amount.

9. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 8, wherein, The deviation between the actual output of the detector and the output of the nonlinear response model is used to dynamically update the parameters of the nonlinear response model of each pixel, and further includes the following steps: After the parameter adjustment is completed, the predicted output value of the pixel is calculated by the nonlinear response model, the new residual value of the new predicted output value and the actual output value of the detector is compared, if the new residual value meets the residual judgment standard and the output value of the pixel is in the effective response interval of the detector, it is confirmed that the parameter updating is effective; if the new residual value does not meet the residual judgment standard or the output exceeds the effective interval, the model parameters before adjustment are returned, the adjustment amplitude is reduced, and the nonlinear response model parameters are adjusted again until the new residual value meets the residual judgment standard and the output value of the pixel is in the effective response interval of the detector.

10. The integration time adjustment based infrared focal plane array non-uniformity correction method of claim 1, wherein, The optimal integration time is used to compensate for the difference in response nonlinearity between different pixels, and the compensated signal is further linearly compensated based on a two-point correction formula, including the following steps: For each pixel of the infrared focal plane array, the corresponding optimal integration time is called, and the output signal of the pixel is adjusted through the adaptation of the optimal integration time and the pixel response characteristic. The correction gain and bias are calculated by setting high and low reference radiation points, and the output signal of all the adjusted pixels is linearly adjusted.

Images