Relative radiation correction method for multi-channel spliced large-area-array infrared camera

Abstract

A multi-channel spliced large-area array infrared camera relative radiation correction method belongs to the technical field of spaceflight optical remote sensing, and comprises the following steps: calculating a non-uniform correction coefficient based on onboard blackbody calibration data; acquiring an on-orbit image, correcting the on-orbit image, acquiring a corrected image, and acting on the corrected image by using the secondary correction coefficient and the channel correction coefficient calibrated in the laboratory to obtain a channel corrected image; a plurality of in-orbit images containing different time periods and scenes are accumulated, pixel gray value sequences corresponding to a preset reference channel and a to-be-corrected channel are extracted as sample data according to design values of overlapping pixels between channels of the in-orbit images, and the sample data of the to-be-corrected channel and the reference channel are taken as an x axis and a y axis respectively, so that the to-be-corrected channel is obtained. And according to the quadratic function correction model, correction based on the overlapping area is carried out on the image after channel correction, long-term calibration and correction of the on-orbit detector pixel response relative change under the condition that the inner black body is stable are realized, and the on-orbit image radiation quality is improved.

Description

Technical Field

[0001] This invention relates to a relative radiation correction method for a multi-channel stitched large-area infrared camera, belonging to the field of aerospace optical remote sensing technology. Background Technology

[0002] Given the limitations of detector size and optical system technology, multi-channel stitching is one of the feasible ways to achieve large-area infrared detection. The radiation characteristics of each channel and the differences between them are directly related to the overall array radiation measurement performance.

[0003] For single-channel infrared cameras, the main factors affecting radiometric correction of blackbody-based infrared cameras include the correction algorithm and the calibration source (surface uniformity). Regarding algorithms, the biggest drawback of existing infrared camera radiometric correction algorithms using least squares regression is that they aim to minimize the absolute difference between the fitted values ​​and the target value at all points. This results in high-end calibration points having a much greater impact on the fitting results than low-end points, inevitably leading to significant differences in non-uniformity between the high and low ends. To overcome the instability of the blackbody standard source's own temperature and fluctuations in ambient temperature and humidity during calibration, which cause differences in the standard source's radiance and the infrared camera's response, thus affecting the accuracy of the response curve fitting, some existing algorithms essentially reduce the impact of measurement fluctuations by controlling the confidence level of calibration points. However, the algorithm model representation is not suitable for optimizing radiometric correction accuracy. Regarding calibration sources, on-orbit radiometric calibration of infrared external array cameras mainly relies on on-board blackbodies. Due to limitations such as on-board weight and temperature control capabilities, their surface uniformity cannot reach the accuracy of laboratory-scale large-area blackbody sources. However, current on-board blackbody calibration generally does not consider the surface uniformity of the on-board calibration source.

[0004] For inter-channel radiometric correction, the differences in radiation originate from variations in the channel's optics and detector systems, manifesting as response differences among pixels of the same name in the overlapping area. Histogram statistics are a primary method for inter-channel relative radiometric correction in linear pushbroom cameras, offering advantages in characterizing nonlinear responses. However, this requires a sample size that ensures approximately identical radiometric probability distributions for each pixel. Existing techniques largely rely on large sample sizes; for large-area array cameras with high quantization bits, the sample size is relatively small and generally insufficient to meet statistical requirements. Summary of the Invention

[0005] The technical problem solved by this invention is to overcome the shortcomings of the prior art and provide a relative radiation correction method for a multi-channel stitched large-area infrared camera, so as to realize long-term calibration and correction of the relative changes in the response of on-orbit detector pixels under the condition of internal blackbody stability, thereby improving the radiation quality of on-orbit images.

[0006] The technical solution of this invention is: a method for relative radiometric correction of a multi-channel stitched large-area infrared camera, comprising:

[0007] Calculate the non-uniformity correction coefficient based on on-board blackbody calibration data;

[0008] Acquire in-orbit images, correct the in-orbit images using the non-uniform correction coefficients, obtain corrected images, and apply the secondary correction coefficients and channel correction coefficients from laboratory calibration to the corrected images to obtain channel-corrected images.

[0009] Accumulate a number of on-orbit images containing different time periods and scenes. According to the design value of overlapping pixels between channels, extract the pixel gray value sequences corresponding to the preset reference channel and the channel to be corrected as sample data. Take the sample data of the channel to be corrected and the reference channel as the x-axis and y-axis, respectively, and obtain a quadratic function correction model between the x-axis coordinates and the y-axis coordinates by fitting the least squares method. Then, perform correction on the image after channel correction based on the overlapping area according to the quadratic function correction model. x and y are the independent variable and the dependent variable, respectively.

[0010] Furthermore, the quadratic function correction model is y = a0 + a1x + a2x 2 Let the least squares objective function be... a0, a1, and a2 are calculated iteratively until Δ converges; a0, a1, and a2 are correction coefficients, x is the gray value of the pixel in the channel to be corrected, y is the gray value of the pixel in the reference channel, n is the number of pixels in the overlapping area, and p is the pixel number in the overlapping area.

[0011] Furthermore, the step of correcting the channel-corrected image using the quadratic function correction model includes: DN correct (i,j)=a0+a1DN(i,j)+a2DN(i,j) 2 DN(i,j) represents the response of pixel (i,j) in the image after channel correction. correct (i,j) represents the response of pixel (i,j) in the image after secondary correction of the overlapping area, where i and j are the row and column numbers of the pixel.

[0012] Furthermore, the laboratory calibration includes:

[0013] For the temperature data of each on-board blackbody calibration point, calculate the average DN value of each pixel in the calibration area. and multi-frame average

[0014]

[0015] Where M, N, and F represent the number of pixels in the row and column directions of a single channel of the area array camera and the number of continuously acquired calibration data frames, respectively; DN t (i,j,f) represents the response of pixel (i,j) in the f-th frame of calibration data at temperature t. Multiple sets of data are obtained based on all calibration temperature points t = 1, 2, ..., K, where K ≥ 2, representing the number of temperature points. The correspondence; for cell (i,j), assume and The relationship y = f(x, α) is satisfied, where y represents represent α is a polynomial coefficient, representing the non-uniform correction coefficient for pixel (i,j), using... Normalization is performed to construct the least squares objective function. Iterative optimization is used to calculate the non-uniform correction coefficient α for the on-board blackbody calibration in the laboratory until Δ converges, t0, t end These are the lowest and highest temperature points, respectively.

[0016] Furthermore, the laboratory calibration also includes:

[0017] Based on the non-uniformity correction coefficients for on-board blackbody calibration in the laboratory, relative radiation correction is performed on the data at various temperature points of the external blackbody calibration using the following formula:

[0018]

[0019] In the formula, α(i,j) is the non-uniformity correction coefficient for pixel (i,j), and DN 外 (i,j) These are the response values ​​of pixel (i,j) in the temperature data of each external blackbody calibration before and after correction.

[0020] Furthermore, the non-uniform quadratic correction coefficients of the corrected outer blackbody images at different temperatures are obtained by least-squares linear fitting, which are used to correct the non-uniform correction coefficients during on-orbit operation.

[0021] Furthermore, the laboratory calibration also includes full array consistency correction based on the external blackbody calibration data of each channel after secondary correction, including:

[0022] Calculate the mean value of each channel at different temperatures. s represents the channel number that makes up the area array camera, s = 1, 2, ... C, where C represents the number of channels; The response of channel s to pixel (i,j) in the external blackbody calibration data at calibration temperature t after secondary correction;

[0023] Calculate the mean of all channels at different temperatures.

[0024] Calculate the difference between the mean of each channel and the mean of all channels, and take the channel with the smallest difference as the reference channel and record it as the reference channel for on-orbit correction.

[0025] Using the mean value of each temperature in the baseline channel as the target value, the mean values ​​of each channel are fitted with least squares to obtain the consistency correction coefficient of each channel.

[0026] The advantages of this invention compared to the prior art are:

[0027] (1) Based on the analysis of the principle and shortcomings of the least squares non-uniform correction algorithm, this invention proposes a response normalized least squares characterization model, which fundamentally improves the relative radiation correction accuracy of infrared cameras in the full dynamic range.

[0028] (2) In view of the difficulties in large-sample statistics of large-area infrared cameras in orbit and the fluctuation of sample point spatial and temporal response, this invention proposes a full dynamic range correction processing model based on the overlap area between channels, realizes multi-channel consistency correction based on the in-orbit scene, and supports rapid dynamic updates in orbit.

[0029] (3) This invention considers the difference in surface uniformity between the star-borne blackbody and the laboratory large-area source blackbody and its phenomenon representation in the image. By decoupling the blackbody and pixel non-uniformity, a secondary correction model for the star-borne blackbody is obtained.

[0030] (4) This invention fully explores the intrinsic factors affecting the on-orbit radiation uniformity of spliced ​​large-area infrared cameras, and proposes a comprehensive solution for the relative radiation correction scheme of spliced ​​large-area infrared cameras using multiple methods such as laboratory, on-board blackbody, and scene, so as to meet the radiation consistency measurement requirements within and between channels to the greatest extent. Attached Figure Description

[0031] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0032] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0033] To better understand the above technical solutions, the technical solutions of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solutions of the present invention, rather than limitations on the technical solutions of the present invention. In the absence of conflict, the embodiments of the present invention and the technical features in the embodiments can be combined with each other.

[0034] The following description, in conjunction with the accompanying drawings, provides a more detailed explanation of the relative radiation correction method for a multi-channel stitched large-area infrared camera provided by an embodiment of the present invention. Figure 1 Specific implementation methods may include:

[0035] Calculate the non-uniformity correction coefficient based on on-board blackbody calibration data;

[0036] Acquire in-orbit images, correct the in-orbit images using the non-uniform correction coefficients, obtain corrected images, and apply the secondary correction coefficients and channel correction coefficients from laboratory calibration to the corrected images to obtain channel-corrected images.

[0037] Accumulate a number of on-orbit images containing different time periods and scenes. According to the design value of overlapping pixels between on-orbit image channels, extract the pixel gray value sequences corresponding to the preset reference channel and the channel to be corrected as sample data. Use the sample data of the channel to be corrected and the reference channel as the x-axis and y-axis, respectively. Fit a quadratic function correction model between the x-axis coordinates and the y-axis coordinates using the least squares method. Then, perform correction on the channel-corrected image based on the overlapping area according to the quadratic function correction model. x and y are the independent variable and dependent variable, respectively.

[0038] In the solutions provided in the embodiments of the present invention, such as Figure 1 Specifically, this includes the laboratory calibration phase and the on-orbit calibration phase.

[0039] (1) In the laboratory calibration stage, the non-uniform separation of blackbody and coefficient calculation

[0040] 1) Statistical analysis of on-board blackbody calibration data.

[0041] For the calibration of the blackbody on the satellite at various temperature points t i For data (i = 1, 2, ..., K) (K ≥ 2, representing the number of temperature points), calculate the mean DN value of pixel (i, j) in the calibration area. and multi-frame average

[0042]

[0043] M, N, and F represent the number of pixels in the row and column directions of a single channel of the area scan camera and the number of consecutively acquired calibration data frames, respectively. Considering all calibration temperature points, multiple sets of data are obtained. The correspondence.

[0044] 2) Construct a normalized least squares fitting model.

[0045] For each pixel, in order to improve the fitting accuracy of the camera response nonlinear characteristics, let... and The relationship y = f(x, α) is satisfied, where y represents x represents α is a polynomial coefficient representing the non-uniformity correction coefficient for pixel (i,j). To minimize the non-uniformity after correction, the following method is used: Normalization should be performed so that... To minimize computational complexity, f(x,α) can be set as a linear function. However, to improve the fitting accuracy of the nonlinear characteristics of the camera response, a high-order polynomial is chosen for f(x,α).

[0046] 3) Iterative optimization solution of non-uniform correction coefficients.

[0047] ① Initialize parameters

[0048] The initial value of the α parameter is set based on prior knowledge. Usually, the coefficient of the first term is 1 and the coefficients of other terms are 0.

[0049] ② Linearized model

[0050] At the current parameter estimate α, perform a first-order Taylor expansion of f(x,α):

[0051]

[0052] in It is the i-th row of the Jacobian matrix;

[0053] ③ Construct and solve the weighted nonlinear least squares problem

[0054] Define the residual vector r (k) =yf(x,α) (k) ), weight matrix Solution:

[0055]

[0056] Solving for:

[0057] △α (k) =(J (k)T WJ (k) ) -1 J (k)T Wr (k)

[0058] ④ Parameter update and iteration

[0059] α (k+1) =α (k) +△α (k)

[0060] Repeat until convergence.

[0061] 4) Relative radiation correction is performed on the temperature data of each temperature point for external blackbody calibration based on the non-uniform correction coefficient.

[0062]

[0063] α(i,j) is the non-uniformity correction coefficient for pixel (i,j), DN 外 (i,j) These are the response values ​​of pixel (i,j) in the temperature data of each external blackbody calibration before and after correction.

[0064] 5) The non-uniform quadratic correction coefficients of the outer blackbody are obtained by least-squares linear fitting of the corrected external calibration images at different temperatures.

[0065] 6) Perform full array consistency correction based on the blackbody calibration data of each channel after secondary correction.

[0066] ① Calculate the mean value of each channel at different temperatures:

[0067] Let represent the response of pixel (i,j) in the t-th frame of the image in channel s, where s = 1, 2, ..., C, and C represents the number of channels.

[0068] ② Calculate the mean value of all channels at different temperatures:

[0069]

[0070] ③ Calculate the difference between the mean of each channel and the mean of all channels, take the channel with the smallest difference as the reference channel, and record it as the reference channel for on-orbit correction.

[0071] ④ Using the mean value of each temperature in the reference channel as the target value, perform least squares fitting on the mean value of each channel to obtain the consistency correction coefficient of each channel.

[0072] (2) Calculation of on-orbit non-uniformity coefficients during the on-orbit phase

[0073] 1) Calculation of non-uniformity correction coefficients based on on-board blackbody

[0074] Radiation calibration coefficients were calculated based on on-board blackbody calibration data. For each temperature point data point in the on-board blackbody calibration, the mean DN value of all pixels in the calibration area was calculated. and the mean of pixel (i,j)

[0075] The number of sampling frames for the on-board blackbody calibration data is given, where M and N are the number of pixels in the row and column directions of a single channel of the area array camera, respectively.

[0076] Obtain effective values ​​at each temperature point After determining the correspondence, the non-uniform correction coefficients are fitted using the aforementioned normalized least squares fitting method. Considering the ease of calculation, f(x,β) can be set as a linear function. Considering the fitting accuracy for the nonlinear characteristics of the camera response, a high-order polynomial is chosen for f(x,β).

[0077] 2) Secondary correction of non-uniformity correction coefficient

[0078] The on-orbit image is corrected using the on-board blackbody correction coefficient, and then the secondary correction coefficient and channel correction coefficient are applied to the corrected image.

[0079] 3) Channel correction based on overlap area

[0080] ①Accumulate on-orbit images from different time periods and scenarios, and extract the reference channels x1, x2, ..., x according to the designed overlap pixel values ​​between channels. n And the gray value sequence y1, y2, ..., y of the corresponding pixels of the channel to be corrected n ;

[0081] ②With the channel to be corrected as the x-axis and the reference channel as the y-axis, a quadratic function correction model is constructed as follows:

[0082] y = a0 + a1x + a2x 2

[0083] ③ Let the least squares objective function be set. The iterative optimization is used to calculate a0, a1, and a2 until Δ converges.

[0084] ④ Correcting the channel-corrected image using a quadratic function model includes: DN correct (i,j)=a0+a1DN(i,j)+a2DN(i,j) 2 DN(i,j) represents the response of pixel (i,j) in the image after channel correction. correct (i,j) represents the response of pixel (i,j) in the image after the overlap area correction, where i and j are the pixel row and column numbers.

[0085] ⑤ After calibration, this channel is transformed from the channel to be calibrated into a reference channel;

[0086] ⑥ Repeat steps ① through ⑤ to iterate through all channels to be calibrated.

[0087] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

[0088] The contents not described in detail in this specification are common knowledge to those skilled in the art.

Claims

1. A method for relative radiometric correction of a multi-channel stitched large-area infrared camera, characterized in that, include: Calculate the non-uniformity correction coefficient based on on-board blackbody calibration data; Acquire in-orbit images, correct the in-orbit images using the non-uniform correction coefficients, obtain corrected images, and apply the secondary correction coefficients and channel correction coefficients from laboratory calibration to the corrected images to obtain channel-corrected images. Accumulate a number of on-orbit images containing different time periods and scenes. According to the design value of overlapping pixels between on-orbit image channels, extract the pixel gray value sequences corresponding to the preset reference channel and the channel to be corrected as sample data. Use the sample data of the channel to be corrected and the reference channel as the x-axis and y-axis, respectively. Fit a quadratic function correction model between the x-axis coordinates and the y-axis coordinates using the least squares method. Then, perform correction on the channel-corrected image based on the overlapping area according to the quadratic function correction model. x and y are the independent variable and dependent variable, respectively.

2. The relative radiometric correction method for a multi-channel stitched large-area infrared camera according to claim 1, characterized in that, The quadratic function correction model is y = a0 + a1x + a2x 2 Let the least squares objective function be... a0, a1, and a2 are calculated iteratively until Δ converges; a0, a1, and a2 are correction coefficients, x is the gray value of the pixel in the channel to be corrected, y is the gray value of the pixel in the reference channel, n is the number of pixels in the overlapping area, and p is the pixel number in the overlapping area.

3. The relative radiometric correction method for a multi-channel stitched large-area infrared camera according to claim 1, characterized in that, The correction of the channel-corrected image using the quadratic function correction model includes: DN correct (i,j)=a0+a1DN(i,j)+a2DN(i,j) 2 DN(i,j) represents the response of pixel (i,j) in the image after channel correction. correnct (i,j) represents the response of pixel (i,j) in the image after secondary correction of the overlapping area, where i and j are the pixel row and column numbers, and a0, a1, and a2 are the correction coefficients.

4. The relative radiometric correction method for a multi-channel stitched large-area infrared camera according to claim 1, characterized in that, The laboratory calibration includes: For the temperature data of each on-board blackbody calibration point, calculate the average DN value of each pixel in the calibration area. and multi-frame average Where M, N, and F represent the number of pixels in the row and column directions of a single channel of the area array camera and the number of continuously acquired calibration data frames, respectively; DN t (i,j,f) represents the response of pixel (i,j) in the f-th frame of calibration data at temperature t. Multiple sets of data are obtained based on all calibration temperature points t = 1, 2, ..., K, where K ≥ 2, representing the number of temperature points. The correspondence; for cell (i,j), assume and The relationship y = f(x, α) is satisfied, where y represents x represents α is a polynomial coefficient, representing the non-uniform correction coefficient for pixel (i,j), using... Normalization is performed to construct the least squares objective function. Iterative optimization is used to calculate the non-uniform correction coefficient α for the on-board blackbody calibration in the laboratory until Δ converges, t0, t end These are the lowest and highest temperature points, respectively.

5. The relative radiometric correction method for a multi-channel stitched large-area infrared camera according to claim 4, characterized in that, The laboratory calibration also includes: Based on the non-uniformity correction coefficients for on-board blackbody calibration in the laboratory, relative radiation correction is performed on the data at various temperature points of the external blackbody calibration using the following formula: In the formula, α(i,j) is the non-uniformity correction coefficient for pixel (i,j), and DN 外 (i,j) These are the response values ​​of pixel (i,j) in the temperature data of each external blackbody calibration before and after correction.

6. The relative radiometric correction method for a multi-channel stitched large-area infrared camera according to claim 5, characterized in that, The non-uniform quadratic correction coefficients of the corrected outer blackbody images at different temperatures are obtained by least-squares linear fitting, which are used to correct the non-uniform correction coefficients during on-orbit operation.

7. The relative radiation correction method for a multi-channel stitched large-area infrared camera according to claim 5, characterized in that, The laboratory calibration also includes full array consistency correction based on the external blackbody calibration data of each channel after secondary correction, including: Calculate the mean value of each channel at different temperatures. s represents the channel number that makes up the area array camera, s = 1, 2, ... C, where C represents the number of channels; The response of channel s to pixel (i,j) in the external blackbody calibration data at calibration temperature t after secondary correction; Calculate the mean of all channels at different temperatures. Calculate the difference between the mean of each channel and the mean of all channels, and take the channel with the smallest difference as the reference channel and record it as the reference channel for on-orbit correction. Using the mean value of each temperature in the baseline channel as the target value, the mean values ​​of each channel are fitted with least squares to obtain the consistency correction coefficient of each channel.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 7.

9. A relative radiation correction device for a multi-channel stitched large-area infrared camera, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 7.

Images