A method and system for autonomous identification and locking of aerial targets by an air defense unmanned aerial vehicle

Abstract

The application discloses a method and system for autonomous identification and locking of an aerial target by an air interception unmanned aerial vehicle, wherein the method comprises optimizing a YOLOv8 photoelectric target identification algorithm based on an aerial unmanned aerial vehicle data set, and making targeted debugging on target confidence threshold values in different scenes; a distance-pixel mapping constraint is established to eliminate interference targets; a light field scene partition differentiation confidence correction strategy is designed to constrain the target to a field center area; an L1 guidance algorithm with time delay position compensation is used to complete external information guidance and approach; and the timing of issuing an identification-to-locking instruction is controlled, and the target locking instruction is only allowed to be issued when the corrected effective confidence meets a preset threshold value and the interception vehicle reaches a photoelectric theoretical effective identification distance. The application improves the reliability of low-cost photoelectric identification in complex environments, reduces the onboard computing power and mechanical energy loss, improves the approach tracking accuracy and prolongs the photoelectric stable tracking time, and reduces the risk of invalid impact and aircraft loss.

Description

Technical Field

[0001] This invention belongs to the field of counter-drone technology, and more specifically, it relates to a method and system for autonomously identifying and locking onto aerial targets by anti-aircraft drones. Background Technology

[0002] Countering drones with drones is a technological means of achieving aerial interception. Its advantages of low cost, long range and high interception efficiency are gradually becoming more prominent. The intercepting drone relies on the long-distance target position information provided by external detection equipment to complete the coarse guidance to approach. After approaching, it switches to the airborne optoelectronic imaging equipment to autonomously identify and lock onto the aerial target, and then approaches the target along the guidance trajectory to complete the interception operation.

[0003] In existing technologies, interceptor drones receive target position, velocity, and distance information from external radar or wide-area detection equipment. This information is used as input to the L1 guidance algorithm to calculate lateral acceleration commands. The interceptor adjusts its flight attitude and heading according to these commands to approach the target. During the approach, an onboard visible light or infrared electro-optical pod continuously acquires images of the target area and sends them to the YOLOv8 target detection algorithm for inference. YOLOv8 outputs the category, bounding box pixel coordinates, and original confidence score for all candidate targets in the image. The original confidence score is compared with a preset single threshold to filter out targets with confidence scores exceeding a certain threshold. The threshold candidate target is identified as an aerial target and a lock command is immediately issued to the electro-optical servo mechanism. The servo drives the electro-optical pod to continuously aim at, track and lock onto the target. The interceptor continues to adjust its flight attitude and speed to approach the target along the guidance trajectory until the interception is completed. During the long-range phase, YOLOv8 continues inference while the servo lock command is issued simultaneously, and the electro-optical servo mechanism begins to follow up and track, regardless of whether the distance between the aircraft and the target has entered the effective electro-optical recognition range. If a false lock occurs, it can only be detected by the operator afterward and the interception process can be manually restarted to correct it. There is no channel for real-time manual intervention.

[0004] However, the aforementioned existing technologies have the following drawbacks: the basic YOLOv8 model is not optimized for small aerial targets, lacks multi-dimensional effectiveness identification and filtering of candidate targets, and relies solely on a single confidence threshold for screening. High-contrast interference sources such as mountains and clouds result in a high false detection rate. In complex weather or terrain environments, it is extremely easy to mistakenly lock onto ground features or cloud clutter. Once a false lock occurs, the electro-optical servo continuously performs erroneous follow-up, significantly compressing the intersecting window for aerial interception, missing interception opportunities, and significantly reducing the interception success rate. Furthermore, the timing of the identification-to-lock command issuance is poorly managed, issuing lock commands to the servo at long distances regardless of whether the aircraft-target distance is within the effective electro-optical recognition range. YOLOv8 continuously outputs a large number of invalid image inference results, and the servo frequently performs erroneous follow-up and idles, increasing airborne computing power and mechanical energy consumption. External detection information has acquisition and transmission delays, and no position compensation is performed. The L1 guidance input target coordinates lag behind the target's actual position, causing the interceptor's following trajectory to deviate, reducing close-range tracking accuracy. Finally, the lack of a human-machine interaction fault tolerance mechanism means that after a false lock, it is impossible to quickly intervene manually to terminate the interception process, posing a risk of invalid collisions and aircraft damage.

[0005] Therefore, those skilled in the art are dedicated to providing a method and system for autonomously identifying and locking onto aerial targets by unmanned aerial vehicles (UAVs) that can effectively solve the above-mentioned technical problems. Summary of the Invention

[0006] According to a first aspect of the embodiments of this specification, a method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone is provided, comprising the following steps: S110. Optimize the YOLOv8 photoelectric target recognition algorithm based on an aerial drone dataset, and perform targeted adjustments to the target confidence threshold for different scenarios; construct an aerial small target dataset to complete model fine-tuning, and output the target category, bounding box pixel coordinates, and original confidence score. ; S120. Establish distance-pixel mapping constraints, perform pixel size filtering on the candidate targets output by YOLOv8, and eliminate interfering targets; establish the mapping relationship between target distance and imaging pixel interval, and filter out interference clusters whose pixel size deviates from the reasonable interval. S130. Design a field-zone differentiation confidence correction strategy for optical television, constraining the target to the center region of the field of view, and combining the zoning correction coefficient with the original confidence level. A second correction is performed to obtain the effective confidence level after correction, suppressing false triggering of lock by noise at the edge of the screen; S140. Based on the target position information output by the long-distance outer wall detection module, the L1 guidance algorithm with time delay position compensation is used to complete the approach guided by external information; a position compensation model is established to correct the target coordinates input by L1 guidance for the time delay error introduced by detection and transmission. S150. Control the timing of issuing the identification-to-lock command. The target lock command is only allowed to be issued when the corrected effective confidence level meets the preset threshold and the interceptor reaches the photoelectric theoretical effective identification distance.

[0007] In one possible implementation, in S110, the aerial small target dataset includes clear airspace, mountainous terrain, and cloudy scenarios; YOLOv8 raw output confidence. ∈[0,1] Set basic thresholds for different scenarios :

[0008] Only meet The candidate bounding boxes are then processed in the subsequent pixel filtering process. In the formula: The confidence level of the raw YOLOv8 output; Basic thresholds for different scenarios; This serves as a basic threshold for scenarios with clear skies and no stray clouds. This serves as a basic threshold for cloudy, low-clutter scenarios. This sets the basic threshold for complex mountain / thick cloud scenarios.

[0009] In one possible implementation, in S120, a mapping relationship between the target distance and the imaging pixel interval is established through multi-distance flight calibration. The interfering targets include mountains and cloud formations. Let the relative distance between the external detector output device and the target be R. The flight calibration obtains the target pixel width and height interval corresponding to distance R. , ]、[ , YOLOv8 detection box width (pixel width) ,high The filtering criteria are:

[0010] Candidate boxes that do not meet the above formula are directly removed; In the formula: The external detector outputs the relative distance between the device and the eye; Distance Lower limit of target pixel width; Distance Upper limit of target pixel width; Distance Lower limit of target pixel height; Distance Upper limit of target pixel height; The width of the YOLOv8 detection box in pixels; This represents the pixel height of the YOLOv8 detection bounding box.

[0011] In one possible implementation, in step S130, the target is constrained to the area slightly above the center of the field of view, the photoelectric image is divided into three layers, and the target pixel center is defined. Distance from the center of the image Normalized distance:

[0012] The normalized coefficients for the half-diagonal of the image; the partition confidence correction coefficients. :

[0013] Corrected effective confidence level:

[0014] only The target is retained, and targets in the central airspace of the field of view are forcibly prioritized; In the formula: The coordinates of the center of the target pixel; The coordinates of the center of the image; The normalization coefficient for the half-diagonal of the image; Normalized distance to the center of the target pixel; The threshold for the boundary between the central core area and the intermediate ring; The threshold for defining the boundary between the central annular region and the edge region; Correction coefficient for the central core area; This is the correction factor for the middle annular region; For edge region correction coefficients; This represents the corrected effective confidence level.

[0015] In one possible implementation, in S140, the flight speed and altitude are adaptively adjusted within the photoelectric identifiable range to extend the stable tracking time; let the total delay of the detection information be... Target's current true velocity vector Report the target location before the time delay. Real-time target location after compensation:

[0016] Will The target coordinates are used as input for L1 guidance to eliminate trajectory deviations caused by transmission and detection lag; the interceptor's own position. Machine-eye vector Interceptor speed L1 guidance desired lateral acceleration command:

[0017] via the acceleration command Continuously adjust the interceptor's attitude to keep the target stably positioned in the upper-center area of ​​the electro-optical display; automatically reduce airspeed once the target enters the effective electro-optical recognition range. This reduces the height difference between the camera and the target, and extends the duration of stable tracking. in: The total latency of the detected information; The target's current true velocity vector; , , The target is in , , The velocity component in the direction; The target location must be reported before the time delay; , , The target before the time delay is at , , The coordinate components of the direction; The real-time target position after compensation; This refers to the location of the interceptor itself. For machine-eye vector; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. The desired lateral acceleration command for L1 guidance.

[0018] In one possible implementation, in S150, the long-distance coarse guidance stage only performs the YOLOv8 image detection and recognition, and disables servo lock output; the pre-calibrated photoelectric theory effective recognition critical distance is determined. Real-time machine - target line distance Lock command enable flag :

[0019] In the formula: For effective identification of critical distances using photoelectric theory; The real-time straight-line distance between the machine and the target; This is a flag that enables the lock command.

[0020] In one possible implementation, the method further includes: S160. Retain the human-computer interaction intervention channel, configure the right mouse button to de-lock and the left mouse button to manually select the target, forming the dual-redundant control logic of YOLOv8 autonomous recognition and manual correction; the priority of manual commands is higher than the output of YOLOv8 autonomous recognition, and when the left mouse button is used to select the target, the algorithm output coordinates are forced to be overwritten and the corrected effective confidence level is also required. Setting it to 1 immediately issues a lock command; right-clicking to cancel the lock enables the lock command. Forced to 0, the servo stops automatically, and the aircraft switches back to the standby route.

[0021] According to a second aspect of the embodiments of this specification, a system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone is provided, the system comprising: The long-range external detection module is used to output the target's original position, velocity, and distance information, providing a long-range coarse guidance data source for L1 guidance; The photoelectric imaging acquisition module is used to carry the photoelectric pod and output a real-time image stream; The YOLOv8 image recognition and processing module is used to build a custom-tuned YOLOv8 inference model for aerial targets. It performs inference on the real-time image stream output by the photoelectric imaging acquisition module, integrates pixel size filtering and field-of-view partitioning confidence correction algorithms, and outputs candidate target coordinates and corrected confidence scores. ; The time delay compensation L1 guidance flight control module is used to build a time delay position compensation model and an L1 guidance acceleration calculation model, receive the original target coordinates output by the long-distance external detection module, and output flight attitude, speed and altitude control commands after compensating for hysteresis error. The instruction timing control module is used to pre-store the critical distance for effective photoelectric recognition, compare the machine-eye distance in real time, and enable or disable the lock instruction according to the lock instruction enable logic. The servo follow-up control module is used to receive the target coordinates after the instruction timing control module has released them, and drive the photoelectric pod to continuously align with and track the target.

[0022] In one possible implementation, the optoelectronic pod mounted on the optoelectronic imaging acquisition module is a visible light / infrared optoelectronic pod; the YOLOv8 image recognition and calculation module has a built-in distance-pixel mapping lookup table unit and a partition confidence coefficient storage unit, which can dynamically switch filtering parameters according to the environment and the machine-to-eye distance; the system also includes a human-computer interaction control module for mouse interaction on the ground host computer and high-priority issuance of manual intervention commands.

[0023] In one possible implementation, the time-delay compensated L1 guidance flight control module acquires the aircraft's velocity and target motion parameters in real time, and dynamically solves for the compensated position and L1 lateral acceleration commands; the time-delay position compensation model is as follows:

[0024] L1 guidance acceleration calculation model:

[0025] In the formula, The total latency of the detected information; The target's current true velocity vector; The target location must be reported before the time delay; The real-time target position after compensation; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. The desired lateral acceleration command for L1 guidance.

[0026] As described above, the present invention provides a method and system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone, which has the following beneficial effects: 1. By optimizing the YOLOv8 photoelectric target recognition algorithm based on aerial UAV dataset and adjusting the confidence threshold according to different scenarios, distance-pixel mapping constraints are established to filter out interfering targets by pixel size. A differentiated confidence correction strategy for photoelectric field partitioning is designed to suppress false locking caused by clutter at the edge of the screen. This significantly reduces the probability of false locking of ground mountains or cloud gap clutter in complex environments such as mountains and cloudy areas, and improves the reliability of low-cost photoelectric recognition in complex environments. 2. By controlling the timing of the identification-to-lock command issuance, the target lock command is only allowed to be issued after the interceptor reaches the theoretically effective identification distance of photoelectric technology. During the long-range coarse guidance phase, only image detection and identification are performed while the servo lock output is blocked, reducing invalid image inference and frequent servo idle, and reducing the loss of airborne computing power and mechanical energy. 3. By adopting the L1 guidance algorithm with time delay position compensation, a position compensation model is established to correct the target coordinates of the guidance input for the time delay error introduced by detection and transmission, eliminating the trajectory deviation caused by transmission and detection lag, improving the close-range tracking accuracy and extending the photoelectric stable tracking time. 4. By retaining the human-machine interaction intervention channel, configuring the right mouse button to cancel locking and the left mouse button to manually select the target, a dual redundant control logic of autonomous recognition and manual correction is formed. In case of false locking, manual intervention can be quickly performed to terminate the interception process, reducing the risk of invalid collisions and aircraft damage. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating a method for autonomously identifying and locking onto aerial targets by an unmanned aerial vehicle (UAV) according to one embodiment of this specification.

[0028] Figure 2 This is a schematic block diagram of a system for autonomously identifying and locking onto aerial targets by an unmanned aerial vehicle (UAV) according to one embodiment of this specification.

[0029] Figure 3 This is a structural block diagram of a computing device provided in one embodiment of this specification.

[0030] Figure 4 This is a schematic diagram showing the effect before optimization; Figure 5 This is a schematic diagram of the optimized effect. Detailed Implementation

[0031] The present invention will be further described below with reference to the accompanying drawings and embodiments: In the description of this invention, it should be noted that the terms "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0032] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "setting," and "connection" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0033] This specification provides a method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone. It also relates to a system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone, a computing device, and a computer-readable storage medium, which will be described in detail in the following embodiments.

[0034] See Figure 1 , Figure 1 A flowchart is shown of a method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to an embodiment of this specification, specifically including the following steps.

[0035] S110 optimizes the YOLOv8 photoelectric target recognition algorithm based on an aerial drone dataset, and performs targeted adjustments to the target confidence threshold for different scenarios; it constructs an aerial small target dataset to complete model fine-tuning, and outputs the target category, bounding box pixel coordinates, and original confidence score. As an example, in this embodiment or other embodiments, the aerial small target dataset may include clear airspace scenarios, mountainous scenarios, or cloudy scenarios; targeted adjustments to the target confidence threshold can be made by setting different basic threshold values ​​according to the clutter interference level of different scenarios, and no specific numerical limit is specified here.

[0036] S120, establish distance-pixel mapping constraints, filter candidate targets output by YOLOv8 by pixel size, and eliminate interfering targets; establish a mapping relationship between target distance and imaging pixel interval, and filter out interfering clusters whose pixel size deviates from the reasonable interval. As an example, in this embodiment or other embodiments, interfering targets may include mountains, clouds, or other non-target ground features; the mapping relationship between target distance and imaging pixel interval can be established through multi-distance flight calibration, which will not be described in detail here.

[0037] S130, design a field-zone differentiation confidence correction strategy for optical television, constraining the target to the central region of the field of view, and combining the zoning correction coefficient with the original confidence level. A second correction is performed to obtain the corrected effective confidence level, suppressing false triggering of locking by clutter at the edge of the image. As an example, in this embodiment or other embodiments, the target is constrained to the area slightly above the center of the field of view, that is, the center of the target pixel is located slightly above the center of the image; the partition correction coefficient can be set according to the normalized distance of the target pixel center from the center of the image to divide into multiple regions, which is not limited here.

[0038] S140, based on the target position information output by the long-range outer wall detection module, an L1 guidance algorithm with time delay compensation is used to guide the approach using external information. A position compensation model is established to correct the target coordinates input by L1 guidance, addressing the time delay error introduced by detection and transmission. As an example, in this embodiment or other embodiments, the interceptor UAV possesses two core operational characteristics: long-range coarse guidance from external detection information and fully autonomous beyond-visual-range interception. The position compensation model can compensate for the reported target position before the time delay based on the total time delay of the detection information and the target's velocity vector. The L1 guidance algorithm can generate desired lateral acceleration commands to continuously correct the interceptor's attitude, keeping the target stably maintained in the upper-center area of ​​the photoelectric image. Within the photoelectric recognizable range, the flight speed and altitude can be adaptively adjusted to extend the stable tracking time, which is not limited here.

[0039] S150, the timing of issuing the control identification to lock command is controlled. The target lock command is only allowed to be issued when the corrected effective confidence level meets a preset threshold and the interceptor reaches the theoretically effective identification distance. As an example, in this embodiment or other embodiments, the long-range coarse guidance phase can only perform YOLOv8 image detection and recognition while shielding the servo lock output; the theoretically effective identification distance can be obtained through pre-calibration; the timing of issuing the lock command can be determined based on the comparison between the real-time machine-target straight-line distance and the theoretically effective identification critical distance, which will not be elaborated here.

[0040] According to the above description, in an exemplary embodiment, the process of optimizing and debugging the YOLOv8 photoelectric target recognition algorithm in step S110 includes: the aerial small target dataset includes clear airspace, mountainous, and cloudy scenes; the YOLOv8 raw output confidence score... Set basic thresholds for different scenarios :

[0041] Only meet The candidate bounding boxes are then processed in the subsequent pixel filtering process; where: The confidence level of the raw YOLOv8 output; Basic thresholds for different scenarios; This serves as a basic threshold for scenarios with clear skies and no stray clouds. This serves as a basic threshold for cloudy, low-clutter scenarios. This sets the basic threshold for complex scenarios such as mountainous terrain and thick clouds. The onboard computing unit loads a YOLOv8s model finely tuned for the UAV, performs real-time inference on photoelectric images, and outputs all candidate bounding boxes. Retrieve the corresponding information based on the current environment. Remove Low-confidence noise. Therefore, by setting differentiated basic thresholds for different scenarios, the filtering conditions can be adaptively adjusted according to the degree of clutter interference in different scenarios. This effectively reduces the probability of low-confidence noise mistakenly entering subsequent processes under complex backgrounds, while ensuring that targets are not mistakenly filtered out by excessively high thresholds in clear-air scenarios.

[0042] According to the above description, in an exemplary embodiment, the process of filtering candidate targets by pixel size in step S120 includes: establishing a mapping relationship between target distance and imaging pixel interval through multi-distance flight calibration; interfering targets include mountains and clouds; assuming the relative distance between the external detector output device and the target is... The distance was obtained from the actual flight calibration. Corresponding target pixel width and height range , YOLOv8 detection box width (pixel width) ,high The filtering criteria are: Candidate boxes that do not satisfy the above formula are directly removed; where: The external detector outputs the relative distance between the device and the eye; Distance Lower limit of target pixel width; Distance Upper limit of target pixel width; Distance Lower limit of target pixel height; Distance Upper limit of target pixel height; The width of the YOLOv8 detection box in pixels; This represents the pixel height of the YOLOv8 detection bounding box.

[0043] Read the distance of the external probe - eye Look up the table to obtain the target pixel range at that distance. , The YOLOv8 output frame width and height are iterated through to remove interfering targets such as mountains and clouds with mismatched sizes. This demonstrates that by establishing a distance-pixel mapping relationship through multi-distance flight calibration and combining it with pixel width and height range filtering, interfering targets such as mountains and clouds with pixel sizes outside the reasonable range can be effectively removed, thus significantly improving the accuracy of candidate target selection.

[0044] According to the above description, in an exemplary embodiment, the process of field-of-view partitioning differential confidence correction in step S130 includes: constraining the target to the region slightly above the center of the field of view, dividing the photoelectric image into three layers of regions, and defining the target pixel center. Distance from the center of the image Normalized distance:

[0045] The normalized coefficients for the half-diagonal of the image; the partition confidence correction coefficients. :

[0046] Corrected effective confidence level: ;only The target is preserved, and targets in the central airspace of the field of view are forcibly prioritized; where: The coordinates of the center of the target pixel; The coordinates of the center of the image; The normalization coefficient for the half-diagonal of the image; Normalized distance to the center of the target pixel; The threshold for the boundary between the central core area and the intermediate ring; The threshold for defining the boundary between the central annular region and the edge region; Correction coefficient for the central core area; This is the correction factor for the middle annular region; For edge region correction factors; This is the corrected effective confidence level. Calculate the normalized distance between the center of the target bounding box. Matching partition correction coefficient ,calculate Only targets with high effective confidence in the central area are retained; flight control continuously adjusts attitude to keep the target stable in the upper center of the image.

[0047] Please see Figure 4 As shown, before optimization, clutter at the image edges easily triggers the lock; please refer to [link / reference]. Figure 5 As shown, after field-of-view partitioning confidence correction, only the target in the center of the field of view is effectively locked, while edge clutter is effectively suppressed. This demonstrates that by dividing the photoelectric image into three regions and applying a partitioning correction coefficient to the original confidence level, false locking triggered by edge clutter can be suppressed, and the target in the center of the field of view can be prioritized for locking, thereby improving the reliability of low-cost photoelectric recognition.

[0048] According to the above description, in an exemplary embodiment, the process of the L1 guidance algorithm with time delay position compensation in step S140 includes: adaptively adjusting the flight speed and altitude within the photoelectric identifiable range to extend the stable tracking time; assuming the total delay of the detection information is... The target's current true velocity vector Report the target location before the time delay. Real-time target location after compensation: ;Will The target coordinates are used as input for L1 guidance to eliminate trajectory deviations caused by transmission and detection lag; the interceptor's own position. Machine-eye vector Interceptor speed L1 guidance desired lateral acceleration command: ; via the acceleration command Continuously adjust the interceptor's attitude to keep the target stably positioned in the upper-center area of ​​the electro-optical display; automatically reduce airspeed once the target enters the effective electro-optical recognition range. To reduce the height difference between the camera and the target, and extend the stable tracking time; where: The total latency of the detected information; The target's current true velocity vector; , , The target is in , , The velocity component in the direction; The target location must be reported before the time delay; , , The target before the time delay is at , , The coordinate components of the direction; The real-time target position after compensation; This refers to the location of the interceptor itself. For machine-eye vector; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. The desired lateral acceleration command is for L1 guidance. External detection outputs the target position with a time delay. ,speed Substituting the values ​​into the time delay compensation formula yields the real-time target coordinates. ; Solve for the L1 guidance lateral acceleration command by combining the machine's position and velocity. Control the drone to continuously approach the target; when The system automatically reduces flight speed and matches target altitude, extending the electro-optical tracking time. This demonstrates that by correcting external detection lag errors through a time-delay position compensation model, and combining this with the L1 guidance algorithm to generate a smooth approach trajectory, the target can be precisely controlled to remain stably centered in the optical field, effectively extending the electro-optical tracking window. Furthermore, by adaptively adjusting flight speed and altitude, the stable tracking time is further extended.

[0049] According to the above description, in an exemplary embodiment, the process of controlling the timing of issuing the identification-to-lock command in step S150 includes: performing only YOLOv8 image detection and recognition during the long-distance coarse guidance stage, and shielding the servo lock output; pre-calibrating the photoelectric theory effective identification critical distance. Real-time machine-to-eye straight-line distance Lock command enable flag : ; In the formula: For effective identification of critical distances using photoelectric theory; The real-time straight-line distance between the machine and the target; Enable flag for locking command. Real-time computer - eye distance ;like , It only performs YOLOv8 identification and does not send lock commands to the server; if , The target coordinates are then released to the servo tracking module. During the long-range phase, only image inference is performed, and no tracking commands are output, reducing onboard computing power and servo power consumption. Therefore, by controlling the timing of locking commands through distance thresholds, and performing only YOLOv8 image detection without triggering servo locking at long distances, invalid image inference and frequent servo idling can be reduced, thus lowering onboard computing power and power consumption.

[0050] According to the above description, in an exemplary embodiment, the method may further include step S160: retaining the human-computer interaction intervention channel, configuring the right mouse button to unlock and the left mouse button to manually select the target, forming a dual-redundant control logic of YOLOv8 autonomous recognition and manual correction; the priority of manual commands is higher than the output of YOLOv8 autonomous recognition, and when the left mouse button is used to select the target, the algorithm output coordinates are forcibly covered and the corrected effective confidence level is also covered. Set directly to 1 to immediately issue a lock command; right-click to cancel the lock command enable flag. Forced reset to 0, servo-driven shutdown, and aircraft return to standby flight path. Ground operators can left-click on any target within the image to force lock; if a target is mistakenly locked onto a cloud or ground object, right-click to immediately reset. The servo stops following the movement, and the interceptor exits the interception route to stand by. This demonstrates that by setting a high-priority channel for manual intervention, the lock can be canceled with a single click in case of accidental locking or during testing scenarios, avoiding unnecessary collisions that could damage the intercepted drone. This creates a dual-redundant control logic of autonomous identification and manual correction, improving system safety and reliability.

[0051] In summary, this invention provides a method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone. First, it optimizes the YOLOv8 electro-optical target recognition algorithm based on an aerial drone dataset and adjusts the confidence threshold for different scenarios, constructing an aerial small target dataset to fine-tune the model. Then, it establishes distance-pixel mapping constraints to filter candidate targets by pixel size, eliminating interfering targets such as mountains and cloud cover. Next, it designs a differentiated confidence correction strategy for differentiating the optical field of view, performing a secondary correction on the original confidence to suppress false locking triggered by edge clutter. Simultaneously, it employs an L1 guidance algorithm with time-delay position compensation to guide approach using external information, correcting the target coordinates input to L1 guidance. Finally, it controls the timing of issuing the identification-to-lock command, allowing the target lock command to be issued only when the corrected effective confidence meets a preset threshold and the interceptor reaches the theoretically effective identification distance for electro-optical targets. Therefore, this method, through pixel size filtering and secondary field-of-view confidence correction with a two-layer anti-interference logic, significantly reduces the probability of false locking onto cloud gaps and ground targets in mountainous or cloudy environments, improving the reliability of low-cost electro-optical recognition. Furthermore, this method corrects external detection lag errors by introducing a time-delay position compensation model, and generates a smooth approach trajectory in conjunction with the L1 guidance algorithm, accurately controlling the target to remain stably at the center of the optical television field, effectively extending the photoelectric tracking window. Simultaneously, this method controls the timing of locking commands through distance threshold management; at long distances, only YOLOv8 image detection is performed without triggering servo locking, reducing invalid image inference and frequent servo idle cycles, thus lowering airborne computing power and energy consumption. Moreover, when setting up a high-priority channel for manual intervention, this method allows for one-click de-locking in case of accidental locking or test scenarios, avoiding unnecessary collisions and wasting drone interception resources. In addition, the entire solution is compatible with low-cost hardware, featuring lightweight YOLOv8 inference and low L1 guidance computation, without relying on high-end photoelectric or airborne computing power, and even reducing the original computational load. It can fully realize a completely autonomous closed-loop interception process, including long-distance external information coarse guidance, time-delay compensated L1 approach, YOLOv8 air target identification and filtering, distance threshold determination and locking, photoelectric servo tracking, autonomous collision interception, or manual intervention exit.

[0052] like Figure 2 As shown, in another exemplary embodiment of the present invention, a system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone is also provided, comprising: The long-range external detection module 210 is used to output the target's original position, velocity, and distance information, providing a long-range coarse guidance data source for L1 guidance. As an example, in this embodiment or other embodiments, the long-range external detection module can periodically output the target's original position. ,speed Machine-to-eye distance The data is synchronously sent to the delay compensation L1 guidance flight control module and the command timing management module, which will not be described in detail here.

[0053] The photoelectric imaging acquisition module 220 is used to carry the photoelectric pod and output a real-time image stream. As an example, in this embodiment or other embodiments, the photoelectric pod is a visible light / infrared photoelectric pod; the photoelectric imaging acquisition module continuously outputs the image stream to the YOLOv8 image recognition and processing module, which is not limited here.

[0054] The YOLOv8 image recognition and processing module 230 is used to build a custom-tuned YOLOv8 inference model for aerial targets, perform inference on the real-time image stream output by the photoelectric imaging acquisition module, integrate pixel size filtering and field-of-view partition confidence correction algorithms, and output candidate target coordinates and corrected confidence scores. As an example, in this embodiment or other embodiments, the YOLOv8 image recognition processing module completes YOLOv8 inference, basic confidence filtering, pixel size filtering, and partition confidence correction, outputting the valid aerial target coordinates and... The YOLOv8 image recognition and processing module can have a built-in distance-pixel mapping lookup table unit, or a built-in partition confidence coefficient storage unit. It can also dynamically switch filtering parameters according to the environment and the machine-to-eye distance, which is not limited here.

[0055] The time-delay compensated L1 guidance flight control module 240 is used to build a time-delay position compensation model and an L1 guidance acceleration calculation model. It receives the original target coordinates output by the long-range external detection module, compensates for hysteresis errors, and outputs flight attitude, speed, and altitude control commands. As an example, in this embodiment or other embodiments, the time-delay compensated L1 guidance flight control module first uses the time-delay position compensation formula... After compensating for transmission delay, the L1 guidance acceleration command is calculated again to drive the interceptor to adjust its attitude, speed, and altitude to approach the target, causing the target to fall into the upper-center position of the photoelectric field; after entering the effective photoelectric identification range, the airspeed is automatically reduced. This reduces the height difference between the machine and the target, and extends the photoelectric tracking time; details will not be elaborated here.

[0056] The instruction timing control module 250 is used to pre-store the effective photoelectric recognition critical distance, compare the machine-to-eye distance in real time, and enable or disable the locking instruction according to the locking instruction enable logic. As an example, in this embodiment or other embodiments, the instruction timing control module compares in real time. and When the distance is not within the target range, the lock command is blocked; once the distance enters the effective recognition range, the target coordinates are sent to the servo follow-up control module, which will not be elaborated here.

[0057] The servo servo control module 260 is used to receive the target coordinates after the timing control module has given permission, and drive the photoelectric pod to continuously align with and track the target. As an example, in this embodiment or other embodiments, the servo servo control module receives the target coordinates after the timing control module has given permission, and drives the photoelectric pod to continuously align with and track the target; details are not elaborated here.

[0058] According to the above description, in an exemplary embodiment, the photoelectric imaging acquisition module 220 is equipped with a visible light / infrared photoelectric pod; the YOLOv8 image recognition and processing module 230 has a built-in distance-pixel mapping lookup table unit and a partitioned confidence coefficient storage unit, which can dynamically switch filtering parameters according to the environment and the machine-to-target distance; the system also includes a human-machine interaction control module for mouse interaction on the ground host computer, and high-priority issuance of manual intervention commands. The human-machine interaction control module transmits photoelectric images back in real time, and mouse operation commands have the highest priority, which can directly cover the autonomous recognition and guidance logic, quickly terminate or manually specify the target. Thus, by setting up a human-machine interaction control module, the high-priority issuance of mouse interaction commands from the ground host computer can quickly terminate the lock in the event of a false lock or in a test scenario, and can also manually specify the target, improving the system's security and flexibility; by building a distance-pixel mapping lookup table unit and a partitioned confidence coefficient storage unit, the filtering parameters can be dynamically switched according to the environment and distance, improving the recognition adaptability and accuracy.

[0059] According to the above description, in an exemplary embodiment, the time-delay compensation L1 guidance flight control module 240 collects the onboard speed and target motion parameters in real time, and dynamically solves for the compensation position and L1 lateral acceleration command; the time-delay position compensation model is as follows: L1 guidance acceleration calculation model: In the formula: The total latency of the detected information; The target's current true velocity vector; The target location was reported before the time delay; The real-time target position after compensation; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. This represents the desired lateral acceleration command for L1 guidance. Therefore, by using a time-delay position compensation model to correct external detection lag errors in real time, and combining this with the L1 guidance acceleration calculation model to generate a smooth approach trajectory, the target can be precisely controlled to remain stably at the center of the optical field, effectively extending the photoelectric tracking window.

[0060] In summary, this invention provides a system for autonomously identifying and locking onto aerial targets by an unmanned aerial vehicle (UAV). First, a long-range external detection module outputs the target's original position, velocity, and distance information. Then, an electro-optical imaging acquisition module equipped with an electro-optical pod outputs a real-time image stream. Next, a YOLOv8 image recognition and processing module infers the real-time image stream, integrating pixel size filtering and field-of-view partitioning confidence correction algorithms to output candidate target coordinates and corrected confidence scores. Simultaneously, a time-delay compensation L1 guidance flight control module compensates for hysteresis errors and outputs flight attitude, velocity, and altitude control commands. Then, a command timing control module enables or disables lock-on commands according to the lock-on command enable logic. Finally, a servo-driven servo control module drives the electro-optical pod to continuously track the target. Therefore, this system, through pixel size filtering and field-of-view partitioning confidence correction with a dual-layer anti-interference logic, significantly reduces the probability of mis-locking onto cloud gaps or ground targets in mountainous or cloudy conditions, improving the reliability of low-cost electro-optical identification. Furthermore, this system corrects external detection lag errors by introducing a time-delay position compensation model, and generates a smooth approach trajectory in conjunction with the L1 guidance algorithm, accurately controlling the target to remain stably centered in the optical field, effectively extending the photoelectric tracking window. Simultaneously, the system controls the timing of locking commands through distance threshold management; at long distances, only YOLOv8 image detection is performed without triggering servo locking, reducing invalid image inference and frequent servo idling, thus lowering onboard computing power and energy consumption. Moreover, when setting up a high-priority channel for manual intervention, the system can cancel locking with a single click in case of accidental locking or test scenarios, avoiding unnecessary collisions and wasting drone interception resources. In addition, the entire system solution is compatible with low-cost hardware, featuring lightweight YOLOv8 inference and low L1 guidance computation, without relying on high-end photoelectric or onboard computing power, and even reducing the original computational load. It can fully realize a completely autonomous closed-loop interception process, including long-distance external information coarse guidance, time-delay compensated L1 approach, YOLOv8 air target identification and filtering, distance threshold determination and locking, photoelectric servo tracking, autonomous collision interception, or manual intervention exit.

[0061] like Figure 3 A structural block diagram of a computing device 400 according to one embodiment of this specification is shown. The components of the computing device 400 include, but are not limited to, a memory 410 and a processor 420. The processor 420 is connected to the memory 410 via a bus 430, and a database 450 is used to store data.

[0062] The computing device 400 also includes an access device 440, which enables the computing device 400 to communicate via one or more networks 460. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 440 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0063] In one embodiment of this specification, the aforementioned components of the computing device 400 and Figure 3 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 3 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0064] Computing device 400 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). Computing device 400 can also be a mobile or stationary server.

[0065] The processor 420 executes computer-executable instructions, which, when executed by the processor, implement the steps of the aforementioned method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone. The above is an illustrative scheme of a computing device according to this embodiment. It should be noted that the technical solution of this computing device and the technical solution of the aforementioned method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone belong to the same concept. Details not described in detail in the technical solution of the computing device can be found in the description of the technical solution of the aforementioned method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone.

[0066] An embodiment of this specification also provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone.

[0067] The above is an illustrative scheme of a computer-readable storage medium according to this embodiment. It should be noted that the technical solution of this storage medium and the technical solution of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor UAV belong to the same concept. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor UAV.

[0068] An embodiment of this specification also provides a computer program, wherein when the computer program is executed in a computer, it causes the computer to perform the steps of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone.

[0069] The above is an illustrative scheme of a computer program according to this embodiment. It should be noted that the technical solution of this computer program and the technical solution of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone belong to the same concept. For details not described in detail in the technical solution of the computer program, please refer to the description of the technical solution of the above-described method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone.

[0070] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0071] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added to or subtracted according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0072] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments in this specification are not limited to the described order of actions, because according to the embodiments in this specification, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in this specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments in this specification.

[0073] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0074] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone, characterized in that, Includes the following steps: S110. Optimize the YOLOv8 photoelectric target recognition algorithm based on an aerial drone dataset, and perform targeted adjustments to the target confidence threshold for different scenarios; construct an aerial small target dataset to complete model fine-tuning, and output the target category, bounding box pixel coordinates, and original confidence score. ; S120. Establish distance-pixel mapping constraints, perform pixel size filtering on the candidate targets output by YOLOv8, and eliminate interfering targets; establish the target distance and imaging pixel interval mapping relationship, and filter out interference clusters whose pixel size deviates from the reasonable interval. S130. Design a field-zone differentiation confidence correction strategy for optical television, constraining the target to the center region of the field of view, and combining the zoning correction coefficient with the original confidence level. A second correction is performed to obtain the effective confidence level after correction, suppressing false triggering of lock by noise at the edge of the screen; S140. The L1 guidance algorithm with time delay position compensation is used to guide the approach based on external information. A position compensation model is established to address the time delay error introduced by detection and transmission, and the target coordinates of the L1 guidance input are corrected. S150. Control the timing of issuing the identification-to-lock command. The target lock command is only allowed to be issued when the corrected effective confidence level meets the preset threshold and the interceptor reaches the photoelectric theoretical effective identification distance.

2. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 1, characterized in that, In step S110, the aerial small target dataset includes clear airspace, mountainous terrain, and cloudy scenarios; YOLOv8 raw output confidence score. ∈[0,1] Set basic thresholds for different scenarios : ， Only meet The candidate bounding boxes are then processed in the subsequent pixel filtering process. In the formula: The confidence level of the raw YOLOv8 output; Basic thresholds for different scenarios; This serves as a basic threshold for scenarios with clear skies and no stray clouds. This serves as a basic threshold for cloudy, low-clutter scenarios. This sets the basic threshold for complex mountain / thick cloud scenarios.

3. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 1, characterized in that, In step S120, a mapping relationship between the target distance and the imaging pixel interval is established through multi-distance flight calibration. The interfering targets include mountains and clouds. Let R be the relative distance between the external detector output device and the target. The flight calibration obtains the target pixel width and height interval corresponding to distance R. , ]、[ , YOLOv8 detection box width (pixel width) ,high The filtering criteria are: ， Candidate boxes that do not meet the above formula are directly removed; In the formula: The external detector outputs the relative distance between the device and the eye; Distance Lower limit of target pixel width; Distance Upper limit of target pixel width; Distance Lower limit of target pixel height; Distance Upper limit of target pixel height; The width of the YOLOv8 detection box in pixels; This represents the pixel height of the YOLOv8 detection bounding box.

4. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 1, characterized in that, In step S130, the target is constrained to the area slightly above the center of the field of view, the photoelectric image is divided into three layers, and the target pixel center is defined. Distance from the center of the image Normalized distance: ， The normalized coefficients for the half-diagonal of the image; the partition confidence correction coefficients. : ， Corrected effective confidence level: ， only The target is retained, and targets in the central airspace of the field of view are forcibly prioritized; In the formula: The coordinates of the center of the target pixel; The coordinates of the center of the image; The normalization coefficient for the half-diagonal of the image; Normalized distance to the center of the target pixel; The threshold for the boundary between the central core area and the intermediate ring; The threshold for defining the boundary between the central annular region and the edge region; Correction coefficient for the central core area; This is the correction factor for the middle annular region; For edge region correction factors; This represents the corrected effective confidence level.

5. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 4, characterized in that, In step S140, the flight speed and altitude are adaptively adjusted within the photoelectric identifiable range to extend the stable tracking time; let the total delay of the detection information be... Target's current true velocity vector Report the target location before the time delay. Real-time target location after compensation: ， Will The target coordinates are used as input for L1 guidance to eliminate trajectory deviation caused by transmission and detection lag. Interceptor's local location Machine-eye vector Interceptor speed L1 guidance desired lateral acceleration command: ， via the acceleration command Continuously adjust the interceptor's attitude to keep the target stably positioned in the upper-center area of ​​the electro-optical display; automatically reduce airspeed once the target enters the effective electro-optical recognition range. This reduces the height difference between the camera and the target, and extends the duration of stable tracking. in: The total latency of the detected information; The target's current true velocity vector; , , The target is in , , The velocity component in the direction; The target location must be reported before the time delay; , , The target before the time delay is at , , The coordinate components of the direction; The real-time target position after compensation; This refers to the location of the interceptor itself. For machine-eye vector; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. The desired lateral acceleration command for L1 guidance.

6. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 5, characterized in that, In S150, during the long-distance coarse guidance phase, only the YOLOv8 image detection and recognition are performed, and the servo lock output is disabled; the critical distance for effective recognition based on photoelectric theory is pre-calibrated. Real-time machine-to-eye straight-line distance Lock command enable flag : ， In the formula: For effective identification of critical distances using photoelectric theory; The real-time straight-line distance between the machine and the target; This is a flag that enables the lock command.

7. The method for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 6, characterized in that, The method further includes: S160. Retain the human-computer interaction intervention channel, configure the right mouse button to de-lock and the left mouse button to manually select the target, forming the dual-redundant control logic of YOLOv8 autonomous recognition and manual correction; the priority of manual commands is higher than the output of YOLOv8 autonomous recognition, and when the left mouse button is used to select the target, the algorithm output coordinates are forced to be overwritten and the corrected effective confidence level is also required. Setting it to 1 immediately issues a lock command; right-clicking to cancel the lock enables the lock command. Forced to 0, the servo stops automatically, and the aircraft switches back to the standby route.

8. A system for autonomously identifying and locking onto aerial targets from unmanned aerial vehicles (UAVs), characterized in that, The system includes: The long-range external detection module is used to output the target's original position, velocity, and distance information, providing a long-range coarse guidance data source for L1 guidance; The photoelectric imaging acquisition module is used to carry the photoelectric pod and output a real-time image stream; The YOLOv8 image recognition and processing module is used to build a custom-tuned YOLOv8 inference model for aerial targets. It performs inference on the real-time image stream output by the photoelectric imaging acquisition module, integrates pixel size filtering and field-of-view partitioning confidence correction algorithms, and outputs candidate target coordinates and corrected confidence scores. ; The time delay compensation L1 guidance flight control module is used to build a time delay position compensation model and an L1 guidance acceleration calculation model, receive the original target coordinates output by the long-distance external detection module, and output flight attitude, speed and altitude control commands after compensating for hysteresis error. The instruction timing control module is used to pre-store the critical distance for effective photoelectric recognition, compare the machine-eye distance in real time, and enable or disable the lock instruction according to the lock instruction enable logic. The servo follow-up control module is used to receive the target coordinates after the instruction timing control module has released them, and drive the photoelectric pod to continuously align with and track the target.

9. The system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 8, characterized in that, The photoelectric imaging acquisition module is equipped with a visible light / infrared photoelectric pod; the YOLOv8 image recognition and calculation module has a built-in distance-pixel mapping lookup table unit and a partition confidence coefficient storage unit, which can dynamically switch filtering parameters according to the environment and the distance between the machine and the eye; the system also includes a human-machine interaction control module for mouse interaction on the ground host computer and high-priority issuance of manual intervention commands.

10. The system for autonomously identifying and locking onto aerial targets by an air-to-air interceptor drone according to claim 9, characterized in that, The time-delay compensation L1 guidance flight control module collects the aircraft's velocity and target motion parameters in real time, and dynamically solves for the compensation position and L1 lateral acceleration command; time-delay position compensation model: ， L1 guidance acceleration calculation model: ， In the formula, The total latency of the detected information; The target's current true velocity vector; The target location must be reported before the time delay; The real-time target position after compensation; For the speed of the interceptor; The forward look-ahead time constant for L1 guidance; This is the L1 guidance reference range; The velocity is a unit vector perpendicular to the line-of-sight direction; The unit vector is the line-of-sight normal. The desired lateral acceleration command for L1 guidance.

Images