Aerial hyperspectral line array combined large field of view camera inner parameter calibration method and system

Abstract

The application provides an aviation hyperspectral line array combined large field of view camera inner parameter calibration method and system, which is used for combining the structural characteristics of a combined line array camera, virtually forming a single camera multi-line array structure from a multi-camera multi-line array, and using the inner parameters of the virtual camera to represent the whole imaging system to realize calibration, and the combined line array camera is obtained by combining and installing three line array cameras. The application equivalently forms a virtual camera from the three cameras, does not consider the specific physical meaning of each geometric distortion, comprehensively describes the influence of various geometric distortions by using an empirical model, calculates the comprehensive influence as the offset of the coordinates of each CCD probe element in the focal plane coordinate system, then creates a digital calibration field by using the data of a high-resolution aviation camera with known parameters, and obtains the control point cloud by matching, so that the application is free from the dependence on a high-precision ground calibration field, effectively reduces the calibration cost, and improves the calibration precision and reliability.

Description

Technical Field

[0001] This invention belongs to the field of aerial photogrammetry, and specifically relates to a method and system for calibrating the intrinsic parameters of an aerial hyperspectral linear array combined large field-of-view camera. Background Technology

[0002] Hyperspectral imagery possesses rich spectral information and has wide applications in agriculture, vegetation monitoring, and environmental monitoring. Currently, most mainstream hyperspectral cameras employ linear scan imaging. Linear scan cameras often offer higher spatial resolution than frame-type cameras and are therefore widely used in photogrammetry and 3D reconstruction. To maximize the field of view, a popular approach is to combine multiple linear CCD elements and then stitch the images together. Mainstream combined linear scan cameras fall into two main categories: one arranges linear CCDs in parallel and staggered positions on the focal plane, and the other combines multiple cameras in a single mounting configuration.

[0003] GFHK is a line scan camera imaging system consisting of three line scan cameras mounted together. The three sub-cameras are mounted as follows: Figure 1 As shown, the center points of the three cameras are denoted as O. R O M O L For example, O R The linear CCD of the camera is denoted as CCD. R The three cameras were positioned so that their principal optical axes intersected at a point P. During aerial photography, all three cameras captured images simultaneously.

[0004] For combined line-scan cameras, camera calibration is fundamental to photogrammetry, and the accuracy of the calibration results significantly impacts subsequent measurements and 3D reconstruction. Two common methods for calibrating combined line-scan cameras are: one is to calibrate each line-scan CCD separately and then stitch the geometrically corrected images together; the second method involves constructing a virtual line array, assuming that rotation and translation between the line-scan CCDs are part of systematic errors such as lens distortion, CCD installation errors, and atmospheric refraction. Then, based on a rigorous geometric imaging model, a coordinate mapping relationship is established between the image points of the original image and the stitched image, thus achieving camera calibration. However, these methods only consider the characteristics of multiple CCDs in a single camera, primarily addressing the stitching problem of images from multiple CCDs in a single camera, and are not suitable for imaging systems composed of multiple cameras. Furthermore, the small overlap area of ​​the three sub-line array images in the GFHK, and the lack of POS data from the left and right sub-cameras, further complicates the calibration process.

[0005] To address the aforementioned technical issues, this invention proposes a method for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined large field-of-view camera, taking into account the structural characteristics of the GFHK camera. Instead of calibrating each of the three sub-cameras separately and then stitching them together, this method treats the three cameras as a single virtual camera. It disregards the specific physical meaning of each geometric distortion and employs a mathematically-based empirical model to comprehensively describe the influence of various geometric distortions. This comprehensive influence is then attributed to the offset of each CCD element's coordinates in the focal plane coordinate system. Finally, a digital calibration field is created using high-resolution airborne camera data with known precise orientation parameters. Control point clouds are obtained through matching, thus eliminating reliance on high-precision ground calibration fields, effectively reducing calibration costs, and improving calibration accuracy and reliability. Summary of the Invention

[0006] This invention proposes a method for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined large field-of-view camera. This method, taking into account the structural characteristics of the GFHK camera, virtualizes a multi-camera, multi-linear array into a single-camera, multi-linear array structure, using the intrinsic parameters of this virtual camera to characterize the entire imaging system. It then utilizes a rigorous geometric model based on the virtual camera to perform efficient and high-precision stitching of images from the sub-cameras, treating image distortion and geometric transformation parameters as parameters to be solved for calibration, and incorporating them into the overall calibration model. Finally, based on cloud control theory, it constructs a cloud control digital calibration field using images with known orientation parameters to solve for the calibration parameters of each detector element, thus completing the camera calibration.

[0007] The technical solution of this invention is a method for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined large field-of-view camera. This method utilizes the structural characteristics of a combined linear array camera to virtually transform a multi-camera, multi-linear array into a single-camera, multi-linear array structure. The intrinsic parameters of this virtual camera characterize the entire imaging system for calibration. The combined linear array camera is obtained by assembling three linear array cameras. The implementation process includes the following steps:

[0008] Step 1: Use a combined line array camera to acquire images of a certain survey area. For each flight strip, stitch together the images acquired by the three sub-cameras into a whole image according to the camera design parameters. This is regarded as the push-broom imaging result of a virtual line array camera. The initial interior orientation parameters of the virtual camera are calculated based on the laboratory orientation parameters of a single camera and the combination relationship of the three cameras.

[0009] Step 2: The POS of the middle camera in the combined line array camera is used as the POS of each scan line in the stitched image. Using the initial interior orientation parameters and POS data of the virtual camera, and setting the ground elevation of the survey area, the push-broom imaging result image of the virtual line array camera in Step 1 is geometrically corrected to initially eliminate the geometric deformation of the image caused by instability during flight. The generated image is called L1 image.

[0010] Step 3: Use calibrated linear or area array cameras to acquire images in the same survey area as reference images, perform tie point matching and aerial triangulation processing, and adjust and calculate the interior and exterior orientation parameters of all reference images.

[0011] Step 4: Extract a large number of dense Harris feature points from the L1 image of each flight strip in Step 2. Use point matching to match the extracted Harris feature points with the reference image. Select the flight strip L1 image with the most successful matching points for subsequent cloud control calibration processing.

[0012] Step 5: Using the interior and exterior orientation elements of the reference image, and based on the principle of forward intersection, solve for the ground coordinates of the matching point as the cloud control point;

[0013] Step 6: Using the cloud control points and the initial interior orientation parameters of the virtual camera, solve for the exterior orientation elements of each scan line of the linear array image using bundle adjustment.

[0014] Step 7: Calculate the reprojection error of each feature point in the linear array image in the row direction and column direction of the image, respectively, where the row direction is denoted as the x direction and the column direction is denoted as the y direction;

[0015] Step 8: Establish a polynomial model to perform curve fitting on the reprojection errors in the x and y directions respectively, calculate the interior orientation parameters of all probes based on the fitted residual curves, and complete the calibration of the linear array camera.

[0016] Furthermore, the implementation process of step 3 is as follows:

[0017] Step 2.1: Construct a rigorous geometric imaging model using the camera's initial design parameters and the POS during flight.

[0018]

[0019] In the formula, This represents the target's position vector in the Earth-centered, Earth-fixed coordinate system. This represents the platform's position in the Earth-centered Earth-fixed coordinate system, where λ represents the scale factor. This represents the rotation matrix from the sensor coordinate system to the platform coordinate system. This represents the rotation matrix from the platform to the inertial coordinate system. Then, it represents the rotation matrix from the inertial coordinate system to the geocentric coordinate system, f is the camera focal length, (x,y) represents the focal plane coordinates of the target point. Since the linear array image CCD is placed in a straight line, x equals 0 in the formula, and y is obtained according to the image point coordinates and the size of the probe; (Δx,Δy) represents the comprehensive influence of various factors on the image point coordinates.

[0020] Step 2.2: Given the average elevation of the ground in the survey area, calculate the coordinates of each pixel on the average elevation surface according to the correspondence of the rigorous geometric imaging model in Equation (1) for each pixel in the stitched overall image, and generate the L1 image according to the orthophoto conversion method to complete the geometric correction of the linear array image.

[0021] Furthermore, step 6, which calculates the exterior orientation element of each scan line, is implemented as follows:

[0022] Step 6.1: Calculate the focal plane coordinates of the feature points in the L1 image from Step 2.2 using back projection.

[0023] Step 6.2: Based on the rigorous geometric imaging model, establish the error equation. Using the POS of each scan line as the initial value, iteratively calculate the correction value of the exterior orientation element of the scan line until the correction value of the exterior orientation element is less than the threshold, at which point the exterior orientation element calculation is considered complete.

[0024] Furthermore, the formula for calculating the reprojection error in step 7 is as follows:

[0025]

[0026] In the formula, Δx and Δy are the reprojection errors of each feature point in the linear array image in the x and y directions, respectively, and (X,Y,Z) are the ground coordinates corresponding to the feature point, which are obtained by forward intersection of the reference image in step 5.

[0027] Furthermore, step 8, which involves solving for the interior orientation parameters of all elements of the linear array camera using a polynomial model, is implemented as follows:

[0028] Step 8.1, establish the polynomial model of the reprojection error between the detector element number and the corresponding image point:

[0029]

[0030] In the formula, u i ,v j ) represents the polynomial coefficients that need to be solved for camera calibration, Δx and Δy are the reprojection errors of each feature point in the linear array image in the x and y directions, and s is the detector element number, which is determined by the detector element in which the image point is located. Each detector element corresponds to a unique detector element number. When fitting curves, the larger i and j are, the more complex the curve is.

[0031] Step 8.2: Substitute the reprojection errors of each image point obtained in Step 6 into the polynomial model, fit the optimal curve, and calculate the polynomial coefficients u. i ,v j ;

[0032] Step 8.3: Substitute the detector number into the polynomial to calculate the offset of each detector on the focal plane, correct the focal plane coordinates of each detector, and complete the calibration of the linear array camera.

[0033] The present invention also provides an intrinsic parameter calibration system for an airborne hyperspectral linear array combined large field-of-view camera, used to implement the intrinsic parameter calibration method for an airborne hyperspectral linear array combined large field-of-view camera as described above.

[0034] Moreover, it includes the following modules,

[0035] The first module is used to acquire images of a certain survey area using a combined line array camera. For each flight strip, the images acquired by the three sub-cameras are stitched together into a whole image according to the camera design parameters, which is regarded as the push-broom imaging result of a virtual line array camera. The initial interior orientation parameters of the virtual camera are calculated based on the laboratory interior orientation parameters of a single camera and the combination relationship of the three cameras.

[0036] The second module is used to take the POS of the middle camera in the combined line array camera as the POS of each scan line in the stitched image, use the initial interior orientation parameters and POS data of the virtual camera, and set the ground elevation of the survey area to perform geometric correction on the push-broom imaging result image of the virtual line array camera in the first module, and initially eliminate the geometric deformation of the image caused by the instability during flight. The generated image is called L1 image.

[0037] The third module is used to acquire images in the same survey area using calibrated linear or area array cameras as reference images, perform tie point matching and aerial triangulation processing, and adjust and calculate the interior and exterior orientation parameters of all reference images.

[0038] The fourth module is used to extract a large number of dense Harris feature points from the L1 image of each flight strip in the second module. The extracted Harris feature points are matched with the reference image using the point matching method. The flight strip L1 image with the most successful matching points is selected for subsequent cloud control calibration processing.

[0039] The fifth module is used to use the interior and exterior orientation elements of the reference image and the principle of forward intersection to solve for the ground coordinates of the matching point as the cloud control point.

[0040] The sixth module is used to solve the exterior orientation elements of each scan line of the linear array image by using cloud control points and the initial interior orientation parameters of the virtual camera and the bundle adjustment method.

[0041] The seventh module is used to calculate the reprojection error of each feature point in the linear array image in the row direction and column direction of the image, where the row direction is denoted as the x direction and the column direction is denoted as the y direction.

[0042] The eighth module is used to establish a polynomial model to perform curve fitting on the reprojection errors in the x and y directions, calculate the interior orientation parameters of all probes based on the fitted residual curves, and complete the calibration of the linear array camera.

[0043] Alternatively, it may include a processor and a memory, with the memory used to store program instructions and the processor used to call the stored instructions in the memory to execute the intra-parameter calibration method for an airborne hyperspectral linear array combined large field-of-view camera as described above.

[0044] Alternatively, it may include a readable storage medium storing a computer program that, when executed, implements the method for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined large field-of-view camera as described above.

[0045] Compared with existing technologies, this invention has the following advantages: First, this invention treats three cameras as a single virtual camera, disregarding the specific physical meaning of each geometric distortion. It employs a mathematically meaningful empirical model to comprehensively describe the effects of various geometric distortions, and attributes this comprehensive effect to the offset of each CCD element's coordinates in the focal plane coordinate system. Compared with the traditional method of calibrating and stitching images after individual camera calibration, the algorithm is simpler, the results are more reliable, and the method is more robust. Second, this method proposes using aerial cameras with known parameters to acquire image data, creating a digital calibration field, and obtaining control point clouds through automatic matching. This eliminates the reliance on high-precision ground calibration fields, significantly reducing camera calibration costs and improving calibration accuracy and reliability. Attached Figure Description

[0046] Figure 1 This is the installation method of three sub-cameras in the combined line scan camera used in the embodiment of the present invention.

[0047] Figure 2 This is a technical flowchart of an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram of the focal plane coordinate system of the virtual camera in an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram of the cloud control principle in an embodiment of the present invention.

[0050] Figure 5 This is a schematic diagram of the projection difference and curve fitting results of the focal plane in the x and y directions before and after calibration according to an embodiment of the present invention. Part a is a schematic diagram of the projection difference in the x direction before calibration, part b is a schematic diagram of the projection difference in the y direction before calibration, part c is a schematic diagram of the projection difference in the x direction after calibration, and part d is a schematic diagram of the projection difference in the y direction after calibration. Detailed Implementation

[0051] This invention provides a method for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined wide field-of-view camera. This method, taking into account the structural characteristics of the GFHK camera, virtualizes a multi-camera, multi-linear array into a single-camera, multi-linear array structure, using the intrinsic parameters of this virtual camera to characterize the entire imaging system. It then utilizes a rigorous geometric model based on the virtual camera to perform efficient and high-precision stitching of images from the sub-cameras, treating image distortion and geometric transformation parameters as parameters to be solved for calibration, and incorporating them into the overall calibration model. Finally, based on cloud control theory, it uses images with known orientation parameters to construct a cloud control digital calibration field to solve for the calibration parameters of each detector element, thus completing the camera calibration.

[0052] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0053] like Figure 2 As shown, the procedure for calibrating the intrinsic parameters of an airborne hyperspectral linear array combined large field-of-view camera provided by this embodiment of the invention includes the following steps:

[0054] Step 1: The combined line array camera used in this embodiment is the GFHK, which is composed of three line array cameras assembled together. Images of a specific survey area are acquired using the combined line array camera. The installation and imaging method of the combined line array camera is as follows: Figure 1 As shown. For each flight strip (the linear array camera uses pushbroom imaging, and the image acquired in one photograph is in strip shape, which is called the flight strip), the images acquired by the three sub-cameras are stitched together into a whole image according to the camera design parameters, which is regarded as the pushbroom imaging result of a virtual linear array camera. The initial interior orientation parameters of the virtual camera are calculated based on the laboratory orientation parameters of a single camera and the combination relationship of the three cameras.

[0055] Step 2: Using the POS of the middle camera in the combined line array camera as the POS of each scan line in the stitched image, and using the initial interior orientation parameters and POS data of the virtual camera, and setting the ground elevation of the survey area, the push-broom imaging result image of the virtual line array camera in Step 1 is geometrically corrected to initially eliminate the geometric distortion of the image caused by instability during flight.

[0056] In this embodiment, the image with distortion corrected using POS is referred to as the L1 image. In this embodiment of the invention, the specific steps of step 2 are as follows:

[0057] Step 2.1: Construct a rigorous geometric imaging model using the camera's initial design parameters and the POS during flight.

[0058]

[0059] In the formula, This represents the target's position vector in the Earth-centered, Earth-fixed coordinate system. This represents the platform's position in the Earth-centered Earth-fixed coordinate system, where λ represents the scale factor. This represents the rotation matrix from the sensor coordinate system to the platform coordinate system. This represents the rotation matrix from the platform to the inertial coordinate system. This represents the rotation matrix from the inertial coordinate system to the geocentric coordinate system, f is the camera focal length, and (x, y) represents the focal plane coordinates of the target point. The focal plane coordinate system is as follows: Figure 3 As shown. Since the linear array CCD is placed in a straight line, x equals 0 in the formula, and y can be obtained from the image point coordinates and the detector size; (Δx, Δy) represents the combined influence of various factors on the image point coordinates.

[0060] Step 2.2: Given the average elevation of the ground in the survey area, calculate the coordinates of each pixel on the average elevation surface according to the correspondence of the rigorous geometric imaging model in Equation (1) for each pixel in the stitched overall image, and generate the L1 image according to the orthophoto conversion method to complete the geometric correction of the linear array image.

[0061] Step 3: Use calibrated linear or area array cameras to acquire images in the same survey area as reference images, perform tie point matching and aerial triangulation processing, and adjust and calculate the interior and exterior orientation parameters of all reference images.

[0062] Step 4: Extract a large number of dense Harris feature points from the L1 image of each flight strip in Step 2. Use correlation coefficient and other point matching methods to match the extracted Harris feature points with the reference image. Select the flight strip virtual orthophoto image with the most successful matches for subsequent cloud control calibration processing. The cloud control principle is as follows: Figure 4 As shown.

[0063] Step 5: Using the interior and exterior orientation elements of the reference image, and based on the principle of forward intersection, solve for the ground coordinates of the matching point as the cloud control point.

[0064] Step 6: Using the cloud control points and the initial interior orientation parameters of the virtual camera, solve for the exterior orientation elements of each scan line of the linear array image using bundle adjustment.

[0065] In this embodiment of the invention, step 6 is implemented as follows:

[0066] Step 6.1: Calculate the focal plane coordinates of the feature points in the L1 image from Step 2.2 using back projection.

[0067] Step 6.2: Based on the rigorous geometric imaging model, establish the error equation. Using the POS of each scan line as the initial value, iteratively calculate the correction value of the exterior orientation element of the scan line until the correction value of the exterior orientation element is less than the corresponding preset threshold, at which point the exterior orientation element calculation is considered complete.

[0068] Step 7: Calculate the reprojection error of each feature point in the linear array image in the row direction (x-direction) and column direction (y-direction). The formula for calculating the reprojection error is as follows:

[0069]

[0070] In the formula, Δx and Δy are the reprojection errors of each feature point in the linear array image in the x and y directions, respectively, and (X,Y,Z) are the ground coordinates corresponding to the feature point, which are obtained by forward intersection of the reference image in step 5.

[0071] Step 8: Establish a polynomial model to perform curve fitting on the reprojection errors in the x and y directions respectively. Based on the fitted residual curves, the interior orientation parameters of all probe elements can be calculated, and the calibration of the linear array camera can be completed.

[0072] Step 8.1, establish the polynomial model of the reprojection error between the detector element number and the corresponding image point:

[0073]

[0074] In the formula, u i ,v j (i,j≤5) represents the polynomial coefficients that need to be solved for camera calibration. Δx and Δy are the reprojection errors of each feature point in the linear array image in the x and y directions, respectively. s is the detector element number, determined by the detector element to which the image point belongs; each detector element corresponds to a unique detector element number. During curve fitting, the larger i and j are, the more complex the curve. The curve fitting result is as follows: Figure 5 As shown.

[0075] Step 8.2: Substitute the reprojection errors of each image point obtained in Step 6 into Equation (3), fit the optimal curve, and calculate all polynomial coefficients u. i ,v j .

[0076] Step 8.3: Substitute the detector number into the polynomial to calculate the offset of each detector on the focal plane, correct the focal plane coordinates of each detector, and complete the calibration of the linear array camera.

[0077] 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.

[0078] In some possible embodiments, an intrinsic parameter calibration system for an airborne hyperspectral linear array combined large field-of-view camera is provided, comprising the following modules:

[0079] The first module is used to acquire images of a certain survey area using a combined line array camera. For each flight strip, the images acquired by the three sub-cameras are stitched together into a whole image according to the camera design parameters, which is regarded as the push-broom imaging result of a virtual line array camera. The initial interior orientation parameters of the virtual camera are calculated based on the laboratory interior orientation parameters of a single camera and the combination relationship of the three cameras.

[0080] The second module is used to take the POS of the middle camera in the combined line array camera as the POS of each scan line in the stitched image, use the initial interior orientation parameters and POS data of the virtual camera, and set the ground elevation of the survey area to perform geometric correction on the push-broom imaging result image of the virtual line array camera in the first module, and initially eliminate the geometric deformation of the image caused by the instability during flight. The generated image is called L1 image.

[0081] The third module is used to acquire images in the same survey area using calibrated linear or area array cameras as reference images, perform tie point matching and aerial triangulation processing, and adjust and calculate the interior and exterior orientation parameters of all reference images.

[0082] The fourth module is used to extract a large number of dense Harris feature points from the L1 image of each flight strip in the second module. The extracted Harris feature points are matched with the reference image using the point matching method. The flight strip L1 image with the most successful matching points is selected for subsequent cloud control calibration processing.

[0083] The fifth module is used to use the interior and exterior orientation elements of the reference image and the principle of forward intersection to solve for the ground coordinates of the matching point as the cloud control point.

[0084] The sixth module is used to solve the exterior orientation elements of each scan line of the linear array image by using cloud control points and the initial interior orientation parameters of the virtual camera and the bundle adjustment method.

[0085] The seventh module is used to calculate the reprojection error of each feature point in the linear array image in the row direction and column direction of the image, where the row direction is denoted as the x direction and the column direction is denoted as the y direction.

[0086] The eighth module is used to establish a polynomial model to perform curve fitting on the reprojection errors in the x and y directions, calculate the interior orientation parameters of all probes based on the fitted residual curves, and complete the calibration of the linear array camera.

[0087] In some possible embodiments, an intrinsic parameter calibration system for an airborne hyperspectral linear array combined large field-of-view camera 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 intrinsic parameter calibration method for an airborne hyperspectral linear array combined large field-of-view camera as described above.

[0088] In some possible embodiments, an intrinsic parameter calibration system for an airborne hyperspectral linear array combined large field-of-view camera is provided, including a readable storage medium on which a computer program is stored. When the computer program is executed, it implements the intrinsic parameter calibration method for an airborne hyperspectral linear array combined large field-of-view camera as described above.

[0089] 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 method for calibrating inner parameters of an aerial hyperspectral line array combined large field of view camera, characterized in that: The application relates to a structure feature for combining a combined linear array camera, a multi-camera multi-linear array virtual single-camera multi-linear array structure, and a virtual camera internal parameter for representing the whole imaging system to realize calibration, wherein the combined linear array camera is obtained by combining and installing three linear array cameras; the realization process comprises the following steps, Step 1, collecting images of a certain survey area by using the combined linear array camera, for each flight strip, the images obtained by the three sub-cameras are spliced into an integral image according to the camera design parameters, the integral image is regarded as the push-broom imaging result of a virtual linear array camera, and the initial internal orientation parameters of the virtual camera are calculated according to the laboratory internal orientation parameters of a single camera and the combined relationship of the three cameras; Step 2, taking the POS of the middle camera of the combined linear array camera as the POS of each scanning line of the spliced image, using the initial internal orientation parameters of the virtual camera and the POS data, and setting the ground average height of the survey area, the push-broom imaging result image of the virtual linear array camera in step 1 is geometrically corrected to preliminarily eliminate the image geometric deformation caused by the instability during flight, and the generated image is called L1 image; Step 3, collecting images in the same survey area by using a calibrated linear array or area array camera as reference images, performing connection point matching and aerial triangulation processing, and adjusting and solving the internal and external orientation parameters of all the reference images; Step 4, extracting a large number of dense Harris feature points from the L1 image of each flight strip in step 2, matching the extracted Harris feature points with the reference images by using the point matching mode, and selecting the L1 image of the flight strip with the most matched points for subsequent cloud control calibration processing; Step 5, using the internal and external orientation elements of the reference images, solving the ground coordinates of the matched points as cloud control points according to the forward intersection principle; Step 6, using the cloud control points and the initial internal orientation parameters of the virtual camera, solving the external orientation elements of each scanning line of the linear array image according to the bundle adjustment; Step 7, respectively calculating the re-projection errors of each feature point in the linear array image in the image row direction and the image column direction, wherein the image row direction is denoted as the x direction, and the image column direction is denoted as the y direction; Step 8, establishing a polynomial model to curve-fit the re-projection errors in the x and y directions, calculating the internal orientation parameters of all the pixels according to the fitted residual curve, and completing the calibration of the linear array camera.

2. The method of claim 1, wherein the method comprises: The realization process of step 3 is as follows, Step 2.1, constructing a rigorous geometric imaging model by using the initial design parameters of the camera and the POS during flight; wherein, represents the position vector of the target in the earth-fixed coordinate system, represents the position of the platform in the earth-fixed coordinate system, and λ represents a scale factor, represents a rotation matrix of the sensor coordinate system to the platform coordinate system, represents a rotation matrix of the platform to the inertial coordinate system, represents a rotation matrix of the inertial coordinate system to the earth-fixed coordinate system, f is the focal length of the camera, and (x, y) represents the focal plane coordinates of the target point. Since the linear array image CCD is placed in a straight line, x is equal to 0 in the formula, and y is obtained according to the image point coordinates and the size of the probe element; (Δx, Δy) represents the comprehensive influence of various factors on the image point coordinates; Step 2.2, giving the ground average height of the survey area, calculating the coordinates of each pixel on the average height plane according to the corresponding relationship of the rigorous geometric imaging model of formula (1) in the spliced integral image, and generating the L1 image according to the orthographic projection conversion mode, so as to complete the geometric correction of the linear array image.

3. The method of claim 2, wherein the method further comprises: determining the internal parameters of the hyperspectral line array camera based on the first and second sets of images. The realization process of step 6 is as follows, Step 6.1, calculating the focal plane coordinates of the feature points in the L1 image in step 2.2 in the form of back projection; Step 6.2, according to the rigorous geometric imaging model, an error equation is established, taking the POS of each scanning line as the initial value, iteratively solving the exterior orientation element correction value of the scanning line, until the exterior orientation element correction value is less than the threshold value, that is, the exterior orientation element is considered to be solved.

4. The method of claim 3, wherein: The calculation formula for calculating the re-projection error in step 7 is as follows: In the formula, Δx, Δy is the re-projection error of each feature point in the linear array image in the x and y directions, and (X, Y, Z) is the ground point coordinates corresponding to the feature point, which is obtained by forward intersection of the reference image in step 5.

5. The method of claim 4, wherein: Step 8, according to the polynomial model, the interior orientation parameters of all elements of the linear array camera are solved, and the steps are as follows, Step 8.1, a polynomial model of the element serial number and the corresponding image point re-projection error is established: wherein u i ,v j are polynomial coefficients to be solved for camera calibration, Δx, Δy are the re-projection errors of each feature point in the linear array image in the x and y directions, s is the detector element number determined by the detector element where the point is located, and each detector element corresponds to a unique detector element number; the greater i and j are, the more complex the curve is during curve fitting; Step 8.2, the re-projection error of each pixel obtained in step 6 is brought into the polynomial model, the best curve is fitted, and the polynomial coefficient u i j is obtained​ Step 8.3, the element serial number is substituted into the polynomial to obtain the offset of each element on the focal plane, and the focal plane coordinates of each element are corrected to complete the calibration of the linear array camera.

6. An airborne hyperspectral line array combined large field of view camera interior parameter calibration system, characterized in that: A method for calibrating the interior parameters of an aerial hyperspectral linear array combined large field of view camera according to any one of claims 1-5.

7. The system of claim 6, wherein: Comprise the following modules, The first module is used for collecting images of a certain survey area by the combined linear array camera. For each flight strip, the images obtained by the three sub-cameras are spliced into an overall image according to the camera design parameters, which is regarded as the push-broom imaging result of a virtual linear array camera, and the initial interior orientation parameters of the virtual camera are calculated according to the laboratory interior orientation parameters of a single camera and the combination relationship of the three cameras; The second module is used for taking the POS of the intermediate camera of the combined linear array camera as the POS of each scanning line of the spliced image, using the initial interior orientation parameters of the virtual camera and the POS data, and setting the ground elevation of the survey area, to perform geometric correction on the push-broom imaging result image of the virtual linear array camera in the first module, to preliminarily eliminate the image geometric distortion caused by instability during flight, and generate an image referred to as L1 image; The third module is used for collecting images in the same survey area by using the calibrated linear array or area array camera as reference images, performing connection point matching and aerial triangulation processing, and adjusting and solving the interior and exterior orientation parameters of all reference images; The fourth module is used for extracting a large number of dense Harris feature points from the L1 image of each flight strip in the second module, matching the extracted Harris feature points with the reference images by using point matching, and selecting the flight strip L1 image with the most matched points for subsequent cloud control calibration processing; The fifth module is used for solving the ground coordinates of the matched points as cloud control points according to the principle of forward intersection using the interior and exterior orientation elements of the reference images; The sixth module is used for solving the exterior orientation elements of each scanning line of the linear array image according to the bundle adjustment using the cloud control points and the initial interior orientation parameters of the virtual camera; The seventh module is used for calculating the re-projection error of each feature point in the linear array image in the image row direction and column direction, wherein the image row direction is referred to as the x direction, and the column direction is referred to as the y direction. The eighth module is used for establishing a polynomial model to respectively perform curve fitting on the re-projection errors in x and y directions, and calculating the interior orientation parameters of all the detectors according to the fitted residual curves, so as to complete the calibration of the linear array camera.

8. The system of claim 6, wherein: The application relates to an aerial hyperspectral linear array combined large field-of-view camera interior parameter calibration method, comprising a processor and a memory, the memory is used for storing program instructions, and the processor is used for calling the stored instructions in the memory to execute the method.

9. The system of claim 6, wherein: The application relates to an aerial hyperspectral linear array combined large field-of-view camera interior parameter calibration method, comprising a readable storage medium, and the readable storage medium stores a computer program, the computer program is executed to realize the method.

Images