Unmanned aerial vehicle day and night image matching navigation method and system in satellite denial environment

Abstract

The invention discloses an unmanned aerial vehicle day and night image matching navigation method and system in a satellite denial environment. The method comprises the following steps: acquiring a visible light aerial image and a thermal infrared aerial image in real time through a visible light and thermal infrared camera carried by an unmanned aerial vehicle; performing preprocessing and geometric registration on the visible light aerial image and the thermal infrared aerial image to obtain an orthographic image consistent with the satellite remote sensing image in direction; and performing image matching on the orthoimage and the satellite remote sensing image through a multi-scale attention mechanism generation type image matching model, determining the accurate position of the unmanned aerial vehicle in real time, and realizing day and night image matching navigation of the unmanned aerial vehicle. The method can realize the day and night image matching navigation of the unmanned aerial vehicle in the satellite rejection environment, has extremely high application value, is the key for improving the navigation performance of the unmanned aerial vehicle, enhancing the flexibility and improving the task execution efficiency especially in the special task flight of the unmanned aerial vehicle, and has the beneficial effects of high precision, strong autonomy and high feasibility.

Description

Technical Field

[0001] The present invention relates to the technical field, and in particular to a day and night image matching navigation method and system for unmanned aerial vehicles in a satellite denial environment. Background Art

[0002] With the rapid development of science and technology, drone technology has been widely used in many fields such as military reconnaissance, environmental monitoring, agricultural management, logistics distribution, etc. The navigation and positioning technology of drones is one of its core functions, which usually relies on satellite navigation systems (such as GPS) and inertial navigation systems (INS). However, in complex terrain, densely populated urban areas or special circumstances, satellite signals are susceptible to electromagnetic interference, obstruction or deception, resulting in the failure of satellite navigation. In this context, it is particularly important to develop a method that can achieve precise navigation in a satellite-denied environment.

[0003] At present, the navigation technologies of UAVs in satellite-denied environments mainly include combined navigation technology based on inertial navigation systems and vision-based navigation technology. Combined navigation technology improves navigation accuracy and reliability by integrating data from multiple sensors (such as magnetometers, barometers, etc.). Vision-based navigation technology uses the camera onboard the UAV to collect images and navigates through computer vision algorithms. These methods have alleviated the problem of satellite navigation failure to a certain extent, but they still face many challenges in practical applications.

[0004] Although existing technologies have made some progress in UAV navigation, there are still some shortcomings. First, although combined navigation technology can make up for the shortcomings of satellite navigation to a certain extent, it is still difficult to avoid the accumulation of positioning errors in complex environments, especially in long-term navigation tasks. Secondly, vision-based navigation technology faces difficulties such as multi-phase, multi-perspective, multi-scale and large-scale search in the image matching process, resulting in low matching accuracy and efficiency. In addition, existing technologies have limited navigation capabilities at night, making it difficult to achieve all-weather navigation. Summary of the invention

[0005] In view of the above defects or improvement needs of the prior art, the purpose of the present invention is to provide a method and system for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment. The method matches the image of the unmanned aerial vehicle with the satellite remote sensing image with geographic location information, so that the unmanned aerial vehicle can also obtain the current position in the case of satellite denial and realize autonomous navigation. This method can further promote the development of unmanned aerial vehicles, especially can effectively improve the anti-interference ability of unmanned aerial vehicles. In addition, the present invention adopts visible light and thermal infrared cameras to take aerial photos of the ground, and uses the characteristics of thermal infrared imaging to enable the unmanned aerial vehicle to obtain aerial images of the ground at night, and uses the geometric information of the image boundary for constraints, and realizes the precise matching of thermal infrared images and satellite remote sensing images through a multi-scale attention mechanism generative image matching model, so that the present invention can realize the day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment, showing extremely high application value, especially in the special mission flight of unmanned aerial vehicles, the present invention has become the key to improving the navigation performance of unmanned aerial vehicles, enhancing flexibility and improving the efficiency of task execution, and has the characteristics of high precision, strong autonomy, high feasibility and high economic benefits.

[0006] According to a first aspect of the present invention, a method for day and night image matching navigation of a UAV in a satellite denial environment is provided, comprising: S100, obtains visible light aerial images and thermal infrared aerial images in real time through the visible light and thermal infrared cameras carried by the drone; S200, preprocessing and geometrically registering the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image that is consistent with the direction of the satellite remote sensing image; S300, performing image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determining the precise position of the UAV in real time, and realizing day and night image matching navigation of the UAV.

[0007] Furthermore, the multi-scale attention mechanism generative image matching model is a self-supervised deep learning model that adopts a fully convolutional neural network design and can perform pixel-level predictions on the input image. The key point detector adopts a bidirectional recurrent neural network to generate a dense key point probability map, and the descriptor generator extracts features around each detected key point to ensure the independence of each key point.

[0008] Furthermore, the multi-scale attention mechanism generative image matching model includes: a feature extraction module for extracting multi-scale features from satellite remote sensing images, visible light and thermal infrared aerial images; a multi-scale attention mechanism module for strengthening the extracted multi-scale features; and a generative module for reducing the color, texture and style differences between satellite images and drone images.

[0009] Furthermore, the generative network core in the generative module is mainly composed of two parts: a generator and a discriminator. During the training process, the goal of the generator is to use the pictures it generates to deceive the discriminator, while the goal of the discriminator is to distinguish the pictures generated by the generator from the real pictures, so that the two constitute a dynamic game process. By inputting data into the generator and then generating new style data, the data and the real data are input into the discriminator at the same time, and the model is continuously trained so that the generator tries its best to generate fake data that can deceive the discriminator, so that the generator module can perform style transfer and color transformation on the data, so that the matching performance between heterogeneous or heterogeneous images is higher and the matching accuracy is improved.

[0010] Furthermore, step S200 includes: calibrating the internal parameters of the camera carried on the drone and correcting the image deformation; using the monocular camera carried by the drone to collect ground images, and using the overlapping images of the front and rear frames to calculate the flight attitude parameters of the drone in real time; filtering and interpolating the flight attitude parameters to eliminate noise and improve the continuity of the parameters; intensively processing the image data and posture data; constructing an image geometric transformation model to calculate the coordinate position of each pixel in the image under the new perspective; based on the calculated new coordinate position, positively resampling and interpolating the original image to complete the geometric correction of the aerial image and obtain an orthophoto image with the same direction as the satellite image; finally, color normalization is performed on the orthophoto image to reduce the color difference between the satellite image and the aerial image, thereby improving the matching accuracy.

[0011] Furthermore, step S300 includes: after denoising and enhancing the orthophoto image and the satellite remote sensing image, important areas or features in the image are extracted through a feature extraction module, a multi-scale attention mechanism module and a generative module, and image matching is completed in combination with an automatic encoder and an automatic decoder.

[0012] Furthermore, it also includes: taking the drone position at the moment before the satellite navigation anomaly as the center of the image matching approximate area, and then calculating the area of ​​the image matching approximate area according to the drone flight speed at the previous moment, and determining the image matching approximate area, thereby narrowing the search range and improving matching efficiency and accuracy.

[0013] According to a second aspect of the present invention, a day and night image matching navigation system for unmanned aerial vehicles in a satellite denial environment is provided, comprising: The first module is used to obtain visible light aerial images and thermal infrared aerial images in real time through visible light and thermal infrared cameras carried by the drone; The second module is used to preprocess and geometrically register the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image with the same direction as the satellite remote sensing image; The third module is used to perform image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determine the precise position of the UAV in real time, and realize the day and night image matching navigation of the UAV.

[0014] According to a third aspect of the present invention, there is provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned day and night image matching navigation method for unmanned aerial vehicles in a satellite denial environment when loading and executing the computer program.

[0015] According to a fourth aspect of the present invention, a computer-readable storage medium is provided, characterized in that a computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the above-mentioned day and night image matching navigation method for unmanned aerial vehicles in a satellite denial environment.

[0016] Beneficial effects of the present invention: 1. This method matches the drone image with the satellite remote sensing image with geographic location information, so that the drone can obtain the current position even in the case of satellite denial, and uses visible light and thermal infrared cameras to achieve autonomous navigation day and night in a satellite denial environment. This method can further promote the development of drones, especially effectively improve the anti-interference ability of drones. In addition, the present invention uses visible light and thermal infrared cameras for aerial photography of the ground, and uses the thermal infrared imaging characteristics to enable drones to obtain aerial images of the ground at night, and uses the image boundary geometry information for constraints. The multi-scale attention mechanism generative image matching model is used to achieve accurate matching of thermal infrared images and satellite remote sensing images, so that the present invention can achieve day and night image matching navigation of drones in a satellite denial environment, showing extremely high application value, especially in the special mission flight of drones. The present invention has become the key to improving the navigation performance of drones, enhancing flexibility and improving the efficiency of task execution, and has the characteristics of high precision, strong autonomy, high feasibility and high economic benefits; 2. The multi-scale attention mechanism generative image matching model in this method is composed of a feature extraction module, a multi-scale attention mechanism module and a generative module. The feature extraction module is used to extract multi-scale features from satellite remote sensing images, visible light and thermal infrared aerial images; the multi-scale attention mechanism module is used to strengthen the extracted multi-scale features; the generative module is used to reduce the color, texture and style differences between satellite images and drone images. The combined application of the above three modules can effectively improve the matching efficiency and accuracy. 3. This method can effectively solve the problem of color and ground feature differences between satellite images and drone images in different phases and seasons, which leads to the problem of multi-phase ground feature feature differences caused by changes in vegetation and lighting conditions on the ground; 4. This method uses the drone position at the previous moment to determine the approximate image matching area at the next moment, effectively improving the matching accuracy and calculation efficiency, and realizing fast search; 5. This method uses visible light and thermal infrared cameras mounted on drones to shoot videos of the ground both during the day and at night, thereby realizing day and night image matching navigation of drones and expanding the scope of application.

[0017] 6. This system has wide applicability. With the continuous improvement of my country's remote sensing satellite earth observation technology, it can obtain images with geographic location information on a large scale, large range, high spatial resolution and high timeliness. It only needs unmanned visible light and thermal infrared cameras, which can add a new navigation function to drones at a low cost. Additional aspects and advantages of the present application will be partially given in the following description, which will become apparent from the following description, or will be understood through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 This is a flow chart of a method for day and night image matching navigation of a UAV in a satellite denial environment in an embodiment of the present invention; Figure 2 It is a flow chart of image preprocessing and geometric registration in the day and night image matching navigation method of a UAV in a satellite denial environment according to an embodiment of the present invention; Figure 3 It is an image matching flow chart of the day and night image matching navigation method of a UAV in a satellite denial environment according to an embodiment of the present invention; Figure 4 It is a schematic diagram of the operation flow of a generative model in a multi-scale attention mechanism generative image matching model according to an embodiment of the present invention; Figure 5 It is a basic flow chart of the drone image matching positioning search approximate area in the drone day and night image matching navigation method in a satellite denial environment according to an embodiment of the present invention. DETAILED DESCRIPTION

[0019] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only parts related to the present invention, rather than all structures, are shown in the accompanying drawings.

[0020] Those skilled in the art will appreciate that, unless otherwise stated, the singular forms "a", "an", "said" and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0021] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical terms and scientific terms) used herein have the same meaning as the general understanding of those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have the same meaning as in the context of the prior art, and will not be interpreted with an idealized or overly formal meaning unless specifically defined as in the embodiments of this application.

[0022] The present invention provides a method and system for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment. The method matches the image of the unmanned aerial vehicle with the satellite remote sensing image with geographic location information, so that the unmanned aerial vehicle can obtain the current position and realize autonomous navigation in the case of satellite denial. The method can further promote the development of unmanned aerial vehicles, especially can effectively improve the anti-interference ability of unmanned aerial vehicles. In addition, the present invention adopts visible light and thermal infrared cameras for aerial photography of the ground, and utilizes the characteristics of thermal infrared imaging to enable the unmanned aerial vehicle to obtain aerial images of the ground at night, and utilizes the geometric information of the image boundary for constraints, and realizes accurate matching of thermal infrared images and satellite remote sensing images through a multi-scale attention mechanism generative image matching model, so that the present invention can realize day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment, showing extremely high application value, especially in the special mission flight of unmanned aerial vehicles. The present invention has become the key to improving the navigation performance of unmanned aerial vehicles, enhancing flexibility and improving the efficiency of task execution, and has the characteristics of high precision, strong autonomy, high feasibility and high economic benefits; Embodiment 1: like Figure 1 As shown, an embodiment of the present invention provides a method for day and night image matching navigation of a UAV in a satellite denial environment, comprising: S100, obtains visible light aerial images and thermal infrared aerial images in real time through the visible light and thermal infrared cameras carried by the drone; S200, preprocessing and geometrically registering the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image that is consistent with the direction of the satellite remote sensing image; like Figure 2As shown, the image preprocessing and geometric registration method provided in this step is to use the flight posture parameters (pitch angle, yaw angle, roll angle) obtained by the UAV in real time during the flight, combined with the camera's internal parameters (focal length, optical center position, etc.) and external parameters (relative position and direction of the camera and the UAV body), to accurately align the direction, correct the direction, and orthographic projection of the collected image to eliminate the image distortion and direction deviation caused by the change of flight posture, specifically including: calibrating the camera internal parameters of the UAV on the UAV to correct the image deformation; using the UAV equipped with a monocular camera to collect ground images, and using the overlapping images of the front and rear frames to calculate the UAV flight posture parameters in real time; filtering and interpolating the flight posture parameters to eliminate noise and improve the continuity of the parameters; performing intensive processing on the image data and posture data to better capture the ground details; constructing an image geometric transformation model to calculate the coordinate position of each pixel in the image under the new perspective; According to the calculated new coordinate position, the original image is resampled and interpolated to eliminate the geometric distortion caused by the shooting angle and posture changes, ensuring that the shape and position of the objects in the image can be accurately reflected in the satellite image; the aerial image geometric correction is completed to obtain an orthophoto with the same direction as the satellite image; finally, color normalization is performed to reduce the color difference between the satellite image and the aerial image, thereby improving the accuracy of matching.

[0023] S300, performing image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determining the precise position of the UAV in real time, and realizing day and night image matching navigation of the UAV.

[0024] A multi-scale attention mechanism generative image matching model provided in an embodiment of the present invention. Image matching is a research hotspot in the field of computer vision. Traditional methods are sensitive to interference such as illumination, rotation, scaling and noise due to their weak feature extraction capabilities, and it is difficult to fully and stably capture complex information and detail changes in images. At the same time, thermal infrared videos taken by drones and satellite remote sensing images are heterogeneous images, which require higher matching. Therefore, the present invention proposes a multi-scale attention mechanism generative image matching model, which aims to improve the accuracy and robustness of image matching by integrating multi-scale features and attention mechanisms, and combining image geometric edge information, so as to achieve accurate matching of satellite remote sensing images with visible light and thermal infrared images taken by drones. In addition, due to the different imaging heights and times of satellite images and drone images, there are differences in color and texture between the two images. The generative network can generate new data samples similar to real data by learning and simulating the intrinsic distribution laws of data, which can effectively weaken the color, texture and style differences between the two images. Therefore, the present invention extracts the feature representation of images at different scales through a convolutional neural network (CNN), then uses the attention mechanism to enhance these features, and then uses a generative network to gradually approximate the features between satellite images and drone images, and finally extracts feature points for matching.

[0025] Specifically, Figure 3 As shown in the figure, after the satellite remote sensing images, visible light and thermal infrared aerial images are denoised and enhanced, the feature extraction module, multi-scale attention mechanism module and generative module are constructed and embedded in the autoencoder, so as to focus on the important areas or features in the image, effectively extract key information, and combine the autoencoder and decoder for image matching. This model is a self-supervised deep learning model. It adopts the fully convolutional neural network (FCN) design and can predict the input image at the pixel level. The key point detector adopts the bidirectional recurrent neural network (Bi-GRU) to generate dense key point probability maps, and the descriptor generator extracts features around each detected key point to ensure the independence of each key point. In addition, the multi-scale attention mechanism module is integrated into the model to improve the efficiency and accuracy of matching and simplify the training process. The model combines the advantages of the multi-scale attention mechanism and the generative module to construct a high-precision and high-robust image matching model.

[0026] Furthermore, in order to accurately obtain the location information of the drone, the center of the drone image is usually used as the location information. Therefore, it is necessary to associate the central features of the drone image with the satellite image, and the matching model feature extraction will focus on the features of the center position. Therefore, this project intends to design an attention mechanism to strengthen the salient features of the central area and suppress the boundary features. When multiple local feature blocks are extracted simultaneously, the extracted features are redundant in each dimension. In order to remove these redundant features and make each local feature block more representative, an attention module is added after the feature map of the local branch. The formula of this module is as follows: , , In the formula, represents the feature concatenation operation, It means multiplying corresponding elements between matrices.

[0027] like Figure 4 As shown, the core of the generative network is to learn and simulate the inherent distribution law of data so as to generate new data with different colors and styles. It is widely used in image generation, text-to-image conversion, style transfer, data enhancement and other fields. The network structure mainly consists of two parts: the generator and the discriminator. During the training process, the goal of the generator is to use the pictures it generates to deceive the discriminator, while the goal of the discriminator is to distinguish the pictures generated by the generator from the real pictures, so that the two constitute a dynamic game process; this embodiment inputs data into the generator, then generates new style data, and inputs the data and the real data into the discriminator at the same time, and continuously trains the model so that the generator tries its best to generate fake data that can deceive the discriminator, so that the generator can transfer the style and color of the data.

[0028] An embodiment of the present invention also provides a method for using the drone position at the moment before the satellite navigation anomaly as the center of the image matching approximate area, and then calculating the area of ​​the image matching approximate area according to the drone flight speed at the previous moment, and determining the image matching approximate area, thereby narrowing the search range and improving matching efficiency and accuracy.

[0029] Specifically, Figure 5 As shown in the figure, before the UAV flies, the flight mission range is determined, and then the latest satellite remote sensing images within the mission range are downloaded to update the UAV base map. Then the UAV is equipped with visible light and thermal infrared cameras to take real-time images of the ground, and determine whether the satellite navigation signal is normal. If an abnormality occurs, image matching navigation is used. First, the UAV position at the moment before the satellite navigation abnormality is used as the center of the image matching approximate area, and then the image matching approximate area S is calculated based on the flight speed at the previous moment. The calculation formula is as follows: , In the formula, Indicates the flight speed of the drone. Indicates the time from the previous moment to the image matching and positioning at the next moment.

[0030] Then, according to the image matching approximate area, the image matching approximate area will be determined, and the satellite image of the area will be used as the base map for matching with the drone image at the next moment, and finally the position of the drone at the next moment will be calculated. When calculating the next moment, repeat until it is determined whether the satellite signal is normal, so as to find the position of the drone and complete the drone navigation task.

[0031] Embodiment 2: An embodiment of the present invention provides a day and night image matching navigation system for unmanned aerial vehicles in a satellite denial environment, comprising: The first module is used to obtain visible light aerial images and thermal infrared aerial images in real time through visible light and thermal infrared cameras carried by the drone; The second module is used to preprocess and geometrically register the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image with the same direction as the satellite remote sensing image; The third module is used to perform image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determine the precise position of the UAV in real time, and realize the day and night image matching navigation of the UAV.

[0032] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned day and night image matching navigation method for unmanned aerial vehicles in a satellite denial environment when loading and executing the computer program.

[0033] An embodiment of the present invention also provides a computer-readable storage medium, in which a computer program is stored. The computer program is loaded and executed by a processor to implement the above-mentioned day and night image matching navigation method for unmanned aerial vehicles in a satellite denial environment.

[0034] In summary, the present application provides a method and system for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment. By matching the unmanned aerial vehicle image with the satellite remote sensing image with geographic location information, the unmanned aerial vehicle can also obtain the current position in the case of satellite denial and realize autonomous navigation. The method and system can further promote the development of unmanned aerial vehicles, especially effectively improve the anti-interference ability of unmanned aerial vehicles. In addition, the method and system use visible light and thermal infrared cameras for aerial photography of the ground, and use the thermal infrared imaging characteristics to enable the unmanned aerial vehicle to obtain aerial images of the ground at night, and use the image boundary geometric information for constraints, and realize the precise matching of thermal infrared images and satellite remote sensing images through the multi-scale attention mechanism generative image matching model, so that the unmanned aerial vehicle day and night image matching navigation can be realized in a satellite denial environment, showing extremely high application value, especially in the special mission flight of unmanned aerial vehicles. The present application has become the key to improving the navigation performance of unmanned aerial vehicles, enhancing flexibility and improving the efficiency of task execution, and has the characteristics of high precision, strong autonomy, high feasibility and high economic benefits.

[0035] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0036] The above is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment, characterized in that: include: S100, obtains visible light aerial images and thermal infrared aerial images in real time through the visible light and thermal infrared cameras carried by the drone; S200, preprocessing and geometrically registering the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image that is consistent with the direction of the satellite remote sensing image; S300, performing image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determining the precise position of the UAV in real time, and realizing day and night image matching navigation of the UAV.

2. According to the method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to claim 1, it is characterized in that: The multi-scale attention mechanism generative image matching model is a self-supervised deep learning model that adopts a fully convolutional neural network design and can perform pixel-level predictions on input images. The key point detector adopts a bidirectional recurrent neural network to generate a dense key point probability map, and the descriptor generator extracts features around each detected key point to ensure the independence of each key point.

3. The method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to claim 2 is characterized in that: The multi-scale attention mechanism generative image matching model includes: a feature extraction module for extracting multi-scale features from satellite remote sensing images, visible light and thermal infrared aerial images; a multi-scale attention mechanism module for strengthening the extracted multi-scale features; and a generative module for reducing the color, texture and style differences between satellite images and drone images.

4. The method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to claim 3 is characterized in that: The generative network core in the generative module is mainly composed of two parts: a generator and a discriminator. During the training process, the goal of the generator is to use the pictures it generates to deceive the discriminator, while the goal of the discriminator is to distinguish the pictures generated by the generator from the real pictures, so that the two constitute a dynamic game process. By inputting data into the generator and then generating new style data, the data and the real data are input into the discriminator at the same time, and the model is continuously trained to make the generator try its best to generate fake data that can deceive the discriminator, so that the generator module can perform style transfer and color transformation on the data, so that the matching performance between heterogeneous or heterogeneous images is higher and the matching accuracy is improved.

5. A method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to any one of claims 1 to 4, characterized in that: Step S200 includes: calibrating the internal parameters of the camera on the drone and correcting the image deformation; using the drone-mounted monocular camera to collect ground images, and using the overlapping images of the front and rear frames to calculate the drone flight attitude parameters in real time; filtering and interpolating the flight attitude parameters to eliminate noise and improve the continuity of the parameters; intensively processing the image data and posture data; constructing an image geometric transformation model to calculate the coordinate position of each pixel in the image under the new perspective; based on the calculated new coordinate position, positively resampling and interpolating the original image to complete the geometric correction of the aerial image and obtain an orthophoto image with the same direction as the satellite image; finally, color normalizing the orthophoto image to reduce the color difference between the satellite image and the aerial image, thereby improving the matching accuracy.

6. A method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to any one of claims 1 to 4, characterized in that: Step S300 includes: after denoising and enhancing the orthophoto image and the satellite remote sensing image, extracting important areas or features in the image through a feature extraction module, a multi-scale attention mechanism module and a generative module, and completing image matching in combination with an automatic encoder and an automatic decoder.

7. A method for day and night image matching navigation of unmanned aerial vehicles in a satellite denial environment according to any one of claims 1 to 4, characterized in that: Also includes: The position of the UAV at the moment before the satellite navigation anomaly is taken as the center of the image matching approximate area, and then the area of ​​the image matching approximate area is calculated according to the flight speed of the UAV at the previous moment to determine the image matching approximate area, thereby narrowing the search range and improving the matching efficiency and accuracy.

8. A day and night image matching navigation system for unmanned aerial vehicles in a satellite denial environment, characterized in that: include: The first module is used to obtain visible light aerial images and thermal infrared aerial images in real time through visible light and thermal infrared cameras carried by the drone; The second module is used to preprocess and geometrically register the visible light aerial image and the thermal infrared aerial image according to the flight posture parameters of the UAV and the internal and external parameters of the camera to obtain an orthophoto image with the same direction as the satellite remote sensing image; The third module is used to perform image matching between the orthophoto image and the satellite remote sensing image through a multi-scale attention mechanism generative image matching model, determine the precise position of the UAV in real time, and realize the day and night image matching navigation of the UAV.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor loads and executes the computer program, the method for day and night image matching and navigation of a UAV in a satellite-denied environment according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which is loaded and executed by a processor to implement the day and night image matching navigation method for a UAV in a satellite-denied environment according to any one of claims 1 to 7.

Images