Air-to-ground real-time mapping analysis method and system based on aerospace vehicle

Abstract

The invention discloses an air-to-ground real-time mapping analysis method and system based on an aerospace vehicle. The method comprises the following steps: receiving images in different directions transmitted by an unmanned aerial vehicle; extracting the same feature points of the plurality of different azimuth images, and calculating the coordinates of the feature points in the plurality of different azimuth images by combining the height of the unmanned aerial vehicle; splicing a plurality of images in different directions, fusing the coordinates of the feature points and the spliced image through a registration technology, and generating a high-precision image with ground height information; extracting depth information in different azimuth images, generating three-dimensional point cloud data, and converting the three-dimensional point cloud data into a three-dimensional grid model; accurate three-dimensional reconstruction is carried out on the ground and the irregular object, the limitation that the complex contour of the object cannot be captured at a single view angle is avoided through multi-directional real-time image acquisition, and the three-dimensional coordinates of the feature points are accurately calculated by using a back projection technology, so that the surface non-uniformity and height difference of the irregular object are solved, and the accuracy and the reliability of the object are improved. High-quality image data is provided.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle map construction, and in particular to an air-to-ground real-time map construction and analysis method and system based on an aerospace vehicle. Background Art

[0002] The construction technology of drone image map needs to focus on solving three problems: first, due to the increase in the number of drones, the bandwidth of a single drone reconnaissance data is limited when it is transmitted, resulting in a low data bit rate, which brings difficulty to calculation; second, multiple drone images are usually not connected, and low or no overlap between images is the norm, which brings difficulties to image matching; finally, the construction of multi-drone image map has a large amount of data and high real-time requirements, which brings challenges to how to efficiently integrate multiple drone images. Summary of the invention

[0003] The present invention overcomes the deficiencies of the prior art and provides an air-to-ground real-time mapping analysis method and system based on an aerospace vehicle.

[0004] To achieve the above object, the technical solution adopted by the present invention is: a real-time air-to-ground mapping analysis method based on an aerospace vehicle, comprising the following steps:

[0005] Step S1, receiving real-time images of different directions transmitted by the drone;

[0006] Step S2, extracting the same feature points of a number of images in different orientations, and calculating the coordinates of the feature points in the images in different orientations in combination with the altitude of the drone;

[0007] Step S3, stitching a number of images in different orientations, fusing the feature point coordinates and the stitched images through registration technology, and generating a high-precision image with ground height information;

[0008] Step S4: extract depth information from images of different orientations, obtain a depth map from a two-dimensional image, generate three-dimensional point cloud data, and convert it into a three-dimensional mesh model.

[0009] In a preferred embodiment of the present invention, in step S1, drones in the same area take real-time images at different locations;

[0010] Different directions include: looking down, looking sideways, and looking horizontally.

[0011] In a preferred embodiment of the present invention, in step S2, the same feature points of several images in different orientations are extracted, specifically, the images are divided into grids, each grid contains several feature points; an interpolation algorithm is used to increase the height value of the feature points in the grid; a feature matching algorithm is used to find the corresponding feature points between different images; and a transformation matrix is ​​calculated based on the matched feature points to align the images.

[0012] In a preferred embodiment of the present invention, in step S2, the coordinates of feature points in a plurality of images in different orientations are calculated in combination with the altitude of the drone, including the following steps:

[0013] Step S21: Real-time acquisition of the drone's location information P = [x u ,y u ,h], where x u ,y u is the plane coordinate of the drone relative to the ground, h is the flight altitude, and the drone attitude is obtained;

[0014] Step S22: extracting feature points in images of different orientations by image processing methods. The feature points have pixel coordinates (x i ,y i );

[0015] Step S23: perform back projection based on the camera internal and external parameters and the height of the drone, and calculate the three-dimensional coordinates of the feature points using the pixel coordinates of the feature points in the image and the corresponding depth information. where d i is the depth information obtained from the depth map,

[0016] K -1 is the inverse of the internal parameter matrix;

[0017] Step S24: Based on the flight altitude of the UAV, the vertical position of the feature point is corrected. i =hd i , Z i Indicates the true height of the feature point.

[0018] In a preferred embodiment of the present invention, in step S21, the drone posture includes: the pitch, roll and yaw angles of the drone.

[0019] In a preferred embodiment of the present invention, step S3 includes the following sub-steps:

[0020] Step S31, using a brute force matching algorithm to compare feature points in images of different orientations one by one, and determine identical feature point pairs;

[0021] Step S32: Select feature points and use the RANSAC algorithm to calculate the transformation matrix H. Among them, (x, y) and (x a ,y a ) are the coordinates of the matching feature points in the source image and the target image respectively;

[0022] Step S33, using the perspective transformation matrix to map each pixel coordinate in the source image to the target image for alignment;

[0023] Step S34: fuse the aligned images, and take the average value of all images for the pixel values ​​in the overlapping area.

[0024] In a preferred embodiment of the present invention, step S4 includes the following sub-steps:

[0025] Step S41, calculating the depth information of the object by the disparity of the same feature point in a plurality of images in different orientations, to obtain a depth map in the two-dimensional image;

[0026] Step S42: convert the depth information of each pixel in the depth map coordinate system into point coordinates (X, Y, Z) in three-dimensional space;

[0027] Step S43: Use a triangulation algorithm to convert the point cloud data into a triangular mesh and then into a three-dimensional mesh model.

[0028] The air-to-ground real-time mapping and analysis system based on aerospace vehicles includes: image receiving module, feature point extraction module, coordinate calculation module, image registration module, stereo vision and depth information calculation module, data storage and output module.

[0029] The image receiving module is used to receive images of different orientations transmitted by the drone in real time and monitor the altitude position of the drone in real time;

[0030] The feature point extraction module is used to extract the same feature points in images of different orientations;

[0031] The coordinate calculation module calculates the coordinates of the feature points of the objects in the image by combining the height position of the drone obtained in the image receiving module;

[0032] The image registration module is used to stitch images of different orientations through registration technology and fuse feature point coordinates to generate a high-precision image with ground height information;

[0033] The stereoscopic vision and depth information calculation module is used to perform parallax calculation on the same feature points in images of different orientations, thereby obtaining the depth information of the object, generating a depth map in the two-dimensional image, and converting the depth information of each pixel point in the depth map coordinate system into point coordinates in three-dimensional space to obtain three-dimensional point cloud data, converting the point cloud data into a triangular mesh, and further generating a three-dimensional mesh model;

[0034] The data storage and output module is used to store and output the finally generated high-precision image with ground height information, three-dimensional point cloud data and three-dimensional grid model.

[0035] In a preferred embodiment of the present invention, the feature point extraction module includes: a feature point extraction unit, a grid division unit, and an interpolation unit;

[0036] The feature point extraction unit extracts the same feature points from images of different orientations using an image processing algorithm;

[0037] The grid division unit is used to divide the image into grids, each grid containing a number of feature points;

[0038] The interpolation unit calculates the height values ​​of the unobtained feature points in the grid by using bilinear interpolation.

[0039] In a preferred embodiment of the present invention, the image registration module includes: a feature point matching unit, a transformation matrix calculation unit, and an image alignment unit;

[0040] The feature point matching unit compares the feature points in the images of different orientations one by one using a brute force matching algorithm to determine the same feature point pairs;

[0041] The transformation matrix calculation unit calculates the transformation matrix between images using the RANSAC algorithm, including the coordinates of matching feature points in the source image and the target image;

[0042] The image alignment unit maps each pixel coordinate in the source image to the target image using a perspective transformation matrix.

[0043] The present invention solves the defects existing in the background technology and has the following beneficial effects:

[0044] (1) The present invention provides an air-to-ground real-time mapping and analysis method and system based on aerospace vehicles. Through multi-directional real-time image acquisition, feature point matching, image registration and depth information extraction, the ground and irregular objects can be accurately reconstructed in three dimensions. The multi-directional real-time image acquisition avoids the limitation that a single perspective cannot capture the complex contours of objects. The back-projection technology is used to accurately calculate the three-dimensional coordinates of feature points, thereby solving the surface unevenness and height differences of irregular objects, providing high-quality image data, and avoiding the low-precision reconstruction of complex geometric shapes by traditional methods. Objects with complex and irregular contours can be accurately reconstructed to ensure high precision, integrity and detail restoration.

[0045] (2) The present invention extracts the same feature points in images of different orientations through image processing technology, and combines the UAV flight altitude information to calculate accurate three-dimensional coordinates through back projection technology. This not only effectively solves the problem of object position and scale changes in multi-orientation images, but also can accurately locate the three-dimensional spatial position of each feature point, thereby providing accurate basic data and providing precise support for subsequent depth map generation and three-dimensional reconstruction.

[0046] (3) The present invention extracts depth information through the principle of stereoscopic vision and combines the parallax of images in different orientations to generate an accurate depth map. The depth information obtained from the depth map is used to convert the two-dimensional pixel coordinates into point coordinates in three-dimensional space, thereby obtaining high-precision three-dimensional point cloud data. This avoids the inadequate processing of complex and irregular object surfaces by traditional methods and can accurately capture the surface details of objects. It is particularly suitable for processing objects with complex shapes and irregular surfaces.

[0047] (4) The present invention uses drones to take real-time images from multiple directions, which can capture irregular objects from multiple angles and in a comprehensive manner, helping to overcome the limitations of the single viewing angle in traditional methods. It ensures that no matter how complex or irregular the geometric structure of the object is, its full appearance can be captured from multiple angles, avoiding information loss or reconstruction errors, thereby improving the accuracy and completeness of the reconstruction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art are briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 It is a flow chart of a method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 are within the scope of protection of the present invention.

[0051] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0052] Application Overview

[0053] There are currently two technical means for human-computer video image map stitching technology. One is to build a map based on image stitching based on image feature point matching, and the other is to build a scene structure based on SFM technology to guide image map mapping.

[0054] The method of constructing maps by stitching together images based on matching image feature points has the problem of drift caused by cumulative errors. At the same time, when dealing with multi-machine video images, the non-connectivity of the reconnaissance area means that there are no feature points between images, which may cause algorithm matching failure. The image construction method based on SFM technology can often obtain the best map construction effect when the image quality is high, but the defects are also obvious: the SFM technology itself has poor real-time calculation performance, and the reconnaissance images of multiple drones are limited by data bandwidth. The quality of real-time image transmission often cannot meet the SFM structure calculation, and irregular objects in the image cannot be extracted, resulting in calculation failure.

[0055] Therefore, it is necessary to study the air-to-ground real-time mapping and analysis method and system based on aerospace vehicles to solve the shortcomings of existing technical problems.

[0056] like Figure 1 As shown, the real-time air-to-ground mapping and analysis method based on aerospace vehicles includes the following steps:

[0057] Step S1, receiving real-time images of different directions transmitted by the drone;

[0058] Step S2, extracting the same feature points of a number of images in different orientations, and calculating the coordinates of the feature points in the images in different orientations in combination with the altitude of the drone;

[0059] Step S3, stitching a number of images in different orientations, fusing the feature point coordinates and the stitched images through registration technology, and generating a high-precision image with ground height information;

[0060] Step S4: extract depth information from images of different orientations, obtain a depth map from a two-dimensional image, generate three-dimensional point cloud data, and convert it into a three-dimensional mesh model.

[0061] In the present invention, in step S1, different orientations include: looking down, looking sideways, and looking horizontally, and the drone takes real-time images in different orientations in the same area;

[0062] Acquiring images from different perspectives can provide information about different dimensions and structures of an object. For irregular objects, a single perspective may result in the failure to capture certain details. Multi-directional image acquisition can effectively make up for these deficiencies and ensure a comprehensive understanding and accurate reconstruction of the object. Therefore, drones use the sensors and cameras they carry to collect high-definition images of aerial target areas in real time, and take images of the same target area from different orientations. For example, for ground targets, you can choose to shoot from a bird's-eye view, an oblique view, or a horizontal view, so that you can obtain detailed information about the object from multiple angles.

[0063] In the present invention, in step S2, the same feature points of several images in different orientations are extracted, specifically, the images are divided into grids, each grid contains several feature points; an interpolation algorithm is used to increase the height value of the feature points in the grid; a feature matching algorithm is used to find the corresponding feature points between different images; and a transformation matrix is ​​calculated based on the matched feature points to align the images.

[0064] Calculating the coordinates of feature points in several images of different orientations based on the altitude of the drone includes the following steps:

[0065] Step S21: Real-time acquisition of the drone's location information P = [x u ,y u ,h], where x u ,y u is the plane coordinate of the drone relative to the ground, h is the flight altitude, and the drone attitude is obtained. The drone attitude includes: the pitch, roll and yaw angles of the drone, which are collected in real time through the drone's navigation system (such as GPS, inertial measurement unit IMU) to provide a basis for the subsequent geometric positioning of feature points;

[0066] Step S22: extract feature points in images of different orientations by image processing methods, and record the two-dimensional pixel positions of the feature points in the image plane. The feature points have their pixel coordinates (x i ,y i );

[0067] Step S23: perform back projection based on the camera internal and external parameters and the height of the drone, and calculate the three-dimensional coordinates of the feature points using the pixel coordinates of the feature points in the image and the corresponding depth information. where d i is the depth information obtained from the depth map,

[0068] K -1 It is the inverse of the intrinsic parameter matrix. Through back-projection calculation, the feature points are transformed from the image plane into three-dimensional coordinates in the real space, reflecting the position of the object in the actual environment;

[0069] Step S24: Based on the flight altitude of the UAV, the vertical position of the feature point is corrected. i =hd i , Z i It represents the true height of the feature point. The corrected feature point height is closer to the actual ground height, which reduces the error caused by changes in the UAV flight path or terrain undulations, and ensures the accuracy of 3D reconstruction.

[0070] By extracting the same feature points in images of different orientations through image processing technology and combining the UAV's flight altitude information, accurate three-dimensional coordinates are calculated through back projection technology. This not only effectively solves the problem of object position and scale changes in multi-orientation images, but also can accurately locate the three-dimensional spatial position of each feature point, thereby providing accurate basic data and providing precise support for subsequent depth map generation and three-dimensional reconstruction. By extracting depth information through the principle of stereo vision and combining the parallax of images of different orientations, an accurate depth map can be generated. Through the depth information obtained from the depth map, the two-dimensional pixel coordinates are converted into point coordinates in three-dimensional space, thereby obtaining high-precision three-dimensional point cloud data, avoiding the insufficient processing of complex and irregular object surfaces by traditional methods, and being able to accurately capture the surface details of objects, which is especially suitable for processing objects with complex shapes and irregular surfaces.

[0071] In the present invention, step S3 includes the following sub-steps:

[0072] Step S31, using a brute force matching algorithm to compare feature points in images of different orientations one by one, determine the same feature point pairs, and obtain a feature point pair set;

[0073] Step S32: Select one set of feature point pairs, remove the mismatched points, and use the RANSAC algorithm to calculate the transformation matrix H. Among them, (x, y) and (x a ,y a ) are the coordinates of the matching feature points in the source image and the target image respectively. A certain number of matching point pairs are randomly selected, and the initial transformation matrix is ​​calculated. The position of the matching point in the target image is predicted according to the transformation matrix. It is determined whether it falls within the tolerance range and marked as an "inlier point". Specifically, an initial tolerance range is set (e.g., 3 pixels), and the source image point (x, y) and the target image point (x a ,y a ) to calculate the Euclidean distance error, Compare the error with the tolerance range. If the error is less than or equal to the tolerance range, mark the point as an "inside point". If the error is greater than the tolerance range, mark the point as an "outside point".

[0074] Step S33, using the perspective transformation matrix to map each pixel coordinate in the source image to the target image for alignment;

[0075] Step S34: fuse the aligned images, and take the average value of all images for the pixel values ​​in the overlapping area.

[0076] In the present invention, step S4 includes the following sub-steps:

[0077] Step S41, calculating the depth information of the object by disparity of the same feature point in several images with different orientations, and obtaining a depth map in the two-dimensional image. Specifically, in step S2, the same feature point position between images with different orientations is obtained by a feature point matching algorithm, and the matched feature point pairs are used to calculate the depth of each matching point by triangulation, that is, the disparity value is set as the pixel offset of the same feature point in images with different orientations, and the depth value of each feature point is calculated by triangulation principle according to the internal and external parameters of the camera (such as focal length, baseline distance, camera position, etc.); wherein the depth map provides the distance information between the surface of the object and the camera, and the pixels in the depth map are converted into actual three-dimensional coordinates by back projection, thereby realizing accurate mapping from two-dimensional to three-dimensional;

[0078] Step S42: convert the depth information of each pixel in the depth map coordinate system into point coordinates (X, Y, Z) in three-dimensional space, combine the two-dimensional coordinates of each pixel in the depth map with the depth value according to the internal and external parameters of the camera, and reversely calculate the corresponding three-dimensional space coordinates;

[0079] Step S43, using the triangulation algorithm, converting the point cloud data into a triangular mesh and then into a three-dimensional mesh model, specifically, generating a number of triangular facets from the point cloud data, the edges of which are composed of points in the point cloud, and finally forming a polygonal mesh, wherein the key to triangulation is to maintain the quality of the triangles, ensuring that the generated mesh facets have a reasonable shape and size, and avoiding long triangular facets, which will affect the final three-dimensional model accuracy and performance; after completing the triangulation, a three-dimensional mesh model composed of triangular facets is obtained, which are based on the point cloud and can accurately reflect the three-dimensional shape of the object or scene.

[0080] The air-to-ground real-time mapping and analysis system based on aerospace vehicles includes: image receiving module, feature point extraction module, coordinate calculation module, image registration module, stereo vision and depth information calculation module, data storage and output module.

[0081] An image receiving module is used to receive images of different orientations transmitted by the drone in real time and monitor the altitude position of the drone in real time;

[0082] A feature point extraction module is used to extract the same feature points in images of different orientations;

[0083] The coordinate calculation module calculates the coordinates of the feature points of the objects in the image by combining the height position of the drone obtained in the image receiving module;

[0084] Image registration module, used to stitch images from different directions through registration technology and fuse feature point coordinates to generate high-precision images with ground height information;

[0085] The stereo vision and depth information calculation module is used to calculate the disparity of the same feature points in images of different orientations, thereby obtaining the depth information of the object, generating a depth map in the two-dimensional image, and converting the depth information of each pixel point in the depth map coordinate system into point coordinates in three-dimensional space to obtain three-dimensional point cloud data, converting the point cloud data into a triangular mesh, and further generating a three-dimensional mesh model;

[0086] The data storage and output module is used to store and output the final generated high-precision images with ground height information, three-dimensional point cloud data and three-dimensional mesh models.

[0087] In the present invention, the feature point extraction module includes: a feature point extraction unit, a grid division unit, and an interpolation unit;

[0088] A feature point extraction unit, which uses an image processing algorithm to extract the same feature points from images in different orientations;

[0089] A grid division unit, used to divide the image into grids, each grid containing a number of feature points;

[0090] The interpolation unit uses bilinear interpolation to calculate the height values ​​of feature points that are not obtained in the grid.

[0091] In the present invention, the image registration module includes: a feature point matching unit, a transformation matrix calculation unit, and an image alignment unit;

[0092] The feature point matching unit uses a brute force matching algorithm to compare the feature points in images of different orientations one by one to determine the same feature point pairs;

[0093] A transformation matrix calculation unit, which uses the RANSAC algorithm to calculate the transformation matrix between images, including the coordinates of matching feature points in the source image and the target image;

[0094] The image alignment unit maps each pixel coordinate in the source image to the target image using the perspective transformation matrix.

[0095] The above is based on the ideal embodiment of the present invention. Through the above description, relevant personnel can make various changes and modifications without departing from the technical concept of the present invention. The technical scope of the present invention is not limited to the content in the specification, and the technical scope must be determined according to the scope of the claims.

Claims

1. A real-time air-to-ground mapping and analysis method based on aerospace vehicles, characterized in that: The following steps are involved: Step S1, receiving real-time images of different directions transmitted by the drone; Step S2, extracting the same feature points of a number of images in different orientations, and calculating the coordinates of the feature points in the images in different orientations in combination with the altitude of the drone; Step S3, stitching a number of images in different orientations, fusing the feature point coordinates and the stitched images through registration technology, and generating a high-precision image with ground height information; Step S4: extract depth information from images of different orientations, obtain a depth map from a two-dimensional image, generate three-dimensional point cloud data, and convert it into a three-dimensional mesh model.

2. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1 is characterized by: In step S1, the drone takes real-time images of the same area at different locations; Different directions include: looking down, looking sideways, and looking horizontally.

3. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1 is characterized by: In the step S2, the same feature points of a plurality of images in different orientations are extracted, specifically, the image is divided into grids, each grid containing a plurality of feature points; and the height value of the feature point is increased within the grid using an interpolation algorithm; Use feature matching algorithm to find corresponding feature points between different images; calculate transformation matrix based on matched feature points to align the images.

4. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1, characterized in that: In step S2, the coordinates of feature points in a number of images in different orientations are calculated in combination with the altitude of the drone, including the following steps: Step S21: Real-time acquisition of the drone's location information P = [x u ,y u ,h], where x u ,y u is the plane coordinate of the drone relative to the ground, h is the flight altitude, and the drone attitude is obtained; Step S22: extracting feature points in images of different orientations by image processing methods. The feature points have pixel coordinates (x i ,y i ); Step S23: perform back projection based on the camera internal and external parameters and the height of the drone, and calculate the three-dimensional coordinates of the feature points using the pixel coordinates of the feature points in the image and the corresponding depth information. where d i is the depth information obtained from the depth map, K -1 is the inverse of the internal parameter matrix; Step S24: Based on the flight altitude of the UAV, the vertical position of the feature point is corrected. i =hd i , Z i Indicates the true height of the feature point.

5. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1, characterized in that: In step S21, the drone posture includes: the pitch, roll and yaw angles of the drone.

6. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1, characterized in that: In the step S3, the following sub-steps are included: Step S31, using a brute force matching algorithm to compare feature points in images of different orientations one by one, and determine identical feature point pairs; Step S32: Select feature points and use the RANSAC algorithm to calculate the transformation matrix H. Among them, (x, y) and (x a ,y a ) are the coordinates of the matching feature points in the source image and the target image respectively; Step S33, using the perspective transformation matrix to map each pixel coordinate in the source image to the target image for alignment; Step S34: fuse the aligned images, and take the average value of all images for the pixel values ​​in the overlapping area.

7. The method for real-time air-to-ground mapping and analysis based on aerospace vehicles according to claim 1, characterized in that: In step S4, the following sub-steps are included: Step S41, calculating the depth information of the object by the disparity of the same feature point in a plurality of images in different orientations, to obtain a depth map in the two-dimensional image; Step S42: convert the depth information of each pixel in the depth map coordinate system into point coordinates (X, Y, Z) in three-dimensional space; Step S43: Use a triangulation algorithm to convert the point cloud data into a triangular mesh and then into a three-dimensional mesh model.

8. Air-to-ground real-time mapping and analysis system based on aerospace vehicles, including: An image receiving module, a feature point extraction module, a coordinate calculation module, an image registration module, a stereoscopic vision and depth information calculation module, and a data storage and output module are characterized in that: The image receiving module is used to receive images of different orientations transmitted by the drone in real time and monitor the altitude position of the drone in real time; The feature point extraction module is used to extract the same feature points in images of different orientations; The coordinate calculation module calculates the coordinates of the feature points of the objects in the image by combining the height position of the drone obtained in the image receiving module; The image registration module is used to stitch images of different orientations through registration technology and fuse feature point coordinates to generate a high-precision image with ground height information; The stereoscopic vision and depth information calculation module is used to perform parallax calculation on the same feature points in images of different orientations, thereby obtaining the depth information of the object, generating a depth map in the two-dimensional image, and converting the depth information of each pixel point in the depth map coordinate system into point coordinates in three-dimensional space to obtain three-dimensional point cloud data, converting the point cloud data into a triangular mesh, and further generating a three-dimensional mesh model; The data storage and output module is used to store and output the finally generated high-precision image with ground height information, three-dimensional point cloud data and three-dimensional grid model.

9. The air-to-ground real-time mapping and analysis system based on aerospace vehicles according to claim 8, characterized in that: The feature point extraction module includes: a feature point extraction unit, a grid division unit, and an interpolation unit; The feature point extraction unit extracts the same feature points from images of different orientations using an image processing algorithm; The grid division unit is used to divide the image into grids, each grid containing a number of feature points; The interpolation unit calculates the height values ​​of the unobtained feature points in the grid by using bilinear interpolation.

10. The air-to-ground real-time mapping and analysis system based on aerospace vehicles according to claim 8, characterized in that: The image registration module includes: a feature point matching unit, a transformation matrix calculation unit, and an image alignment unit; The feature point matching unit compares the feature points in the images of different orientations one by one using a brute force matching algorithm to determine the same feature point pairs; The transformation matrix calculation unit calculates the transformation matrix between images using the RANSAC algorithm, including the coordinates of matching feature points in the source image and the target image; The image alignment unit maps each pixel coordinate in the source image to the target image using a perspective transformation matrix.

Images