Vector map-based unmanned aerial vehicle visual geographic positioning method and device
Through multi-layer encoding and decoding processing based on vector maps, combined with convolutional neural network feature extraction, efficient, low-cost and high-precision positioning of drone visual geolocation is achieved, solving the high storage and high maintenance problems caused by reliance on satellite images.
Patent Information
- Application Number
- CN202510804288.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-10-17
AI Technical Summary
In existing technologies, drone visual geolocation relies on satellite images or 3D point clouds, which is costly and takes up a lot of storage space, and requires frequent updates to adapt to changes in the visual appearance of the environment.
A UAV visual geolocation method based on vector maps is adopted. By performing multi-layer encoding and decoding on the environmental images collected by the UAV, combining the vectorized vector map features, using convolutional neural networks to extract map features, and realizing the visual geolocation of the UAV through rotation template matching.
It reduces data processing complexity and storage costs, improves the accuracy and robustness of drone visual geolocation, and adapts to environmental changes across seasons and scenarios.
Smart Images
Figure CN120800370A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of unmanned aerial vehicle positioning and navigation, and particularly relates to an unmanned aerial vehicle visual geographic positioning method and device based on a vector map. BACKGROUND
[0002] Unmanned aerial vehicles (UAVs) have been widely used in the fields of inspection, agriculture, and even military affairs. However, the UAVs frequently suffer from interference such as occlusion and multipath interference when acquiring global navigation satellite system (GNSS) signals, which seriously reduces the positioning accuracy. In the environment where the GNSS signals are limited, visual geographic positioning gradually shows its great potential by matching the images taken by the UAVs with geographic reference images (such as satellite maps) to achieve accurate positioning.
[0003] The related art usually relies on satellite images or 3D point clouds to compare the environment images acquired by the UAVs with the satellite images or 3D point clouds to achieve visual geographic positioning of the UAVs. The satellite images or 3D point clouds are costly in data collection and need to be frequently updated to adapt to the changes in the visual appearance of the environment, which leads to an increase in maintenance costs and also requires a large amount of storage space. SUMMARY
[0004] In order to solve the problem of high cost and large storage space in the prior art, the present application provides an unmanned aerial vehicle visual geographic positioning method and device based on a vector map.
[0005] In the first aspect, the present application provides an unmanned aerial vehicle visual geographic positioning method based on a vector map, which can include: extracting image features from environment images collected by a UAV. Vectorizing a vector map and extracting map features from the vectorized vector map. Visual geographic positioning of the UAV is performed according to the image features and the map features.
[0006] In some possible implementation manners, the image features are extracted from the environment images collected by the UAV, which includes:
[0007] The environment images are multi-layer encoded to obtain first image encoding information and a plurality of second image encoding information. The first image encoding information is used to indicate the image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate the image encoding information obtained by each layer of encoding.
[0008] The first image coding information and the plurality of second image coding information are subjected to multi-layer decoding to obtain first image decoding information and a plurality of second image decoding information. The first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by decoding each layer.
[0009] The first image decoding information and the plurality of second image decoding information are fused to obtain image features.
[0010] In some possible implementation manners, the vector map is subjected to vectorization processing, including:
[0011] The vector map is subjected to rasterization processing to obtain a raster map.
[0012] The raster map is subjected to feature transformation to obtain vector features of the raster map.
[0013] The weight corresponding to each type of ground surface element in the raster map is subjected to normalization processing to obtain a normalized weight.
[0014] The vector features of the raster map and the normalized weight are multiplied to obtain a vector map subjected to vectorization processing.
[0015] Optionally, the vector features of the raster map and the normalized weight are multiplied to obtain a vector map subjected to vectorization processing, including:
[0016] The vector features of the raster map and the normalized weight are multiplied by Hadamard product to obtain a vector map subjected to vectorization processing.
[0017] In some possible implementation manners, a map feature is extracted from the vector map subjected to vectorization processing, including:
[0018] The map feature is extracted from the vector map subjected to vectorization processing by a convolutional neural network.
[0019] In some possible implementation manners, the unmanned aerial vehicle is subjected to visual geolocation according to the image feature and the map feature, including:
[0020] The image feature is rotated a plurality of times according to a preset rotation angle to obtain a plurality of rotation templates.
[0021] The map feature is matched with each rotation template in the plurality of rotation templates to obtain a plurality of matching probabilities.
[0022] A coordinate corresponding to a maximum value of the plurality of matching probabilities is selected as a pose of the image feature relative to the map feature.
[0023] The pose of the UAV is obtained according to the pose of the image feature relative to the map feature and the vector map, wherein the pose of the UAV includes a position (which can be represented by latitude and longitude) and an attitude (which can be represented by a heading angle) of the UAV.
[0024] Exemplarily, the map feature is matched with each of the plurality of rotation templates to obtain a plurality of matching probabilities, including:
[0025] The Fourier transform of the map feature is multiplied by the complex conjugate of the Fourier transform of each rotation template to obtain the correlation of the map feature and each rotation template in the frequency domain.
[0026] The Fourier inverse transform is performed on the correlation of the map feature and each rotation template in the frequency domain to obtain the plurality of matching probabilities.
[0027] In a second aspect, the present application provides a vector map-based visual geolocation device for a UAV, which can include:
[0028] A first extraction module configured to extract an image feature from an environment image collected by the UAV.
[0029] A second extraction module configured to perform vectorization processing on a vector map and extract a map feature from the vectorized vector map.
[0030] A positioning module configured to perform visual geolocation on the UAV according to the image feature and the map feature.
[0031] In some possible implementation manners, the first extraction module is specifically configured to:
[0032] The environment image is multi-layer encoded to obtain first image encoding information and a plurality of second image encoding information. The first image encoding information is used to indicate image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate image encoding information obtained by each layer of encoding.
[0033] The first image encoding information and the plurality of second image encoding information are multi-layer decoded to obtain first image decoding information and a plurality of second image decoding information. The first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by each layer of decoding.
[0034] The first image decoding information and the plurality of second image decoding information are fused to obtain the image feature.
[0035] In other possible implementation manners, the second extraction module is specifically configured to:
[0036] The vector map is rasterized to obtain a raster map.
[0037] The grid map is subjected to feature transformation to obtain vector features of the grid map.
[0038] The weight corresponding to each type of ground surface element in the grid map is normalized to obtain a normalized weight.
[0039] The vector features of the grid map are multiplied by the normalized weight to obtain a vector map subjected to vectorization processing.
[0040] Optionally, the second extraction module is specifically configured to:
[0041] The vector features of the grid map are multiplied by the normalized weight by using Hadamard product to obtain a vector map subjected to vectorization processing.
[0042] Optionally, the second extraction module is specifically configured to:
[0043] The map features are extracted from the vector map subjected to vectorization processing by using a convolutional neural network.
[0044] In still some possible implementation manners, the positioning module is specifically configured to:
[0045] The image features are rotated multiple times according to a preset rotation angle to obtain multiple rotation templates.
[0046] The map features are matched with each rotation template in the multiple rotation templates to obtain multiple matching probabilities.
[0047] The maximum value of the multiple matching probabilities is selected as a coordinate of a pose of the image features relative to the map features.
[0048] The pose of the unmanned aerial vehicle is obtained according to the pose of the image features relative to the map features and the vector map, wherein the pose of the unmanned aerial vehicle includes a position and an attitude of the unmanned aerial vehicle.
[0049] Optionally, the positioning module is specifically configured to:
[0050] The Fourier transform of the map features is multiplied by the complex conjugate of the Fourier transform of each rotation template to obtain the correlation of the map features and each rotation template in the frequency domain.
[0051] The Fourier inverse transform is performed on the correlation of the map features and each rotation template in the frequency domain to obtain the multiple matching probabilities.
[0052] In still another aspect, the present application further provides a computer device, comprising: one or more processors.
[0053] The processor is configured to execute one or more programs.
[0054] When the one or more programs are executed by the one or more processors, the evaluation method as described above is implemented.
[0055] In still another aspect, the application also provides a computer readable storage medium having a computer program stored thereon. The computer program, when executed, implements the evaluation method as described above.
[0056] Compared with the prior art, the application has the following beneficial effects:
[0057] The application realizes the visual geographic positioning of the UAV based on the vector map and the environment image. The data amount of the vector map is relatively small, and the vector map is easier to store and process. Moreover, the vector map can provide environmental features, which facilitates the UAV to quickly match and judge in the visual geographic positioning process. The application can not only improve the visual geographic positioning accuracy of the UAV, but also effectively reduce the complexity and cost of data processing, that is, reduce the cost of visual geographic positioning of the UAV and save storage space.
[0058] The application obtains first image encoding information and a plurality of second image encoding information by performing multi-layer encoding on the environment image, and obtains first image decoding information and a plurality of second image decoding information by performing multi-layer decoding on the first image encoding information and the plurality of second image encoding information. The image features can be obtained by fusing the first image decoding information and the plurality of second image decoding information. As can be seen, the application effectively copes with the scale difference caused by the change of flight height by multi-layer encoding and multi-layer decoding, while meeting the low delay requirement.
[0059] The application performs vectorization processing on the vector map, extracts map features from the vector map after vectorization processing, and can dynamically distinguish the contribution weight of key landmarks such as buildings and roads and secondary elements, thereby significantly improving the discriminability of the map features and each rotation template matching.
[0060] The application matches the map features with each rotation template in the plurality of rotation templates, selects the coordinate corresponding to the maximum value of the plurality of matching probabilities as the pose of the image features relative to the map features, and then obtains the pose of the UAV according to the pose of the image features relative to the map features and the vector map, thereby obtaining the parameter-free and efficient 3-DOF (latitude, longitude and heading angle) pose of the UAV.
[0061] The application makes full use of the vector map, and realizes the positioning of the UAV through geometric topological relationship instead of pixel-level visual features, thereby significantly reducing the data storage cost and improving the positioning robustness across seasons and scenes. BRIEF DESCRIPTION OF DRAWINGS
[0062] In order to more clearly illustrate the technical solutions in the application or the prior art, the accompanying drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced. Obviously, the accompanying drawings in the following description are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.
[0063] Figure 1 A schematic flow chart of the UAV visual geolocation method based on the vector map in the embodiments of the present application;
[0064] Figure 2 A schematic flow chart of the image feature extraction from the environment image collected by the UAV in the embodiments of the present application;
[0065] Figure 3 A schematic flow chart of the vectorization processing of the vector map and the map feature extraction from the vector map after the vectorization processing in the embodiments of the present application;
[0066] Figure 4 A schematic flow chart of the UAV visual geolocation according to the image feature and the map feature in the embodiments of the present application;
[0067] Figure 5 A schematic structural diagram of the UAV visual geolocation device based on the vector map in the embodiments of the present application. DETAILED DESCRIPTION
[0068] The technical solutions in the present application will be described below with reference to the accompanying drawings.
[0069] The terms "first", "second", etc. in the description of the embodiments of the present application and the claims and the accompanying drawings are only used for distinguishing the purposes of description, and cannot be understood as indicating or implying relative importance, and cannot be understood as indicating or implying sequence. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, inclusion of a series of steps or units. The method, system, product or device is not necessarily limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.
[0070] It should be understood that in this application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.
[0071] Example 1:
[0072] The embodiment of the present application provides a method for visual geolocation of drones based on vector maps. Figure 1 As shown, the positioning method 100 includes the following steps:
[0073] Step S1: Extract image features from the environment image collected by the UAV.
[0074] Step S2: performing vectorization processing on the vector map, and extracting map features from the vectorized vector map.
[0075] Step S3: Visually geolocate the UAV based on image features and map features.
[0076] In some possible implementations, in step S1, extracting image features from an environment image captured by a drone includes:
[0077] like Figure 2 As shown, the environment image is multi-layer encoded to obtain first image encoding information and multiple second image encoding information, wherein the first image encoding information is used to indicate the image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate the image encoding information obtained by encoding each layer.
[0078] Multi-layer decoding is performed on the first image coding information and the plurality of second image coding information to obtain first image decoding information and the plurality of second image decoding information, wherein the first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by decoding each layer.
[0079] The first image decoding information and the plurality of second image decoding information are fused to obtain image features.
[0080] In some possible implementation manners, the vectorization processing on the vector map in step S2 comprises:
[0081] As shown in Figure 3 , the rasterization processing is performed on the vector map to obtain a raster map. The raster map R map may be represented as:
[0082]
[0083] wherein k∈{1,2,...,K} represents different map element category indexes, covering common geographical elements such as buildings, roads, trees, rivers, and the like, K represents a total number of map elements, and M represents an element set.
[0084] The feature transformation is performed on the raster map to obtain vector features of the raster map. The vector features V′ map of the raster map may be represented as V′ map = f embed (R map (k)), f embed represents an embedding function, which is used to map the discrete category index R map (k) to a continuous vector space, thereby providing a basic representation for subsequent feature weighting.
[0085] The weight corresponding to each type of surface element in the raster map is normalized to obtain a normalized weight. The normalized weight W map may be represented as W map = softmax(W grid ,R map (k)), soft represents a normalization function, and W grid represents a learnable element weight matrix.
[0086] The vector features of the raster map and the normalized weight are multiplied to obtain a vector map after vectorization processing.
[0087] In the embodiment of the application, the vector features V′ map of the raster map and the normalized weight W map may be multiplied by using Hadamard product to obtain a vector map V map after vectorization processing. The vector map V map after vectorization processing may be represented as V map = V′ map ⊙W map .
[0088] Of course, the vector features of the raster map and the normalized weight may also be multiplied by using other manners, which are not limited in the embodiment of the application.
[0089] In some possible implementation ways, the step S2 of extracting the map feature from the vectorized vector map comprises:
[0090] The map feature is extracted from the vectorized vector map by using a convolutional neural network. Of course, other ways can also be used to extract the map feature from the vectorized vector map, which are not limited in the embodiments of the present application.
[0091] In some possible implementation ways, the step S3 of performing the visual geolocation on the unmanned aerial vehicle according to the image feature and the map feature comprises:
[0092] As shown in Figure 4 , the image feature is rotated multiple times according to a preset rotation angle to obtain multiple rotation templates. The map feature is matched with each rotation template in the multiple rotation templates to obtain multiple matching probabilities. A coordinate corresponding to a maximum value of the multiple matching probabilities is selected as a pose of the image feature relative to the map feature. The pose of the unmanned aerial vehicle is obtained according to the pose of the image feature relative to the map feature and the vector map, wherein the pose of the unmanned aerial vehicle comprises a position and an attitude of the unmanned aerial vehicle.
[0093] Exemplarily, the step of matching the map feature with each rotation template in the multiple rotation templates to obtain multiple matching probabilities comprises:
[0094] The Fourier transform of the map feature is multiplied by a complex conjugate of the Fourier transform of each rotation template to obtain a correlation of the map feature and each rotation template in the frequency domain.
[0095] The Fourier inverse transform is performed on the correlation of the map feature and each rotation template in the frequency domain to obtain the multiple matching probabilities.
[0096] Embodiment 2:
[0097] Based on the same inventive concept, the embodiments of the present application further provide an unmanned aerial vehicle visual geolocation device based on a vector map. As shown in Figure 5 , the geolocation device 200 comprises:
[0098] A first extraction module 201 is configured to extract an image feature from an environment image collected by an unmanned aerial vehicle.
[0099] A second extraction module 202 is configured to perform vectorization processing on a vector map, and extract a map feature from the vectorized vector map.
[0100] A positioning module 203 is configured to perform visual geolocation on the unmanned aerial vehicle according to the image feature and the map feature.
[0101] In some possible implementation ways, the first extraction module 201 is specifically configured to:
[0102] The environment image is multi-layer encoded to obtain first image encoding information and a plurality of second image encoding information. The first image encoding information is used to indicate image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate image encoding information obtained by each layer encoding.
[0103] The first image encoding information and the plurality of second image encoding information are multi-layer decoded to obtain first image decoding information and a plurality of second image decoding information. The first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by each layer decoding.
[0104] The first image decoding information and the plurality of second image decoding information are fused to obtain image features.
[0105] In some possible implementation manners, the second extraction module 202 is specifically configured to:
[0106] The vector map is rasterized to obtain a raster map.
[0107] The vector features of the raster map are transformed to obtain vector features of the raster map.
[0108] The weights corresponding to each type of ground surface element in the raster map are normalized to obtain normalized weights.
[0109] The vector features of the raster map are multiplied by the normalized weights to obtain a vector map after vectorization processing.
[0110] Optionally, the second extraction module 202 is specifically configured to:
[0111] The vector features of the raster map are multiplied by the normalized weights by using Hadamard product to obtain a vector map after vectorization processing.
[0112] Illustratively, the second extraction module 202 is specifically configured to:
[0113] The map features are extracted from the vector map after vectorization processing by using a convolutional neural network.
[0114] In some possible implementation manners, the positioning module 203 is specifically configured to:
[0115] The image features are rotated a plurality of times according to a preset rotation angle to obtain a plurality of rotation templates.
[0116] The map features are matched with each rotation template in the plurality of rotation templates to obtain a plurality of matching probabilities.
[0117] Select the maximum value of the plurality of matching probabilities corresponding to the coordinates as the pose of the image feature relative to the map feature.
[0118] According to the pose of the image feature relative to the map feature and the vector map, the pose of the UAV is obtained, wherein the pose of the UAV includes the position and the attitude of the UAV.
[0119] Optionally, the positioning module is specifically configured to:
[0120] The Fourier transform of the map feature is multiplied by the complex conjugate of the Fourier transform of each rotation template to obtain the correlation of the map feature and each rotation template in the frequency domain.
[0121] The Fourier inverse transform is performed on the correlation of the map feature and each rotation template in the frequency domain to obtain a plurality of matching probabilities.
[0122] Embodiment 3:
[0123] Based on the same inventive concept, the embodiments of the present application further provide a computer device, which comprises a processor and a memory. The memory is used to store a computer program, and the computer program comprises program instructions. The processor is used to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The processor is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method process or a corresponding function, so as to implement the steps of the positioning method provided in the above embodiments.
[0124] Embodiment 4:
[0125] Based on the same inventive concept, the embodiment of the present application further provides a computer readable storage medium, specifically, a computer readable storage medium (Memory). The computer readable storage medium is a memory device in a computer device, and is used to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can include an extended storage medium supported by the computer device. The computer readable storage medium provides a storage space, and the storage space stores an operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the steps of the positioning method provided in the above embodiment.
[0126] Those skilled in the art should understand that the embodiments of the application can be provided as a method, a system, or a computer program product. Therefore, the application can adopt a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the application can adopt a computer program product in the form of being implemented on one or more computer usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer usable program codes.
[0127] The application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of the flows and / or blocks in the flowcharts and / or block diagrams can be implemented by computer program instructions. These computer program instructions can be provided to a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to produce a machine, so that the instructions executed by the computer or other programmable data processing devices generate a device for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1 The devices for implementing the functions specified in one or more flows and / or blocks.
[0128] These computer program instructions can also be stored in a computer readable storage medium, which can guide the computer or other programmable data processing devices to work in a specific manner, so that the instructions stored in the computer readable storage medium produce a manufactured product including instruction devices, which implement the functions specified in the flowcharts and / or block diagrams. Figure 1 The functions specified in one or more flows and / or blocks. Figure 1the function specified in the one or more blocks.
[0129] These computer program instructions can also be loaded into computer or other programmable data processing devices, so that a series of operation steps are performed on the computer or other programmable data processing devices to generate computer-implemented processes, thus the instructions executed on the computer or other programmable data processing devices provide processes for implementing the flows Figure 1 the flows or the flows and / or blocks Figure 1 the steps of the function specified in the one or more blocks.
[0130] The above merely provides the embodiments of the application and is not intended to limit the application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the application shall be included in the scope of the claims of the application to be granted.
Claims
1. A method for visual geolocation of unmanned aerial vehicles based on vector maps, characterized in that: include: Extract image features from environmental images collected by drones; Performing vectorization processing on the vector map, and extracting map features from the vectorized vector map; Visual geolocation of the drone is performed based on the image features and the map features.
2. The positioning method according to claim 1, wherein: The step of extracting image features from the environment image collected by the drone includes: Performing multi-layer encoding on the environment image to obtain first image encoding information and a plurality of second image encoding information; wherein the first image encoding information is used to indicate image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate image encoding information obtained by encoding each layer; performing multi-layer decoding on the first image coding information and the plurality of second image coding information to obtain first image decoding information and a plurality of second image decoding information; wherein the first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by decoding on each layer; The first image decoding information and the plurality of second image decoding information are fused to obtain the image feature.
3. The positioning method according to claim 1, wherein: The vectorization processing of the vector map includes: Performing rasterization processing on the vector map to obtain a raster map; Performing feature transformation on the grid map to obtain vector features of the grid map; Normalizing the weights corresponding to each type of surface element in the grid map to obtain normalized weights; The vector features of the raster map are multiplied by the normalized weights to obtain the vectorized vector map.
4. The positioning method according to claim 3, characterized in that: The multiplying the vector features of the raster map by the normalized weights to obtain the vectorized vector map includes: The vector map after the vectorization processing is obtained by multiplying the vector features of the grid map and the normalized weights using a Hadamard product.
5. The positioning method according to claim 1, wherein: The extracting of map features from the vectorized vector map includes: The map features are extracted from the vectorized vector map through a convolutional neural network.
6. The positioning method according to claim 1, characterized in that: The performing visual geolocation of the drone according to the image features and the map features includes: Rotating the image feature multiple times according to a preset rotation angle to obtain multiple rotation templates; Matching the map feature with each of the multiple rotation templates to obtain multiple matching probabilities; Selecting the coordinates corresponding to the maximum values of the multiple matching probabilities as the position of the image feature relative to the map feature; The pose of the UAV is obtained according to the pose of the image feature relative to the map feature and the vector map, wherein the pose of the UAV includes the position and attitude of the UAV.
7. The positioning method according to claim 6, characterized in that: The matching of the map feature with each of the plurality of rotation templates to obtain a plurality of matching probabilities includes: Multiplying the Fourier transform of the map feature by the complex conjugate of the Fourier transform of each rotation template to obtain a correlation between the map feature and each rotation template in the frequency domain; Perform inverse Fourier transform on the correlation between the map feature and each rotation template in the frequency domain to obtain the multiple matching probabilities.
8. A UAV visual geographic positioning device based on vector map, characterized in that: include: The first extraction module is used to extract image features from the environment image collected by the drone; The second extraction module is used to perform vectorization processing on the vector map and extract map features from the vectorized vector map; A positioning module is used to perform visual geographic positioning of the drone based on the image features and the map features.
9. The positioning device according to claim 8, characterized in that The first extraction module is specifically configured to: Performing multi-layer encoding on the environment image to obtain first image encoding information and a plurality of second image encoding information; wherein the first image encoding information is used to indicate image encoding information obtained by multi-layer encoding, and the second image encoding information is used to indicate image encoding information obtained by encoding each layer; performing multi-layer decoding on the first image coding information and the plurality of second image coding information to obtain first image decoding information and a plurality of second image decoding information; wherein the first image decoding information is used to indicate image decoding information obtained by multi-layer decoding, and the second image decoding information is used to indicate image decoding information obtained by decoding on each layer; The first image decoding information and the plurality of second image decoding information are fused to obtain the image feature.
10. The positioning device according to claim 8, characterized in that The second extraction module is specifically used for: Performing rasterization processing on the vector map to obtain a raster map; Performing feature transformation on the grid map to obtain vector features of the grid map; Normalizing the weights corresponding to each type of surface element in the grid map to obtain normalized weights; The vector features of the raster map are multiplied by the normalized weights to obtain the vectorized vector map.
11. The positioning device according to claim 10, characterized in that The second extraction module is specifically used for: The vector map after the vectorization processing is obtained by multiplying the vector features of the grid map and the normalized weights using a Hadamard product.
12. The positioning device according to claim 8, characterized in that The second extraction module is specifically used for: The map features are extracted from the vectorized vector map through a convolutional neural network.
13. The positioning device according to claim 8, characterized in that The positioning module is specifically used for: Rotating the image feature multiple times according to a preset rotation angle to obtain multiple rotation templates; Matching the map feature with each of the multiple rotation templates to obtain multiple matching probabilities; Selecting the coordinates corresponding to the maximum values of the multiple matching probabilities as the position of the image feature relative to the map feature; The pose of the UAV is obtained according to the pose of the image feature relative to the map feature and the vector map, wherein the pose of the UAV includes the position and attitude of the UAV.
14. The positioning device according to claim 13, characterized in that The positioning module is specifically used for: Multiplying the Fourier transform of the map feature by the complex conjugate of the Fourier transform of each rotation template to obtain a correlation between the map feature and each rotation template in the frequency domain; Perform inverse Fourier transform on the correlation between the map feature and each rotation template in the frequency domain to obtain the multiple matching probabilities.
15. A computer device, characterized in that: include: one or more processors; The processor is configured to store one or more programs; When the one or more programs are executed by the one or more processors, the positioning method according to any one of claims 1 to 7 is implemented.
16. A computer-readable storage medium, characterized in that A computer program is stored thereon, and when the computer program is executed, the positioning method according to any one of claims 1 to 7 is implemented.