Method and apparatus for multi-drone cooperative mapping in wide-area environments

Abstract

To provide a multi-drone cooperative map creation method and device that can combine compositions of a plurality of drones totally to create a wide-range environmental map.SOLUTION: A method includes the steps of: automatically dividing multi-drone images, in which matching of identical points among images has been completed, into a plurality of sub-areas so that image nodes in the plurality of divided sub-areas include in-area points appearing only in one sub-area and area connection points appearing simultaneously in different divided areas; and obtaining an aerial triangulation calculation result by using image nodes in the respective divided sub-areas to unify coordinates of the divided sub-areas through parallel aerial triangulation calculation using an area network cooperatively set by multiple drones, performing error integration compensation after eliminating issues of internal parameters and distortion of cameras, and automatically integrating the divided sub-areas thereafter.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to the technical field of drone mapping, and more particularly to a method and apparatus for multi-drone cooperative mapping in a wide-area environment. [Background technology]

[0002] The map area that a single drone takes photos of using its onboard camera cannot cover a wide area due to the camera's angle of view.

[0003] It is technically necessary to expand the coverage area by combining images taken by multiple drones, and it is technically difficult to achieve multi-drone collaborative mapping in wide-area environments, especially in the absence of GNSS-assisted positioning. Summary of the Invention [Problem to be solved by the invention]

[0004] In view of the above analysis, the objective of the present invention is to provide a method and apparatus for multi-drone collaborative map construction in a wide-area environment, thereby integrating the compositions of multiple drones as a whole to realize the generation of a wide-area environmental map. [Means for solving the problem]

[0005] The present invention discloses a method for multi-drone cooperative mapping in a wide-area environment, the method comprising: Step S1: automatically dividing the images of the multi-drone camera into a plurality of sub-areas after matching of identical points between the images is completed; In the automatic division process, a topology graph between images is created based on points with the same name that appear multiple times between images, and after graph joining is performed on the topology graph, the multi-drone image is divided into multiple divided sub-areas so that image nodes in the divided sub-areas include area internal points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas. Step S2 of automatically integrating the divided sub-areas and obtaining an aerial triangulation calculation result of an area network set by the multiple drones in cooperation; The automatic integration process includes a step S2 in which the image nodes of each divided sub-area are used to perform aerial triangulation calculations in parallel, and different three-dimensional coordinates obtained by performing aerial triangulation calculations on different divided sub-areas that have the same area connection point are used to unify the coordinate systems of the divided sub-areas, and then the integration error compensation process is performed by eliminating the internal parameters and distortion problems of the drone camera, and then the aerial triangulation calculation results of the area network set up in cooperation by the multiple drones are obtained.

[0006] Furthermore, in step S1, a process of creating a topology graph between images based on points with the same name that appear multiple times between images and performing graph connection on the topology graph is performed as follows: Step S101: Based on the results of the same-name point matching between the images of the multiple drones, an undirected weighted graph G = (V, E) of image nodes is created, where V is a point set consisting of image nodes, and E is a set of weight values ​​of graph edges consisting of distances between image nodes; Step S102: Initialize a graph G0=G, and perform graph joining of the multi-drone images based on k adjacent nodes. In the graph joining, neighboring nodes are joined based on a set neighboring node threshold k to obtain a new node; The new nodes are used to create a new undirected weighted graph G * Step S103: G0=G * Steps S102 and S103 are repeated to continue combining the graph nodes to obtain the final graph G * Step S104 to obtain Final graph G *and step S105 of determining the images to be included in each divided sub-area based on the images included in each node in the image, dividing connection points with the same name that appear multiple times, and dividing a set of internal points included in each divided sub-area and a set of connection points connecting different sub-areas.

[0007] Furthermore, step S102 1) Initialize a graph G0 = G, and 2) Take the first node V0 of graph G0 and find the k neighbors of V0; TIFF0007787953000001.tif51170

[0008] TIFF0007787953000002.tif32170

[0009] Furthermore, step S2 specifically includes the following steps: Step S201: normalizing the area coordinate systems of the respective sub-areas into one unified coordinate system to complete the unification as the coordinate system of the entire area network; Step S202: integrate the intrinsic camera parameters and distortions calculated for each sub-area, and first eliminate the intrinsic camera parameters and distortion problems; and step S203 of compensating for integration errors, compensating for integration errors that occurred during the integration processes of steps S201 and S202, and obtaining aerial triangulation calculation results for the area network set up by the multiple drones in cooperation.

[0010] Furthermore, step S201 1) Each sub-area is treated as a node, and two adjacent sub-areas are connected by an area connection point, and a graph is created as an edge between the two nodes. The weight value of the graph edge is determined based on the number of valid connection points between the two nodes; 2) Finding the node with the highest degree in the graph and creating a minimum spanning tree as the root node of the spanning tree, and then determining the optimal path for coordinate system transformation by following the minimum spanning tree method in hierarchical order; 3) Along the optimal route, the coordinate system of the divided sub-area corresponding to the root node is taken as the reference coordinate system, and the aerial triangulation calculation results of other sub-areas are sequentially converted into the reference coordinate system, thereby completing the unification of the coordinate system of the entire area network.

[0011] TIFF0007787953000003.tif37170b.The rotation matrix between the two point sets is calculated based on SVD decomposition, and the large error points of the point set are removed using the RANSAC algorithm. i and C j and calculating coordinate system transformation parameters between CD i and D j Based on the coordinate system transformation parameters between j The results of aerial triangulation calculations are i and converting it to the coordinate system of d. Scanning the optimal route and unifying the aerial triangulation calculation results of all sub-areas into a reference coordinate system; e. Adding all area connection points and corresponding image points into a set of connection points that connect different areas to the integrated data.

[0012] Furthermore, in step S202, each sub-area undergoes self-calibration and bundle adjustment individually, and then obtains the camera internal parameters and distortion, and the average values ​​of the camera internal parameters and distortion of each sub-area are used as the overall camera internal parameters and distortion.

[0013] Furthermore, in step S203, steps S201 and S202 are combined to perform the overall self-calibration and bundle adjustment again, and the result after this adjustment is used as the final aerial triangulation calculation result for the entire area network.

[0014] The present invention also discloses an apparatus for realizing the above-mentioned method for multi-drone cooperative mapping in a wide-area environment, the apparatus including an automatic segmentation module and an automatic integration module; The automatic segmentation module is used to automatically segment the multi-drone images into multiple sub-areas after matching of the same points between the images is completed; In the automatic division process, an inter-image topology graph is created based on the same points that appear multiple times between images, and after graph joining is performed on the topology graph, the multi-drone image is divided into multiple divided sub-areas, and the image nodes in the divided sub-areas include area internal points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas; The automatic integration module is used to automatically integrate the divided sub-areas and obtain aerial triangulation calculation results of an area network set up by multiple drones in cooperation; In the automatic integration process, the image nodes in each divided sub-area are used to perform aerial triangulation calculations in parallel, and the different three-dimensional coordinates obtained by performing aerial triangulation calculations at the same area connection points in different divided sub-areas are used to unify the coordinates of the divided sub-areas.Furthermore, after performing integration error compensation processing by processing the internal parameters of the drone camera and eliminating distortion problems, the aerial triangulation calculation results of the area network set up in cooperation by multiple drones are obtained. [Effects of the Invention]

[0015] The present invention can achieve at least the following beneficial effects: The method and device for multi-drone cooperative map creation in a wide-area environment disclosed in the present invention realizes map creation by combining screen information only, without relying on navigation information, and automatically divides and integrates sub-areas of the screens captured by multiple drones, integrates the screens of each drone as a whole, and eliminates combining errors, thereby creating a map of a wide-area environment, improving the data processing speed and the accuracy and reliability of the map; The data collected by different drones is spatially matched to avoid data conflicts and error accumulation, and parallel computing is optimized to improve the computing efficiency during the map creation process, improving the robustness of the map creation process, reducing manual intervention, reducing operational complexity, and improving work efficiency.

[0016] Based on the above, the present invention can be applied to mapping a wide range of environments, whether urban, forest or other complex terrain, and can provide an effective solution. [Brief explanation of the drawings]

[0017] The drawings are only for purposes of illustrating particular embodiments and are not to be construed as limiting the invention, and like reference numerals refer to like parts throughout the drawings. [Figure 1] 1 is a flow chart of a method for multi-drone cooperative mapping in a wide-area environment according to an embodiment of the present invention; [Figure 2] 10 is a flowchart showing a process for creating a topology graph between images based on points with the same name that appear multiple times between the images, and performing graph connection on the topology graph, in an embodiment of the present invention. [Figure 3] FIG. 10 is a diagram showing the result after automatic division of subareas in an embodiment of the present invention. [Figure 4] 1 is a flowchart of an automatic integration process according to an embodiment of the present invention. [Figure 5] 1 is a schematic block diagram of a multi-drone cooperative mapping system in a wide-area environment according to an embodiment of the present invention; FIG. DETAILED DESCRIPTION OF THE INVENTION

[0018] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will now be described in detail with reference to the drawings, which form a part of this application and, together with the embodiments of the present invention, serve to explain the principles of the present invention.

[0019] Example 1 One embodiment of the present invention discloses a multi-drone cooperative mapping method in a wide-area environment, as shown in FIG. 1, which includes: Step S1: automatically dividing the images of the multi-drone camera into a plurality of sub-areas after matching of identical points between the images is completed; In the automatic division process, a topology graph between images is created based on points with the same name that appear multiple times between images, and after performing graph joining on the topology graph, the multi-drone image is divided into multiple divided sub-areas so that image nodes in the divided sub-areas include area internal points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas. Step S2 of automatically integrating the divided sub-areas and obtaining an aerial triangulation calculation result of an area network set by the multiple drones in cooperation; The automatic integration process includes step S2, in which aerial triangulation calculations are performed in parallel using the image nodes of each divided subarea, and different three-dimensional coordinates obtained by performing aerial triangulation calculations on different divided subareas that have the same area connection point are used to unify the coordinate systems of the divided subareas, and then performing integration error compensation processing by eliminating the internal parameters and distortion problems of the drone cameras, and then obtaining the aerial triangulation calculation results of the area network set up in cooperation by multiple drones.

[0020] Specifically, the multiple drones each take images of the ground using their onboard cameras, and then use existing technology to perform similarity and consistency analysis based on correspondences such as image content, features, structure, relationships, texture, and grayscale, thereby achieving matching of identical points between images.

[0021] Specifically, in step S1, a topology graph between images is created based on points with the same name that appear multiple times between images, and the process of performing graph joining on the topology graph is as follows, as shown in FIG. 2: Based on the results of the same-name point matching between the images of the multiple drones, an undirected weighted graph G = (V, E) of image nodes is created, where V is a point set consisting of image nodes, and E is a set of weight values ​​of graph edges consisting of distances between image nodes. Step S101: After matching the same points, set edge weights between the nodes of different images to lay the foundation for subsequent merging; Step S102: Initialize a graph G0=G and perform graph joining of the multi-drone images based on k adjacent nodes. In the graph joining, the adjacent nodes are joined based on a set adjacent node threshold k to obtain a new node. Specifically, step S102 is 1) Initialize a graph G0 = G, and 2) Take the first node V0 of graph G0 and find the k neighbors of V0; TIFF0007787953000004.tif51170

[0022] TIFF0007787953000005.tif32170

[0023] The distance calculation method between two nodes used to calculate the distance between a small node and a non-small node is to calculate the sum of the distances from all image nodes of one node of the graph to all image nodes of the other node, and use this sum as the distance between the two nodes.

[0024] In step S103, the new node is used to create a new undirected weighted graph G * The weight value of the graph edge is set to the distance between the new nodes, and the calculation method uses the distance calculation method between two nodes in step S102. In step S104, G0=G * Then, steps S102 and S103 are repeated to combine the graph nodes to form the final graph G * and Final graph G * The number of nodes in is closest to the number of subareas to be divided and is equal to or greater than the number of subareas to be divided, In step S105, the images included in the divided sub-areas are determined based on the images included in each node in the final graph G*, and connection points with the same name that appear multiple times are divided to separate the set of internal points included in each divided sub-area and the set of connection points connecting different sub-areas.

[0025] The method for dividing connection points with the same name that appear multiple times is to determine the sub-area to which each connection point belongs for the image corresponding to that connection point, and if the connection points belong to the same sub-area, the connection points are grouped into one set.If the connection points belong to multiple different sub-areas, those sub-areas are grouped into one set.

[0026] As shown in Figure 3, the result after automatic segmentation of sub-areas is shown. In the figure, there are seven images I1,···,I7 and x1,···,x 11 There are a total of 11 multiple-appearing same-name connection points, and Images I1 and I2 contain points with the same name x1 and x2, and image I1 also contains a point with the same name x4, just like image I6, and image I2 also contains a point with the same name x3, just like image I3. The correspondence between points with the same name in other images can be intuitively obtained from the diagram.

[0027] After the automatic segmentation in step S1, the seven images are divided into three areas D1={I1,I2}, D2={I3,I4,I5}, D3={I6,I7}. There are a total of 11 object points corresponding to the same-named connected points that appear multiple times. By segmenting the same-named connected points that appear multiple times, four sets of point sets are obtained: O1={x1,x2} is the internal point set of D1, O2={x5,x6,x7} is the internal point set of D2, and O3={x 10 ,x 11} is the interior point set of D3, and O L ={x3,x4,x8,x9} is a set of connection points that connect different areas.

[0028] Specifically, in step S2, the automatic integration process performs aerial triangulation calculations using image nodes of each divided sub-area; By dividing the image and connection points in step S1, the entire area network can be divided into multiple independent subareas and the connection points between the areas. For each independent subarea, independent aerial triangulation calculations are performed in parallel according to the serial aerial triangulation method, making full use of the computer's multi-core CPU and memory performance at this stage. The area connection points are the input required for automatic area network integration after the aerial triangulation of each subarea is completed.

[0029] In this embodiment, the aerial triangulation calculation method is not improved, but the existing aerial triangulation calculation method is adopted to realize the aerial triangulation calculation.

[0030] As shown in FIG. 4, the automatic integration process in step S2 is as follows: Step S201: normalizing the area coordinate systems of each sub-area into one unified coordinate system to complete the unification of the coordinate system of the entire area network; Area connection point O L is the key for subarea integration. L Since the points in are visible in multiple sub-areas, independent aerial triangulation was performed in each sub-area, and then O L Many of the points correspond to multiple 3D coordinates in different sub-areas, and some points are invalid due to calculation failures in some sub-areas (corresponding image points are mismatched or there is insufficient overlap in a single sub-area).

[0031] TIFF0007787953000006.tif591702) Find the node with the highest degree in the graph and create a minimum spanning tree as the root node of the spanning tree, then follow the minimum spanning tree method in hierarchical order to determine the optimal path for coordinate system transformation; 3) Along the optimal route, the coordinate system of the divided sub-area corresponding to the root node is used as the reference coordinate system, and the aerial triangulation results (image external parameters, object point coordinates) of other sub-areas are sequentially converted into the reference coordinate system, thereby completing the unification of the coordinate system of the entire area network.

[0032] TIFF0007787953000007.tif35170b. The rotation matrix between the two point sets is calculated based on SVD decomposition, and the large error points of the point set are removed using the RANSAC algorithm. i and C j Calculate the coordinate system transformation parameters Δλ, ΔR, and Δt between the two, where Δλ is a translation parameter, ΔR is a rotation parameter, and Δt is a scale parameter. The SVD decomposition is used to create an initial estimate of the rotation matrix, and then the RANSAC algorithm is used to iterate and find the interior point set to reliably estimate the final rotation matrix. The translation parameters, rotation parameters, and scale parameters can also be estimated using the least squares method. Here, SVD decomposition can be performed using a linear algebra library (e.g., NumPy in Python), and an existing RANSAC algorithm can be used to iteratively select point pairs, fit the model, evaluate the error, and select interior points. Least-squares optimization can be performed using a nonlinear least-squares method (e.g., CeresSolver or the Levenberg-Marquardt algorithm) to optimize the model parameters. c.D i and D j Based on the coordinate system transformation parameters Δλ, ΔR, and Δt between j The results of aerial triangulation calculations are i and converting it to the coordinate system of d. Scanning the optimal route and unifying the aerial triangulation calculation results of all sub-areas into a reference coordinate system; e.O L and adding all area connection points and corresponding image points in

[0033] In step S202, the camera internal parameters and distortions calculated for each sub-area are integrated to first eliminate the problems of the camera internal parameters and distortions; After each sub-area undergoes self-calibration and bundle adjustment individually, different camera intrinsic parameters and distortions are obtained.

[0034] When integrating into the entire area network, the average values ​​of the camera internal parameters and distortions of each sub-area are used as the overall camera internal parameters and distortions.

[0035] The camera internal parameters and distortion obtained by taking the average values ​​in this step are used as initial values, and corrections are made in the next step, "integrated error compensation" processing, to obtain accurate results.

[0036] In step S203, the integration error is compensated for, and the integration error that occurred during the integration process in step S201 and step S202 is compensated for, to obtain the aerial triangulation calculation result of the area network that is set up in cooperation by the multiple drones.

[0037] In the integration process of step S201, the accuracy of the transformation parameters between the two coordinate systems is affected by the accuracy of the area connection points used in the calculation. If the accuracy of the connection points is not good, the results for the entire sub-area will be biased and will be further propagated between the sub-areas. Therefore, in the integration process of step S202, the camera internal parameters and distortion are also rough estimates.

[0038] To compensate for errors that occur during the integration process, the method adopted in this specification performs global self-calibration and bundle adjustment again after the integration in steps S201 and S202, and the result after this adjustment is used as the final aerial triangulation calculation result for the entire area network. Because locally high-accuracy results are obtained through bundle adjustment within subareas, convergence is facilitated during the global bundle adjustment after integration.

[0039] In summary, the multi-drone collaborative map creation method for a wide-area environment according to an embodiment of the present invention creates maps solely by combining image information, without relying on navigation information. It automatically divides and combines subareas based on the images captured by multiple drones, combines the images of each drone as a whole, and eliminates combining errors, thereby creating a map of a wide-area environment.

[0040] <Example 2> One embodiment of the present invention discloses a multi-drone collaborative mapping device for a wide-area environment, as shown in FIG. 5, the device includes an automatic segmentation module and an automatic integration module; The automatic segmentation module is used to automatically segment the multi-drone images into multiple sub-areas after matching the same points between the images is completed; In the automatic division process, an inter-image topology graph is created based on points with the same name that appear multiple times between images, and after graph joining is performed on the topology graph, the multi-drone image is divided into multiple divided sub-areas, so that image nodes in the divided sub-areas include area internal points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas; The automatic integration module is used to automatically integrate the divided sub-areas and obtain aerial triangulation calculation results of an area network set up by multiple drones in cooperation; In the automatic integration process, aerial triangulation calculations are performed in parallel using the image nodes of each divided subarea, and the different three-dimensional coordinates obtained by performing aerial triangulation calculations on different divided subareas that have the same area connection point are used to unify the coordinate systems of the divided subareas.Furthermore, after performing integration error compensation processing by eliminating the internal parameters and distortion problems of the drone camera, the aerial triangulation calculation results of the area network set up in cooperation by multiple drones are obtained.

[0041] The specific implementation details and beneficial effects of this embodiment are the same as those disclosed in the first embodiment, so please refer to the first embodiment for details and will not be described here.

[0042] The above contents are merely preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any modifications or replacements that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention are included in the protection scope of the present invention.

Claims

1. 1. A method for multi-drone cooperative mapping in a wide-area environment, comprising: Step S1: automatically dividing the multi-drone images into a plurality of sub-areas after matching of identical points between the images is completed; In the automatic division process, a topology graph between images is created based on points with the same name that appear multiple times between images, and after graph joining is performed on the topology graph, the multi-drone image is divided into multiple divided sub-areas, and image nodes in the divided sub-areas include area interior points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas. Step S2: automatically integrating the divided sub-areas and obtaining an aerial triangulation calculation result of an area network set up by the multiple drones in cooperation; The automatic integration process includes step S2, in which aerial triangulation calculations are performed in parallel using the image nodes of each divided subarea, and different three-dimensional coordinates obtained by performing aerial triangulation calculations on different divided subareas that have the same area connection point are used to integrate the coordinate systems of the divided subareas, and then performing an integration error compensation process by eliminating the internal parameters and distortion problems of the drone camera, and then obtaining the aerial triangulation calculation results of the area network set up in cooperation by multiple drones. A multi-drone collaborative map creation method characterized by:

2. In step S1, a topology graph between images is created based on the same points that appear multiple times between the images, and the process of performing graph connection on the topology graph is as follows: Step S101: Based on the results of matching the same points between the images of the multiple drones, an undirected weighted graph G = (V, E) of image nodes is created, where V is a point set consisting of image nodes, and E is a set of weight values ​​of graph edges composed of distances between image nodes; Graph G ０ Step S102: Initialize = G, and perform graph joining of multi-drone images based on k adjacent nodes. In the graph joining, neighboring nodes are joined based on the set neighboring node threshold k to obtain a new node; The new nodes are used to create a new undirected weighted graph G ＊ Step S103: G ０ = G ＊ Steps S102 and S103 are repeated to continue combining the graph nodes to form the final graph G ＊ Step S104 to obtain Final Graph G ＊ and step S105 of determining an image included in each divided sub-area based on an image included in each node in the image, dividing connection points with the same name that appear multiple times, and dividing a set of internal points included in each divided sub-area and a set of connection points connecting different sub-areas. The multi-drone cooperative map building method according to claim 1 .

3. Step S102 is 1) Graph G ０ = Initializing G, and 2) Graph G ０ The first node V ０ and V ０ Finding the k neighbors of

4.

5. Specifically, step S2 is Step S201: normalizing the area coordinate systems of the respective sub-areas into one unified coordinate system to complete the unification as the coordinate system of the entire area network; Step S202: Integrating the camera intrinsic parameters and distortions calculated for each sub-area, and first eliminating the camera intrinsic parameters and distortion problems; and step S203 of compensating for integration errors, compensating for integration errors that occurred during the integration processes of steps S201 and S202, and obtaining aerial triangulation calculation results for the area network set up by the multiple drones in cooperation. The multi-drone cooperative map building method according to claim 1 .

6. Step S201 1) Each sub-area is treated as one node, and two adjacent sub-areas are connected by an area connection point, and a graph is created as an edge between the two nodes. The weight value of the graph edge is determined based on the number of valid connection points between the two nodes; 2) Finding the node with the highest degree in the graph and creating a minimum spanning tree as the root node of the spanning tree, and then determining the optimal path for coordinate system transformation by following the minimum spanning tree method in hierarchical order; 3) along the optimal route, the coordinate system of the divided sub-area corresponding to the root node is taken as the reference coordinate system, and the aerial triangulation calculation results of other sub-areas are sequentially converted into the reference coordinate system, thereby completing the unification of the coordinate system of the entire area network. The multi-drone cooperative map building method according to claim 5 .

7. b. A rotation matrix between two point sets is calculated based on SVD decomposition, and large error points of the point sets are removed using the RANSAC algorithm, and the point set C i and C j and calculating coordinate system transformation parameters between c.D i and D j Based on the coordinate system transformation parameters between j The results of the aerial triangulation calculation are i and converting it to the coordinate system of d. Scanning the optimal route and unifying the aerial triangulation calculation results of all sub-areas into a reference coordinate system; e. Adding the connection points and corresponding image points of all areas into a set of connection points that connect different areas to the integrated data. The multi-drone cooperative map building method according to claim 6 .

8. In step S202, each sub-area undergoes self-calibration and bundle adjustment individually, and then acquires the camera internal parameters and distortion. The average values ​​of the camera internal parameters and distortion of each sub-area are set as the overall camera internal parameters and distortion. The multi-drone cooperative map building method according to claim 5 .

9. In step S203, steps S201 and S202 are combined to perform the whole self-calibration and bundle adjustment again, and the adjusted result is used as the final aerial triangulation calculation result for the whole area network. The method for multi-drone cooperative mapping according to any one of claims 5 to 8, characterized in that:

10. An apparatus for implementing the multi-drone cooperative mapping method of claim 1, the apparatus including: an automatic segmentation module and an automatic integration module; The segmentation module is used to automatically segment the multi-drone images into multiple sub-areas after matching the same points between the images; In the automatic division process, an inter-image topology graph is created based on points with the same name that appear multiple times, and after graph joining is performed on the topology graph, the multi-drone image is divided into multiple divided sub-areas, and the image nodes in the divided sub-areas include area internal points that appear only in one sub-area and area connection points that appear simultaneously in different divided areas; The automatic integration module is used to automatically integrate the divided sub-areas and obtain aerial triangulation calculation results of an area network set up by multiple drones in cooperation; In the automatic integration process, the image nodes in each divided sub-area are used to perform aerial triangulation calculations in parallel, and the coordinates of the divided sub-areas are unified using the different three-dimensional coordinates obtained by performing aerial triangulation calculations at the same area connection points in different divided sub-areas.Furthermore, after performing integration error compensation processing by eliminating the internal parameters and distortion problems of the drone camera, the aerial triangulation calculation results of the area network set up by the multi-drones are obtained. A multi-drone collaborative map creation device characterized by:

Images