Imaging signal-to-noise ratio prediction method based on real image iterative correction

Abstract

The invention relates to the technical field of image processing, in particular to an imaging signal-to-noise ratio prediction method based on real image iterative correction, which comprises the following steps of: inputting parameters of a defined reference imaging scene into an imaging system model to obtain a plurality of scene images and corresponding reference signal-to-noise ratios; predicting a to-be-optimized parameter of the imaging system model and a corresponding predicted signal-to-noise ratio by using an optimization algorithm; based on the predicted signal-to-noise ratio and the reference signal-to-noise ratio, constructing a target function related to the to-be-optimized parameters of the imaging system model; performing updating iteration on the to-be-optimized parameter by taking the target function smaller than a preset value as a target to obtain an optimal to-be-optimized parameter; and adjusting the imaging system model according to the optimal to-be-optimized parameter, inputting the parameter of the new target scene needing to be predicted into the adjusted imaging system model, performing signal-to-noise ratio calculation, and predicting the signal-to-noise ratio corresponding to the target new scene. The method can be used for carrying out high-reliability signal-to-noise ratio prediction on any new scene.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of image processing, and particularly relates to an imaging signal-to-noise ratio prediction method based on real image iterative correction. BACKGROUND

[0002] With the rapid development of aerial photoelectric imaging technology and its wide application in military and civilian fields, the performance requirements of imaging systems are increasingly stringent. Especially in new platforms such as near-space vehicles and long-range wide-area surveillance and extreme application scenarios, photoelectric imaging systems face unprecedented challenges. These scenarios usually have characteristics such as long imaging distance, severe atmospheric influence, and low target-to-background contrast, making the imaging signal-to-noise ratio a core and key indicator for measuring system imaging quality and determining target detection and recognition capability.

[0003] In the demonstration, design and mission planning stage of an aerial photoelectric imaging system, it is of great theoretical and application value to accurately predict the image signal-to-noise ratio under specific imaging conditions in future missions. Currently, the mainstream method for predicting the imaging signal-to-noise ratio under specific scenarios mainly relies on pure theoretical simulation. This method usually uses mature atmospheric radiation transfer models such as MODTRAN or CART, inputs the pre-set sun-target-sensor geometric relationship, atmospheric parameters and ground reflectivity, calculates the radiance at the sensor entrance pupil, and then combines the sensor's own theoretical design parameters such as quantum efficiency, optical system transmittance, pixel size, focal length, etc. to calculate the signal electron number, and combines it with the calculation results of the theoretical noise model, including photon shot noise, dark current noise, readout noise, and photoresponse non-uniformity PRNU, to finally obtain the theoretical signal-to-noise ratio. However, this pure theoretical simulation method has inherent and difficult-to-overcome limitations: (1) Deviation between idealized model parameters and actual situation: Many noise parameters used in the theoretical model, such as the actual fluctuation characteristics of the sensor's dark current under real working temperature and irradiation environment, the correction residual of PRNU after laboratory calibration, the actual point spread function of the optical lens and the level of stray light, etc., all have significant differences with the real working state of the sensor. Small deviations in these parameters will be amplified under long-link and low-signal imaging conditions, resulting in distorted final signal-to-noise ratio prediction results; (2) Simplification of system link complexity: The real imaging link is a complex optical-electrical system, including non-uniformity of optical system transmittance, time-domain and spatial-domain noise crosstalk in electronic system, etc. Theoretical models often simplify these complex physical processes, introducing unavoidable model errors.

[0004] The above factors jointly cause that the pure theoretical simulation result often has large difference with the actual imaging condition, and the credibility of the prediction result is not high. For the maturely operated aircraft platform, a large amount of measured data can be accumulated, and the statistical correction and calibration are performed on the theoretical model, so that the prediction precision is improved. However, for the new platform which is still in the research and development, demonstration or test stage, and for the special scene which is difficult to frequently implement the photographing task, the actual image data which can be used for model verification and calibration are seriously lacked. The predicament of the data shortage makes it extremely difficult to accurately predict the imaging capability in the future task, so that the optimization design of the key indicators of the system, the accuracy of the task feasibility analysis are directly affected, and even the waste of the research and development resources and the increase of the task risk are caused. SUMMARY

[0005] Therefore, the present application aims to provide an imaging signal-to-noise ratio prediction method based on real image iterative correction, which uses an optimization algorithm to predict the to-be-optimized parameters of the imaging system model, compares the corresponding predicted signal-to-noise ratio with the actual imaging data from a known scene, and through reverse iterative correction, makes the to-be-optimized parameters of the imaging system model approach the “equivalent value” in the real imaging system model, so as to “anchor” the entire imaging system model. The imaging system model after the correction can be used for high-credibility signal-to-noise ratio prediction for any new scene.

[0006] To achieve the above object, the technical scheme of the present application is as follows: An imaging signal-to-noise ratio prediction method based on real image iterative correction, comprising: S1: inputting the parameters of a defined reference imaging scene into the imaging system model to obtain a plurality of scene images, and calculating the reference signal-to-noise ratio of each scene image; S2: using an optimization algorithm to predict the to-be-optimized parameters of the imaging system model, and calculating the predicted signal-to-noise ratio of each scene under the predicted to-be-optimized parameters of the imaging system model; S3: constructing a target function related to the to-be-optimized parameters of the imaging system model according to the relative error between the predicted signal-to-noise ratio obtained in step S2 and the reference signal-to-noise ratio obtained in step S1; S4: repeating steps S2-S3 until the target function is less than a preset value, at which time the to-be-optimized parameters corresponding to the target function obtained are the optimal to-be-optimized parameters; S5: adjusting the imaging system model according to the optimal to-be-optimized parameters obtained in step S4, inputting the parameters of a new target scene which needs to be predicted into the adjusted imaging system model, and calculating the signal-to-noise ratio to predict the signal-to-noise ratio corresponding to the target new scene.

[0007] Further, the imaging system model in step S1 is as follows: ; wherein N e represents photoelectron signal of the imaging sensor, L represents entrance radiance of the imaging sensor, F represents relative aperture of the lens of the imaging sensor, t int represents integration time, τ opt represents optical transmittance, and R represents sensitivity of the imaging sensor.

[0008] Further, the entrance radiance includes target radiance and background radiance, and the target radiance and the background radiance are obtained by the following formula: ; ; wherein L a and L b represent the target radiance and the background radiance respectively, ρ t and ρ b represent target reflectivity and background reflectivity respectively, and A, B, C and D represent atmospheric complex constants of an environment in which the imaging system model is located.

[0009] Further, the atmospheric complex constant A is path radiance when the surface reflectivity is 0.0; the atmospheric complex constant B is: ; wherein TOP 100 , TOP 50 and TOP0 represent total radiance when the surface reflectivity is 1.0, 0.5 and 0.0 respectively; the atmospheric complex constant C is: ; the atmospheric complex constant D is: ; wherein GRFL 100 represents ground direct radiance when the surface reflectivity is 1.0.

[0010] Further, the step S2 includes: initializing the to-be-optimized parameters; searching the to-be-optimized parameters in a preset physically reasonable range by using a constrained nonlinear optimization algorithm, to find locally optimal to-be-optimized parameters; and calculating the predicted signal-to-noise ratio of each scene under the locally optimal to-be-optimized parameters of the imaging system model.

[0011] Further, the objective function in the step S3 is: ; wherein J(p) represents an objective function related to the to-be-optimized parameters p, represents the predicted signal-to-noise ratio of the i th scene. a reference signal-to-noise ratio of the i-th scene.

[0012] Compared with the prior art, the present application can achieve the following beneficial effects: In the imaging signal-to-noise ratio prediction method based on real image iterative correction provided by the present application: (1) greatly improve the prediction accuracy: the present application innovatively introduces real image data as a "scale" or "anchor point" to accurately correct the pure theoretical model, forcing the key parameters in the model to fit the real physical world, thereby effectively overcoming the inherent errors caused by the idealization of theoretical model parameters and the simplification of system link; the corrected model inherits the physical completeness of the theoretical model and the authenticity of the actual data, and its prediction result is more consistent with the future actual imaging than pure theoretical simulation; (2) reduce the dependence on measured data: the method provided by the present application does not need to actually shoot the target new scene, but only needs to use one or a few image data obtained by the photoelectric remote sensing system under known conditions to complete the "calibration" of the entire prediction model, effectively reducing the difficulty and cost of data acquisition, and is particularly suitable for performance evaluation of new and research aircraft platforms; (3) wide application scenarios and high practical value: the corrected imaging system model has strong generalization ability and can be used to predict the imaging performance in various new scenes that have not been experienced, which provides a strong and reliable decision support tool for the demonstration design, actual imaging performance estimation and feasibility analysis of complex imaging tasks of new photoelectric remote sensing systems, effectively reducing the risk of research and development and the probability of task failure, and saving research and development cost and time. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, which form a part of the present application, are used to provide further understanding of the present application, and the schematic embodiments of the present application and their descriptions are used to explain the present application, and do not constitute improper limitations on the present application. In the drawings: Fig. 1 a flowchart of the imaging signal-to-noise ratio prediction method based on real image iterative correction according to the embodiment of the present application; Fig. 2 a flowchart of the imaging signal-to-noise ratio prediction method based on real image iterative correction according to the embodiment of the present application. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application will be further described in detail below with reference to the drawings and specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and do not constitute limitations on the present application.

[0015] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0016] In the description of the present application, it should be understood that the terms "center", "longitudinal", "transverse", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation of the present application. In addition, the terms "first", "second" and the like are only for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined with "first", "second" and the like can explicitly or implicitly include one or more features. In the description of the present application, unless otherwise specified, the meaning of "a plurality of" is two or more.

[0017] In the description of the present application, it should be noted that unless otherwise explicitly specified and limited, the terms "mounting", "connection", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be connected inside two elements. For those skilled in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0018] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0019] As Figs. 1-2 shown, the imaging signal-to-noise ratio prediction method based on real image iterative correction according to the embodiments of the present application comprises: S1: input the parameters of the defined reference imaging scene into the imaging system model, obtain a plurality of scene images, and calculate the reference signal-to-noise ratio of each scene image.

[0020] In the present application, a reference imaging scene is first defined, and all the environmental and geometric parameters of the scene should be known and accurately described, including but not limited to platform height, imaging slant range, sun angle, atmospheric model, visibility, and reflectivity of scene objects. On this basis, a full-link imaging system model is constructed, which serves as a programmable framework and integrates at least three core parts: atmospheric model parameterization, in-come radiance calculation, and sensor photoelectric conversion. Specifically, in some embodiments: The atmospheric model parameterization part includes: Three independent atmospheric radiation transfer simulations are performed: (1) Set the surface albedo to 0.0, and obtain the total radiance TOT0 and path radiance PATH0 corresponding to the albedo 0.0; (2) Set the surface albedo to 0.5, and obtain the total radiance TOT 50 and path radiance PATH 50 corresponding to the albedo 0.5; (3) Set the surface albedo to 1.0, and obtain the total radiance TOT 500 , path radiance PATH 100 , and ground direct radiance GRFL a corresponding to the albedo 1.0.

[0021] The atmospheric composite constants obtained by inversion from the results of the three simulations are: A = PATH0; ; ; ; Wherein, A, B, C and D represent the atmospheric composite constants of the environment in which the imaging system model is located.

[0022] For the incoming radiance calculation part, the atmospheric composite constants A, B, C and D obtained by the atmospheric model parameterization part are used to calculate the target radiance and background radiance: ; ; Wherein, L a and L b represent the target radiance and background radiance respectively, and ρ t and ρ b represent the target albedo and background albedo respectively.

[0023] For the sensor photoelectric conversion part, according to the optical and electronic parameters of the imaging sensor, the target radiance L a and background radiance L b are substituted into the photoelectron signal calculation formula respectively, and the target photoelectron number N e_target and background photoelectron number N e_background are obtained. The photoelectron signal calculation formula is: ; Wherein, N e represents the photoelectron signal of the imaging sensor, L represents the incoming radiance of the imaging sensor, F represents the relative aperture of the lens of the imaging sensor, t int represents the integration time, and τopt Let L represent optical transmittance, and R represent the sensitivity of the imaging sensor. Correspondingly, when the incident radiance L of the imaging sensor is equal to the target radiance L... a When L=L a Substituting into the above formula, we can obtain the target number of photoelectrons N. e_target When the incident radiance L of the imaging sensor is equal to the background radiance L... b When L=L b Substituting into the above formula, we can obtain the number of background photoelectrons N. e_background The photoelectronic signals from the target and background will be used for subsequent signal-to-noise ratio calculations.

[0024] In this embodiment of the invention, a target region and one or more uniform background regions are selected from each captured scene image. The mean of the pixel grayscale values ​​within the target and background regions is calculated as a measure of the signal, and the standard deviation of the background region is calculated as a measure of the noise. The baseline signal-to-noise ratio is thus calculated. ; Among them, SNR base Represents the reference signal-to-noise ratio, μ tgt and μ bkg σ represents the average gray level of the target region and the selected neighboring uniform background region, respectively. bkg This represents the standard deviation of gray levels in the selected neighboring uniform background region.

[0025] In this embodiment of the invention, step S1 further includes: Based on some adjustable key parameters of the imaging sensor, a comprehensive noise model is constructed as follows: ; Where, σ totle σ represents the combined noise. shot σ represents the photon shot noise in the imaging sensor. dark σ represents the dark current noise in the imaging sensor. read This represents the readout noise; this value is generally constant and can be obtained from the detector manual. σ quant To quantize noise, σ is determined by the number of bits in the analog-to-digital converter (ADC) and the dynamic range of the input signal. PRNU This indicates non-uniformity in photoelectric response.

[0026] Specifically, photon shot noise σ shot As shown in the following formula: ; Where, k shot This represents the correction coefficient used to correct the difference between the theoretical Poisson distribution and the actual distribution for photon shot noise. Dark current noise σ darkThe following formula: ; Wherein, t int represents the integral time, t int represents the integral time I dark (T) represents the actual characterization parameter of the deviation dark current noise for correcting the deviation between the laboratory calibration value and the actual working value, and the dark current related to the temperature T; Quantization noise σ quant The following formula: ; Wherein, n represents the number of bits of the analog-to-digital converter, N full-scale is the total number of electrons corresponding to the full scale range of the analog-to-digital converter.

[0027] Photoelectric response non-uniformity σ PRNU The following formula: ; Wherein, k PRNU represents the equivalent correction residual coefficient of the PRNU for quantifying the residual noise after non-uniformity correction.

[0028] S2: predicting the to-be-optimized parameters of the imaging system model by using an optimization algorithm, and calculating the predicted signal-to-noise ratio of each scene under the predicted to-be-optimized parameters of the imaging system model.

[0029] In some embodiments, step S2 comprises: S21: initializing the to-be-optimized parameters.

[0030] Wherein, the to-be-optimized parameters include one or more of the correction coefficient k shot of the photon shot noise, the actual characterization parameter I dark (T) of the deviation dark current noise, and the equivalent correction residual coefficient k PRNU .

[0031] S22: searching the to-be-optimized parameters in a preset physically reasonable range by using a constrained nonlinear optimization algorithm, and finding the locally optimal to-be-optimized parameters.

[0032] In the embodiments of the present application, the constrained nonlinear optimization algorithm can be a self-interior point method, a sequential quadratic programming method, an effective set method or a trust region method. The to-be-optimized parameters are iteratively adjusted by the constrained nonlinear optimization method, so as to ensure that the to-be-optimized parameters are searched in the preset physically reasonable range.

[0033] S23: calculating the predicted signal-to-noise ratio of each scene under the locally optimal to-be-optimized parameters of the imaging system model.

[0034] S3: constructing a target function related to the to-be-optimized parameter of the imaging system model according to the relative error between the predicted signal-to-noise ratio obtained in step S2 and the reference signal-to-noise ratio obtained in step S1.

[0035] In some embodiments, the target function is: ; wherein J(p) represents the target function related to the to-be-optimized parameter p, represents the predicted signal-to-noise ratio of the i-th scene, represents the reference signal-to-noise ratio of the i-th scene.

[0036] S4: repeating steps S2-S3 until the target function is less than a preset value, considering that the to-be-optimized parameter approaches an effective true value close to a physical real situation, and the to-be-optimized parameter corresponding to the target function obtained at this time is the optimal to-be-optimized parameter.

[0037] S5: adjusting the imaging system model according to the optimal to-be-optimized parameter obtained in step S4, inputting the parameters of a new target scene to be predicted into the adjusted imaging system model, and performing signal-to-noise ratio calculation to predict the signal-to-noise ratio corresponding to the target new scene.

[0038] To clearly describe the imaging signal-to-noise ratio prediction method based on real image iterative correction provided by the embodiments of the present application, a specific embodiment is given: S1: inputting the parameters of a defined reference imaging scene into the imaging system model to obtain two scene images, and calculating the reference signal-to-noise ratio of each scene image.

[0039] In this specific embodiment, two scenes are included: Scene 1 (aircraft shooting airport scene): flight height 5.8 km, imaging slant range 9.27 km; target reflectivity 0.5 (0.5 for civil aircraft), airport ground reflectivity 0.15; atmospheric condition: mid-latitude summer, visibility 15 km, solar zenith angle 60 degrees, imaging environment temperature 5℃; Scene 2 (ground shooting building scene): height 0.948 km, imaging slant range 10 km; target reflectivity 0.4, background reflectivity 0.2; atmospheric condition: mid-latitude summer, visibility 15 km, solar zenith angle 30 degrees, imaging environment temperature 20℃.

[0040] The atmospheric composite constants A, B, C and D in scene 1 are A=3.1718, B=0.05934, C=41.228, D=27.57731 respectively; the target radiance and background radiance are 19.1503 and 9.4115 respectively.

[0041] The atmospheric composite constants A, B, C and D in scenario 2 are A = 8.5518, B = 0.025414, C = 48.75806, and D = 13.92469, respectively; the target radiance and the background radiance are 22.55198026 and 18.35323, respectively. .

[0042] The incident radiance is converted into photoelectron signals according to the optical and electronic parameters of the imaging sensor. The model parameters are as follows: the relative aperture F of the lens is 3.6, τ opt The optical transmittance is 0.5, and the sensitivity R of the imaging sensor is 33300000. At this time, the following can be obtained: The integration time of scenario 1 is 0.2ms, and the calculated target radiation electron number is 3864e-, and the background radiation electron number is 1899; the integration time of scenario 2 is 0.13ms, and the calculated target radiation electron number is 2958e-, and the background radiation electron number is 2407.

[0043] The gray mean μ tgt_1 of the target region in scenario 1 is 836DN, the gray mean μ bkg_1 of the background region is 403DN, the standard deviation σ bkg_1 of the background region is 8.256DN, and the calculated actual measured reference signal-to-noise ratio SNR base_1 is 52.45. The gray mean μ tgt_2 of the target region in scenario 2 is 631DN, the gray mean μ bkg_2 of the background region is 510DN, the standard deviation σ bkg_2 of the background region is 8.431DN, and the calculated actual measured reference signal-to-noise ratio SNR base_2 is 14.35. These parameters become the "anchor points" for correcting the imaging system model.

[0044] S2: predicting the to-be-optimized parameters of the imaging system model using an optimization algorithm, and calculating the predicted signal-to-noise ratio of each scenario under the predicted to-be-optimized parameters of the imaging system model.

[0045] In this specific embodiment, step S2 includes: S21: initializing the to-be-optimized parameters. Among them, considering that the shot noise is the main noise parameter, and in addition, the system corrected non-uniform correction error is also the main error term, therefore, the correction coefficient k shot of the photon shot noise and the equivalent correction residual coefficient k PRNU of the PRNU are mainly adjusted. The correction coefficient k shot of the photon shot noise is initialized to 1.0, and the equivalent correction residual coefficient k PRNU ​is initialized to 0. In addition, according to the detector manual, the calibration curve expression of the dark current is: 0.9066e 0.05228T Therefore, the dark current of scenario 1 is calculated as 1.177e- / s, and the dark current of scenario 2 is calculated as 3.3499e- / s. In addition, the readout noise σ read = 6.3e-, and the quantization noise σ quant = 1.163e-. It can be obtained that the total number of noise electrons of the target region in scenario 1 is 62, the total number of noise electrons of the background region in scenario 1 is 43, the total number of noise electrons of the target region in scenario 2 is 54, and the total number of noise electrons of the background region in scenario 2 is 49.

[0046] S22: using a constrained nonlinear optimization algorithm, searching for the to-be-optimized parameters in a preset physically reasonable range to find locally optimal to-be-optimized parameters. In the specific embodiment, the correction coefficient k shot of the photon shot noise ranges from [0, 1], and the equivalent correction residual coefficient k PRNU ranges from [0, 0.2].

[0047] In the specific embodiment, the interior point method is used as the constrained nonlinear optimization algorithm, which searches inside the feasible region by converting the constraint conditions into barrier functions, and effectively handles the boundary constraints of the correction parameters. The interior point method iteratively solves a series of unconstrained optimization problems to gradually approach the optimal solution of the original constrained problem, and has the characteristics of fast convergence speed and good numerical stability.

[0048] S23: calculating the predicted signal-to-noise ratio of each scenario under the locally optimal to-be-optimized parameters of the imaging system model.

[0049] S3: constructing a target function related to the to-be-optimized parameters of the imaging system model according to the relative error between the predicted signal-to-noise ratio obtained in step S2 and the reference signal-to-noise ratio obtained in step S1.

[0050] In the specific embodiment, the initial predicted signal-to-noise ratio of scenario 1 is calculated as 44.56, and the relative error between the initial predicted signal-to-noise ratio of scenario 1 and the reference predicted signal-to-noise ratio is 15.03%; the initial predicted signal-to-noise ratio of scenario 2 is calculated as 1.12, and the relative error between the initial predicted signal-to-noise ratio of scenario 2 and the reference predicted signal-to-noise ratio is 22.52%, which indicates that the initial parameters do not completely reflect the actual situation of the imaging system model.

[0051] S4: repeating steps S2-S3 until the target function is less than a preset value, and the to-be-optimized parameters corresponding to the target function at this time are the optimal to-be-optimized parameters.

[0052] In the specific embodiment, the error threshold is set to 10% to minimize the error between the predicted signal-to-noise ratio and the reference predicted signal-to-noise ratio, and the optimization problem is defined as the target function provided by the present application.

[0053] After iterative optimization, the photon shot noise correction coefficient k shot The adjusted value is 0.8445, and the equivalent correction residual coefficient k PRNU When the adjusted value is 0.0012, the new predicted signal-to-noise ratio of scene 1 is 52.43, and at this time, the relative error between the predicted signal-to-noise ratio and the benchmark predicted signal-to-noise ratio is only 0.04%; the new predicted signal-to-noise ratio of scene 2 is 13.088, and at this time, the relative error between the predicted signal-to-noise ratio and the benchmark predicted signal-to-noise ratio is only 8.8%, and the maximum value of the scene error is taken as the comparison error, which is less than the error threshold 10%, at this time, the imaging system model conforming to the real physical condition is obtained, and the model is saved.

[0054] S5: Adjust the imaging system model according to the optimal parameter to be optimized obtained in step S4, input the parameters of the new target scene to be predicted into the adjusted imaging system model, and calculate the signal-to-noise ratio to predict the signal-to-noise ratio corresponding to the new target scene.

[0055] In summary, the specific embodiment shows how to calibrate the theoretical model with the measured data of the benchmark scene, and to predict the signal-to-noise ratio of the new scene with high credibility. The method provided by the present application ensures the consistency of the prediction result with the real physical process, and significantly improves the accuracy of the signal-to-noise ratio prediction.

[0056] It should be understood that the various forms of flow shown above can be used to reorder, add or delete steps. For example, the steps described in the present disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions of the present disclosure can be achieved, which is not limited herein.

[0057] The above specific embodiments do not constitute a limitation on the protection scope of the present application. Those skilled in the art should understand that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent replacement and improvement within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. An imaging signal-to-noise ratio prediction method based on iterative correction of real images, characterized in that, include: S1: Input the parameters of the defined reference imaging scene into the imaging system model to obtain multiple scene images, and calculate the reference signal-to-noise ratio of each scene image; S2: Use optimization algorithms to predict the parameters to be optimized for the imaging system model, and calculate the predicted signal-to-noise ratio for each scene under the predicted parameters to be optimized. S3: Using the relative error between the predicted signal-to-noise ratio obtained in step S2 and the baseline signal-to-noise ratio obtained in step S1, construct an objective function related to the parameters to be optimized in the imaging system model; S4: Repeat steps S2 to S3 until the objective function is less than the preset value. The parameters to be optimized corresponding to the objective function obtained at this time are the optimal parameters to be optimized. S5: Adjust the imaging system model according to the optimal parameters to be optimized obtained in step S4, input the parameters of the new target scene to be predicted into the adjusted imaging system model, and calculate the signal-to-noise ratio to predict the signal-to-noise ratio corresponding to the new target scene.

2. The imaging signal-to-noise ratio prediction method based on iterative correction of real images according to claim 1, characterized in that, The imaging system model in step S1 is as follows: ； Where, N e The image sensor's photoelectronic signal is represented by L, which represents the incident radiance of the image sensor. F represents the relative aperture of the image sensor's lens. int τ represents the integration time. opt R represents optical transmittance, and R represents the sensitivity of the imaging sensor.

3. The imaging signal-to-noise ratio prediction method based on iterative correction of real images according to claim 2, characterized in that, Entrance radiance includes target radiance and background radiance, which are obtained by the following formula: ； ； Among them, L a and L b ρ represents the target radiance and background radiance, respectively. t and ρ b Let A, B, C, and D represent the target reflectance and background reflectance, respectively, and let D represent the atmospheric composite constants of the environment in which the imaging system model is located.

4. The imaging signal-to-noise ratio prediction method based on iterative correction of real images according to claim 3, characterized in that, The atmospheric composite constant A is the path radiance when the surface reflectance is 0.0; The atmospheric composite constant B is: ； Among them, TOP 100 TOP 50 TOP0 and TOP0 represent the total radiance when the surface reflectance is 1.0, 0.5 and 0.0, respectively; The atmospheric composite constant C is: ; The atmospheric composite constant D is: ； Among them, GRFL 100 This represents the direct radiance of the ground when the surface reflectance is 1.

0.

5. The imaging signal-to-noise ratio prediction method based on iterative correction of real images according to claim 1, characterized in that, Step S2 includes: Initialize the parameters to be optimized; By using a constrained nonlinear optimization algorithm, the parameters to be optimized are searched within a preset physical reasonable range, and the locally optimal parameters to be optimized are found. The predicted signal-to-noise ratio for each scene in a computational imaging system model under locally optimal parameters to be optimized.

6. The imaging signal-to-noise ratio prediction method based on iterative correction of real images according to claim 1, characterized in that, The objective function in step S3 is: ； Where J(p) represents the objective function related to the parameter p to be optimized. Let represent the predicted signal-to-noise ratio for the i-th scene. This represents the baseline signal-to-noise ratio for the i-th scene.

Images