Precise guidance unmanned aerial vehicle

Abstract

The invention discloses a precise guidance unmanned aerial vehicle, and relates to the technical field of precise guidance of unmanned aerial vehicles, and the precise guidance unmanned aerial vehicle comprises an unmanned aerial vehicle body, a ground control station, a multi-mode anti-interference navigation unit, an airborne AI control unit, a target identification and tracking unit, a visual control loop, a trajectory calculation unit and a throwing execution mechanism. The multi-modal navigation is fused with GPS, inertial measurement and visual / infrared feature matching, so that the problem of poor interference resistance of single navigation is solved; the target identification unit locks a target through deep learning and solves a three-dimensional coordinate; the visual control loop finely adjusts the path, the trajectory resolving unit combines real-time parameters and meteorological data to calculate the throwing preposition amount, and the throwing executing mechanism accurately triggers and releases. And a damage evaluation module can be arranged to realize strike effect evaluation and autonomous return. The method improves the navigation reliability and the striking precision, adapts to a complex environment, and meets the requirement of precise guidance.

Description

Technical Field

[0001] This invention relates to the field of precision guidance technology for unmanned aerial vehicles (UAVs), and particularly to a precision-guided UAV. Background Technology

[0002] With the rapid development of UAV technology, precision-guided UAVs are increasingly widely used in national defense, emergency response, and precision strikes. Their core requirements are stable navigation in complex environments, accurate target identification, and efficient strike delivery, while simultaneously achieving rapid assessment of strike effectiveness to improve mission reliability and accuracy. Currently, navigation systems overly rely on a single Global Positioning System (GPS). In complex battlefield environments or scenarios with strong electromagnetic interference, satellite signals are easily interfered with, blocked, or even completely interrupted, leading to significantly increased positioning deviations, loss of flight attitude control, and inability to reach the target area along the predetermined route. While quantum navigation possesses high precision and anti-interference characteristics, its complex structure, high cost, and sensor sensitivity to the environment, coupled with challenging data processing, make it difficult to adapt to various UAV platforms. Celestial navigation, while simple in construction and highly concealable, is severely limited by weather conditions and day / night cycles, only applicable to specific high-altitude environments and unable to meet all-weather, all-scenario navigation needs. Although a combination of inertial navigation and GPS is partially adopted, inertial navigation suffers from cumulative errors and lacks an effective error calibration mechanism, making it difficult to maintain long-term stable positioning accuracy after satellite signal loss.

[0003] In target recognition and tracking, traditional image recognition algorithms suffer from low accuracy in identifying targets against complex backgrounds, struggle to quickly lock onto dynamic targets, and lack precise miss distance calculation and 3D spatial coordinate resolution capabilities, resulting in significant target positioning deviations. Even with the introduction of deep learning models, slow recognition speed and weak anti-interference capabilities are common, failing to meet the rapid response requirements of battlefield environments. During drop strikes, ballistic calculations only consider basic flight parameters, failing to fully integrate real-time meteorological data for pre-emptive correction, leading to significant deviations in the impact point of the projectile. Drop execution mechanisms often employ fixed-timing triggering methods, unable to dynamically adjust the release timing based on the real-time matching status of the impact point and target coordinates, further weakening the accuracy of guided strikes. Furthermore, existing systems generally lack damage assessment capabilities, failing to perform rapid secondary imaging and strike effect analysis of the target area after drop, making it difficult for operators to monitor strike results in real time. Return path planning relies on manual intervention and lacks coordination with the navigation system, resulting in low mission execution efficiency. These technical bottlenecks severely restrict the reliability and accuracy of precision-guided missions in complex environments. Summary of the Invention

[0004] The purpose of this application is to provide a precision-guided unmanned aerial vehicle (UAV) with advantages such as improved navigation reliability, enhanced strike accuracy, autonomous target recognition, and accurate ballistic calculation.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows: A precision-guided unmanned aerial vehicle (UAV), characterized in that it comprises: The drone itself is equipped with an AI control module, a GPS module on one side of its upper part, and a throwing device at its lower part. Ground control station, used to mark targets on digital maps and calculate and generate a baseline mission path that includes bomb drop point coordinates and entry route; A multimodal anti-interference navigation unit is mounted on the UAV body and includes a GPS receiver, an inertial measurement unit, and an optoelectronic pod connected to the GPS module. The optoelectronic pod is used to acquire ground images in real time. An airborne AI control unit is installed on the UAV body and serves as the core of the AI ​​control module. It is connected to the multimodal anti-interference navigation module and the flight control interface. The airborne AI control unit has an embedded integrated navigation and positioning module, which is used to match the real-time ground images collected by the optoelectronic pod with the pre-stored ground station image features when the GPS signal is lost, and calculate the absolute position of the UAV to maintain stable flight. The target recognition and tracking unit is connected to the optoelectronic pod and the airborne AI control unit. It is used to identify and lock the predetermined target through a deep learning model after the UAV enters the target area, and to calculate the three-dimensional spatial coordinates of the target relative to the UAV. The visual control loop, integrated into the airborne AI control unit and connected to the target recognition and tracking unit, is used to generate a locally fine-tuned path based on the target's relative coordinates, guiding the drone to the dynamically adjusted bombing point. The ballistic calculation unit, connected to the vision control loop, is used to calculate the lead-out amount based on real-time flight parameters and meteorological data. The throwing mechanism is connected to both the ballistic calculation unit and the throwing device, and is used to trigger the throwing device to perform a release action when the current impact point coincides with the target coordinates, so as to throw the projectile.

[0006] Furthermore, the integrated navigation and positioning module includes a judgment subunit. When the GPS signal quality is lower than a preset threshold, it automatically switches to the visual / infrared feature matching navigation mode, and uses the real-time ground image obtained by the down-looking optoelectronic pod to match the feature points of the ground station image stored on the airborne in real time, and combines the data of the inertial measurement unit to maintain the stability of the flight attitude.

[0007] Furthermore, the target recognition and tracking unit includes a miss distance calculation subunit, which is used to calculate the miss distance of the target in the photoelectric pod image coordinate system, and to obtain the three-dimensional spatial coordinates of the target relative to the UAV using laser ranging or image size estimation.

[0008] Furthermore, the ballistic calculation unit is connected to the target identification and tracking unit to calculate the lead-out amount in real time by integrating the current flight speed, flight altitude, flight attitude angle, real-time wind speed and direction, and the accurate coordinates of the target.

[0009] Furthermore, the UAV body is also equipped with a damage assessment module, which is connected to the optoelectronic pod and the airborne AI control unit. It is used to perform secondary imaging of the target area after the drop, compare the changes in the target image before and after the strike through the AI ​​model, generate a strike effect assessment report, and trigger the multimodal anti-interference navigation module to plan the return path.

[0010] Furthermore, the precision guidance includes the following steps: Step S1: Mark the target on the digital map of the ground control station and calculate and generate a baseline mission path that includes the coordinates of the bomb drop point and the entry route. Step S2: After the drone takes off, the onboard AI control unit performs integrated navigation through the integrated navigation and positioning module. When the GPS signal is lost, the drone's absolute position is calculated by matching the real-time ground images collected by the photoelectric pod with the pre-stored ground station image features to maintain stable flight. Step S3: After the UAV enters the target area, the predetermined target is identified and locked through a deep learning model, and the three-dimensional spatial coordinates of the target relative to the UAV are calculated. Step S4: The visual control loop generates a locally fine-tuned path based on the target's relative coordinates, guiding the UAV to the dynamically adjusted bomb drop point; Step S5: The ballistic calculation unit calculates the lead-out amount based on real-time flight parameters and meteorological data. When the calculated future impact point coincides with the current target coordinates, the autonomous drop-out actuator is triggered to perform the release action. Step S6: After the launch, the target area is imaged a second time. The changes in the target image before and after the strike are compared using an AI model to generate a strike effect assessment report and trigger the multimodal anti-interference navigation module to plan the return path.

[0011] Furthermore, the specific steps of the integrated navigation and positioning module performing integrated navigation in step S2 include: When the GPS signal is good, use satellite data to calibrate the cumulative error of the visual inertial odometry. When the GPS signal is interfered with or its quality is below a preset threshold, it automatically switches to the visual / infrared feature matching navigation mode. It uses the real-time ground images acquired by the down-looking electro-optical pod to match the feature points of the ground station images stored on the airborne in real time, calculates its own position, and combines the data from the inertial measurement unit to maintain stable flight attitude.

[0012] Furthermore, the specific steps in step S3 for calculating the target's three-dimensional spatial coordinates relative to the UAV include: Real-time analysis of the video stream of the photoelectric / infrared payload is performed using a deep learning model to identify the preset targets. Calculate the target's miss distance in the image coordinate system; The three-dimensional spatial coordinates of the target relative to the UAV are obtained by using laser ranging or image size estimation.

[0013] Furthermore, the specific steps for the ballistic calculation unit to calculate the pre-launch amount in step S5 include: Based on the current flight speed, flight altitude, flight attitude angle, real-time wind speed and direction data, and the accurate target coordinates measured in step S3; Real-time calculation of the pre-release quantity to determine the optimal release time.

[0014] Furthermore, the specific steps for generating the strike effect evaluation report in step S6 include: The drone continues to fly or hover according to a preset trajectory; Secondary imaging of the target area is performed using an optoelectronic pod; AI models compare changes in target images before and after the strike, including changes in smoke features or target shape. Generate a strike effectiveness assessment report and transmit it back to the ground station.

[0015] The beneficial effects of this invention are as follows: by setting up a multimodal anti-interference navigation unit and integrating GPS, inertial measurement and visual / infrared feature matching navigation, continuous and reliable positioning and attitude stabilization are achieved in complex environments such as strong electromagnetic interference and satellite signal blockage, which significantly improves the navigation reliability and mission adaptability of UAVs in complex environments. The target recognition and tracking algorithm, which combines miss distance calculation and laser ranging / image size estimation, enables rapid target locking, high-precision positioning and three-dimensional coordinate calculation. It effectively solves the problems of low recognition rate and lack of depth information in traditional image recognition under complex backgrounds, and improves the stability and accuracy of target tracking. By integrating a vision control loop, a ballistic calculation unit, and an adaptive throwing mechanism, dynamic ballistic correction and precise throwing based on real-time target coordinates, flight parameters, and meteorological data are achieved, significantly reducing the impact point deviation and improving the hit rate against both fixed and dynamic targets. Attached Figure Description

[0016] Fig. 1 This is a schematic diagram of a precision-guided unmanned aerial vehicle (UAV) according to the present invention; Fig. 2 This is a system architecture diagram of a precision-guided unmanned aerial vehicle according to the present invention.

[0017] As shown in the figure: 1. Drone body; 2. AI control module; 3. GPS module; 4. Drop device. Detailed Implementation

[0018] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0019] like Figs. 1-2 As shown, a precision-guided drone includes a drone body 1, an AI control module 2 on which is provided, a GPS module 3 on one side of its upper part, and a throwing device 4 on its lower part. Ground control station, used to mark targets on digital maps and calculate and generate a baseline mission path that includes bomb drop point coordinates and entry route; A multimodal anti-interference navigation unit is installed on the UAV body 1, including a GPS receiver, an inertial measurement unit and an optoelectronic pod connected to the GPS module 3. The optoelectronic pod is used to acquire ground images in real time. An airborne AI control unit is installed on the UAV body 1 and serves as the core of the AI ​​control module 2. It is connected to the multimodal anti-interference navigation unit and the flight control interface. The airborne AI control unit has an embedded integrated navigation and positioning module, which is used to match the real-time ground images collected by the optoelectronic pod with the pre-stored ground station image features when the GPS signal is lost, and calculate the absolute position of the UAV to maintain stable flight. The target recognition and tracking unit, connected to the optoelectronic pod and the airborne AI control unit, is used to identify and lock onto the predetermined target through a deep learning model after the UAV enters the target area, and to calculate the three-dimensional spatial coordinates of the target relative to the UAV. The vision control loop, integrated into the airborne AI control unit and connected to the target recognition and tracking unit, is used to generate a locally fine-tuned path based on the target's relative coordinates, guiding the drone to the dynamically adjusted bombing point. The ballistic calculation unit, connected to the vision control loop, is used to calculate the lead-out amount based on real-time flight parameters and meteorological data; The throwing mechanism is connected to the ballistic calculation unit and the throwing device 4 respectively, and is used to trigger the throwing device 4 to perform a release action when the current impact point coincides with the target coordinates, so as to throw the ball.

[0020] This application further proposes that the integrated navigation and positioning module includes a judgment subunit, with a preset GPS signal quality threshold of T. When the detected GPS signal quality S < T, it automatically switches to the visual / infrared feature matching navigation mode, using real-time ground images acquired by the down-looking optoelectronic pod to perform real-time matching with pre-stored ground station image feature points on the airborne system, and combining this with inertial measurement unit data to maintain flight attitude stability. The attitude stability judgment condition is: attitude angle deviation. ≤0.5° ≤0.5° For pitch angle deviation, (This refers to the roll angle deviation).

[0021] Specifically, the judgment subunit is a processing unit within the integrated navigation and positioning module, whose main function is to continuously monitor and evaluate the quality of GPS signals. It is responsible for acquiring various quality indicators of GPS signals in real time, such as carrier-to-noise ratio (C / N0), horizontal accuracy factor (HDOP), number of visible satellites, signal strength, etc., and comparing them with a preset GPS signal quality threshold T.

[0022] The preset GPS signal quality threshold T is a key configuration parameter used to define the availability and reliability of the GPS signal. This threshold T can be preset according to the UAV's mission requirements, the complexity of the flight environment, and the performance characteristics of the GPS receiver. When the carrier-to-noise ratio is lower than a certain value, or the horizontal accuracy factor is higher than a certain value, the GPS signal quality can be considered poor or lost. The setting of this threshold aims to ensure that the system can switch to the backup navigation mode in a timely and accurate manner when the GPS navigation accuracy cannot meet the mission requirements.

[0023] When the judgment subunit detects that the GPS signal quality S is lower than a preset threshold T, the system will automatically switch to visual / infrared feature matching navigation mode. This automatic switching mechanism ensures the continuity and robustness of navigation, avoiding navigation failure due to GPS signal interruption or interference. During the switching process, the airborne AI control unit will instruct the electro-optical pod in the multimodal anti-interference navigation unit to start the downward view image acquisition function and activate the corresponding image processing and matching algorithms.

[0024] In visual / infrared feature matching navigation mode, the UAV uses real-time ground images acquired by a downward-looking electro-optical pod to perform real-time matching with pre-stored ground station image feature points. The electro-optical pod continuously acquires high-resolution images of the ground below the UAV, which are then transmitted to the onboard AI control unit for processing. The onboard AI control unit extracts identifiable feature points from the real-time images using image processing algorithms. Simultaneously, the UAV has a pre-stored database of ground station image feature points for the target area. These feature points are typically geocoded and contain precise geographical location information. By matching the real-time extracted feature points with feature points in the pre-stored database, the system can calculate the UAV's precise position and attitude information relative to the ground.

[0025] The visual / infrared feature matching navigation mode also incorporates inertial measurement unit (IMU) data to maintain flight attitude stability. The IMU provides high-frequency acceleration and angular velocity data. By fusing the relative position information obtained through visual matching with the attitude and motion information provided by the IMU (using Kalman filtering or extended Kalman filtering algorithms), the system can obtain smoother and more accurate UAV attitude and position estimates. IMU data can effectively compensate for transient errors that may occur in visual matching when image texture is insufficient or during rapid movement, ensuring that the UAV maintains a stable flight attitude even in visual navigation mode. ≤0.5° ≤0.5° For pitch angle deviation, (This refers to the roll angle deviation).

[0026] In this process, the attitude stability determination condition is set as: attitude angle deviation. ≤0.5° ≤0.5° For pitch angle deviation, (This refers to roll angle deviation). This means that, in visual / infrared feature matching navigation mode, the onboard AI control unit will continuously monitor the drone's pitch angle deviation. and roll angle deviation The UAV is considered to be in a stable state only when both attitude angle deviations remain within 0.5°. This criterion is crucial for ensuring the accuracy of visual matching, as unstable attitudes can lead to image blurring or distortion, affecting the accuracy of feature extraction and matching. If the attitude deviation exceeds this range, the flight control system will immediately adjust to restore the UAV to a stable attitude, thereby ensuring the reliability of visual navigation.

[0027] Through the above-described solution, this application effectively addresses the reliability and stability issues of UAV navigation when GPS signals are interfered with or lost. The introduction of a judgment subunit and a preset GPS signal quality threshold T enables the UAV to intelligently and autonomously assess the GPS signal condition and promptly switch to a backup navigation mode, preventing navigation failure due to GPS signal interruption. The visual / infrared feature matching navigation mode, combined with inertial measurement unit data, provides high-precision absolute position and attitude information, ensuring continuous and stable flight in GPS-free environments. Clear attitude stability determination conditions guarantee the quality of visual image acquisition and feature matching, thereby improving the overall accuracy and robustness of navigation. This multimodal fusion and intelligent switching mechanism enhances the UAV's mission execution and survivability in complex electromagnetic environments.

[0028] This application further proposes that the target recognition and tracking unit includes a miss distance calculation subunit, assuming that the theoretical center coordinates of the target in the photoelectric pod image coordinate system are... The actual identified coordinates are The formula for calculating the miss distance D is: The system uses laser ranging or image size estimation to obtain the target's three-dimensional spatial coordinates relative to the UAV.

[0029] To obtain the precise three-dimensional spatial coordinates of a target relative to a UAV, this application provides two optional technical approaches. First, laser ranging technology can be used. A laser beam is emitted towards the target using a laser rangefinder integrated into the electro-optical pod, and the reflected signal is received to directly measure the slant distance between the UAV and the target. Combining the UAV's attitude information and the electro-optical pod's field of view, the relative height and horizontal distance of the target can be accurately calculated, thus obtaining the three-dimensional spatial coordinates. Second, an image size estimation method can be used. If the actual physical dimensions of the target (length, width, or height) are known, the distance between the target and the UAV can be deduced by using the pixel size occupied by the target in the real-time image acquired by the electro-optical pod, combined with the camera's focal length, pixel size, and other internal parameters, using the principle of perspective projection. Both methods effectively overcome the limitations of single-vision recognition in obtaining depth information, providing reliable three-dimensional positioning data for precision guidance.

[0030] Through the above scheme, this application introduces a miss distance calculation subunit in the target recognition and tracking process, and clarifies the calculation formula for the miss distance D. This enables the UAV to evaluate the accuracy of the deep learning model in target recognition in real time and quantitatively, thereby effectively solving the problem of insufficient positioning accuracy that may be caused by relying solely on model recognition. When the miss distance D is detected to exceed the preset range, the system can adjust the target recognition parameters in a timely manner or trigger the vision control loop for more precise path fine-tuning to ensure that the UAV always stays aligned with the target. By using two more accurate methods, laser ranging or image size estimation, to obtain the three-dimensional spatial coordinates of the target relative to the UAV, the robustness and accuracy of target positioning are further improved, making up for the shortcomings of pure vision recognition in depth information acquisition. This technical solution, which combines quantitative error assessment and multi-source three-dimensional positioning, improves the ability of precision-guided UAVs to lock onto and track dynamic targets in complex environments, providing more reliable and accurate input data for subsequent ballistic calculation and drop execution, thereby ensuring the success rate and accuracy of strike missions.

[0031] This application further proposes that the ballistic calculation unit is connected to the target recognition and tracking unit, and is used to integrate the current flight speed v, flight altitude h, and flight attitude angle. Real-time wind speed and direction (v wind) and the target's accurate coordinates Real-time calculation of the pre-cast quantity L, where the basic calculation formula for the pre-cast quantity is: Where t is the estimated time from release to impact with the target, adjusted for wind speed and direction: .

[0032] Specifically, the ballistic calculation unit establishes a connection with the target recognition and tracking unit to obtain the accurate target coordinates provided by the target recognition and tracking unit. This is the fundamental data for ballistic calculations. This connection ensures that ballistic solutions are based on the latest and most accurate target position information. When calculating the pre-launch trajectory, the ballistic calculation unit needs to consider several key parameters, including the current flight speed. The horizontal flight speed of the UAV at the moment of release is a key parameter for calculating the horizontal inertial distance of the projectile. Flight altitude *h* refers to the height of the UAV relative to the target at the moment of release. This altitude determines the time required for the projectile to free fall under gravity, thus affecting its horizontal displacement. Flight attitude angle *α* refers to the pitch and roll angles of the UAV at the moment of release. These attitude angles affect the initial velocity vector of the projectile when it leaves the UAV, thus having a small but significant impact on the trajectory. Real-time wind speed and direction *v_wind* refers to the real-time wind force affecting the projectile during its flight. Wind force causes additional horizontal drift, therefore requiring accurate measurement and calculation for compensation. Target accurate coordinates. This refers to the precise position of the target in three-dimensional space provided in real time by the target recognition and tracking unit; this is the final target point in ballistic calculation. The ballistic calculation unit can comprehensively evaluate the entire process of the projectile from release to impact with the target, thereby performing accurate ballistic prediction.

[0033] Using the above method, the ballistic calculation unit can comprehensively analyze the real-time flight parameters of the UAV (flight speed, etc.). Flight altitude Flight attitude angle Real-time environmental data (wind speed, wind direction, and wind speed v) and accurate target coordinates provided by the target recognition and tracking unit. The system performs real-time calculation of the lead time L for the projectile. Specifically, it first calculates the basic lead time based on the UAV's flight speed and the estimated flight time of the projectile, and then corrects this lead time according to real-time wind speed and direction. This real-time, comprehensive ballistic calculation mechanism overcomes the insufficient accuracy problem of relying solely on the visual control loop to guide the UAV to the drop point, and effectively compensates for the ballistic deviation caused by factors such as gravity, UAV inertia, and wind during the projectile's flight. This allows the delivery mechanism to trigger the delivery device 4 to execute the release action at the moment when the impact point precisely coincides with the target coordinates, thereby significantly improving the accuracy and hit rate of the precision-guided UAV in striking targets in complex dynamic environments and ensuring the successful execution of the mission.

[0034] This application further proposes that a damage assessment module is also provided on the UAV body 1. This damage assessment module is connected to the electro-optical pod and the onboard AI control unit, and is used to perform secondary imaging of the target area after deployment. Specifically, the damage assessment module is a hardware and / or software unit specifically designed to assess the degree of damage to the target. It can be an independent processor unit with embedded image processing algorithms and AI models, or it can be implemented as part of the functionality of the onboard AI control unit 2. Through its connection with the electro-optical pod, it can acquire visual information of the target area after impact, serving as raw data for assessing the degree of damage. After the deployment is completed, the damage assessment module controls the electro-optical pod to photograph the target area again. This can be done by hovering above the original deployment point or by flying along a preset path over the target area to scan and photograph.

[0035] After acquiring secondary imaging data, the damage assessment module sets the target characteristic parameters before the attack as follows: The target's characteristic parameters after the strike are And according to the public Calculate the extent of damage Here, the target feature parameter A can be the changes in the target's area, outline, color, texture, etc., before and after the impact. The damage assessment module extracts these feature parameters through image processing and pattern recognition technology, and substitutes them into the formula to calculate the K value, thereby providing a mathematical model for quantitatively assessing the degree of damage.

[0036] Through the above-described scheme, this application achieves real-time, quantitative assessment of the strike effect, avoiding the subjectivity and lag of manual judgment. In-depth analysis of image changes before and after the strike using an AI model improves the accuracy and reliability of the assessment, providing crucial decision-making support for mission command. Simultaneously, automatic triggering of return path planning significantly enhances the autonomy of the UAV system and mission execution efficiency, reduces the operator's workload, and ensures the UAV can return safely and efficiently after completing its mission.

[0037] This application further proposes that precision guidance includes the following steps: First, the target is marked on the digital map of the ground control station, and a baseline mission path containing the coordinates of the bomb drop point and the entry route is calculated and generated. Secondly, after the drone takes off, the onboard AI control unit performs integrated navigation through the integrated navigation and positioning module. When the GPS signal is lost, it matches the real-time ground images collected by the photoelectric pod with the pre-stored ground station image features to calculate the absolute position of the drone in order to maintain stable flight. Next, once the drone enters the target area, it identifies and locks onto the predetermined target using a deep learning model, and calculates the target's three-dimensional spatial coordinates relative to the drone. Subsequently, the vision control loop generates a locally fine-tuned path based on the target's relative coordinates, guiding the UAV to the dynamically adjusted bombing point; Furthermore, the ballistic calculation unit calculates the lead time for the ballistic missile based on real-time flight parameters and meteorological data. When the calculated future impact point coincides with the current target coordinates, the autonomous ballistic missile release mechanism is triggered to perform the release action. Finally, after the missile is launched, the target area is imaged a second time. The AI ​​model is used to compare the changes in the target images before and after the strike, generate a strike effect assessment report, and trigger the multimodal anti-jamming navigation module to plan the return path.

[0038] Through the above-described scheme, this application provides a complete, closed-loop operational process for precision-guided unmanned aerial vehicles (UAVs), encompassing mission planning, flight navigation, target identification, precision strike, and damage assessment. This process organically integrates various advanced functional modules of the UAV body 1, such as the multimodal anti-jamming navigation unit, target identification and tracking unit, visual control loop, ballistic calculation unit, and damage assessment module, forming an efficient and collaborative working system. Clearly defined steps ensure the systematic and coherent execution of missions, effectively resolving issues of inefficiency or poor coordination that may arise when modules operate independently. In complex environments where GPS signals are interfered with, the integrated navigation capability ensures the UAV's continuous and stable flight; the visual identification and control mechanism enables the UAV to perform real-time, high-precision tracking and path fine-tuning of dynamic targets; real-time ballistic calculation ensures optimal delivery timing; and automated damage assessment provides timely and objective feedback for mission decision-making. Overall, this process significantly improves the mission success rate, strike accuracy, and combat effectiveness of precision-guided UAVs in complex battlefield environments.

[0039] This application further proposes that the specific steps for the integrated navigation positioning module to perform integrated navigation include: when the GPS signal is good (S≥T, T is a preset threshold), using satellite data to calibrate the cumulative error of the visual inertial odometry, and the error calibration formula is: in, This is the error after calibration. This is the original cumulative error. To compensate for satellite calibration errors; when the GPS signal is interfered with or its quality is below a preset threshold (S<T), it automatically switches to visual / infrared feature matching navigation mode, uses real-time ground images acquired by the down-looking optoelectronic pod to match the feature points of the ground station images stored on the airborne in real time, calculates its own position, and combines the data from the inertial measurement unit to maintain stable flight attitude.

[0040] Through the above technical solution, this application effectively solves the problem of difficulty in ensuring the navigation accuracy and flight stability of UAVs when GPS signal quality fluctuates or is interfered with. When the GPS signal is good, its high-precision absolute positioning capability is used to correct the cumulative error of the visual inertial odometry, ensuring the long-term accuracy and reliability of the navigation system. When the GPS signal quality deteriorates or is interfered with, the system can intelligently and quickly switch to visual / infrared feature matching navigation mode, use ground image features for autonomous positioning, and combine inertial measurement unit data to maintain flight attitude stability. This achieves seamless connection and high-precision maintenance of UAV navigation in complex electromagnetic environments, greatly improving the adaptability and success rate of UAVs in precision guidance missions.

[0041] This application further proposes that the specific steps for calculating the target's three-dimensional spatial coordinates relative to the UAV include: real-time analysis of the video stream of the electro-optical / infrared payload using a deep learning model to identify the preset target; and calculating the target's miss distance in the image coordinate system. , ( , Using the theoretical center coordinates, , (For actual identification coordinates); the three-dimensional spatial coordinates of the target relative to the UAV are obtained using laser ranging or image size estimation. If laser ranging is used, let the laser ranging value be d, and the pitch angle be... Then the relative height of the target Horizontal distance .

[0042] This study employs a deep learning model to perform real-time analysis of the video stream of an electro-optical / infrared payload. The aim is to automatically and accurately identify pre-defined targets from consecutive video frames. After target identification, the miss distance of the target in the image coordinate system needs to be calculated. Off-target amount It is a key indicator for measuring target recognition accuracy and positioning deviation, and its calculation formula is: ,in( , ) represents the theoretical center coordinates of the target in the image coordinate system, while ( , The coordinates represent the actual center coordinates of the target identified by the deep learning model. The theoretical center coordinates can be determined based on a preset attack strategy or target type. The actual identified coordinates are the center position of the target on the image plane in the target detection results output by the deep learning model. By calculating the miss distance, the accuracy of target recognition can be quantified in real time, providing a precise correction basis for subsequent vision control loops to guide the UAV to more accurately align with the target.

[0043] This application utilizes laser ranging or image size estimation to obtain the three-dimensional spatial coordinates of the target relative to the UAV. When laser ranging is used, the laser rangefinder, typically integrated into the electro-optical pod, emits a laser beam towards the target and receives the reflected signal, thereby accurately measuring the straight-line distance from the UAV to the target. Meanwhile, the attitude sensor (inertial measurement unit) inside the electro-optical pod can provide the pitch angle of the electro-optical pod relative to the UAV body 1. Combined with the UAV's own flight attitude, the pitch angle of the target relative to the UAV can be obtained. Based on this data, the vertical height difference between the target and the drone can be calculated using simple trigonometric functions. and horizontal distance If image size estimation is used, the actual size of the target needs to be known in advance. By measuring the pixel size of the target in the image and combining it with parameters such as the camera's focal length, pixel size, and the drone's flight altitude, the distance between the target and the drone can be deduced using the principle of similar triangles or camera calibration parameters. This method is usually used as a backup or auxiliary means for laser ranging to improve the robustness of ranging.

[0044] Through the above technical solution, after the UAV enters the target area, the video stream of the electro-optical / infrared payload can be analyzed in real time using a deep learning model, achieving high-precision identification and locking of the preset target. By calculating the miss distance of the target in the image coordinate system, the accuracy of target identification can be quantified in real time, providing a precise basis for correction in subsequent vision control loops. Using laser ranging or image size estimation, the three-dimensional spatial coordinates of the target relative to the UAV can be accurately obtained. When using laser ranging, by combining the ranging value and the pitch angle, the relative height and horizontal distance of the target can be accurately calculated. This improves the accuracy and reliability of target positioning, provides high-quality input data for the ballistic calculation unit, and effectively solves the challenge of accurately obtaining the target's three-dimensional spatial coordinates, ensuring the strike accuracy and mission success rate of precision-guided UAVs in complex dynamic environments.

[0045] This application further proposes that the specific steps for the ballistic calculation unit to calculate the lead-out amount include: Based on current flight speed Flight altitude Flight attitude angle Real-time wind speed and direction And the aforementioned accurate coordinates of the target, ; First, calculate the fall time of the projectile using the formula for free fall. ( Let be the acceleration due to gravity, and take . Then calculate the pre-cast quantity in real time. Determine the optimal time to release.

[0046] The ballistic calculation unit can comprehensively analyze the real-time flight speed of the UAV. Flight altitude Flight attitude angle Real-time wind speed and direction and the above-mentioned accurate coordinates of the target This multi-dimensional data provides a solid foundation for calculating the pre-launch amount. The fall time of the launched object is calculated based on the free-fall motion formula. And combined with the horizontal flight speed of the drone Wind speed and wind direction It calculates the precise lead-out amount L in real time. This overcomes the accuracy problems that traditional methods may suffer from due to incomplete data or model simplification, ensuring the accuracy of trajectory prediction for projectiles in complex dynamic environments, thereby significantly improving the hit rate and mission success rate of precision-guided UAVs.

[0047] This application further proposes that the specific steps for generating a strike effectiveness assessment report include: The drone continues to fly or hover according to a preset trajectory; Secondary imaging of the target area is performed using an optoelectronic pod; AI models compare changes in target images before and after the attack, including changes in smoke features or target morphology, to calculate the extent of damage. ( For the characteristic parameters before the attack, (Post-strike characteristic parameters). Generate a strike effectiveness assessment report and transmit it back to the ground station.

[0048] After completing its delivery mission, the drone can autonomously conduct secondary observations of the target area and utilize advanced AI models to intelligently analyze and quantitatively evaluate the image data before and after the strike. This method overcomes the subjectivity and lag of traditional manual assessments, achieving an objective, rapid, and accurate evaluation of the strike's effectiveness. By calculating the damage severity K-value, the ground control station can obtain clear and quantitative mission feedback, enabling it to promptly determine mission success, the need for a secondary strike, or adjustments to subsequent operational strategies. This significantly improves the automation level and decision-making efficiency of precision-guided missions, ensuring closed-loop management and optimization of the mission.

[0049] Example: First, the user precisely marks the moving target on the digital map of the ground control station. The ground control station then calculates and generates a baseline mission flight path that includes the coordinates of the bomb drop point and the entry route. The UAV body 1 takes off after receiving the mission command.

[0050] The onboard AI control unit, located on the UAV body 1, serves as the core of the AI ​​control module 2, and performs integrated navigation through a multimodal anti-interference navigation unit. This multimodal anti-interference navigation unit includes a GPS receiver connected to the GPS module 3, an inertial measurement unit, and an electro-optical pod. During UAV flight, the integrated navigation and positioning module embedded in the onboard AI control unit continuously monitors the GPS signal quality. When the UAV enters a densely populated area with tall buildings, and the detected GPS signal quality S falls below a preset threshold T, the judgment subunit of the integrated navigation and positioning module automatically switches to visual / infrared feature matching navigation mode. At this time, the electro-optical pod acquires real-time downward-looking ground images and performs real-time matching with pre-stored ground station image feature points to calculate the UAV's absolute position. Simultaneously, combined with data from the inertial measurement unit, the UAV's flight attitude is maintained stably, ensuring accurate pitch angle deviation. and roll angle deviation All angles were maintained within 0.5°. This multimodal navigation method effectively overcomes the problems of single GPS navigation being susceptible to interference and experiencing a decrease in positioning accuracy in complex environments, ensuring stable flight and accurate position awareness for UAVs in the absence of GPS signals.

[0051] Once the drone enters the target area, the target recognition and tracking unit begins operation. This unit connects to the electro-optical pod and the onboard AI control unit, using a deep learning model to analyze the video stream acquired by the electro-optical pod in real time, identifying and locking onto the predetermined target. Compared to traditional image recognition algorithms, the deep learning model demonstrates higher recognition accuracy and stronger anti-interference capabilities in such complex environments. The target recognition and tracking unit also includes a miss distance calculation subunit, which calculates the target's miss distance in the image coordinate system. Subsequently, laser ranging is used to obtain the target's three-dimensional spatial coordinates relative to the UAV. For example, if the laser ranging value is d and the pitch angle is β, then the target's relative height... Horizontal distance This provides accurate target location information, solving the problem of large target positioning deviations in existing technologies.

[0052] The vision control loop within the airborne AI control unit receives the target's relative coordinates from the target recognition and tracking unit. Based on these real-time coordinates, the vision control loop generates a locally fine-tuned path, dynamically guiding the drone to the adjusted bombing point to adapt to the target's movement.

[0053] The ballistic calculation unit is connected to the vision control loop, taking into account the current flight speed. Flight altitude Flight attitude angle Real-time wind speed and direction And the aforementioned accurate coordinates of the target, The ballistics calculation unit first calculates the descent time of the projectile based on the free fall motion formula. Then, calculate the pre-cast quantity in real time. Compared with existing ballistic calculation methods that do not fully consider meteorological data, this scheme significantly improves the calculation accuracy of the pre-launch quantity by introducing real-time wind speed and direction correction. When the ballistic calculation unit determines that the current impact point coincides with the target coordinates, the launching actuator immediately triggers the launching device 4 to perform the release action and launch the ballistic missile.

[0054] After the drop is completed, the damage assessment module on the UAV body 1 begins to function. The UAV continues to fly or hover according to a preset trajectory, performing secondary imaging of the target area via an electro-optical pod. The AI ​​model compares changes in the target images before and after the impact, such as changes in smoke features or target shape, to calculate the degree of damage. Subsequently, a strike effect assessment report is generated and transmitted back to the ground station, allowing user A to monitor the strike results in real time. Simultaneously, the damage assessment module triggers the multimodal anti-jamming navigation module to plan the return path, achieving closed-loop management and automated return of the mission, thus improving mission efficiency.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A precision-guided unmanned aerial vehicle (UAV), characterized in that, include: The drone body (1) is equipped with an AI control module (2), a GPS module (3) on one side of its upper part, and a throwing device (4) on its lower part. Ground control station, used to mark targets on digital maps and calculate and generate a baseline mission path that includes bomb drop point coordinates and entry route; A multimodal anti-interference navigation unit is installed on the UAV body (1) and includes a GPS receiver, an inertial measurement unit and an optoelectronic pod connected to the GPS module (3). The optoelectronic pod is used to collect ground images in real time. An airborne AI control unit is set on the UAV body (1) and serves as the core of the AI ​​control module (2). It is connected to the multimodal anti-interference navigation module and the flight control interface respectively. The airborne AI control unit is embedded with a combined navigation and positioning module, which is used to match the real-time ground images collected by the photoelectric pod with the pre-stored ground station image features when the GPS signal is lost, and calculate the absolute position of the UAV to maintain stable flight. The target recognition and tracking unit is connected to the optoelectronic pod and the airborne AI control unit. It is used to identify and lock the predetermined target through a deep learning model after the UAV enters the target area, and to calculate the three-dimensional spatial coordinates of the target relative to the UAV. The visual control loop, integrated into the airborne AI control unit and connected to the target recognition and tracking unit, is used to generate a locally fine-tuned path based on the target's relative coordinates, guiding the drone to the dynamically adjusted bombing point. The ballistic calculation unit, connected to the vision control loop, is used to calculate the lead-out amount based on real-time flight parameters and meteorological data. The throwing mechanism is connected to the ballistic calculation unit and the throwing device (4) respectively, and is used to trigger the throwing device (4) to perform a release action when the current impact point coincides with the target coordinates, so as to throw the ball.

2. The precision-guided unmanned aerial vehicle according to claim 1, characterized in that, The integrated navigation and positioning module includes a judgment subunit. When the GPS signal quality is lower than a preset threshold, it automatically switches to the visual / infrared feature matching navigation mode. It uses the real-time ground images acquired by the down-looking optoelectronic pod to match the feature points of the pre-stored ground station images on the airborne platform in real time, and combines the data from the inertial measurement unit to maintain stable flight attitude.

3. The precision-guided unmanned aerial vehicle according to claim 1, characterized in that, The target recognition and tracking unit includes a miss distance calculation subunit, which is used to calculate the miss distance of the target in the photoelectric pod image coordinate system and obtain the three-dimensional spatial coordinates of the target relative to the UAV using laser ranging or image size estimation.

4. A precision-guided unmanned aerial vehicle according to claim 1, characterized in that, The ballistic calculation unit is connected to the target identification and tracking unit and is used to calculate the lead-out amount in real time by integrating the current flight speed, flight altitude, flight attitude angle, real-time wind speed and direction, and the accurate coordinates of the target.

5. A precision-guided unmanned aerial vehicle according to claim 1, characterized in that, The UAV body (1) is also equipped with a damage assessment module, which is connected to the optoelectronic pod and the airborne AI control unit. It is used to perform secondary imaging of the target area after the drop, compare the changes in the target image before and after the strike through the AI ​​model, generate a strike effect assessment report, and trigger the multimodal anti-interference navigation module to plan the return path.

6. A precision-guided unmanned aerial vehicle according to claims 1-5, characterized in that, The precision guidance includes the following steps: Step S1: Mark the target on the digital map of the ground control station and calculate and generate a baseline mission path that includes the coordinates of the bomb drop point and the entry route. Step S2: After the drone takes off, the onboard AI control unit performs integrated navigation through the integrated navigation and positioning module. When the GPS signal is lost, the drone's absolute position is calculated by matching the real-time ground images collected by the photoelectric pod with the pre-stored ground station image features to maintain stable flight. Step S3: After the UAV enters the target area, the predetermined target is identified and locked through a deep learning model, and the three-dimensional spatial coordinates of the target relative to the UAV are calculated. Step S4: The visual control loop generates a locally fine-tuned path based on the target's relative coordinates, guiding the UAV to the dynamically adjusted bomb drop point; Step S5: The ballistic calculation unit calculates the lead-out amount based on real-time flight parameters and meteorological data. When the calculated future impact point coincides with the current target coordinates, the autonomous drop-out actuator is triggered to perform the release action. Step S6: After the launch, the target area is imaged a second time. The changes in the target image before and after the strike are compared using an AI model to generate a strike effect assessment report and trigger the multimodal anti-interference navigation module to plan the return path.

7. A precision-guided unmanned aerial vehicle according to claim 6, characterized in that, The specific steps for the integrated navigation and positioning module to perform integrated navigation in step S2 include: When the GPS signal is good, use satellite data to calibrate the cumulative error of the visual inertial odometry. When the GPS signal is interfered with or its quality is below a preset threshold, it automatically switches to the visual / infrared feature matching navigation mode. It uses the real-time ground images acquired by the down-looking electro-optical pod to match the feature points of the ground station images stored on the airborne in real time, calculates its own position, and combines the data from the inertial measurement unit to maintain stable flight attitude.

8. A precision-guided unmanned aerial vehicle according to claim 6, characterized in that, The specific steps for calculating the target's three-dimensional spatial coordinates relative to the UAV in step S3 include: Real-time analysis of the video stream of the photoelectric / infrared payload is performed using a deep learning model to identify the preset targets. Calculate the target's miss distance in the image coordinate system; The three-dimensional spatial coordinates of the target relative to the UAV are obtained by using laser ranging or image size estimation.

9. A precision-guided unmanned aerial vehicle according to claim 6, characterized in that, The specific steps for the ballistic calculation unit to calculate the pre-launch amount in step S5 include: Based on the current flight speed, flight altitude, flight attitude angle, real-time wind speed and direction data, and the accurate target coordinates measured in step S3; Real-time calculation of the pre-release quantity to determine the optimal release time.

10. A precision-guided unmanned aerial vehicle according to claim 6, characterized in that, The specific steps for generating the strike effect evaluation report in step S6 include: The drone continues to fly or hover according to a preset trajectory; Secondary imaging of the target area is performed using an optoelectronic pod; AI models compare changes in target images before and after the strike, including changes in smoke features or target shape. Generate a strike effectiveness assessment report and transmit it back to the ground station.

Images