Unmanned aerial vehicle guided artillery sighting method based on photogrammetry and differential pos constraint

Abstract

The application provides a kind of unmanned aerial vehicle guiding gun aiming method based on photogrammetry and differential POS constraint, utilizes unmanned aerial vehicle with differential positioning system to collect target area image sequence and synchronously obtains differential POS data, provides absolute coordinate constraint by differential POS, first establishes sparse three-dimensional point cloud model with absolute coordinate reference, in subsequent monitoring stage, re-registers new collected target image with existing model, obtains the exterior orientation parameter of new image, then combines artificial intelligence target recognition result and photogrammetry space intersection principle, obtains the absolute space coordinate of target, realizes the automatic acquisition of target coordinate.The application improves the automation degree and positioning accuracy of target absolute coordinate determination in the environment where satellite positioning signal exists.The application does not rely on the use of laser ranging to obtain gun aiming coordinates, avoids the gun aiming indication error caused by laser ranging interference, and improves the long-distance target unmanned aerial vehicle positioning capability in the gun aiming scene.

Description

Technical Field

[0001] This invention belongs to the fields of UAV remote sensing, photogrammetry, computer vision and target monitoring technology, and in particular relates to a UAV-guided gun aiming method based on photogrammetry and differential POS constraints. Background Technology

[0002] With the development of unmanned aerial vehicle (UAV) platforms, airborne video acquisition equipment, and satellite positioning technology, target positioning based on UAV imagery has been widely applied.

[0003] In existing technologies, common methods for drone target monitoring mainly include the following categories:

[0004] 1. Target position estimation relies solely on UAV-borne GNSS / IMU data;

[0005] 2. Perform manual interpretation and relative position analysis using ordinary images;

[0006] 3. After reconstructing the scene model using photogrammetry, target localization is performed;

[0007] 4. Combine artificial intelligence recognition algorithms to automatically detect targets in images.

[0008] However, existing technologies still have the following shortcomings:

[0009] If only ordinary GNSS positioning is used, the absolute coordinate accuracy is limited, making it difficult to meet the monitoring requirements of meter-level or even higher accuracy;

[0010] If a target is identified using only a single image, it is usually only possible to obtain the target's two-dimensional image position, and it is difficult to directly obtain the target's three-dimensional absolute coordinates.

[0011] If newly added images cannot be stably registered with existing models, it will be difficult to achieve continuous and repeatable target localization and monitoring.

[0012] Manual identification methods are inefficient and highly subjective, making it difficult to meet the needs of automated monitoring.

[0013] Therefore, it is necessary to propose a UAV-guided gun aiming method that integrates differential POS, photogrammetric reconstruction, image re-registration, and AI target recognition to improve the automation and positioning accuracy of target absolute coordinate determination. Summary of the Invention

[0014] In view of this, the present invention aims to overcome the shortcomings of the above-mentioned problems in the prior art and proposes a UAV-guided gun aiming method based on photogrammetry and differential POS constraints, which is used to realize the absolute spatial position monitoring and positioning of the target in an environment with satellite positioning signals.

[0015] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0016] In a first aspect, the present invention provides a UAV-guided gun-aiming method based on photogrammetry and differential POS constraints, comprising the following steps:

[0017] Step S1: Under the condition of satellite positioning signal, use a drone equipped with video acquisition equipment and differential positioning system to collect video data of the target area or background scene, and simultaneously acquire differential POS data;

[0018] Step S2: Extract frames from the video data to obtain multiple images that are time-synchronized with the differential POS data;

[0019] Step S3: Perform SfM reconstruction based on the image and differential POS data to establish a sparse 3D point cloud model and a digital surface model (DSM) with absolute coordinate reference.

[0020] Step S4: Perform artificial intelligence target recognition on the target location in the keyframe image to obtain the two-dimensional pixel position of the target in the image;

[0021] Step S5: Based on the target's two-dimensional pixel position, the exterior orientation elements of the keyframe image, and the camera's intrinsic parameters, backproject the target image points into spatial rays, and intersect them with the Digital Surface Model (DSM) to obtain the target's absolute three-dimensional position and upload it to the tactical cloud.

[0022] Step S6: In the target monitoring stage, the UAV collects video data containing the target area again and extracts frames to generate the second stage UAV image. The target in the second stage UAV image is the bomb point.

[0023] Step S7: Perform feature matching and re-registration between the second-stage UAV imagery and the sparse 3D point cloud model with absolute coordinate reference, and calculate and obtain the absolute exterior orientation elements of the second-stage UAV imagery.

[0024] Step S8: Perform artificial intelligence recognition on the targets in the second-stage UAV imagery to obtain the two-dimensional pixel positions of the targets in the image;

[0025] Step S9: Based on the target's two-dimensional pixel position, the absolute exterior orientation elements of the second-stage UAV image, and the camera's intrinsic parameters, backproject the target image points into spatial rays, and intersect them with the Digital Surface Model (DSM) to obtain the absolute three-dimensional position of the explosion point. Then, upload the tactical cloud to the artillery terminal to calculate the firing correction parameters.

[0026] Furthermore, step S3 includes:

[0027] The global reprojection error is minimized by bundle adjustment, and the exterior orientation elements of all images are jointly solved. and the coordinates of the three-dimensional structural points of the scene :

[0028] ;

[0029] in, Let be the rotation matrix of the j-th frame image. Let j be the translation vector of the j-th frame image. As an indicator variable, For perspective projection functions, Here, K is the robust kernel function, K is the camera intrinsic parameter matrix, N is the total number of 3D points, and M is the total number of images. This represents the projection of the i-th 3D point onto the j-th frame of the image;

[0030] The calculated camera pose and coordinates of all three-dimensional structural points By combining synchronously acquired differential POS data for absolute orientation and adjustment, a sparse 3D point cloud model and a digital surface model (DSM) with absolute geographic coordinate reference are established.

[0031] Furthermore, the perspective projection function Defined as the coordinates of three-dimensional structure points The perspective transformation mapped to a two-dimensional pixel plane is calculated in a manner that satisfies:

[0032] ;

[0033] in For homogeneous pixel coordinates, by... Perform dehomogenization to obtain the theoretical reprojection pixel coordinates. By minimizing the actual observed image points With reprojected pixel coordinates The Euclidean distance between them is used to iteratively optimize the solution of the exterior orientation elements of all cameras. , and coordinates of all three-dimensional structural points .

[0034] Furthermore, in step S3, based on the coordinates of all three-dimensional structural points... A high-precision DSM surface is generated by locally fitting the sparse point cloud, and the equation is denoted as: Where (X, Y) represents the absolute planar coordinates of the target point. The elevation function representing the digital surface model.

[0035] Furthermore, in step S5, the target image point is back-projected into a spatial ray and its intersection with the digital surface model (DSM) is determined, specifically including:

[0036] Establish the space ray equation starting from the optical center C of the absolute camera:

[0037]

[0038] in, , For absolute rotation matrix, The absolute translation vector, the ray direction vector , For distance parameters, Here, K represents the normalized homogeneous pixel coordinates, and K is the camera intrinsic parameter matrix.

[0039] Find the intersection of the ray equation and the DSM equation:

[0040] ;

[0041] The distance on the space ray from the optical center is... The Z-coordinate of the point, The distance on the space ray from the optical center is... The X coordinate of the point is The distance on the space ray from the optical center is... The Y-coordinate of the point;

[0042] The optimal parameters are obtained by solving the problem. This allows us to obtain the target's absolute three-dimensional coordinates.

[0043] Furthermore, in step S7, re-registration solves for the absolute exterior orientation element of the new frame image by minimizing the reprojection error. :

[0044] ;

[0045] in, For two-dimensional image pixel coordinates, Represents three-dimensional absolute geographic coordinates. For perspective projection functions, Here, K is the robust kernel function, K is the camera intrinsic parameter matrix, and R is the robust kernel function. new For the rotation matrix of the new frame image, t new Let m be the translation vector of the new frame image, and m represent the index of the m-th matching point pair. This represents the total number of successfully matched 2D-3D point pairs.

[0046] Furthermore, the differential POS data includes position information and attitude information, which are used to provide absolute coordinate constraints for SfM reconstruction.

[0047] Furthermore, the method is applied to artillery bombardment point capture and firing correction. By comparing the initial target position with the bombardment point, it provides firing correction deviation parameters for the artillery terminal.

[0048] Secondly, the present invention provides an electronic device, including a processor and a memory communicatively connected to the processor and used to store executable instructions of the processor, wherein the processor is used to execute the above-mentioned UAV-guided gun aiming method based on photogrammetry and differential POS constraints.

[0049] Thirdly, the present invention provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the aforementioned UAV-guided gun-aiming method based on photogrammetry and differential POS constraints.

[0050] Compared with existing technologies, the UAV-guided gun-aiming method based on photogrammetry and differential POS constraints described in this invention has the following advantages:

[0051] (1) It does not rely on laser ranging and has strong environmental adaptability;

[0052] (2) By using differential POS to provide absolute coordinate constraints for the photogrammetric reconstruction process, the geographic reference accuracy of the model is improved;

[0053] (3) Using the sparse three-dimensional point cloud model established in advance as a reference, the rapid re-registration of subsequent monitoring images can be realized;

[0054] (4) Automatic target detection is achieved through AI recognition, which improves monitoring efficiency and reduces the cost of manual interpretation;

[0055] (5) The three-dimensional absolute coordinates are calculated from the two-dimensional image position by using the spatial intersection method of photogrammetry, which improves the scientificity and interpretability of target positioning;

[0056] (6) Under the conditions of image quality, positioning accuracy and spatial geometry, meter-level absolute positioning accuracy of targets can be achieved. Attached Figure Description

[0057] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0058] Figure 1 This is a schematic diagram of the method flow in Embodiment 1 of the present invention. Detailed Implementation

[0059] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0060] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0061] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0062] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0063] Example 1

[0064] refer to Figure 1 This invention provides a UAV-guided gun aiming method based on photogrammetry and differential POS constraints, comprising the following:

[0065] First drone data acquisition: Under the condition of satellite positioning signal, use a drone equipped with video acquisition equipment and differential positioning system to collect video data of the target area or background scene, and simultaneously acquire differential POS data;

[0066] Video frame extraction and SfM mapping: Frame extraction is performed on the video data to obtain multiple images that are time-synchronized with the differential POS data; based on the images and differential POS data, SfM reconstruction is performed to establish a sparse 3D point cloud model and a digital surface model (DSM) with absolute coordinate reference.

[0067] AI-powered target image recognition: AI-powered target recognition is performed on keyframe images to obtain the two-dimensional pixel position of the target in the image;

[0068] Image point back projection and DSM intersection to output the absolute coordinates of the target point: Based on the target's two-dimensional pixel position, the exterior orientation elements of the keyframe image and the camera's intrinsic parameters, the target image point is back projected into a spatial ray and intersected with the digital surface model (DSM) to obtain the target's absolute three-dimensional position and upload it to the tactical cloud.

[0069] Second drone data acquisition: During the target monitoring phase, the drone once again acquires video data containing the target area and extracts frames to generate the second-stage drone imagery;

[0070] Target image (explosion point) re-registration: Perform feature matching and re-registration between the second-stage UAV image and the sparse 3D point cloud model with absolute coordinate reference, and calculate and obtain the absolute exterior orientation elements of the second-stage UAV image.

[0071] AI-powered bomb detection: AI-powered target identification in the second-stage drone imagery to obtain the two-dimensional pixel position of the target in the image.

[0072] Image point back projection and DSM intersection: Based on the target's two-dimensional pixel position, the absolute exterior orientation elements of the second-stage UAV image, and the camera intrinsic parameters, the target image points are back projected into spatial rays and intersected with the digital surface model (DSM).

[0073] Output the absolute coordinates of the explosion point, and calculate the firing correction parameters based on the absolute coordinates of the target point and the explosion point: obtain the absolute three-dimensional position of the explosion point and upload it to the tactical cloud for the artillery terminal to calculate the firing correction parameters.

[0074] As a further solution, in this embodiment, when establishing the sparse 3D point cloud model and digital surface model (DSM), after the UAV initially acquires scene sequence images and simultaneously obtains high-precision differential POS data, matching feature points between each frame of the image are extracted to obtain multiple sets of cross-view 2D observation point pairs. (This represents the projection of the i-th 3D point onto the j-th frame image). The global reprojection error is minimized through bundle adjustment, and the exterior orientation elements of all images are jointly calculated. and the coordinates of the three-dimensional structural points of the scene :

[0075] ;

[0076] in, Let be the rotation matrix of the j-th frame image, representing the three-dimensional pose (angle) of the camera in the air when the j-th frame image was captured. Let be the translation vector of the j-th frame image, representing the absolute position (coordinates) of the camera center in three-dimensional space when the j-th frame image was captured. As an indicator variable (when a three-dimensional point) (1 if visible in image j, 0 otherwise). For perspective projection functions, This is a robust kernel function (used to suppress outliers caused by mismatches).

[0077] In this embodiment, the perspective projection function Defined as the coordinates of three-dimensional structure points The perspective transformation mapped to a two-dimensional pixel plane is calculated in a manner that satisfies:

[0078]

[0079] in, Here, K represents the homogeneous pixel coordinates, and K is the camera intrinsic parameter matrix. By... Perform homogenization to obtain the theoretical reprojected pixel coordinates. That is, the theoretical (calculated) two-dimensional pixel coordinates of the i-th three-dimensional spatial point projected onto the j-th image. This is achieved by minimizing the two-dimensional observation point pairs. With reprojected pixel coordinates The Euclidean distance between them is used to iteratively optimize and solve for the optimal exterior orientation elements of all cameras. , and the coordinates of all three-dimensional structural points .

[0080] The calculated camera pose and coordinates of all three-dimensional structural points By combining synchronously acquired differential POS data for absolute orientation and adjustment, a sparse 3D point cloud model and digital surface model (DSM) with absolute geographic coordinate reference are established, providing a spatial reference for the re-registration of new images and the intersection of absolute rays of the target in the subsequent monitoring stage.

[0081] As a further solution, this embodiment, when performing image point back projection and DSM intersection, is based on the coordinates of all three-dimensional structure points. A high-precision DSM surface is generated by locally fitting the sparse point cloud, and the equation is denoted as follows: Where (X, Y) represents the absolute planar coordinates of the target point. The elevation function of the digital surface model is represented by the equation, which obtains the elevation on the DSM using the absolute screen coordinates of the DSM and the known target point.

[0082] AI identifies targets and obtains two-dimensional pixels. Subsequently, the optical center of the absolute camera was established. Space ray equation:

[0083]

[0084] in, , This is the absolute rotation matrix, which defines the direction of rotation from the world coordinate system to the camera coordinate system. The absolute translation vector represents the linear offset of the origin of the time coordinate system in the camera coordinate system, and the ray direction vector. , For distance parameters, These are normalized homogeneous pixel coordinates.

[0085] Find the intersection of the ray equation and the DSM equation:

[0086] ;

[0087] The distance on the space ray from the optical center is... The Z-coordinate of the point, The distance on the space ray from the optical center is... The X coordinate of the point is The distance on the space ray from the optical center is... The Y-coordinate of the point;

[0088] The optimal parameters are obtained by solving the problem. Then the optimal parameters Substituting back into the space ray equation In this process, the spatial vector representation of the target point in the absolute coordinate system is obtained:

[0089]

[0090] Known optical center of absolute camera Ray direction vector Expand the above equation according to the corresponding axial components. At this point, the coordinates of the intersection point simultaneously satisfy the equations of the space ray and the ground. Surface equations (i.e.) Thus, the absolute three-dimensional coordinates of the target can be obtained without error. .

[0091] As a further solution, in this embodiment, during the re-registration of target images (explosion points), after the UAV re-acquires images of the explosion points, it extracts new feature points and matches them with the established absolute reference point cloud to obtain... Group 2D-3D pairs Where N is the number of matching point pairs, representing the total number of successfully matched feature point sets. Since the camera's exterior orientation elements have 6 degrees of freedom (3 positional degrees of freedom X, Y, Z and 3 attitude angles α, β, γ), theoretically, at least 3 sets of nonlinear point pairs are needed for the solution. However, in practical engineering (especially high-precision measurements), to combat environmental noise and image distortion, dozens or even hundreds of point pairs are usually matched. ). Here, represents the pixel coordinates of a two-dimensional image, specifically the pixel coordinates of the m-th feature point extracted from the latest bomb-bombed image acquired by the UAV in a secondary acquisition. Representing three-dimensional absolute geographic coordinates, the absolute exterior orientation elements of the new frame image are calculated by minimizing the reprojection error. :

[0092]

[0093] in For perspective projection functions, Here, K is the robust kernel function, K is the camera intrinsic parameter matrix, and R is the robust kernel function. new For the rotation matrix of the new frame image, t new Let m be the translation vector of the new frame image, and m represent the index of the m-th matching point pair. This represents the total number of successfully matched 2D-3D point pairs.

[0094] Among them, perspective projection function Defined as three-dimensional absolute geographic coordinates The perspective transformation mapped to a two-dimensional pixel plane is calculated in a manner that satisfies: ,in, For homogeneous pixel coordinates, by... Perform homogenization to obtain the reprojected pixel coordinates. By minimizing the observed image points With reprojected pixel coordinates The Euclidean distance between them is used to iteratively optimize the optimal exterior orientation element R of the camera. new and t new .

[0095] The calculated Substitute the spatial ray model into the DSM again to quickly calculate the three-dimensional coordinates of the explosion point, compare it with the initial target position, and provide firing correction parameters for the artillery terminal.

[0096] As a further solution, in this embodiment, AI recognition is used to automatically detect targets and provide their two-dimensional positioning results in the image; photogrammetry is used to provide image pose and spatial geometry calculation mechanisms. The combination of the two forms a complete closed loop of "image detection—geometric solution—spatial positioning," enabling the automatic determination of the target's absolute coordinates.

[0097] This invention moves the computationally intensive SfM 3D reconstruction to the pre-reconnaissance stage, and only performs lightweight single-frame image "re-registration / re-location" during actual combat (bomb point capture stage), which greatly shortens the computation delay of secondary strikes.

[0098] This invention cleverly utilizes a pre-generated DSM to replace a real physical rangefinder (such as a laser). While a single-lens 2D image can only provide ray direction, the DSM provides "depth / elevation" constraints for ray truncation, enabling low-cost extraction of absolute coordinates from a single lens.

[0099] This invention uses artificial intelligence to solve the problem of "where is the target (2D pixels)" and photogrammetry to solve the problem of "where is the target in reality (3D coordinates)". The combination of the two forms an automated closed loop.

[0100] This embodiment presents the experimental results of using a Pegasus V500R vertical takeoff and landing fixed-wing UAV equipped with a TIRV510 tri-light pod.

[0101] The absolute coordinates of the target point were measured by extracting frames from the video data of the V500R+TIRV510, where the absolute coordinates of the target point were known. The UAV flew around the target point in one circle, acquiring 24 photos, and then flew in a half-circle to acquire 12 photos. The standard error between the UAV and the known target point coordinates was calculated, as shown in Table 1.

[0102] Additionally, the continuous pointing function of the laser rangefinder of the TIRV510 (continuously measuring the distance to the target) is used to calculate the real-time coordinates of the target, and then the absolute position error with respect to the target is calculated. The specific accuracy is shown in Table 1.

[0103] In Table 1, the planar accuracy represents the standard error of the planar coordinates (x, y), and the three-dimensional accuracy represents the standard error of the three-dimensional coordinates (x, y, h). Here, x, y, and h represent the north, east, and altitude coordinates, respectively.

[0104] Table 1

[0105]

[0106] As shown in Table 1, the measured data indicates that compared to the traditional gun-aiming positioning method using a UAV equipped with a laser pod for continuous pointing, the positioning method of this invention, based on photogrammetry and differential POS constraints, significantly improves the accuracy of both planar and three-dimensional coordinate measurements. Furthermore, reducing the number of aerial photographs (12 for half a circle) not only does not decrease positioning accuracy but also shortens data processing time. Experimental results demonstrate that this invention can achieve meter-level high-precision target / explosion point absolute coordinate calculation without relying on laser rangefinders. It also possesses advantages such as fast calculation speed, minimal aerial work, low hardware cost, and strong resistance to environmental interference, thus meeting the practical needs of rapid artillery firing correction.

[0107] This invention's method can be applied to scenarios requiring rapid, accurate, and automated target localization. It does not rely on expensive sensors such as laser rangefinders; it only requires a conventional UAV equipped with a visible light camera and a differential POS, making it economical and worthy of widespread adoption.

[0108] Example 2

[0109] An electronic device includes a processor and a memory communicatively connected to the processor for storing processor-executable instructions, the processor being used to execute the aforementioned UAV-guided gun-aiming method based on photogrammetry and differential POS constraints.

[0110] Example 3

[0111] A computer-readable storage medium stores a computer program that, when executed by a processor, implements the aforementioned UAV-guided gun-aiming method based on photogrammetry and differential POS constraints.

[0112] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A UAV-guided gun-aiming method based on photogrammetry and differential POS constraints, characterized in that: Includes the following steps: Step S1: Under the condition of satellite positioning signal, use a drone equipped with video acquisition equipment and differential positioning system to collect video data of the target area or background scene, and simultaneously acquire differential POS data; Step S2: Extract frames from the video data to obtain multiple images that are time-synchronized with the differential POS data; Step S3: Perform SfM reconstruction based on the image and differential POS data to establish a sparse 3D point cloud model and a digital surface model (DSM) with absolute coordinate reference. Step S4: Perform artificial intelligence target recognition on the target location in the keyframe image to obtain the two-dimensional pixel position of the target in the image; Step S5: Based on the target's two-dimensional pixel position, the exterior orientation elements of the keyframe image, and the camera's intrinsic parameters, backproject the target image points into spatial rays, and intersect them with the Digital Surface Model (DSM) to obtain the target's absolute three-dimensional position and upload it to the tactical cloud. Step S6: In the target monitoring stage, the UAV collects video data containing the target area again and extracts frames to generate the second stage UAV image. The target in the second stage UAV image is the bomb point. Step S7: Perform feature matching and re-registration between the second-stage UAV imagery and the sparse 3D point cloud model with absolute coordinate reference, and calculate and obtain the absolute exterior orientation elements of the second-stage UAV imagery. Step S8: Perform artificial intelligence recognition on the targets in the second-stage UAV imagery to obtain the two-dimensional pixel positions of the targets in the image; Step S9: Based on the target's two-dimensional pixel position, the absolute exterior orientation elements of the second-stage UAV image, and the camera's intrinsic parameters, backproject the target image points into spatial rays, and intersect them with the Digital Surface Model (DSM) to obtain the absolute three-dimensional position of the explosion point. Then, upload the tactical cloud to the artillery terminal to calculate the firing correction parameters.

2. The UAV-guided gun-aiming method based on photogrammetry and differential POS constraints according to claim 1, characterized in that: Step S3 includes: The global reprojection error is minimized by bundle adjustment, and the exterior orientation elements of all images are jointly solved. and the coordinates of the three-dimensional structural points of the scene : ； in, Let be the rotation matrix of the j-th frame image. Let j be the translation vector of the j-th frame image. As an indicator variable, For perspective projection functions, Here, K is the robust kernel function, K is the camera intrinsic parameter matrix, N is the total number of 3D points, and M is the total number of images. This represents the projection of the i-th 3D point onto the j-th frame of the image; The calculated camera pose and coordinates of all three-dimensional structural points By combining synchronously acquired differential POS data for absolute orientation and adjustment, a sparse 3D point cloud model and a digital surface model (DSM) with absolute geographic coordinate reference are established.

3. The UAV-guided gun aiming method based on photogrammetry and differential POS constraints according to claim 2, characterized in that: The perspective projection function Defined as the coordinates of three-dimensional structure points The perspective transformation mapped to a two-dimensional pixel plane is calculated in a manner that satisfies: ； in For homogeneous pixel coordinates, by... Perform dehomogenization to obtain the theoretical reprojection pixel coordinates. By minimizing the actual observed image points With reprojected pixel coordinates The Euclidean distance between them is used to iteratively optimize the solution of the exterior orientation elements of all cameras. , and coordinates of all three-dimensional structural points .

4. The UAV-guided gun-aiming method based on photogrammetry and differential POS constraints according to claim 1, characterized in that: In step S3, the coordinates of all three-dimensional structural points are used as the basis. A high-precision DSM surface is generated by locally fitting the sparse point cloud, and the equation is denoted as: Where (X, Y) represents the absolute planar coordinates of the target point. The elevation function representing the digital surface model.

5. The UAV-guided gun-aiming method based on photogrammetry and differential POS constraints according to claim 4, characterized in that: In step S5, the target image point is back-projected into a spatial ray and its intersection with the digital surface model (DSM) is determined. This specifically includes: Establish the space ray equation starting from the optical center C of the absolute camera: ； in, , For absolute rotation matrix, The absolute translation vector, the ray direction vector , For distance parameters, For normalized homogeneous pixel coordinates, K is the camera intrinsic parameter matrix; Find the intersection of the ray equation and the DSM equation: ； The distance on the space ray from the optical center is... The Z-coordinate of the point, The distance on the space ray from the optical center is... The X coordinate of the point is The distance on the space ray from the optical center is... The Y-coordinate of the point; The optimal parameters are obtained by solving the problem. This allows us to obtain the target's absolute three-dimensional coordinates.

6. The UAV-guided gun aiming method based on photogrammetry and differential POS constraints according to claim 1, characterized in that: In step S7, re-registration calculates the absolute exterior orientation element of the new frame image by minimizing the reprojection error. : ； in, For two-dimensional image pixel coordinates, Represents three-dimensional absolute geographic coordinates. For perspective projection functions, Here, K is the robust kernel function, K is the camera intrinsic parameter matrix, and R is the robust kernel function. new For the rotation matrix of the new frame image, t new Let m be the translation vector of the new frame image, and m represent the index of the m-th matching point pair. This represents the total number of successfully matched 2D-3D point pairs.

7. The UAV-guided gun aiming method based on photogrammetry and differential POS constraints according to claim 1, characterized in that: The differential POS data includes position and attitude information, which is used to provide absolute coordinate constraints for SfM reconstruction.

8. The UAV-guided gun-aiming method based on photogrammetry and differential POS constraints according to claim 1, characterized in that: The method is applied to artillery bomb point capture and firing correction. By comparing the initial target position with the bomb point position, it provides firing correction deviation parameters for the artillery terminal.

9. An electronic device comprising a processor and a memory communicatively connected to the processor and used for storing processor-executable instructions, characterized in that: The processor is used to execute the UAV-guided gun aiming method based on photogrammetry and differential POS constraints as described in any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by the processor, it implements the UAV-guided gun aiming method based on photogrammetry and differential POS constraints as described in any one of claims 1-8.

Images