A multi-source data fusion cooperative target positioning method and system based on infrared vision

Abstract

The application discloses a multi-source data fusion cooperative target positioning method and system based on infrared vision, and the method comprises the following steps: deploying an infrared point light source beacon on the ground and starting when the signal is disturbed; acquiring an infrared point light source beacon image by using an infrared camera carried by a UAV; and analyzing the infrared point light source beacon image by using a constructed visual navigation relative positioning algorithm to complete the UAV target positioning in the state that the signal is disturbed. The application adopts infrared vision to support all-weather positioning and is not limited by illumination conditions; meanwhile, the positioning error is less than or equal to 5% by fusing IMU and visual data through EKF. In addition, the application adopts a unique layout and waveband design of the infrared beacon, can effectively resist electromagnetic interference, and can realize 24-hour continuous operation without cumulative error.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) navigation and positioning technology, specifically to a multi-source data fusion cooperative target positioning method and system based on infrared vision. Background Technology

[0002] With the widespread application of drones in military reconnaissance, disaster relief, and other fields, their positioning reliability in complex environments faces severe challenges. Existing technologies primarily rely on a combination of satellite navigation (such as GPS) and inertial navigation systems (IMUs) for positioning. However, when satellite signals are interfered with by enemy forces or obstructed by terrain, drones cannot obtain effective positioning information, leading to mission failure or even crashes. Traditional visual navigation relies on visible light cameras, which struggle to extract effective features at night or in rainy or foggy weather. Furthermore, inertial navigation systems (IMUs) accumulate significant errors over long periods, failing to meet the demands of 24-hour continuous flight. Current solutions such as lidar or millimeter-wave radar are expensive and require high computational power. Therefore, there is an urgent need for a low-cost, highly robust positioning method to achieve accurate positioning in GPS-denied, complex lighting, and high-altitude environments. Summary of the Invention

[0003] This invention aims to provide a positioning method to solve the following problems: UAVs cannot obtain continuous positioning information in environments where GPS signals are blocked; traditional visual positioning fails under complex lighting conditions such as nighttime, rain, and fog; long-term operation of inertial navigation systems (IMUs) leads to accumulated errors that reduce positioning accuracy; and UAVs can reliably identify and track ground cooperative targets at altitudes above 300 meters.

[0004] To achieve the above objectives, this invention provides a multi-source data fusion cooperative target localization method based on infrared vision, comprising the following steps:

[0005] Deploy infrared point light source beacons on the ground and activate them when the signal is interfered with;

[0006] Using an infrared camera mounted on a drone, images of infrared point light source beacons are acquired;

[0007] The infrared point light source beacon image is analyzed using the constructed visual navigation relative positioning model to complete the UAV target positioning under signal interference.

[0008] Preferably, seven asymmetrically distributed infrared point light source beacons are deployed in the ground cooperation target area, with four located at the apex of the helipad and three at a height of 2.5 meters above the ground. The spacing between the beacons is greater than or equal to 2 meters. The three-dimensional coordinate matrix of the beacons is defined as follows:

[0009]

[0010] in, G represents the preset distance between beacons; G represents the geometric constraint function between the three-dimensional coordinates of the infrared point light source beacon and the preset distance; X represents the infrared point light source beacon. i , j This indicates the infrared point light source beacon number.

[0011] Preferably, the visual navigation relative positioning model includes: an image processing part, a relative pose calculation part, and a data fusion part;

[0012] The image processing section is used to process the infrared point light source beacon image captured by the UAV;

[0013] The relative pose calculation part is used to obtain the relative pose of the UAV based on the infrared point light source beacon image;

[0014] The data fusion section is used to obtain the relative coordinates of the tethered UAV relative to the ground mobile platform based on the relative pose.

[0015] Preferably, the image processing section uses the FAST algorithm to extract beacon corner points and introduces a gradient direction weighting factor:

[0016]

[0017] in, Represents pixels The corner response value; Represents pixels The gradient vector; I represents the Gaussian kernel parameters; I represents the indicator function. This represents the average intensity of neighboring pixels.

[0018] Preferably, the relative pose calculation part combines the geometric features of the seven infrared point light source beacons to optimize the Hungarian matching cost matrix:

[0019]

[0020] in, Represents the SIFT descriptor for image points; Represents a real-point predefined descriptor; This represents the maximum matching distance threshold; I Indicates an indicator function; , , β , γ These represent the geometric distance weight, geometric layout constraint weight, descriptor similarity weight, and distance constraint penalty weight, respectively; T represents the transpose matrix. Xj Indicates the first A three-dimensional world coordinate system for a ground beacon; xiIndicates the detected first in the image Two-dimensional pixel coordinates of a feature point.

[0021] Preferably, the data fusion section introduces an extended Kalman filter, which uses the relative pose as an observation value and fuses it with the inertial navigation system (IMU) data of the tethered UAV itself to finally obtain the relative coordinates of the tethered UAV relative to the ground mobile platform.

[0022] The present invention also provides a multi-source data fusion cooperative target localization system based on infrared vision. The system is used to implement the above method and includes: a signal source providing module, an acquisition module, and a localization module.

[0023] The signal source providing module is an infrared point light source beacon deployed on the ground, which is activated when the signal is interfered with;

[0024] The acquisition module is used to acquire infrared point light source beacon images;

[0025] The positioning module is used to analyze the infrared point light source beacon image using the constructed visual navigation relative positioning model to complete the UAV target positioning under signal interference.

[0026] Preferably, the visual navigation relative positioning model includes: an image processing part, a relative pose calculation part, and a data fusion part;

[0027] The image processing section is used to process the infrared point light source beacon image captured by the UAV;

[0028] The relative pose calculation part is used to obtain the relative pose of the UAV based on the infrared point light source beacon image;

[0029] The data fusion section is used to obtain the relative coordinates of the tethered UAV relative to the ground mobile platform based on the relative pose.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0031] This invention utilizes infrared vision to support all-weather positioning, unaffected by lighting conditions; simultaneously, by fusing IMU and visual data through EKF, the positioning error is ≤5%. Furthermore, the invention employs a unique infrared beacon layout and band design, effectively resisting electromagnetic interference; and can achieve 24-hour continuous operation without cumulative error. The ground beacon structure of this invention is simple, requiring only a low-cost infrared camera and computing module from the UAV. Attached Figure Description

[0032] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0034] Figure 2 This is a schematic diagram of the cooperative target infrared point light source according to an embodiment of the present invention;

[0035] Figure 3 This is a schematic diagram of FAST feature extraction according to an embodiment of the present invention;

[0036] Figure 4 This is a schematic diagram of the P3P algorithm in an embodiment of the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0039] Example 1

[0040] This embodiment provides a multi-source data fusion cooperative target localization method based on infrared vision, the steps of which include:

[0041] S1. Deploy infrared point light source beacons on the ground and activate them when the signal is interfered with.

[0042] Seven asymmetrically distributed infrared point light source beacons were deployed in the ground cooperation target area. Four of them were located at the apex of the helipad, and three were located at a height of 2.5 meters above the ground. The beacon spacing was ≥2 meters, forming a unique geometric feature. A schematic diagram of the infrared point light sources is attached. Figure 2 The infrared beacon uses a specific wavelength (e.g., 850nm) light source to enhance penetration and anti-interference capabilities, ensuring clear imaging at an altitude of 300 meters. The beacon's three-dimensional coordinate matrix is ​​defined as follows:

[0043]

[0044] in, G represents the preset distance between beacons; G represents the geometric constraint function between the three-dimensional coordinates of the infrared point light source beacon and the preset distance; X represents the infrared point light source beacon. i , j This indicates the infrared point light source beacon number.

[0045] S2. Use the infrared camera mounted on the drone to acquire images of infrared point light source beacons.

[0046] When the drone is interfered with by satellite signals, the infrared point light source beacon is activated, and the infrared camera on the drone is used to acquire the infrared point light source beacon image.

[0047] S3. Analyze the infrared point light source beacon image using the constructed visual navigation relative positioning model to complete the UAV target localization under signal interference.

[0048] The visual navigation relative positioning model constructed in this embodiment includes: an image processing part, a relative pose calculation part, and a data fusion part.

[0049] The infrared image acquired by the UAV at the current moment is input into the image processing part, and through a deep learning algorithm, the acquired infrared beacon image is matched with the real coordinates of the infrared beacon to obtain the real coordinates of the infrared beacon and its corresponding infrared image coordinates.

[0050] Specifically, the image processing section uses the FAST algorithm to extract beacon corner points, and enhances the robustness of feature points by introducing a gradient direction weighting factor:

[0051]

[0052] in, Represents pixels The corner response value; Represents pixels The gradient vector; I represents the Gaussian kernel parameters; I represents the indicator function. This represents the average intensity of neighboring pixels.

[0053] If 12 consecutive neighboring pixels meet this condition, it is determined to be a corner point. A schematic diagram of FAST feature extraction is attached. Figure 3 Combining Focal Loss and CIoU loss to optimize object detection:

[0054]

[0055] Where Ltotal represents the total loss function for model training; λ1, λ2, and λ3 represent the weight coefficients used to control the contribution ratio of each loss term; LFocal represents the classification loss function; LCIoU represents the regression loss function; and Lobj represents the confidence loss function.

[0056] Classification Loss (Focal Loss):

[0057]

[0058] Among them, L Focal Represents the classification loss function; α represents the class balance factor; p t γ represents the confidence level of the model's prediction; γ represents the adjustment factor.

[0059] Regression Loss (CloU Loss):

[0060]

[0061] in, For aspect ratio consistency, Weights for geometric layout constraints; c This represents the diagonal length of the smallest bounding rectangle between the predicted bounding box and the ground truth bounding box. bpred Indicates the center point of the prediction box; bgt Indicates the center point of the true bounding box; ρ This represents the Euclidean distance between the center point of the predicted bounding box and the center point of the ground truth bounding box.

[0062] The relative pose calculation part is based on the infrared point light source beacon image to obtain the relative pose of the UAV. According to the coordinates of the infrared image obtained by the front end and the real beacon, the relative pose of the camera is estimated by the P3P algorithm, thereby obtaining the coordinates of the UAV relative to the ground mobile platform coordinate system where the ground platform is located, and thus obtaining the relative pose.

[0063] The Hungarian matching cost matrix is ​​optimized by combining the geometric features of the seven beacons:

[0064]

[0065] in, Represents the SIFT descriptor for image points; Represents a real-point predefined descriptor; This represents the maximum matching distance threshold; I Indicates an indicator function; , , β , γThe weights represent the weights used to balance the contributions of different cost terms, namely geometric distance weight, geometric layout constraint weight, descriptor similarity weight, and distance constraint penalty weight; T represents the transpose matrix. Xj Indicates the first A three-dimensional world coordinate system for a ground beacon; xi Indicates the detected first in the image Two-dimensional pixel coordinates of a feature point.

[0066] After minimizing the total matching cost, the P3P algorithm is used to calculate the UAV pose. Three ground points are considered. The coordinates in the camera coordinate system are The corresponding image projection point is A schematic diagram of the P3P algorithm is attached. Figure 4 Equation can be established using the Law of Cosines:

[0067] ,

[0068] in, Represents the rotation matrix. It is a translation vector.

[0069] Camera pose is solved using algebraic elimination, and the accuracy of P3P pose calculation is improved by utilizing the redundant information from seven beacons. Let the optimization objective of the projection points in the camera coordinate system be... The optimization objective is:

[0070]

[0071] in, For camera projection functions; The weights of the geometric constraints are denoted by xi; xi represents the weights of the first geometric constraint. The two-dimensional pixel coordinates of the i-th beacon in the image; Xi represents the i-th beacon. The three-dimensional world coordinates of an infrared beacon.

[0072] Because the relative pose of the UAV at each moment is independent, the error between each moment may be large, making it unsuitable as positioning information for control. Therefore, in the back-end data fusion section, an extended Kalman filter is introduced to fuse the relative pose obtained by the aforementioned method with the inertial navigation system (IMU) data of the tethered UAV itself, ultimately obtaining the relative coordinates of the tethered UAV relative to the ground mobile platform.

[0073] (1) Definition of state vector

[0074] The state vector x contains the UAV's position, velocity, attitude (quaternion), IMU accelerometer bias, and gyroscope bias:

[0075]

[0076] in, Indicates the position of the drone in the world coordinate system (3×1); This represents the velocity of the drone in the world coordinate system (3×1). The attitude quaternion (4×1) represents the distance from the body coordinate system to the world coordinate system. This indicates zero bias of the accelerometer (3×1); This indicates that the gyroscope has zero bias (3×1).

[0077] (2) State transition equation

[0078] IMU prediction model: via acceleration and angular velocity Estimated state:

[0079]

[0080] in, express The derived rotation matrix; g represents gravitational acceleration; This represents the quaternion multiplication operator; Let Q represent process noise, where Q is the covariance matrix that includes IMU noise characteristics.

[0081] Discretization form (sampling time) ):

[0082]

[0083] in, Indicates the first Prior state estimation at time step; Indicates the first Posterior state estimation at time 1.

[0084] (3) Observation equation (based on P3P positioning results):

[0085]

[0086] in, denoted as observation noise, where R is the covariance matrix containing the P3P positioning error; The 3D position of the drone calculated using the P3P algorithm; This represents the quaternion pose calculated by P3P. Convert to Euler angles (roll, pitch, yaw).

[0087] Mapping relationship between observation model and state:

[0088]

[0089] Among them, the observation matrix for:

[0090]

[0091] (4) Kalman gain and covariance update:

[0092] Kalman gain:

[0093]

[0094] Where Kk represents the Kalman gain matrix; Let represent the covariance matrix of the prior state estimate.

[0095] Status Update:

[0096]

[0097] in, Indicates the first Posterior state estimation at time 1; Indicates the first Prior state estimation at time step; This indicates actual observation.

[0098] Covariance update:

[0099]

[0100] in, Indicates the first The covariance matrix of the posterior state estimate at time t; Indicates the first The covariance matrix of the prior state estimate at time t.

[0101] Finally, the fused high-precision positioning information is obtained. The method flow of this invention is as follows: Figure 1 As shown.

[0102] Example 2

[0103] This embodiment also provides a multi-source data fusion cooperative target localization system based on infrared vision, including: a signal source providing module, an acquisition module, and a positioning module; the signal source providing module is an infrared point light source beacon deployed on the ground, which is activated when the signal is interfered with; the acquisition module is used to acquire images of the infrared point light source beacon; the positioning module is used to analyze the infrared point light source beacon images using a constructed visual navigation relative positioning model to complete the UAV target localization under the condition of signal interference.

[0104] The following will describe in detail, with reference to this embodiment, how the present invention solves the technical problems in practical work.

[0105] Seven asymmetrically distributed infrared point light source beacons were deployed in the ground cooperation target area. Four of them were located at the apex of the helipad, and three were located at a height of 2.5 meters above the ground. The beacon spacing was ≥2 meters, forming a unique geometric feature. A schematic diagram of the infrared point light sources is attached. Figure 2 The infrared beacon uses a specific wavelength (e.g., 850nm) light source to enhance penetration and anti-interference capabilities, ensuring clear imaging at an altitude of 300 meters. The beacon's three-dimensional coordinate matrix is ​​defined as follows:

[0106]

[0107] in, G represents the preset distance between beacons; G represents the geometric constraint function between the three-dimensional coordinates of the infrared point light source beacon and the preset distance; X represents the infrared point light source beacon. i , j This indicates the infrared point light source beacon number.

[0108] When the drone is interfered with by satellite signals, the infrared point light source beacon is activated, and the acquisition module (in this embodiment, the infrared camera mounted on the drone) is used to acquire the infrared point light source beacon image.

[0109] Finally, the positioning module is used to analyze the infrared point light source beacon image using the constructed visual navigation relative positioning model to complete the UAV target positioning under signal interference conditions.

[0110] The visual navigation relative positioning model constructed in this embodiment includes: an image processing part, a relative pose calculation part, and a data fusion part.

[0111] The infrared image acquired by the UAV at the current moment is input into the image processing part, and through a deep learning algorithm, the acquired infrared beacon image is matched with the real coordinates of the infrared beacon to obtain the real coordinates of the infrared beacon and its corresponding infrared image coordinates.

[0112] Specifically, the image processing section uses the FAST algorithm to extract beacon corner points, and enhances the robustness of feature points by introducing a gradient direction weighting factor:

[0113]

[0114] in, Represents pixels The corner response value; Represents pixels The gradient vector; I represents the Gaussian kernel parameters; I represents the indicator function. This represents the average intensity of neighboring pixels.

[0115] If 12 consecutive neighboring pixels meet this condition, it is determined to be a corner point. A schematic diagram of FAST feature extraction is attached. Figure 3 Combining Focal Loss and CIoU loss to optimize object detection:

[0116]

[0117] Where Ltotal represents the total loss function for model training; λ1, λ2, and λ3 represent the weight coefficients used to control the contribution ratio of each loss term; LFocal represents the classification loss function; LCIoU represents the regression loss function; and Lobj represents the confidence loss function.

[0118] Classification Loss (Focal Loss):

[0119]

[0120] Among them, L Focal Represents the classification loss function; α represents the class balance factor; p t γ represents the confidence level of the model's prediction; γ represents the adjustment factor.

[0121] Regression Loss (CloU Loss):

[0122]

[0123] in, For aspect ratio consistency, Weights for geometric layout constraints; c This represents the diagonal length of the smallest bounding rectangle between the predicted bounding box and the ground truth bounding box. bpred Indicates the center point of the prediction box; bgt Indicates the center point of the true bounding box; ρ This represents the Euclidean distance between the center point of the predicted bounding box and the center point of the ground truth bounding box.

[0124] The relative pose calculation part is based on the infrared point light source beacon image to obtain the relative pose of the UAV. According to the coordinates of the infrared image obtained by the front end and the real beacon, the relative pose of the camera is estimated by the P3P algorithm, thereby obtaining the coordinates of the UAV relative to the ground mobile platform coordinate system where the ground platform is located, and thus obtaining the relative pose.

[0125] The Hungarian matching cost matrix is ​​optimized by combining the geometric features of the seven beacons:

[0126]

[0127] in, Represents the SIFT descriptor for image points; Represents a real-point predefined descriptor; This represents the maximum matching distance threshold; I Indicates an indicator function; , , β , γ The weights represent the weights used to balance the contributions of different cost terms, namely geometric distance weight, geometric layout constraint weight, descriptor similarity weight, and distance constraint penalty weight; T represents the transpose matrix. Xj Indicates the first A three-dimensional world coordinate system for a ground beacon; xi Indicates the detected first in the image Two-dimensional pixel coordinates of a feature point.

[0128] After minimizing the total matching cost, the P3P algorithm is used to calculate the UAV pose. Three ground points are considered. The coordinates in the camera coordinate system are The corresponding image projection point is A schematic diagram of the P3P algorithm is attached. Figure 4 Equation can be established using the Law of Cosines:

[0129] ,

[0130] in, Represents the rotation matrix. It is a translation vector.

[0131] Camera pose is solved using algebraic elimination, and the accuracy of P3P pose calculation is improved by utilizing the redundant information from seven beacons. Let the optimization objective of the projection points in the camera coordinate system be... The optimization objective is:

[0132]

[0133] in, For camera projection functions; The weights of the geometric constraints are denoted by xi; xi represents the weights of the first geometric constraint. The two-dimensional pixel coordinates of the i-th beacon in the image; Xi represents the i-th beacon. The three-dimensional world coordinates of an infrared beacon.

[0134] Because the relative pose of the UAV at each moment is independent, the error between each moment may be large, making it unsuitable as positioning information for control. Therefore, in the back-end data fusion section, an extended Kalman filter is introduced to fuse the relative pose obtained by the aforementioned method with the inertial navigation system (IMU) data of the tethered UAV itself, ultimately obtaining the relative coordinates of the tethered UAV relative to the ground mobile platform.

[0135] (1) Definition of state vector

[0136] The state vector x contains the UAV's position, velocity, attitude (quaternion), IMU accelerometer bias, and gyroscope bias:

[0137]

[0138] in, Indicates the position of the drone in the world coordinate system (3×1); This represents the velocity of the drone in the world coordinate system (3×1). The attitude quaternion (4×1) represents the distance from the body coordinate system to the world coordinate system. This indicates zero bias of the accelerometer (3×1); This indicates that the gyroscope has zero bias (3×1).

[0139] (2) State transition equation

[0140] IMU prediction model: via acceleration and angular velocity Estimated state:

[0141]

[0142] in, express The derived rotation matrix; g represents gravitational acceleration; This represents the quaternion multiplication operator; Let Q represent process noise, where Q is the covariance matrix that includes IMU noise characteristics.

[0143] Discretization form (sampling time) ):

[0144]

[0145] in, Indicates the first Prior state estimation at time step; Indicates the first Posterior state estimation at time 1.

[0146] (3) Observation equation (based on P3P positioning results):

[0147]

[0148] in, denoted as observation noise, where R is the covariance matrix containing the P3P positioning error; The 3D position of the drone calculated using the P3P algorithm; This represents the quaternion pose calculated by P3P. Convert to Euler angles (roll, pitch, yaw).

[0149] Mapping relationship between observation model and state:

[0150]

[0151] Among them, the observation matrix for:

[0152]

[0153] (4) Kalman gain and covariance update:

[0154] Kalman gain:

[0155]

[0156] Where Kk represents the Kalman gain matrix; Let represent the covariance matrix of the prior state estimate.

[0157] Status Update:

[0158]

[0159] in, Indicates the first Posterior state estimation at time 1; Indicates the first Prior state estimation at time step; This indicates actual observation.

[0160] Covariance update:

[0161]

[0162] in, Indicates the first The covariance matrix of the posterior state estimate at time t; Indicates the first The covariance matrix of the prior state estimate at time t.

[0163] The final result is high-precision positioning information after fusion.

[0164] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A multi-source data fusion cooperative target localization method based on infrared vision, characterized by the following steps: include: Deploy infrared point light source beacons on the ground and activate them when the signal is interfered with; Using an infrared camera mounted on a drone, images of infrared point light source beacons are acquired; The constructed visual navigation relative positioning model is used to analyze the infrared point light source beacon image to complete the UAV target localization under signal interference. The visual navigation relative positioning model includes: an image processing part, a relative pose calculation part, and a data fusion part. The image processing section is used to process the infrared point light source beacon image acquired by the UAV; the image processing section uses the FAST algorithm to extract beacon corner points and introduces a gradient direction weighting factor: in, Represents pixels The corner response value; Represents pixels The gradient vector; I represents the Gaussian kernel parameters; I represents the indicator function. This represents the average intensity of neighboring pixels; The relative pose calculation part is used to obtain the relative pose of the UAV based on the infrared point light source beacon image; The data fusion section is used to obtain the relative coordinates of the tethered UAV relative to the ground mobile platform based on the relative pose. The data fusion section introduces an extended Kalman filter to fuse the relative pose as an observation value with the inertial navigation system (IMU) data of the tethered UAV itself, and finally obtains the relative coordinates of the tethered UAV relative to the ground mobile platform.

2. The multi-source data fusion cooperative target localization method based on infrared vision according to claim 1, characterized in that, Seven asymmetrically distributed infrared point light source beacons were deployed in the ground cooperation target area. Four of them were located at the apex of the helipad, and three were located at a height of 2.5 meters above the ground. The spacing between the beacons was greater than or equal to 2 meters. The three-dimensional coordinate matrix of the beacons was defined as follows: in, G represents the preset distance between beacons; G represents the geometric constraint function between the three-dimensional coordinates of the infrared point light source beacon and the preset distance; X represents the infrared point light source beacon. i , j This indicates the infrared point light source beacon number.

3. The multi-source data fusion cooperative target localization method based on infrared vision according to claim 2, characterized in that, The relative pose calculation part combines the geometric features of the seven infrared point light source beacons to optimize the Hungarian matching cost matrix: in, Represents the SIFT descriptor for image points; Represents a real-point predefined descriptor; This represents the maximum matching distance threshold; I Indicates an indicator function; , , β , γ These represent the geometric distance weight, geometric layout constraint weight, descriptor similarity weight, and distance constraint penalty weight, respectively; T represents the transpose matrix. Xj Indicates the first A three-dimensional world coordinate system for a ground beacon; xi Indicates the first detected in the image Two-dimensional pixel coordinates of a feature point.

4. A multi-source data fusion cooperative target localization system based on infrared vision, the system being used to implement the method described in any one of claims 1-3, characterized in that, include: Signal source module, acquisition module, and positioning module; The signal source providing module is an infrared point light source beacon deployed on the ground, which is activated when the signal is interfered with; The acquisition module is used to acquire infrared point light source beacon images; The positioning module is used to analyze the infrared point light source beacon image using the constructed visual navigation relative positioning model to complete the UAV target positioning under signal interference.

5. The multi-source data fusion cooperative target localization system based on infrared vision according to claim 4, characterized in that, The visual navigation relative positioning model includes: an image processing part, a relative pose calculation part, and a data fusion part; The image processing section is used to process the infrared point light source beacon image captured by the UAV; The relative pose calculation part is used to obtain the relative pose of the UAV based on the infrared point light source beacon image; The data fusion section is used to obtain the relative coordinates of the tethered UAV relative to the ground mobile platform based on the relative pose.

Images