Geometric calibration method and system for spaceborne tiled linear array sensor based on RPC model

Abstract

This invention provides a geometric calibration method and system for a spaceborne patched linear array sensor based on an RPC model. Using control points from a matched self-calibration field reference image as observations, and based on the current calibration parameters and an RPC fitted to these parameters, a geometric calibration model is constructed. Through step-by-step optimization of the absolute and relative geometric calibration of the patched CCDs with added angular resolution, high-precision on-orbit geometric calibration of the linear array patched sensor is achieved. This method eliminates the need to construct a complex rigorous geometric imaging model or process complex attitude and orbital auxiliary data, offering simplicity and convenience. The calibration parameters obtained from this technical solution are consistent with the rigorous geometric imaging model in the satellite image geometric processing chain, and the calculated results can be directly used to construct a high-precision rigorous geometric imaging model, conforming to the geometric processing flow of satellite remote sensing images.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of optical satellite remote sensing image processing, and relates to an on-orbit geometric calibration method and system for a satellite-borne split-line array sensor based on an RPC model. BACKGROUND

[0002] On-orbit geometric calibration of a sensor is a necessary link in on-orbit geometric processing of an optical remote sensing satellite. For a satellite-borne split-line array sensor, on-orbit geometric calibration needs to not only compensate for the absolute geometric distortion of each piece of CCD (Charge-coupled Device) detector, but also realize high-precision geometric splicing and registration of the models of each piece of detector. Since the geometric processing of a satellite remote sensing image is based on the construction of a strict geometric imaging model, the traditional on-orbit geometric calibration method constructs an on-orbit geometric calibration model based on the strict geometric imaging model, so as to ensure that the constructed sensor calibration model and the solved calibration parameters can directly compensate for various systematic geometric errors in the strict geometric imaging model. However, the strict geometric imaging model of an optical remote sensing satellite is extremely complex, needs multiple types of observation data such as the attitude, orbit and time of the satellite imaging system on orbit, and involves complex attitude-orbit-time auxiliary data processing and conversion between multiple coordinate systems. In addition, the construction of the strict geometric imaging model is closely related to the design of the satellite and the imaging sensor, and cannot be unified into a standard model, which leads to the fact that the traditional on-orbit calibration based on the strict geometric imaging model is not only complex in itself, but also needs a lot of time-consuming and laborious work to assist the completion of the calibration work.

[0003] In view of the problems of complex and various models and great calibration processing difficulty of the traditional on-orbit geometric calibration of a satellite-borne split-line array sensor based on a strict geometric imaging model, the application provides an on-orbit geometric calibration method for a satellite-borne split-line array sensor based on a prior RPC (Rational Polynomial Coefficient) model. The method takes the current calibration parameters and the RPC fitted based on the calibration parameters as a basic model, takes the ground control points as observation values to solve the calibration parameters, adopts a step-by-step optimization strategy of absolute and relative internal calibration to realize high-precision compensation, geometric splicing and registration of the internal distortion of the split-line sensor, and the solved calibration parameters can be directly used to compensate for the systematic geometric errors of the strict geometric imaging model. The method does not need to construct a complex strict geometric imaging model, and does not need to process complex attitude-orbit-time auxiliary data, and has the characteristics of simple and convenient processing. SUMMARY

[0004] The problem to be solved by the application is the on-orbit geometric calibration of a satellite-borne split-line array sensor.

[0005] The technical scheme of the application is a kind of satellite borne slice linear array sensor geometric calibration method based on RPC model, based on the control points matched with the self-calibration field reference image as observation values, based on the current calibration parameters and the RPC fitted based on the current calibration parameters, a geometric calibration model is constructed, through the absolute and relative geometric calibration step optimization of the slice CCD with additional angular resolution, the high-precision on-orbit geometric calibration of the linear array slice sensor is realized.

[0006] Furthermore, the method comprises the following steps,

[0007] Step 1, according to the area where the reference image is located, satellite remote sensing images used for sensor geometric calibration are selected, and dense distribution of homonymic points are matched as control points;

[0008] Step 2, a sensor element pointing angle calibration model facing the strict geometric imaging model processing is constructed, and the element pointing angles of the multi-piece linear array imaging detector CCD are introduced through a plurality of polynomial fittings, and then the calibration parameters to be solved are determined;

[0009] Step 3, before the calibration parameter solving, based on the correlation characteristics of the interior and exterior orientation element error parameters, an error correction model of the same order as the pointing angle model is introduced in the RPC model image side to correct the imaging model error of the image RPC, and then the control point observation value error coupled with the interior and exterior orientation element error is reflected, and the detection and elimination of the mis-matching gross error control points are carried out through the model error iteration correction;

[0010] Step 4, based on the error parameter characteristics of the sensor exterior orientation elements, an RPC exterior orientation error compensation model with additional translation and rotation transformation is constructed, the exterior orientation element error in the sensor imaging model is corrected by using the control points, and then the virtual image points reflecting the internal geometric error of the sensor are determined;

[0011] Step 5, taking the real image point and the virtual image point of each control point as observation values, taking the virtual image point and the real image point pointing angle equivalence as constraint conditions, the angle resolution is introduced on the basis of the element pointing angle model to construct a satellite sensor geometric calibration adjustment optimization model;

[0012] Step 6, based on the constructed calibration parameter adjustment optimization model, under the assistance of the angle resolution in the two directions along the CCD and perpendicular to the CCD, the row and column two-direction step optimization method is adopted, and the absolute geometric calibration of each CCD slice is carried out piece by piece, and then the absolute geometric distortion of each CCD slice is compensated;

[0013] Step 7, a certain CCD slice is selected as a reference slice, the relative geometric calibration of the non-reference slice relative to the reference slice is carried out based on the angle resolution exterior orientation result, and the relative position distortion of the non-reference slice CCD relative to the reference slice CCD is compensated slice by slice;

[0014] Step 8, according to the corrected relative position distortion, a deflection field distortion compensation model is constructed based on the satellite sensor imaging angle, and the compensation distortion of each CCD detector is introduced into the solved pointing angle model parameters through least square adjustment, and then the deflection field imaging distortion of the non-reference piece CCD caused by the along-track position translation is corrected.

[0015] Furthermore, the reference image is a high-precision digital orthographic image and a digital elevation model.

[0016] Furthermore, all the piece images select the same name points with dense distribution in the image row direction as control points.

[0017] Furthermore, the implementation mode of step 3 is,

[0018] Firstly, based on the control point object coordinates, the position of the object coordinates on the image is calculated according to the RPC of the image, and then the image positioning residual error between the position and the control point image point coordinates is obtained;

[0019] Then, the model error of the inner and outer orientation elements in the image positioning residual error is corrected by using a polynomial error correction model with the same order as the sensor calibration model;

[0020] According to the correlation constraint condition of the image residual error, the error correction model coefficients are calculated by using least square adjustment, and then the image point residual error after the model error of the inner and outer orientation elements is corrected, and the residual error directly reflects the matching error of the control points;

[0021] The mean and root mean square error of all the residual errors are calculated, and the gross error detection and elimination are performed;

[0022] After each gross error elimination, the error correction model coefficient estimation, gross error detection and elimination are performed again based on the remaining control points, until the model coefficients calculated in two consecutive times change less than a set threshold, and then the control point gross error detection and elimination are ended.

[0023] On the other hand, the application provides a satellite-borne piece linear array sensor geometric calibration system based on an RPC model, which is used to realize the satellite-borne piece linear array sensor geometric calibration method based on the RPC model.

[0024] Furthermore, the system comprises the following modules,

[0025] The first module is used to select satellite remote sensing images for sensor geometric calibration according to the area where the reference image is located, and match the same name points with dense distribution as control points;

[0026] The second module is used to construct a sensor element pointing angle calibration model facing a strict geometric imaging model processing, and introduce the element pointing angles of the multiple piece linear array imaging detectors CCD, and then determine the calibration parameters to be solved.

[0027] The third module is configured to introduce an error correction model of the same order as the pointing angle model into the RPC model image side based on the correlation characteristics of the interior and exterior orientation element error parameters before solving the calibration parameters, to correct the imaging model error of the image RPC, to reflect the control point observation value error coupled with the interior and exterior orientation element error, and to detect and eliminate the mismatched coarse error control points through model error iteration correction;

[0028] The fourth module is configured to construct an RPC exterior orientation error compensation model of additional translation and rotation transformation based on the sensor exterior orientation element error parameter characteristics, to correct the exterior orientation element error in the sensor imaging model by using the control points, and to determine the virtual image points reflecting the internal geometric error of the sensor.

[0029] The fifth module is configured to introduce the angle resolution into the satellite sensor geometric calibration adjustment optimization model based on the pointing angle model of the probe, to take the real image points and the virtual image points of each control point as observation values, and to take the equivalence of the virtual image points and the real image points as the constraint conditions.

[0030] The sixth module is configured to perform the absolute geometric calibration of each piece of CCD in a step-by-step optimization method in the row and column directions based on the constructed calibration parameter adjustment optimization model, and to compensate for the absolute geometric distortion of each piece of CCD under the assistance of the angle resolution in the two directions of the CCD and the vertical CCD.

[0031] The seventh module is configured to select a piece of CCD as a reference piece, to perform the relative geometric calibration of the non-reference pieces relative to the reference piece based on the angle resolution exterior orientation result, and to compensate for the relative position distortion of the non-reference piece CCD relative to the reference piece CCD piece by piece.

[0032] The eighth module is configured to construct a distortion compensation model of the partial field of view based on the imaging angle of the satellite sensor according to the corrected relative position distortion, and to compensate for the distortion of each CCD detector in the least square adjustment of the pointing angle model parameters, so as to correct the imaging distortion of the partial field of view in the CCD direction caused by the along-track position translation of the non-reference piece CCD.

[0033] Alternatively, the device comprises a processor and a memory, the memory is configured to store program instructions, and the processor is configured to call the stored instructions in the memory to execute the above-mentioned satellite-borne piecewise linear array sensor geometric calibration method based on the RPC model.

[0034] Alternatively, the device comprises a readable storage medium, and the readable storage medium stores a computer program, and the computer program is configured to execute the above-mentioned satellite-borne piecewise linear array sensor geometric calibration method based on the RPC model.

[0035] The advantages of this invention are as follows: it further unifies the on-orbit geometric calibration technology in the geometric processing of optical remote sensing satellite images into the RPC model. In actual calibration processing, only the current calibration parameters and the RPC fitted based on these calibration parameters are needed, without the need to generate additional attitude and orbital auxiliary data from the operational ground processing system. This avoids the complex auxiliary data processing and multiple coordinate system conversions required in the construction of a rigorous geometric imaging model, thus exhibiting high practicality. Furthermore, the calibration parameters obtained based on this technical solution are consistent with the fundamental model (rigorous geometric imaging model) in the satellite image geometric processing chain, and the solution results can be directly used to construct a high-precision rigorous geometric imaging model, conforming to the geometric processing flow of satellite remote sensing images. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating an embodiment of the present invention;

[0037] Figure 2 This is a schematic diagram of the pointing angle model of the linear array sensor element in an embodiment of the present invention;

[0038] Figure 3 This is a schematic diagram illustrating the relative geometric distortion and correction between CCDs in an embodiment of the present invention;

[0039] Figure 4 This is a schematic diagram of the field distortion correction in an embodiment of the present invention. Detailed Implementation

[0040] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings and examples.

[0041] The method of this invention uses control points matched with a self-calibrated field reference image as observations. Based on the current calibration parameters and the RPC fitted based on the current calibration parameters, a geometric calibration model is constructed. Through step-by-step optimization of absolute and relative geometric calibration of the sliced ​​CCD with additional angular resolution, high-precision on-orbit geometric calibration of the linear array sliced ​​sensor is achieved.

[0042] See Figure 1 The embodiment provides an on-orbit geometric calibration method for a spaceborne segmented linear array sensor based on an RPC model. The implementation process can be divided into 8 steps, as detailed below:

[0043] 1. Preparation of calibration data: Based on the area where the reference image (high-precision digital orthophoto and digital elevation model) is located, select satellite remote sensing images for sensor geometric calibration and match densely distributed corresponding points as control points.

[0044] Based on the location of the ground calibration field, the patch images acquired by the cloudless spaceborne patch linear array sensor are selected as the calibration scene images. A preferred method is to use a least-squares matching method based on triangulation constraints to match a certain number of densely corresponding image points from the calibration field reference images (Digital Orthophoto DOM and Digital Elevation Model DEM) as control point observations. To overcome the time-varying nonlinear differences between the fitted attitude and the actual attitude during RPC generation, it is preferred to perform dense control point matching on the same small segment (approximately 1000 rows) in the row direction of all patch images, thereby obtaining dense control point observations. The ground longitude, latitude, and elevation coordinates of the control points are denoted as (Lon, Lat, Hei), and the column and row numbers of the corresponding image point coordinates are denoted as (s, l).

[0045] 2. Construct a sensor calibration model based on the detector pointing angle. Construct a sensor detector pointing angle calibration model for rigorous geometric imaging model processing, and introduce multiple sets of polynomial fitting to the detector pointing angle of the multi-chip linear array imaging detector (Charge-coupled Device, CCD) to determine the calibration parameters to be solved.

[0046] One advantage of this invention is that the calculated calibration parameters can be directly used in a rigorous geometric imaging model. Therefore, the same pointing angle model as that used in the rigorous geometric imaging model is employed here as the sensor calibration model, as shown in the attached figure. Figure 2 As shown, the pointing angle model determines the pointing angle of each CCD sensor in two directions within the camera coordinate system O-XYZ. To determine the vector V of its light ray. image Considering the design characteristics of spaceborne linear array sensors—long focal length and narrow field of view—high-order distortion accounts for a limited proportion of all errors. Therefore, two univariate cubic polynomials with respect to the CCD detector number *s* (the detector number is the same as the image point column number) are used to describe the detector pointing (v) of each CCD detector in the imaging field of view. x ,v y ):

[0047]

[0048] In the formula, (a i ,b i (i = 0, 1, ..., 3) represents the coefficients of the probe pointing angle model. Each segmented CCD has an independent set of pointing angle model coefficients. Indicates the pointing angle along the CCD direction. This indicates the pointing angle perpendicular to the CCD direction.

[0049] 3. Gross error detection and removal of control point observations: Based on the relevant characteristics of error parameters of interior and exterior orientation elements, an error correction model of the same order as the pointing angle model is introduced to correct the imaging model error, and then the detection and removal of mismatched gross error control points are carried out iteratively.

[0050] Since this invention uses a large number of automatically matched control points as observations, and gross errors are unavoidable among these observations, it is necessary to detect and remove these gross errors to ensure the accuracy and reliability of the calibration solution. However, when calibrating based on the RPC model, the model errors of the sensor's internal and external orientation elements are intertwined and coupled with the gross errors of the observations, making it difficult to distinguish and detect them during the calibration solution. Therefore, this invention proposes that, before calculating the calibration parameters, based on the relevant characteristics of the error parameters of the internal and external orientation elements, an error correction model of equal order to the pointing angle model is introduced into the image side of the RPC model to correct the imaging model error of the image RPC. This reflects the control point observation errors that are intertwined and coupled with the errors of the internal and external orientation elements, and the detection and removal of mismatched gross error control points are achieved through iterative correction of the model error.

[0051] Based on the characteristic that the intrinsic and extrinsic parameters of a spaceborne sensor are fully correlated, this invention employs a gross error detection and elimination method based on prior compensation for model errors to improve the reliability of geometric calibration observations. First, based on the object coordinates (Lon, Lat, Hei) of the control point, the position of the object coordinates on the image is calculated according to the image's RPC, thus obtaining the image-side positioning residual (ds, dl) between this position and the image point coordinates (s, l), where s and l represent the column and row numbers of the image, respectively. Then, a polynomial error correction model (Δs, Δl) of equal order to the sensor calibration model is used to correct the model errors of the intrinsic and extrinsic orientation elements in the image-side positioning residual. The corrected image-side residual (v... s ,v l )as follows:

[0052]

[0053]

[0054] Among them, (ca i ,cb i (i = 0, 1, ... 3) are the coefficients of the polynomial error correction model.

[0055] by Minimum is the constraint condition, and least squares adjustment is used to calculate the error correction model coefficients. Then, according to equation (2), the image point residuals (v) after correcting the internal and external orientation element model errors are calculated. s ,v l This residual directly reflects the control point matching error. The mean of all residuals is calculated. vs ,meanvl ) and root mean square error (RMSE) vs ,rmse vl In this embodiment, a threshold of three times the root mean square error is preferably used for gross error detection and removal. The gross error removal conditions are as follows:

[0056] (|v s -mean vs |>3·rmse vs )||(|v l -mean vl |>3rmse vl (4)

[0057] After each outlier removal, the error correction model coefficients are estimated, outliers are detected and removed again based on the remaining control points, until the change in model coefficients between two consecutive solutions is less than the set threshold, at which point the control point outlier detection and removal ends.

[0058] 4. Based on the error correction of the external orientation element, an external orientation error compensation model is constructed based on the characteristics of the sensor's external orientation element error parameters. This model corrects the external orientation element error in the sensor's imaging model and determines virtual image points that can reflect the sensor's internal geometric errors.

[0059] Accurately separating the interior orientation element error from the exterior orientation element error is crucial for precise sensor calibration. This invention proposes constructing an RPC exterior orientation error compensation model with additional translation and rotation transformations. Based on the geometric characteristics of the exterior orientation element error parameters, an error correction model describing rotation and translation transformations is used for exterior orientation, thereby correcting the orientation error caused by the exterior orientation element in the RPC model. The exterior orientation model is constructed as follows:

[0060]

[0061] Among them, (v es ,v el ) is the control point image-side positioning residual after correcting for exterior orientation element errors, and (s0,l0) and θ are the translation and rotation transformation parameters, respectively.

[0062] For each CCD image, all control points on that image after gross error removal are used as observations. The minimum observation condition is used, and the least squares adjustment is employed to calculate the rotation and translation error correction parameters. Finally, based on the calculated transformation parameters, the corrected residuals obtained after external orientation are calculated using equation (5), thereby obtaining the virtual image point (s',l') that can express the interior orientation element error:

[0063]

[0064] 5. Based on the real and virtual image points of each control point, and by introducing angular resolution, a geometric calibration and adjustment optimization model for satellite sensors is constructed: taking the real and virtual image points of each control point as observations and the equivalence of the pointing angles of the virtual and real image points as constraints, an angular resolution model for satellite sensor geometric calibration and adjustment is constructed based on the detector pointing angle model.

[0065] This invention further proposes to perform sensor geometric calibration based on the real image point coordinates (s,l) and virtual image point (s',l') of each control point. The mathematical constraint for sensor calibration in this invention is based on the newly calculated calibration parameter gc. new It can compensate for the initial calibration parameter gc ori The inherent distortion leads to the following conditional equation for sensor calibration:

[0066]

[0067] Among them, v x (s,gc new ) and v y (s,gc new The new calibration parameters gc are respectively used to determine the calibration parameters gc. new The tangent function of the pointing angle along the CCD and perpendicular to the CCD, determined by equation (1) for the real image point s. x (s',gc ori ) and v y (s',gc ori ) are respectively derived from the initial calibration parameters gc ori The tangent function of the pointing angle of the virtual image point s' is determined according to equation (1) in two directions: along the CCD and perpendicular to the CCD.

[0068] However, since the calibration model based on the pointing angle is only related to the column number of the image point, it cannot reflect the sensor distortion in the row direction. To address this problem, this invention, based on the characteristic that the angular resolution of satellite sensor imaging in a local area perpendicular to the CCD direction is relatively stable, introduces an angular resolution v in this direction. ar_l To reconstruct the image point residual v in that direction in the model. el This leads to the final calibration parameter adjustment model (G). x G y ):

[0069]

[0070] in, and These are respectively derived from the initial calibration parameters gc ori The original pointing angle tangent function of the virtual image point s' determined according to equation (1) in two directions: along the CCD and perpendicular to the CCD. and The calibration parameters gc to be solved are respectively est The tangent function of the pointing angle along the CCD and perpendicular to the CCD, determined by equation (1) for the real image point s, is gc. ori =(oa i ,ob i ) and gc est =(na i ,nb i (i = 0, 1, ..., 3) represent the initial calibration parameters and the new calibration parameters to be solved, respectively. i ,ob i For the initial calibration parameters (initial pointing angle model coefficients), na i ,nb i These are the calibration parameters for the solution (the coefficients of the pointing angle model).

[0071] 6. Piecewise solution of CCD absolute calibration parameters based on least squares adjustment, based on the constructed calibration parameter adjustment optimization model.

[0072] Each CCD is individually calibrated to absolute geometric distortion to compensate for the absolute geometric distortion of each CCD.

[0073] In this embodiment, based on the constructed calibration parameter adjustment optimization model, with the assistance of angular resolution in both the CCD and perpendicular directions, a step-by-step optimization method in both row and column directions is adopted to perform absolute geometric calibration on each CCD, thereby compensating for the absolute geometric distortion of each CCD.

[0074] This invention prioritizes the least squares adjustment algorithm to calculate the calibration parameters of each CCD piece by piece. Since solving the calibration parameters in the row direction requires the angular resolution along the CCD direction determined by the column direction calibration parameters, a step-by-step optimization method is used to solve the adjustment equations in both directions of equation (8), and the column direction adjustment equations need to be solved first. Since both directions are solved using least squares adjustment, only the column direction calibration parameter solution is introduced here, as follows:

[0075] For the error equation constructed from the j-th pair of real and virtual image points, model linearization is used to construct its error equation:

[0076]

[0077] in, Let x be the residual of the observed values. iop These are the corrections for the calibration parameters. L is the coefficient matrix of the error equation obtained by linearizing with respect to the calibration parameters. jP is the residual vector determined based on the initial calibration parameters and the current calibration parameters obtained from the solution. j This is the corresponding weight matrix. To find G x About na i The partial derivatives, To find G x Regarding nb i The partial derivatives of .

[0078] Based on the principle of least squares adjustment, we can derive x iop The solution is:

[0079]

[0080] Where n is the number of control points. Least square adjustment is an iterative process. The current calibration parameters are updated based on the calculated calibration parameter corrections, and these corrections are used as input for the next least squares solution. The iterative calculation continues until the change in the calibration parameter corrections is less than a preset threshold, at which point the iterative solution ends.

[0081] Then, the angular resolution v along the CCD direction is obtained from the calibration parameters calculated in this direction. ar_s Furthermore, based on the equivalence relationship between the ratio of satellite image angular resolution to ground resolution, v ar_l / v ar_s =v gsd_l / v gsd_s Calculate the angular resolution v perpendicular to the CCD direction ar_l The ratio of ground resolution to v gsd_l / v gsd_s The calibration parameters can be determined from the RPC model of the image. Then, based on another adjustment equation in equation (8), the calibration parameters perpendicular to the CCD direction are calculated, thus completing the absolute geometric calibration of the CCD calibration parameters. The parameter solution still uses least squares adjustment optimization, which will not be elaborated here. Each CCD is calibrated individually to achieve the absolute geometric calibration of all CCDs in the imaging sensor.

[0082] 7. Correction of relative geometric calibration error of segmented CCDs: Select a CCD as the reference image and perform relative geometric calibration of the non-reference images relative to the reference image to compensate for the relative positioning distortion of the non-reference images relative to the reference image.

[0083] In this embodiment, a specific CCD is selected as the reference image. Based on the angular resolution out-of-orientation results, relative geometric calibration of the non-reference images relative to the reference image is performed, compensating for the relative positional distortion of the non-reference CCDs relative to the reference CCDs one by one. For sensors with a slab CCD design, geometric calibration needs to compensate for not only the absolute geometric distortion of each CCD but also the relative geometric distortion between CCDs. (See attached...) Figure 3As shown, S is the projection center of the CCD before relative positioning distortion correction, and O base -X base Y base This refers to the focal plane coordinate system referenced to the reference CCD. Since each CCD undergoes absolute geometric calibration based on its own external orientation, each CCD is essentially calibrated within its own focal plane coordinate system determined by its external orientation (e.g., ...). Figure 3 The O3-X3Y3 coordinate system of the CCD3 in the middle causes the calibration parameters of each image to be unable to achieve accurate geometric stitching and registration under a unified camera coordinate system.

[0084] Therefore, relative geometric calibration between the CCD slices is also required. Here, a specific CCD slice is selected as the reference CCD (e.g., ...). Figure 3 If the reference CCD is calibrated to CCD2, then there is a relative positioning distortion (dx, dy) between the absolute calibrated reference CCD and the non-reference CCD. The corresponding relative angular distortion can be expressed as (Δv). x ,Δv y Since the absolute translation (s0, l0) of each CCD has been calculated in the external orientation, and the corresponding angular resolution has also been calculated in the absolute calibration, the sensor calibration parameter correction amount of the k-th non-reference CCD relative to the reference CCD is... as follows:

[0085]

[0086] in, This refers to the relative angular distortion corresponding to the non-reference CCD; and These represent the external orientation translation amounts for the non-reference CCD and the reference CCD, respectively; v or_l The angular resolution remains perpendicular to the CCD direction. Since the angular resolution of the image is basically the same at different locations in this direction, a uniform v is used for both the reference CCD and non-reference CCDs. or_l ; and These are the angular resolutions of the non-reference CCD and the reference CCD along the CCD direction, respectively, determined by the calibration parameters calculated in that direction.

[0087] These are the absolute translations along the CCD and perpendicular to the CCD calculated for the k-th non-reference piece in the external orientation, respectively.

[0088] These represent the absolute translations along the CCD and perpendicular to the CCD, calculated for the reference piece in external orientation, respectively.

[0089] Finally, the calibration parameter correction amount is based on the solution of the k-th non-reference CCD. Update the calibration parameter gc solved in the absolute calibration. est The constant term (na0, nb0) k This can correct the relative positioning distortion (dx, dy) between the reference CCD and the non-reference CCD.

[0090] 8. Non-reference CCD field-of-view imaging distortion correction, including constructing a field-of-view distortion model based on the corrected relative positioning distortion, and correcting the field-of-view imaging distortion along the CCD direction caused by the translation of the non-reference CCD along the track position.

[0091] Because the position of the non-reference CCD is changed in the direction perpendicular to the CCD when correcting the relative positioning distortion between CCDs, its imaging mode is transformed into off-field imaging, thus introducing nonlinear distortion caused by the off-field along the CCD direction. In this embodiment, based on the corrected relative position distortion and the satellite sensor imaging angle, an off-field distortion compensation model is constructed. The compensation distortion of each CCD detector is incorporated into the calculated pointing angle model parameters through least squares adjustment, thereby correcting the off-field imaging distortion along the CCD direction caused by the translation of the non-reference CCD's position along the track. (See attached...) Figure 4 As shown, S remains the projection center of the CCD before relative positioning distortion correction, and O base -X base Y base Still using the focal plane coordinate system with the reference CCD as the reference, pitch is the elevation angle of the sensor relative to the nadir point, perpendicular to the CCD direction, f base Indicates line segment SO in the diagram base , is the principal distance of the reference piece, for Figure 4 The initial principal distance of the non-standard CCD is f. ccd Without distortion correction, the corresponding line segment SO ccd O ccd To correct the principal point of the reference image without distortion correction, the relative positional distortion dy (corresponding to the relative angular distortion Δv) in the direction perpendicular to the CCD was corrected based on the reference CCD. y After that, the imaging principal distance of the CCD changed from the initial f. ccd Transform into f base This is equivalent to the line segment SO ccd 'become SO base O ccd 'For SO base A virtual point on a line segment satisfies condition SO ccd '=SO ccd This will introduce a new field-of-view imaging distortion along the CCD direction, which will not only reduce the absolute calibration accuracy of the CCD image, but also reduce the relative calibration accuracy between CCD images.

[0092] To compensate for this distortion in the calibration parameters, this invention first constructs a distortion model based on the distortion generation mechanism. (See attached...) Figure 4 As shown, in the distortion-corrected imaging direction SO base If a virtual CCD with the same principal distance as the initial CCD is set up, then the positional distortion dx between the virtual and real CCDs in this imaging direction is... i (The corresponding angular distortion is dvx) i This refers to field-of-view imaging distortion. (The text abruptly ends here, likely due to an incomplete sentence or a formatting error.) Figure 4 Based on geometric relationships, the field-of-view distortion dvx of the t-th element of the k-th non-reference CCD is derived. t k for:

[0093]

[0094] in, The positional distortion of the t-th element of the k-th non-reference CCD. This represents the offset of the probe relative to the center position of the reference CCD along the CCD direction.

[0095] Furthermore, for the k-th non-reference CCD, the internal calibration parameters calculated at the current time can be used as a basis. Establish a mathematical model for field distortion compensation:

[0096]

[0097] in, and These are the calibration parameters currently calculated by the CCD and the new calibration parameters after compensating for field distortion; and These are the pointing angle models along the CCD direction calculated by equation (1), where Est is used to identify the parameters calculated by the absolute internal calibration above, and new is used to identify the final calibration result calculated by the relative internal calibration based on Est above; t = 0, 1…m k For the probe, m k This represents the number of detector elements for the non-reference CCD.

[0098] For each non-reference CCD, a set of observation equations can be established based on all its CCD elements according to equation (13), and new calibration parameters can be calculated using least squares adjustment. This compensates for the field-of-view distortion of each non-reference CCD, enabling high-precision geometric splicing and registration of the calibration parameters of each piece of the segmented sensor under the same external orientation geometric reference.

[0099] In specific implementation, the method proposed in the technical solution of this invention can be automatically executed by those skilled in the art using computer software technology. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of this invention and computer equipment including the computer program running the corresponding computer program, should also be within the protection scope of this invention.

[0100] In some possible embodiments, a geometric calibration system for a spaceborne segmented linear array sensor based on an RPC model is provided, including the following modules:

[0101] The first module is used to select satellite remote sensing images for sensor geometric calibration based on the area where the reference image is located, and match densely distributed corresponding points as control points.

[0102] The second module is used to construct a sensor element pointing angle calibration model for rigorous geometric imaging model processing, and introduces multiple sets of polynomial fitting to the element pointing angle of the multi-slice linear array imaging detector CCD, thereby determining the calibration parameters to be solved.

[0103] The third module is used to introduce an error correction model of the same order as the pointing angle model into the image side of the RPC model before the calibration parameter calculation, based on the relevant characteristics of the error parameters of the interior and exterior orientation elements, to correct the imaging model error of the image RPC. This reflects the control point observation error that is intertwined and coupled with the errors of the interior and exterior orientation elements, and detects and removes mismatched gross control points through iterative correction of model error.

[0104] The fourth module is used to construct an RPC external orientation error compensation model with additional translation and rotation transformation based on the error parameter characteristics of the sensor's external orientation elements. It uses control points to correct the external orientation element errors in the sensor imaging model, and then determines the virtual image points that reflect the internal geometric errors of the sensor.

[0105] The fifth module is used to construct a satellite sensor geometric calibration adjustment optimization model based on the probe pointing angle model, using the real and virtual image points of each control point as observations and the equivalence of the pointing angles of the virtual and real image points as constraints.

[0106] The sixth module is used to perform absolute geometric calibration of each CCD based on the constructed calibration parameter adjustment optimization model. With the assistance of angular resolution in both the direction along the CCD and the direction perpendicular to the CCD, the module adopts a step-by-step optimization method in both the row and column directions to compensate for the absolute geometric distortion of each CCD.

[0107] The seventh module is used to select a CCD as a reference image, perform relative geometric calibration of the non-reference images relative to the reference image based on the angular resolution external orientation results, and compensate for the relative positional distortion of the non-reference CCDs relative to the reference CCDs one by one.

[0108] The eighth module is used to construct a field-of-view distortion compensation model based on the satellite sensor imaging angle and the corrected relative position distortion. It also incorporates the compensation distortion of each CCD detector into the calculated pointing angle model parameters through least squares adjustment, thereby correcting the field-of-view imaging distortion along the CCD direction caused by the translation of the non-reference CCD along the track.

[0109] In some possible embodiments, a geometric calibration system for a spaceborne segmented linear array sensor based on an RPC model is provided, including a processor and a memory. The memory is used to store program instructions, and the processor is used to call the stored instructions in the memory to execute the geometric calibration method for a spaceborne segmented linear array sensor based on an RPC model as described above.

[0110] In some possible embodiments, a geometric calibration system for a spaceborne segmented linear array sensor based on an RPC model is provided, including a readable storage medium on which a computer program is stored. When the computer program is executed, it implements the geometric calibration method for a spaceborne segmented linear array sensor based on an RPC model as described above.

[0111] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A geometric calibration method for a spaceborne segmented linear array sensor based on an RPC model, characterized in that: Based on the control points of the matching self-calibrated field reference image as the observation value, and on the basis of the current calibration parameters and the RPC fitted based on the current calibration parameters, a geometric calibration model is constructed. Through step-by-step optimization of the absolute and relative geometric calibration of the piecewise CCD with additional angular resolution, high-precision on-orbit geometric calibration of the linear array piecewise sensor is achieved. Includes the following steps, Step 1: Select satellite remote sensing images for sensor geometric calibration based on the area where the reference image is located, and match densely distributed corresponding points as control points. Step 2: Construct a sensor element pointing angle calibration model for rigorous geometric imaging model processing, and introduce multiple sets of polynomial fitting to the element pointing angle of the multi-slice linear array imaging detector CCD, thereby determining the calibration parameters to be solved. Step 3: Before the calibration parameters are calculated, based on the relevant characteristics of the error parameters of the interior and exterior orientation elements, an error correction model of the same order as the pointing angle model is introduced into the image side of the RPC model to correct the imaging model error of the image RPC. This reflects the control point observation error that is intertwined and coupled with the errors of the interior and exterior orientation elements. The detection and elimination of mismatched gross control points are carried out through iterative correction of model error. Step 4: Based on the characteristics of the sensor's exterior orientation element error parameters, construct an RPC exterior orientation error compensation model with additional translation and rotation transformations. Use control points to correct the exterior orientation element errors in the sensor imaging model, and then determine the virtual image points that reflect the sensor's internal geometric errors. Step 5: Using the real and virtual image points of each control point as observations and the equivalence of the pointing angles of the virtual and real image points as constraints, an angular resolution is introduced to construct a geometric calibration adjustment optimization model for satellite sensors based on the detector pointing angle model. Step 6: Based on the constructed calibration parameter adjustment optimization model, with the assistance of angular resolution in both the CCD and perpendicular directions, a step-by-step optimization method in both row and column directions is adopted to perform absolute geometric calibration of each CCD piece by piece, thereby compensating for the absolute geometric distortion of each CCD piece. Step 7: Select a CCD as the reference film, perform relative geometric calibration of the non-reference film relative to the reference film based on the angular resolution external orientation results, and compensate for the relative positional distortion of the non-reference CCD relative to the reference CCD one by one. Step 8: Based on the corrected relative position distortion, construct a field-of-view distortion compensation model based on the satellite sensor imaging angle, and incorporate the compensation distortion of each CCD detector into the calculated pointing angle model parameters through least squares adjustment, thereby correcting the field-of-view imaging distortion along the CCD direction caused by the translation of the non-reference CCD position along the track.

2. The geometric calibration method for a spaceborne segmented linear array sensor based on the RPC model according to claim 1, characterized in that: The reference images are high-precision digital orthophotos and digital elevation models.

3. The geometric calibration method for a spaceborne segmented linear array sensor based on the RPC model according to claim 1, characterized in that: For all image segments, control points are selected from the same densely distributed points in the same row direction.

4. The geometric calibration method for a spaceborne segmented linear array sensor based on the RPC model according to claim 1, characterized in that: Step 3 is implemented as follows: First, based on the object coordinates of the control points, the position of the object coordinates on the image is calculated according to the RPC of the image, and then the image-space positioning residual between the position and the image point coordinates of the control points is obtained. Then, a polynomial error correction model of the same order as the sensor calibration model is used to correct the model error of the interior and exterior orientation elements in the image-side positioning residual; Based on the image residual related constraints, the least squares adjustment is used to calculate the error correction model coefficients, and then the image point residuals after correcting the internal and external orientation element model errors are calculated. These residuals directly reflect the control point matching errors. Calculate the mean and root mean square error of all residuals, and perform gross error detection and elimination; After each outlier removal, the error correction model coefficients are estimated, outliers are detected and removed again based on the remaining control points, until the change in model coefficients between two consecutive solutions is less than the set threshold, at which point the control point outlier detection and removal ends.

5. A geometric calibration system for a spaceborne segmented linear array sensor based on an RPC model, characterized in that: This method is used to implement the geometric calibration method for a spaceborne segmented linear array sensor based on the RPC model as described in any one of claims 1-4. Includes the following modules, The first module is used to select satellite remote sensing images for sensor geometric calibration based on the area where the reference image is located, and match densely distributed corresponding points as control points. The second module is used to construct a sensor element pointing angle calibration model for rigorous geometric imaging model processing, and introduces multiple sets of polynomial fitting to the element pointing angle of the multi-slice linear array imaging detector CCD, thereby determining the calibration parameters to be solved. The third module is used to introduce an error correction model of the same order as the pointing angle model into the image side of the RPC model before the calibration parameter calculation, based on the relevant characteristics of the error parameters of the interior and exterior orientation elements, to correct the imaging model error of the image RPC. This reflects the control point observation error that is intertwined and coupled with the errors of the interior and exterior orientation elements, and detects and removes mismatched gross control points through iterative correction of model error. The fourth module is used to construct an RPC external orientation error compensation model with additional translation and rotation transformation based on the error parameter characteristics of the sensor's external orientation elements. It uses control points to correct the external orientation element errors in the sensor imaging model, and then determines the virtual image points that reflect the internal geometric errors of the sensor. The fifth module is used to construct a satellite sensor geometric calibration adjustment optimization model based on the probe pointing angle model, using the real and virtual image points of each control point as observations and the equivalence of the pointing angles of the virtual and real image points as constraints. The sixth module is used to perform absolute geometric calibration of each CCD based on the constructed calibration parameter adjustment optimization model. With the assistance of angular resolution in both the direction along the CCD and the direction perpendicular to the CCD, the module adopts a step-by-step optimization method in both the row and column directions to compensate for the absolute geometric distortion of each CCD. The seventh module is used to select a CCD as a reference image, perform relative geometric calibration of the non-reference images relative to the reference image based on the angular resolution external orientation results, and compensate for the relative positional distortion of the non-reference CCDs relative to the reference CCDs one by one. The eighth module is used to construct a field-of-view distortion compensation model based on the satellite sensor imaging angle and the corrected relative position distortion. It also incorporates the compensation distortion of each CCD detector into the calculated pointing angle model parameters through least squares adjustment, thereby correcting the field-of-view imaging distortion along the CCD direction caused by the translation of the non-reference CCD along the track.

6. A geometric calibration system for a spaceborne segmented linear array sensor based on an RPC model, characterized in that: It includes a processor and a memory, the memory being used to store program instructions, and the processor being used to call the stored instructions in the memory to execute the geometric calibration method for a spaceborne segmented linear array sensor based on the RPC model as described in any one of claims 1-4.

7. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which, when executed, implements a geometric calibration method for a spaceborne segmented linear array sensor based on an RPC model as described in any one of claims 1-4.

Images