Orthograph generation method and device, equipment, storage medium and product

Abstract

The invention discloses an orthograph generation method and device, equipment, a storage medium and a product. The method comprises the following steps: determining a corresponding scene sparse point cloud and a camera external parameter according to an acquired scene aerial image sequence of a target scene and a camera internal parameter; performing region segmentation on the target scene according to the camera internal reference and the camera external reference to obtain scene segmentation sub-regions; performing Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region; rendering each local two-dimensional Gaussian scene based on an orthographic camera model to generate a corresponding local scene Gaussian orthographic graph; and splicing the Gaussian orthographs of the local scenes according to the external parameters of the camera to generate an orthograph corresponding to the target scene. The problems that a traditional method is limited by high-precision topographic information requirements and high-precision DSM generation costs too many computing resources are solved, and the high-precision orthograph can be quickly generated.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and in particular to an orthophoto generation method, device, equipment, storage medium and product. Background Art

[0002] An orthophoto is a ground image that is corrected to an undistorted image using aerial or satellite remote sensing images through geometric correction and projection transformation. The current mainstream methods for generating high-precision orthophotos are to generate digital surface models (DSMs) using technologies such as lidar and optical remote sensing (such as drones or aerial photography), and then restore the images using the elevation information in the DSM to ultimately generate a high-precision orthophoto. However, in the above methods, orthophoto generation is difficult due to the geometric distortion caused by changes in the scene terrain height and the high computational cost and time involved in generating high-precision DSMs. Summary of the invention

[0003] The present invention provides an orthophoto generation method, device, equipment, storage medium and product to solve the problem that traditional methods are limited by the demand for high-precision terrain information and that high-precision DSM generation consumes too many computing resources, and can quickly generate high-precision orthophotos.

[0004] According to one aspect of the present invention, a method for generating an orthophoto is provided, the method comprising:

[0005] According to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene, the corresponding scene sparse point cloud and camera extrinsic parameters are determined;

[0006] Performing region segmentation on the target scene according to the camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions;

[0007] Based on a preset two-dimensional Gaussian splashing algorithm, Gaussian splashing processing is performed on the scene sparse point cloud in each scene segmentation sub-region to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region;

[0008] Render each local two-dimensional Gaussian scene based on the orthophoto camera model to generate a corresponding local scene Gaussian orthophoto;

[0009] The Gaussian orthophotos of each local scene are stitched according to the camera extrinsic parameters to generate the orthophoto corresponding to the target scene.

[0010] According to another aspect of the present invention, there is provided an orthophoto generating device, the device comprising:

[0011] The point cloud and camera extrinsic parameter determination module is used to determine the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene;

[0012] A scene segmentation module is used to segment the target scene into regions according to camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions;

[0013] A Gaussian splash processing module is used to perform Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region;

[0014] A rendering module, used to render each local two-dimensional Gaussian scene based on the orthophoto camera model to generate a corresponding local scene Gaussian orthophoto;

[0015] The orthophoto generation module is used to stitch the Gaussian orthophotos of each local scene according to the camera external parameters to generate the orthophoto corresponding to the target scene.

[0016] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the orthophoto generating method described in any embodiment of the present invention.

[0020] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the orthophoto generating method described in any embodiment of the present invention when executed.

[0021] According to another aspect of the present invention, a computer program product is provided. The computer program product comprises a computer program. When the computer program is executed by a processor, the orthophoto generating method according to any one of the embodiments of the present invention is implemented.

[0022] The technical solution of the embodiment of the present invention determines the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene; performs region segmentation on the target scene according to the camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions; performs Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region; renders each local two-dimensional Gaussian scene based on an orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map; and splices each local scene Gaussian orthophoto map according to the camera extrinsic parameters to generate an orthophoto map corresponding to the target scene. This solution performs regional segmentation on the target scene, and then uses a preset two-dimensional Gaussian splashing algorithm to perform Gaussian splashing on the scene sparse point cloud in each scene segmentation sub-region, and renders the two-dimensional Gaussian scene based on the orthophoto camera model to generate the corresponding local scene Gaussian orthophoto map. Finally, the local scene Gaussian orthophoto maps are spliced ​​according to the camera extrinsic parameters to generate a complete orthophoto map corresponding to the target scene. This solves the problem that traditional methods are limited by the need for high-precision terrain information and that high-precision DSM generation consumes too many computing resources. It can quickly generate high-precision orthophoto maps and reduce computing costs.

[0023] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 is a flow chart of a method for generating an orthophoto according to the first embodiment of the present invention;

[0026] Figure 2 is a flow chart of a method for generating an orthophoto according to a second embodiment of the present invention;

[0027] Figure 3 is a schematic diagram comparing the orthographic viewing angle and the perspective viewing angle provided according to the second embodiment of the present invention;

[0028] Figure 4 is a structural schematic diagram of an orthophoto generating device provided according to Embodiment 3 of the present invention;

[0029] Figure 5It is a structural schematic diagram of an electronic device for implementing the orthophoto generating method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0031] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0032] Embodiment 1

[0033] Figure 1 A flowchart of an orthophoto generation method is provided for the first embodiment of the present invention. This embodiment is applicable to the case of quickly generating a high-precision orthophoto. The method can be executed by an orthophoto generation device. The orthophoto generation device can be implemented in the form of hardware and / or software. The orthophoto generation device can be configured in an electronic device. Figure 1 As shown, the first embodiment provides a method for generating an orthophoto, which specifically includes the following steps:

[0034] S110 , determining a corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene.

[0035] The scene aerial image sequence may refer to multiple frames of scene aerial images collected for the target scene. The scene aerial images may be collected by drones, and there needs to be sufficient overlap between adjacent frames to ensure that high-quality scene sparse point clouds can be generated subsequently. The scene sparse point cloud may refer to a set of three-dimensional coordinate point data sets used to describe the three-dimensional structure of the target scene.

[0036] Camera intrinsic parameters may refer to parameters used to describe the internal properties of a camera, such as the focal length of the camera, the coordinates of the principal point (optical center), distortion coefficients, pixel spacing, etc. These parameters are usually determined during camera calibration and are fixed for a specific camera model and do not change over time.

[0037] Camera extrinsics may refer to parameters used to describe the position and posture of a camera in a world coordinate system, such as rotation matrices and translation matrices, where the rotation matrix is ​​used to describe the rotation relationship between the camera coordinate system and the world coordinate system, and the translation matrix represents the position of the camera in the world coordinate system. It should be understood that camera extrinsics may change at different camera positions, such as when the camera is moved or the shooting angle is changed, the camera extrinsics will change accordingly.

[0038] In an embodiment of the present invention, a scene aerial image sequence captured for a target scene and pre-calibrated camera intrinsic parameters are obtained, and then a scene sparse point cloud and camera extrinsic parameters corresponding to the target scene are determined by calling an algorithm such as Structure from Motion (SfM) that uses a two-dimensional image sequence to restore the three-dimensional structure and camera motion.

[0039] S120 , performing region segmentation on the target scene according to camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions.

[0040] It is understandable that since the target scenes in the embodiments of the present invention are usually large scenes, such as urban parks, mountainous areas, canyon areas, etc., the amount of sparse point cloud data generated is large. If the corresponding data is not processed in blocks, it will affect the efficient execution of the subsequent preset two-dimensional Gaussian splash algorithm. Specifically, the camera intrinsic parameters and the aforementioned camera extrinsic parameters can be used to perform regional segmentation processing on the target scene to obtain a number of scene segmentation sub-regions; wherein the regional segmentation method can be equal division according to the area of ​​the region, or can be specified according to user requirements, etc., which is not specifically limited in this embodiment.

[0041] S130 , performing Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region.

[0042] Among them, the preset 2D Gaussian Splatting algorithm (2DGS) is a method for representing 3D scenes and rendering new views. Its core is to fold the 3D volume into a set of 2D oriented planar Gaussian disks (primitives), thereby providing visual Figure 1This embodiment can use the preset 2DGS algorithm to generate a local two-dimensional Gaussian scene corresponding to the scene sparse point cloud in each scene segmentation sub-region.

[0043] The local two-dimensional Gaussian scene may refer to a 2D Gaussian scene (ie, Gaussian representation) corresponding to each scene segmentation sub-region after Gaussian splash processing (training).

[0044] In an embodiment of the present invention, after the target scene is segmented, the preset 2DGS algorithm can be called for the scene sparse point cloud in each scene segmentation sub-region to perform 2DGS training, and after the training is completed, the local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region can be obtained. It can be understood that the camera model used in the 2DGS training process is a perspective camera model.

[0045] S140, rendering each local two-dimensional Gaussian scene based on the orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map.

[0046] The orthophoto camera model may refer to a camera model for generating an orthophoto, in which all objects are projected onto an image plane along parallel lines, and the size of the object is independent of its distance from the camera and is not affected by perspective distortion. The local scene Gaussian orthophoto may refer to a local orthophoto corresponding to a scene segmentation sub-region.

[0047] In an embodiment of the present invention, after obtaining the local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region, the camera model can be modified to an orthographic camera model, and each local two-dimensional Gaussian scene can be rendered into a corresponding local scene Gaussian orthographic map through differentiable rasterization rendering technology.

[0048] S150 , stitching the Gaussian orthophotos of each local scene according to the camera extrinsic parameters to generate an orthophoto corresponding to the target scene.

[0049] In an embodiment of the present invention, after rendering multiple local scene Gaussian orthophotos corresponding to the target scene, the local scene Gaussian orthophotos can be spliced ​​according to the camera pose information recorded in the camera extrinsic parameters to obtain a complete orthophoto corresponding to the target scene.

[0050] The technical solution of the embodiment of the present invention determines the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene; performs region segmentation on the target scene according to the camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions; performs Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region; renders each local two-dimensional Gaussian scene based on an orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map; and splices each local scene Gaussian orthophoto map according to the camera extrinsic parameters to generate an orthophoto map corresponding to the target scene. This solution performs regional segmentation on the target scene, and then uses a preset two-dimensional Gaussian splashing algorithm to perform Gaussian splashing on the scene sparse point cloud in each scene segmentation sub-region, and renders the two-dimensional Gaussian scene based on the orthophoto camera model to generate the corresponding local scene Gaussian orthophoto map. Finally, the local scene Gaussian orthophoto maps are spliced ​​according to the camera extrinsic parameters to generate a complete orthophoto map corresponding to the target scene. This solves the problem that traditional methods are limited by the need for high-precision terrain information and that high-precision DSM generation consumes too many computing resources. It can quickly generate high-precision orthophoto maps and reduce computing costs.

[0051] Embodiment 2

[0052] Figure 2 This is a flowchart of a method for generating an orthophoto provided in the second embodiment of the present invention, which is further optimized and expanded based on the above embodiment and can be combined with various optional technical solutions in the above embodiment. Figure 2 As shown, the second embodiment provides a method for generating an orthophoto, which specifically includes the following steps:

[0053] S210: Acquire a scene aerial image sequence and camera internal parameters for the target scene.

[0054] In an embodiment of the present invention, a scene aerial image sequence collected for a target scene and a pre-calibrated camera internal parameter can be obtained. In one embodiment, the scene aerial image sequence can be obtained by drone collection. For example, the drone can be placed in an area with a flat ground and open on all sides, and the drone can be allowed to fly randomly to collect multiple frames of scene aerial images of the target scene. At the same time, during the collection process, the camera shooting frequency can be reasonably adjusted to ensure that there is enough overlapping area between adjacent frames.

[0055] S220 , using the scene aerial image sequence and the camera intrinsic parameters as inputs of a preset structure-from-motion algorithm to obtain a scene sparse point cloud and camera extrinsic parameters corresponding to the target scene.

[0056] Among them, the preset Structure from Motion (SfM) algorithm is an algorithm that can recover the three-dimensional structure of the scene and the motion state of the camera from multiple two-dimensional images. Its basic principle is to estimate the camera motion between two adjacent images by matching feature points in multiple images. These feature points are usually prominent positions such as corners and edges in the image. By matching these feature points, the SfM algorithm can calculate the motion trajectory of the camera and the coordinates of the three-dimensional points in the scene, thereby constructing the three-dimensional structure of the scene, and finally obtaining the corresponding sparse point cloud and camera extrinsics.

[0057] In an embodiment of the present invention, after acquiring the scene aerial image sequence and the camera intrinsic parameters, the preset SfM algorithm can be called to reconstruct the three-dimensional structure of the target scene to obtain the corresponding scene sparse point cloud and camera extrinsic parameters; wherein the preset SfM algorithm may include but is not limited to: a global SfM algorithm and an incremental SfM algorithm, etc.

[0058] S230: extracting camera distribution from camera extrinsic parameters, and segmenting the camera distribution according to a preset region segmentation rule to obtain corresponding camera segmentation sub-regions.

[0059] Among them, the camera distribution can be used to characterize the location distribution of the camera. It can be understood that the drone usually collects aerial images of the scene at a fixed vertical height, and the camera distribution can usually be approximated as a square (square, rectangle) to facilitate the segmentation of the camera distribution.

[0060] The preset region segmentation rule can be understood as a region segmentation rule for camera distribution. For example, the camera distribution can be evenly divided into several camera segmentation sub-regions, for example, according to 2×2, 3×3, 4×4 segmentation, etc.; the camera distribution segmentation can also be performed according to the distribution of the scene sparse point cloud. For example, the camera distribution area corresponding to the scene sparse point cloud distribution with a denser distribution can be divided into more camera segmentation sub-regions, and the camera distribution area corresponding to the scene sparse point cloud distribution with a sparser distribution can be divided into fewer camera segmentation sub-regions, etc. This embodiment does not make any specific limitations on this.

[0061] In an embodiment of the present invention, the camera distribution can be extracted from the camera extrinsic parameters obtained by sparse reconstruction, and then the camera distribution can be divided into a number of camera segmentation sub-regions according to a preset region segmentation rule, such as a region equal division rule; wherein each camera segmentation sub-region can contain one or more cameras, and each camera corresponds to a local scene.

[0062] S240, for the camera outside the region edge and the nearest camera in the region corresponding to each camera segmentation sub-region, determine the overlapping area of ​​the mapping range of the scene aerial images corresponding to the camera outside the region edge and the nearest camera in the region in the world coordinate system according to the camera intrinsic parameters and the camera extrinsic parameters, and when the overlapping area of ​​the mapping range is greater than a preset area threshold, divide the camera outside the region edge into the corresponding camera segmentation sub-region.

[0063] Among them, the camera outside the area edge may refer to a camera located outside a camera segmentation sub-area. Usually, the camera outside the area edge refers to a camera whose viewing angle or position does not directly cover the interior of the camera segmentation sub-area, but observes the area from the outside.

[0064] The nearest camera in the area refers to a camera that is located in a camera segmentation sub-area and is closest to a reference point or target position. Usually, the nearest camera in the area can be a camera that can provide the clearest and closest distance view.

[0065] In an embodiment of the present invention, after the camera distribution is divided into a plurality of camera segmentation sub-areas, the camera outside the region edge and the nearest camera within the region corresponding to each camera segmentation sub-area can be determined according to the camera posture information recorded in the camera extrinsic parameters; then, for each camera segmentation sub-area, the overlapping area of ​​the mapping range of the scene aerial image corresponding to the camera outside the region edge and the nearest camera within the region in the world coordinate system is determined according to the camera intrinsic parameters and the camera extrinsic parameters; if the overlapping area of ​​the mapping range is greater than the preset area threshold S max , then divide the cameras outside the edge of the above area into the corresponding camera segmentation sub-area.

[0066] Further, based on the above-mentioned embodiment of the invention, determining the overlapping area of ​​the mapping range of the scene aerial images corresponding to the camera outside the edge of the region and the nearest camera in the region in the world coordinate system according to the camera intrinsic parameters and the camera extrinsic parameters in S240 includes:

[0067] Step 1, respectively extract the pixel coordinates of the four corner points of the aerial image of the scene corresponding to the camera outside the edge of the region and record them as the first corner point coordinate group, and extract the pixel coordinates of the four corner points of the aerial image of the scene corresponding to the camera closest to the region and record them as the second corner point coordinate group;

[0068] Step 2: Based on the camera intrinsic parameters and the camera extrinsic parameters, respectively call the preset coordinate system conversion formula to convert the first corner point coordinate group into the world coordinate system to obtain a first mapping range surrounded by the converted first corner point coordinate group, and convert the second corner point coordinate group corresponding to the nearest camera in the area into the world coordinate system to obtain a second mapping range surrounded by the converted second corner point coordinate group;

[0069] Step 3: Determine the overlapping area between the first mapping range and the second mapping range to obtain a mapping range overlapping area.

[0070] In the embodiment of the present invention, the specific process of determining the overlapping area of ​​the mapping range is as follows: the pixel coordinates of the four corner points of the aerial image of the scene corresponding to the camera outside the edge of the region can be first extracted and recorded as the first corner point coordinate group, and the second corner point coordinate group corresponding to the nearest camera in the region can be extracted similarly; and then the first corner point coordinate group and the second corner point coordinate group are respectively converted to the world coordinate system using the following preset coordinate system conversion formula:

[0071]

[0072] Where (x, y) represents the pixel coordinates of the corner point; (x c ,y c ) represents the coordinates in the camera coordinate system; (X w ,Y w ,Z w ) represents the coordinates in the world coordinate system; K represents the camera intrinsic parameter matrix; f x and f y They represent the normalized focal length of the camera in the x-axis direction and the normalized focal length in the y-axis direction respectively; (c x ,c y ) represents the camera principal point coordinates; R and t represent the rotation matrix and translation matrix in the camera extrinsic parameters respectively; Z c Represents the depth information from the camera to the scene.

[0073] Next, after the pixel coordinates of the four corner points in the first corner point coordinate group are converted to the real coordinates in the world coordinate system, the area enclosed by these four real coordinate points, namely the first mapping range, can be determined. Similarly, the second mapping range corresponding to the second corner point coordinate group can be determined. Finally, the overlapping area between the first mapping range and the second mapping range is calculated to obtain the corresponding mapping range overlapping area.

[0074] S250 , segmenting the target scene accordingly according to each camera segmentation sub-region to obtain each scene segmentation sub-region.

[0075] In the embodiment of the present invention, after the camera distribution is segmented into regions, the target scene can be segmented accordingly according to the divided camera segmentation sub-regions, thereby obtaining the scene segmentation sub-regions corresponding to the target scene.

[0076] S260. For the scene sparse point cloud in each scene segmentation sub-area, call a preset two-dimensional Gaussian splash algorithm to convert each point in the scene sparse point cloud into a primitive; wherein the primitive includes the following parameters: position parameter, shape parameter, color parameter and opacity parameter.

[0077] In the preset 2DGS algorithm, primitives are used as the basic units for constructing and representing objects or scenes. Primitives are defined as oriented elliptical disks. Primitives can contain a series of parameters, which together define the shape, position, direction and properties of the primitives. Specifically, primitives can include the following parameters: position parameters, shape parameters, color parameters and opacity parameters, where the position parameters are used to describe the position of the primitives, which can refer to the center point of the primitives, which can be a point in the sparse point cloud of the scene; shape parameters are used to describe the shape and direction of the primitives in three-dimensional space, which can be represented by a covariance matrix, which can be determined based on the rotation matrix and the scaling parameters (scaling vector); color parameters are used to describe the color of the primitives, which can be color values ​​determined by the spherical harmonic function coefficients; opacity parameters are used to describe the transparency or opacity based on, which can usually be mapped to the [0,1] interval by the S-type function, where 0 represents complete transparency and 1 represents complete opacity.

[0078] In the embodiment of the present invention, after the target scene is segmented, the preset 2DGS algorithm can be called for parallel processing for the scene sparse point cloud in each scene segmentation sub-region to improve the efficiency of data processing. Specifically, for the scene sparse point cloud in each scene segmentation sub-region, the preset 2DGS algorithm can be called to convert each point in the scene sparse point cloud into a corresponding primitive representation, and each primitive includes the following parameters: position parameter, shape parameter, color parameter and opacity parameter, and these primitive parameters are the parameters to be optimized in the subsequent 2DGS training.

[0079] In one embodiment, 2DGS uses flat primitives to represent the scene. The radiation field of the 2DGS primitive is perpendicular to its normal, so that it can better fit the surface of the scene. The 2DGS primitive mainly includes a center point p k , two tangent vectors t u ,t v And the scaling vector S that controls the variance of the 2D Gaussians = (s u ,s v ). The normal of 2D Gaussians is defined as t w =t u ×t v The 2DGS is defined in the world space local coordinate system as:

[0080] P(u)=p k +s u t u u+s v t v v=H·(u,v,1,1) T

[0081] Here H∈4×4 is the homogeneous transformation matrix representing the 2DGS geometry. For a point u=(u,v) in the uv plane, its 2D Gaussian value can be calculated using the standard Gaussian:

[0082]

[0083] S270, training the parameters of the primitives; the training objectives include: minimizing the difference between the image rendered by the primitive corresponding to each point in the scene sparse point cloud and the original scene aerial image corresponding to the scene sparse point cloud, while taking into account depth distortion and normal consistency.

[0084] Among them, depth distortion can refer to the fact that when using three-dimensional Gaussian distribution for rendering, different Gaussian distributions may overlap in depth, which will cause confusion in the depth and color of the rendering results. To solve the problem of depth distortion, a depth distortion regularization term L is introduced in 2DGS training. D , by minimizing the depth gap between intersection points, we ensure that these Gaussian distributions are at the correct depth positions, making the reconstructed geometry clearer and more accurate.

[0085] Normal consistency can mean that during the rendering process, if the normals (the direction pointing to the camera) of the two-dimensional Gaussian distribution are inconsistent, the surface will be rough and look unnatural. To solve the problem of normal consistency, the normal consistency regularization term L is introduced in the 2DGS training. n ,By aligning the normal of the 2D Gaussian distribution and the normal of the actual surface, we ensure that the reconstructed surface is smooth and the local geometry is accurate.

[0086] In an embodiment of the present invention, the parameters of each primitive can be continuously optimized iteratively. The training goal is to compare the image rendered by the primitive corresponding to each point in the scene sparse point cloud with the original scene aerial image corresponding to the scene sparse point cloud, and gradually make the two close. At the same time, the depth distortion and normal consistency need to be considered during the training process. In one embodiment, the following loss function can be constructed based on the above training goal:

[0087] L Loss =L c +αL d +βL n

[0088] Where, L c represents the reconstruction loss function, which is used to measure the difference between the image rendered by the primitive corresponding to each point in the scene sparse point cloud and the original scene aerial image corresponding to the scene sparse point cloud; L d represents the depth distortion regularization term, which is used to reduce the depth error between Gaussian distributions and ensure the accuracy of depth information; L nrepresents the normal consistency regularization term, which is used to ensure that the normal of the Gaussian distribution is consistent with the normal of the actual surface, ensuring the smoothness and naturalness of the surface; α and β represent weight coefficients, which are used to control the influence of the regularization term in the total loss.

[0089] In each round of 2DGS training iteration, the model parameters can be updated based on the value of the loss function by using methods such as gradient descent until the preset training end conditions are met; the training end conditions may include that the value of the loss function is less than or equal to a preset loss function threshold, the number of iterations reaches a preset number threshold, etc.

[0090] S280: After the training is completed, a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region is obtained.

[0091] In the embodiment of the present invention, after the scene sparse point cloud in each scene segmentation sub-region is trained with 2DGS, the local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region can be generated using the optimized parameters of each primitive.

[0092] S290, modifying the camera model used in the Gaussian splash processing into an orthographic camera model, and using a differentiable rasterization rendering technique to render each local two-dimensional Gaussian scene into a corresponding local scene Gaussian orthographic map.

[0093] Among them, differentiable rasterization rendering technology is an important technology in the field of 3D reconstruction and rendering. It allows the rendering process to be differentiated, so that the gradient of scene parameters can be calculated, thereby optimizing the 3D scene.

[0094] In an embodiment of the present invention, after obtaining the local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region, the camera model can be modified to an orthographic camera model, and each local two-dimensional Gaussian scene can be rendered into a corresponding local scene Gaussian orthographic map using a differentiable rasterization rendering technique. Under the orthographic camera model, all objects are projected onto the image plane along parallel lines, and the size of the object has nothing to do with its distance from the camera and is not affected by perspective distortion. Figure 3 shown.

[0095] The common strategy for rendering 2D Gaussian is to project the 2DGS primitive into the image space using the affine approximation of perspective projection. W∈4×4 is defined as the transformation matrix from world space to screen space. The coordinates of the point in screen space can be determined by the following formula:

[0096] x=(xz,yz,z,z) T =WP(u)=WH(u,v,1,1) T

[0097] Where x represents a uniform ray emitted from the camera, passing through the pixel (x, y) and intersecting the splat plane at depth z. On this basis, 2DGS calculates the intersection of three non-parallel planes to obtain screen coordinates. Given an arbitrary point (x, y) in the image coordinate system, the ray is parameterized into two orthogonal planes h x =(-1,0,0,x) T and h y =(0,-1,0,y) T The plane is transformed into the world space local coordinate system, where h u =(WH) T h x ,h v =(WH) T h y .

[0098] Further we can get: h u (u,v,1,1) T =h v (u,v,1,1) T =0,

[0099]

[0100] in, and Represents the i-th parameter of the homogeneous coordinate plane. The rasterization process is similar to the 3DGS process, using volume alpha fusion to calculate Gaussian color:

[0101]

[0102] In the formula, c is the center of the projected 2D Gaussian.

[0103] S210 , stitching the Gaussian orthophotos of each local scene according to the camera extrinsic parameters to generate an orthophoto corresponding to the target scene.

[0104] In an embodiment of the present invention, after rendering multiple local scene Gaussian orthophotos corresponding to the target scene, the local scene Gaussian orthophotos can be spliced ​​according to the camera pose information recorded in the camera extrinsic parameters to obtain a complete orthophoto corresponding to the target scene.

[0105] The technical solution of the embodiment of the present invention has at least the following beneficial effects: 1) the step of acquiring terrain information is eliminated, that is, there is no need to generate or process digital surface models, thereby reducing computing costs; 2) the image processing time can be significantly shortened, and high-precision orthophotos can be quickly generated; 3) since it is free from dependence on terrain, the requirements for drone flight path planning and image capture accuracy are reduced, and path planning time is saved in practical applications, making data acquisition more convenient and economical; 4) at the same time, since it is free from dependence on terrain information, the efficiency and effectiveness of emergency response can be greatly improved.

[0106] Embodiment 3

[0107] Figure 4 This is a schematic diagram of the structure of an orthophoto generating device provided in Embodiment 3 of the present invention. Figure 4 As shown, the device comprises:

[0108] The point cloud and camera extrinsic parameter determination module 31 is used to determine the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene;

[0109] A scene segmentation module 32 is used to segment the target scene into regions according to camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions;

[0110] A Gaussian splash processing module 33 is used to perform Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region;

[0111] A rendering module 34 is used to render each local two-dimensional Gaussian scene based on the orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map;

[0112] The orthophoto generation module 35 is used to stitch the Gaussian orthophotos of each local scene according to the camera external parameters to generate an orthophoto corresponding to the target scene.

[0113] The technical solution of the embodiment of the present invention determines the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene through a point cloud and camera extrinsic parameter determination module; the scene segmentation module performs regional segmentation on the target scene according to the camera intrinsic parameters and camera extrinsic parameters to obtain scene segmentation sub-regions; the Gaussian splash processing module performs Gaussian splash processing on the scene sparse point cloud in each scene segmentation sub-region based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region; the rendering module renders each local two-dimensional Gaussian scene based on an orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map; the orthophoto map generation module splices the Gaussian orthophoto maps of each local scene according to the camera extrinsic parameters to generate an orthophoto map corresponding to the target scene. This solution performs regional segmentation on the target scene, and then uses a preset two-dimensional Gaussian splashing algorithm to perform Gaussian splashing on the scene sparse point cloud in each scene segmentation sub-region, and renders the two-dimensional Gaussian scene based on the orthophoto camera model to generate the corresponding local scene Gaussian orthophoto map. Finally, the local scene Gaussian orthophoto maps are spliced ​​according to the camera extrinsic parameters to generate a complete orthophoto map corresponding to the target scene. This solves the problem that traditional methods are limited by the need for high-precision terrain information and that high-precision DSM generation consumes too many computing resources. It can quickly generate high-precision orthophoto maps and reduce computing costs.

[0114] Further, based on the above-mentioned embodiment of the invention, the point cloud and camera extrinsic parameter determination module 31 includes:

[0115] An image sequence and camera intrinsic parameter acquisition unit, used to acquire a scene aerial image sequence and camera intrinsic parameters for a target scene;

[0116] The point cloud and camera extrinsic parameter determination unit is used to use the scene aerial image sequence and the camera intrinsic parameters as inputs of a preset motion recovery structure algorithm to obtain the scene sparse point cloud and camera extrinsic parameters corresponding to the target scene.

[0117] Further, based on the above-mentioned embodiment of the invention, the scene segmentation module 32 includes:

[0118] A camera distribution segmentation unit is used to extract camera distribution from camera extrinsics and segment the camera distribution according to a preset region segmentation rule to obtain corresponding camera segmentation sub-regions;

[0119] The area edge camera division unit is used to determine the overlapping area of ​​the mapping ranges of the scene aerial images respectively corresponding to the area edge camera and the nearest camera in the area in the world coordinate system according to the camera intrinsic parameters and the camera extrinsic parameters for the area edge camera and the nearest camera in the area corresponding to each camera segmentation sub-area, and when the overlapping area of ​​the mapping ranges is greater than a preset area threshold, the area edge camera is divided into the corresponding camera segmentation sub-area;

[0120] The scene segmentation unit is used to segment the target scene accordingly according to each camera segmentation sub-region to obtain each scene segmentation sub-region.

[0121] Further, based on the above-mentioned embodiments of the invention, the overlapping area of ​​the mapping range of the scene aerial images corresponding to the camera outside the edge of the region and the nearest camera in the region in the world coordinate system is determined according to the camera intrinsic parameters and the camera extrinsic parameters, including:

[0122] The pixel coordinates of the four corner points of the aerial image of the scene corresponding to the camera outside the edge of the region are respectively extracted and recorded as the first corner point coordinate group, and the pixel coordinates of the four corner points of the aerial image of the scene corresponding to the camera closest to the region are extracted and recorded as the second corner point coordinate group;

[0123] Based on the camera intrinsic parameters and the camera extrinsic parameters, respectively call the preset coordinate system conversion formula to convert the first corner point coordinate group into the world coordinate system to obtain a first mapping range surrounded by the converted first corner point coordinate group, and convert the second corner point coordinate group corresponding to the nearest camera in the area into the world coordinate system to obtain a second mapping range surrounded by the converted second corner point coordinate group;

[0124] An overlapping area between the first mapping range and the second mapping range is determined to obtain a mapping range overlapping area.

[0125] Further, based on the above-mentioned embodiment of the invention, the Gaussian splatter processing module 33 includes:

[0126] A primitive conversion unit is used to convert each point in the scene sparse point cloud into a primitive by calling a preset two-dimensional Gaussian splash algorithm for the scene sparse point cloud in each scene segmentation sub-region; wherein the primitive includes the following parameters: position parameter, shape parameter, color parameter and opacity parameter;

[0127] A training unit is used to train the parameters of the primitives. The training objectives include: minimizing the difference between the primitive rendered image corresponding to each point in the scene sparse point cloud and the original scene aerial image corresponding to the scene sparse point cloud, while considering the depth distortion and normal consistency;

[0128] The local two-dimensional Gaussian scene generation unit is used to obtain the local two-dimensional Gaussian scene corresponding to each scene segmentation sub-region after the training is completed.

[0129] Further, based on the above-mentioned embodiment of the invention, the rendering module 34 includes:

[0130] The rendering unit is used to modify the camera model used in the Gaussian splash processing into an orthographic camera model, and use a differentiable rasterization rendering technology to render each local two-dimensional Gaussian scene into a corresponding local scene Gaussian orthographic map.

[0131] The orthophoto generating device provided in the embodiment of the present invention can execute the orthophoto generating method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0132] Embodiment 4

[0133] Figure 5 A schematic diagram of the structure of an electronic device 40 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0134] like Figure 5 As shown, the electronic device 40 includes at least one processor 41, and a memory connected to the at least one processor 41, such as a read-only memory (ROM) 42, a random access memory (RAM) 43, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 41 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 42 or the computer program loaded from the storage unit 48 to the random access memory (RAM) 43. In the RAM 43, various programs and data required for the operation of the electronic device 40 can also be stored. The processor 41, the ROM 42, and the RAM 43 are connected to each other through a bus 44. An input / output (I / O) interface 45 is also connected to the bus 44.

[0135] A number of components in the electronic device 40 are connected to the I / O interface 45, including: an input unit 46, such as a keyboard, a mouse, etc.; an output unit 47, such as various types of displays, speakers, etc.; a storage unit 48, such as a disk, an optical disk, etc.; and a communication unit 49, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 49 allows the electronic device 40 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0136] The processor 41 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 41 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 41 executes the various methods and processes described above, such as an orthophoto generation method.

[0137] In some embodiments, the orthophoto generation method may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 48. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 40 via the ROM 42 and / or the communication unit 49. When the computer program is loaded into the RAM 43 and executed by the processor 41, one or more steps of the orthophoto generation method described above may be performed. Alternatively, in other embodiments, the processor 41 may be configured to perform the orthophoto generation method in any other suitable manner (e.g., by means of firmware).

[0138] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0139] In some embodiments, the orthophoto generation method may be implemented as a computer program, which is invisibly included in a computer program product. The computer program implements the orthophoto generation method of the present invention when executed by a processor. The computer program product can be understood as a software product that implements its solution mainly through a computer program. The computer program for implementing the method of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flowchart and / or block diagram are implemented. The computer program can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0140] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0141] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0142] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0143] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0144] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0145] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for generating an orthophoto, characterized in that: The method comprises: According to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene, the corresponding scene sparse point cloud and camera extrinsic parameters are determined; Performing region segmentation on the target scene according to the camera intrinsic parameters and the camera extrinsic parameters to obtain scene segmentation sub-regions; Performing Gaussian splash processing on the scene sparse point cloud in each of the scene segmentation sub-regions based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each of the scene segmentation sub-regions; Rendering each of the local two-dimensional Gaussian scenes based on an orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map; The Gaussian orthophotos of the local scenes are spliced ​​according to the camera extrinsic parameters to generate an orthophoto corresponding to the target scene.

2. The method according to claim 1, characterized in that The step of determining the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene includes: Acquire the scene aerial image sequence and the camera intrinsic parameters for the target scene; The scene aerial image sequence and the camera intrinsic parameters are used as inputs of a preset motion recovery structure algorithm to obtain the scene sparse point cloud and the camera extrinsic parameters corresponding to the target scene.

3. The method according to claim 1, characterized in that The performing region segmentation on the target scene according to the camera intrinsic parameters and the camera extrinsic parameters to obtain scene segmentation sub-regions includes: Extracting camera distribution from the camera extrinsic parameters, and segmenting the camera distribution according to a preset region segmentation rule to obtain corresponding camera segmentation sub-regions; For each camera outside the region edge and the nearest camera in the region corresponding to each camera segmentation sub-region, determine the overlapping area of ​​the mapping ranges of the scene aerial images respectively corresponding to the camera outside the region edge and the nearest camera in the region in the world coordinate system according to the camera intrinsic parameters and the camera extrinsic parameters, and when the overlapping area of ​​the mapping ranges is greater than a preset area threshold, divide the camera outside the region edge into the corresponding camera segmentation sub-region; The target scene is segmented accordingly according to each of the camera segmentation sub-regions to obtain each of the scene segmentation sub-regions.

4. The method according to claim 3, characterized in that The determining, according to the camera intrinsic parameters and the camera extrinsic parameters, the overlapping areas of the mapping ranges of the scene aerial images respectively corresponding to the camera outside the region edge and the nearest camera within the region in the world coordinate system comprises: Extracting pixel coordinates of four corner points of the aerial image of the scene corresponding to the camera outside the edge of the region respectively and recording them as a first corner point coordinate group, and extracting pixel coordinates of four corner points of the aerial image of the scene corresponding to the camera closest to the region and recording them as a second corner point coordinate group; Based on the camera intrinsic parameters and the camera extrinsic parameters, respectively calling preset coordinate system conversion formulas to convert the first corner point coordinate group into the world coordinate system to obtain a first mapping range surrounded by the converted first corner point coordinate group, and converting the second corner point coordinate group corresponding to the nearest camera in the area into the world coordinate system to obtain a second mapping range surrounded by the converted second corner point coordinate group; An overlapping area between the first mapping range and the second mapping range is determined to obtain the mapping range overlapping area.

5. The method according to claim 1, characterized in that The performing Gaussian splash processing on the scene sparse point cloud in each of the scene segmentation sub-regions based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each of the scene segmentation sub-regions includes: For the scene sparse point cloud in each of the scene segmentation sub-areas, calling the preset two-dimensional Gaussian splash algorithm to convert each point in the scene sparse point cloud into a primitive; wherein the primitive includes the following parameters: position parameter, shape parameter, color parameter and opacity parameter; The parameters of the primitives are trained; the training objectives include: minimizing the difference between the image rendered by the primitive corresponding to each point in the sparse point cloud of the scene and the original aerial image of the scene corresponding to the sparse point cloud of the scene, taking into account depth distortion and normal consistency; After the training is completed, a local two-dimensional Gaussian scene corresponding to each of the scene segmentation sub-regions is obtained.

6. The method according to claim 1, characterized in that The rendering of each of the local two-dimensional Gaussian scenes based on the orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map includes: The camera model used in the Gaussian splatting process is modified to the orthophoto camera model, and each of the local two-dimensional Gaussian scenes is rendered into a corresponding Gaussian orthophoto of the local scene using a differentiable rasterization rendering technique.

7. An orthophoto generating device, characterized in that: The device comprises: The point cloud and camera extrinsic parameter determination module is used to determine the corresponding scene sparse point cloud and camera extrinsic parameters according to the acquired scene aerial image sequence and camera intrinsic parameters of the target scene; A scene segmentation module, used to perform region segmentation on the target scene according to the camera intrinsic parameters and the camera extrinsic parameters to obtain scene segmentation sub-regions; A Gaussian splash processing module, used for performing Gaussian splash processing on the scene sparse point cloud in each of the scene segmentation sub-regions based on a preset two-dimensional Gaussian splash algorithm to generate a local two-dimensional Gaussian scene corresponding to each of the scene segmentation sub-regions; A rendering module, used for rendering each of the local two-dimensional Gaussian scenes based on an orthophoto camera model to generate a corresponding local scene Gaussian orthophoto map; The orthophoto generation module is used to stitch the Gaussian orthophotos of the local scenes according to the camera extrinsic parameters to generate an orthophoto corresponding to the target scene.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can perform the orthophoto generating method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the orthophoto generation method according to any one of claims 1 to 6 when executed.

10. A computer program product, characterized in that The computer program product comprises a computer program, which, when executed by a processor, implements the orthophoto generation method according to any one of claims 1 to 6.

Images