Image recognition-based tethered unmanned aerial vehicle pay-off anti-interference control method and system

Abstract

The application discloses a tethered unmanned aerial vehicle pay-off anti-interference control method based on image recognition and relates to the technical field of unmanned aerial vehicle cable anti-interference control systems. The method comprises the following steps: acquiring timestamp-aligned wind speed, wind direction, cable direction, measured tension, capstan speed and cable images, projecting the wind speed along the cable direction to obtain a wind disturbance component along the cable; inputting the measured tension, capstan speed and wind disturbance component into a tension prediction model to output predicted tension at the next moment; comparing the predicted tension with a safety interval to adjust the pay-off speed of the capstan; and combining image recognition cable morphology to visually correct the pay-off speed of the capstan through loose elimination logic and vibration suppression logic, and the corrected pay-off speed of the capstan is executed and synchronized to the unmanned aerial vehicle to adaptively cooperate with the pay-off action of the capstan. The application effectively solves the problem that the unmanned aerial vehicle shakes and drifts in the air due to abnormal tension and unmatched pay-off speed through tension prediction and multi-link evaluation and adjustment.

Description

Technical Field

[0001] This invention relates to the field of anti-interference control system technology for unmanned aerial vehicle (UAV) cables, and in particular to an anti-interference control method and system for tethered UAV cable deployment and retrieval based on image recognition. Background Technology

[0002] A tethered UAV (Tethered Drone) is a drone system that operates continuously in the air, connected to ground equipment via a physical tether cable. It is typically used in scenarios requiring long-term, stable hovering and continuous power / communication. It consists of a ground-based high-voltage DC regulated system, a cable reel, a synchronous winding reel, a tether cable, an in-flight voltage regulator module, and a backup battery. During operation, the tethered drone actively controls a ground winch or a cable winding mechanism on the drone to orderly retract or extend the tether cable. This enables the drone to ascend, descend, maintain altitude, automatically avoid obstacles, and adapt to changes in wind speed. Cameras are deployed on the ground winch, the drone itself, or at key path points to capture real-time images of the cable, the drone, the environment, and the cable routing area. By capturing cable images through the cameras and using image recognition algorithms to identify the cable's actual spatial position, image analysis can infer the cable's actual condition, assess the distance and risk level between the cable and obstacles in real time, and preemptively limit and adjust the cable retraction and extension speeds and tension to prevent entanglement or collisions.

[0003] For example, the controller design method for a ducted multi-rotor tethered drone disclosed in Chinese Invention Patent CN109542110B addresses the shortcomings of existing drone control system simulation platforms in optimizing control parameters and the poor control capability and weak anti-interference ability of drone controllers. This new method employs finite element analysis to analyze the overall aerodynamic characteristics of the ducted multi-rotor tethered drone. Based on these parameters, a dynamic model of the ducted multi-rotor tethered drone body is established, the distribution values ​​of the tension control for each motor are obtained, and the aerodynamic parameters are measured. Combining the established model and the measured aerodynamic parameters, a control simulation platform for the ducted multi-rotor tethered drone control system is built in the Simulink environment. A fuzzy PID controller is used to design the ducted multi-rotor tethered drone control system with a tether cable, and the parameters are improved.

[0004] For example, Chinese invention patent CN118707982B discloses a method and system for the safe landing of a tethered drone. The method includes: after detecting or receiving a power failure signal from the tethered drone, sending a parachute opening command to the parachute to control its opening; detecting the descent speed of the tethered drone and stabilizing it within a preset speed range; detecting the distance between the tethered drone and the ground, and when the distance is less than a preset distance value, activating the backup battery and power system to provide lift to the tethered drone and control its deceleration and descent to the ground.

[0005] The aforementioned technology has at least the following technical problems: In the event of a large-scale communication outage, to ensure uninterrupted communication, tethered drones equipped with communication relays can be deployed to quickly restore long-distance communication signals and continuously hover on the edge of urban areas. However, sudden changes in wind speed or persistent strong winds pose multiple challenges to the continuous hovering operation of tethered drones equipped with communication relays. Strong winds can cause sudden changes in the tension of the tethered cable (such as sudden tightening or lateral swaying), leading to unstable hovering of the drone. In some cases, excessive tension may trigger safety protection or even damage the cable and winch. During the retrieval and deployment process under wind interference, abnormal tension and mismatch between retrieval and deployment speeds can cause the drone to sway and drift in the air.

[0006] In actual operation, from the perspective of the tension change of the mooring cable, when a sudden gust of wind blows along the cable axis (such as an updraft from the ground to the air or a downdraft from the air to the ground), it will instantly change the tension state of the cable. The cable will be suddenly tightened, and the tension will soar from the normal suspension tension to far exceeding the upper limit of the cable's safe tension in a short period of time, which will seriously affect the stability of the communication relay and the safety of the operation. Summary of the Invention

[0007] This invention provides an image recognition-based anti-interference control method and system for tethered drone deployment and take-up lines, which solves the problem of drones swaying and drifting in the air due to abnormal tension and mismatch between deployment and take-up speeds in the prior art, and improves the accuracy of deployment and take-up speed matching with tension.

[0008] To achieve the above-mentioned objectives, the present invention provides the following technical solution:

[0009] On the one hand, an anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition is provided. This method includes: when the UAV is performing aerial operations, acquiring time-series features aligned with timestamps within a set time period, including wind speed, wind direction, cable direction, measured tension, winch speed, and cable image; projecting the wind speed along the cable direction to obtain the wind disturbance component along the cable, which represents the intensity of wind disturbance in the cable direction; inputting the real-time measured tension, winch speed, and wind disturbance component into a tension prediction model to output the predicted tension for the next moment; comparing the predicted tension with a safe range and adjusting the winch deployment and retrieval speed in real time; and combining image recognition of cable morphology to visually correct the winch deployment and retrieval speed using anti-slack logic and vibration suppression logic, where anti-slack logic suppresses cable slack and vibration suppression logic suppresses lateral swaying of the cable caused by external interference or operational dynamics; and sending the corrected winch deployment and retrieval speed down for execution and synchronizing it with the UAV's speed adjustment system to adaptively coordinate with the winch deployment and retrieval actions.

[0010] On the other hand, an anti-interference control system for tethered drone cable deployment and retrieval based on image recognition is provided. This system applies an anti-interference control method for tethered drone cable deployment and retrieval based on image recognition, including: a data acquisition and alignment module, a tension prediction module, a deployment and retrieval speed adjustment module, a visual correction module, and a command execution and drone synchronization module. The data acquisition and alignment module acquires the time-series characteristics of timestamp alignment within a set time period when the drone is performing aerial operations, projects the wind speed along the cable direction, and obtains the wind disturbance component along the cable. The tension prediction module inputs the real-time measured tension, winch speed, and wind disturbance component into the tension prediction model and outputs the predicted tension for the next moment. The deployment and retrieval speed adjustment module compares the predicted tension with the safe range and adjusts the winch deployment and retrieval speed in real time. The visual correction module combines image recognition of the cable shape with loosening and vibration suppression logic to visually correct the winch deployment and retrieval speed. The command execution and drone synchronization module sends the corrected winch deployment and retrieval speed and synchronizes the target speed to the drone. After receiving the target speed command, the drone actively adjusts its flight attitude to adaptively coordinate with the winch deployment and retrieval actions.

[0011] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0012] 1. Existing technologies are mostly based on real-time measured tension and passive adjustment. When sudden changes in wind speed cause a sudden change in tension, the cable may be overloaded and broken or become loose and tangled due to untimely adjustment. However, by calculating the wind disturbance component along the cable direction, the intensity of the directional disturbance of the wind on the cable can be accurately quantified. Combined with the tension prediction model, the predicted tension for the next moment can be output in advance, so that the winch speed adjustment can be changed from passive response to advanced prediction. It can be adjusted in advance before the tension abnormality occurs, reducing the risk of drone attitude instability.

[0013] 2. Directly using ground or drone-based single-point wind speed data to calculate wind disturbance components without considering the wind speed gradient deviation between the ground and air wind fields caused by terrain (such as urban building obstruction) and height differences can easily lead to overestimation or underestimation of wind disturbance components, resulting in inaccurate tension prediction. By obtaining the current cable sag and comparing it with the set sag based on the predicted tension, the wind disturbance components can be corrected. This can accurately reduce the interference of ground and air wind speed deviations on the predicted tension and improve the accuracy of the predicted tension.

[0014] 3. In existing technologies, the tension adjustment of tethered drones' cable reeling and deployment mostly relies on real-time measured tension triggering, which results in adjustment lag. Moreover, most of them use a single adjustment step, and even when the tension is within the safe range, they still use large step adjustments, which can easily cause drastic tension fluctuations. Based on the predicted tension, the adjustment is carried out in advance through graded steps (the first step is for emergency adjustment, and the second step is for precise fine-tuning). Furthermore, a hysteresis range and adjustment maintenance time are set for dual protection. This solves the problems of frequent winch wear, secondary cable oscillation, and drone attitude instability caused by critical tension fluctuations under the traditional mechanism, and improves the timeliness, stability, and equipment reliability of cable reeling and deployment control.

[0015] 4. The corrected winch speed is synchronized to the UAV speed adjustment system, enabling the UAV to adaptively coordinate with the winch's retrieval and release actions (e.g., when the winch accelerates cable release, the UAV adjusts its flight speed appropriately to avoid excessive cable pulling). This avoids tension fluctuations caused by the traditional independent control of the two systems, improves overall control coordination, and solves the reliability issues of command parameters and timing. The UAV performs integrity verification on the speed parameters issued by the winch, which can avoid erroneous actions caused by missing parameters. The handling of timing consistency can eliminate the hidden danger of asynchronous actions caused by timestamp deviations, ensuring precise coordination between winch retrieval and UAV flight in the time dimension. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an anti-interference control method for tethered unmanned aerial vehicle (UAV) cable deployment and retrieval based on image recognition, provided in an embodiment of the present invention.

[0018] Figure 2 This is a flowchart of wind disturbance component correction provided in an embodiment of the present invention;

[0019] Figure 3This is a flowchart of the visual correction provided in an embodiment of the present invention;

[0020] Figure 4 This is a schematic diagram of the anti-interference control system for tethered UAVs based on image recognition, provided in an embodiment of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms “first,” “second,” and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an,” “a,” or “the,” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising,” “including,” or “including,” and similar terms mean that the element or object preceding the word encompasses the element or object listed following the word and its equivalents, without excluding other elements or objects. The terms “connected,” “linked,” or “connected,” and similar terms are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect.

[0023] It should be noted that the terms "up", "down", "left", "right", "front", and "back" used in this invention are only used to indicate relative positional relationships. When the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0024] like Figure 1 The flowchart shown is a process for anti-interference control of tethered drone cable deployment and retrieval based on image recognition, according to an embodiment of this application. The method includes:

[0025] S1. When the drone is performing aerial operations, it acquires the time-series features aligned with the timestamp within a set time period. The time-series features include wind speed, wind direction, cable direction, measured tension, winch speed, and cable image. The wind speed is projected along the cable direction to obtain the wind disturbance component along the cable. The wind disturbance component along the cable is used to represent the disturbance intensity of the wind in the cable direction.

[0026] S2 inputs the real-time measured tension, winch speed, and wind disturbance component into the tension prediction model and outputs the predicted tension for the next moment.

[0027] S3 compares the predicted tension with the safe range and adjusts the winch's winding and unwinding speed in real time.

[0028] S4, combined with image recognition of cable shape, uses loosening logic and vibration suppression logic to visually correct the winding and unwinding speed of the winch. The loosening logic is used to suppress cable slack, and the vibration suppression logic is used to suppress the lateral sway of the cable caused by external interference or dynamic operation.

[0029] S5 sends the corrected winch winding and unwinding speed down and executes it, and synchronizes it with the drone's speed adjustment system to adaptively match the winch winding and unwinding actions.

[0030] In this embodiment, vector operations are performed based on the acquired wind speed, wind direction, and cable direction to obtain the wind speed vector and the cable direction unit vector. The angle between the wind speed vector and the cable direction unit vector is calculated based on the current spatial principal axis direction of the cable, using the following formula:

[0031] ,

[0032] in To represent the dot product between the wind speed vector and the unit vector of the cable direction, |V wind | represents the magnitude of the wind speed vector. Angle.

[0033] The projection of the wind speed onto the cable direction, i.e., the wind disturbance component of the cable, is calculated using the formula V. cable =|V wind |·cos , where V cable Cosγ represents the component of wind speed along the main axis of the cable. Only this part of the wind speed directly causes tension disturbance to the cable. Cosγ reflects that the smaller the angle between the wind and the cable, the greater the wind disturbance component.

[0034] Existing technologies rely on passive tension adjustment based on real-time measured tension, which is prone to cable failure and drone attitude instability during sudden wind speed changes. This is especially problematic when operating in complex outdoor wind fields (such as gusts and turbulence). When tethered drones are inspecting power lines in the field, sudden gusts of wind can disrupt the performance of traditional solutions, which can only detect wind speed magnitude and cannot determine the actual disturbance to the cables. This results in poor anti-interference capabilities and low control precision. This new technology calculates wind disturbance components and outputs predicted tension in advance, enabling proactive adjustment. Simultaneously, it considers wind speed gradient deviations and corrects for wind disturbance components, improving the accuracy of predicted tension. Furthermore, it uses graded step-by-step adjustment of predicted tension, setting hysteresis intervals and holding times to address tension fluctuation issues. Finally, it synchronizes the corrected winch speed to the drone, achieving adaptive coordination, avoiding tension fluctuations, and improving overall control synergy.

[0035] Preferably, the real-time measured tension, winch speed, and wind disturbance component are input into the tension prediction model, and the predicted tension for the next moment is output. The specific process is as follows:

[0036] The collected data is denoised, outliers are removed, and normalized.

[0037] A time-series sliding window is constructed to transform historical tension, velocity, and wind disturbance components into multi-time-series input features;

[0038] Based on the system characteristics and data conditions, a state-space model was selected, and a recursive formula was established. The model inputs included the measured tension, winch speed, and wind disturbance component.

[0039] The model is trained using a large amount of historical operating data to optimize weights or system parameters;

[0040] Validate the model's prediction performance on independent test sets and in extreme scenarios;

[0041] In actual operation, the latest set of time series features is input into the model, and the tension prediction value for the next moment is output in real time.

[0042] In this embodiment, the system state is converted into a mathematical expression:

[0043] ,

[0044] Among them, X k This represents the system state vector (e.g., measured tension) at the k-th sampling time; u k Let w be the input vector at time k. The input vector contains multiple input quantities such as winch speed and wind disturbance component. The physical meaning of each input quantity is defined in sequence; A and B represent the state and input weight matrices, respectively; w kLet k represent the system noise or residual term at the k-th sampling time, k=1,2,3,…,n, where n represents the total number of samples; the formula is expressed as the system's next state equals the natural evolution of the system's current state plus the effects of external control and interference, plus some unpredictable random factors.

[0045] Through the tension prediction model, the system can predict the tension change at the next moment, rather than passively responding after the tension anomaly occurs. This allows the control system to act in advance, predict tension fluctuations in advance, and avoid large oscillations to ensure the stability of the UAV platform and cable. The system does not rely on a single information source, but deeply integrates physical sensor data (wind speed, tension, velocity) and visual perception information (cable image), realizing the transformation of tethered UAV cable release and retrieval control from passive reaction to active anti-interference.

[0046] like Figure 2 As shown in the flowchart of a wind disturbance component correction method provided in this application embodiment, the specific process is as follows: By comparing the actual sag with a set threshold, the system can determine whether the initial wind disturbance component estimation is accurate: if the sag is too large, it indicates that the wind disturbance is underestimated and positive compensation is needed; if the sag is too small, it indicates that the wind disturbance is overestimated or other factors cause the cable to be taut and reverse compensation is needed. The corrected predicted tension also needs to undergo consistency verification to ensure that it meets the system safety constraints before it is finally used as a reliable input to the control system. This process effectively reduces the impact of the difference between ground and air wind speeds on the model accuracy.

[0047] Preferably, the output of the predicted tension at the next moment also includes wind disturbance component correction of the output predicted tension. The specific steps are as follows: after the prediction stage is completed, the current cable sag is obtained, and the current cable sag is compared with the set sag based on the predicted tension. The wind disturbance component correction is performed based on the comparison result. The wind disturbance component correction is used to reduce the impact of the deviation of ground and air wind speed on the predicted tension.

[0048] If the actual sag measured is greater than the maximum value of the set sag threshold range, it is determined that the wind disturbance component is underestimated and the model does not cover some disturbances. In this case, the model is positively compensated by increasing the wind disturbance component. The set sag threshold range represents the closed interval range formed by the minimum and maximum allowable values ​​of cable sag. If the actual sag measured is less than the minimum value of the set sag threshold range, it is determined that the wind disturbance component is overestimated and there are other tight disturbances. In this case, the model is negatively compensated by decreasing the wind disturbance component.

[0049] When the sag recovers to the set sag threshold range, compensation for wind disturbance components stops, and the sag in the next cycle is monitored until the sag remains within the sag threshold range for the set time period. This indicates that the cable has returned to a stable state. If the sag exceeds the sag threshold range continuously within the set time period, a tension prediction warning is issued.

[0050] After each wind disturbance component compensation, the corrected predicted tension is then checked for consistency with the safe range and system constraints. Once it passes the check, the predicted tension is used as the input for the main control adjustment.

[0051] In this embodiment, when it is necessary to increase the wind disturbance component, the formula for calculating the increase adjustment amount is as follows:

[0052]

[0053] Where k is the adjustment ratio coefficient, determined by engineering experience; S obs S represents the actual sag measured at present. max This is the maximum value within the set sag threshold range; conversely, if the wind disturbance component is adjusted downwards, the calculation formula is: S min The minimum value of the set sag threshold range, the absolute value of a single compensation amount shall not exceed the preset maximum allowable step size (usually 5% to 10% of the current wind disturbance component).

[0054] By monitoring sag to identify disturbances not covered by the model, abnormal sag indicates that the wind disturbance component does not reflect the actual force. Real-time compensation is applied to the wind disturbance component, dynamically aligning it with the actual wind field in the air. This reduces the deviation between predicted and actual tension, avoiding the risk of cable overload or slack caused by wind field deviations. On one hand, it accurately captures real disturbances not covered by the initial wind disturbance component, improving the matching degree between predicted tension and actual cable force, providing a precise basis for subsequent winch reel adjustments. On the other hand, dynamic compensation prevents the accumulation of deviations, and combined with timeout warnings, it promptly identifies and eliminates anomalies such as persistent wind disturbances. Consistency verification further prevents over-correction, enhancing the anti-interference capability and operational reliability of tethered UAV cable reel control.

[0055] Preferably, the current cable sag is obtained by: acquiring a cable image with acceptable image quality and stable continuous frames; using edge detection and morphological processing; edge detection is used to detect the edge contour of the cable in the image; and morphological processing is used to connect the edge contour of the cable into continuous line segments, and then refine them into cable skeleton lines.

[0056] The cable skeleton line is extracted as the geometric principal axis. The cable skeleton line is used to reflect the cable's direction, sag, and swing. The cable skeleton is fitted into a catenary model. The catenary model is used to transform the discrete skeleton point cloud into a continuous curve that conforms to physical laws, so as to obtain the cable's sag (reflecting tension) and swing amplitude.

[0057] In this embodiment, the image quality is judged based on the effective proportion of the cable target and the continuity of its edges. The core steps can be summarized as follows: First, define the horizontally centered area of ​​interest for the cable target in the image, covering a set height vertically; then, convert the image to grayscale and extract the effective pixels of the cable target through thresholding and filtering; calculate the proportion of effective pixels to the total pixels in the area of ​​interest to obtain the effective pixel proportion. The Canny operator is used to extract the edges of the effective pixels of the cable target, obtaining the cable edge pixels. The length of continuous segments of the edge pixels is calculated, and the length of these continuous segments is compared with the vertical height of the area of ​​interest to obtain the proportion of continuous edge pixel segments. When the effective pixel proportion is not less than the set effective pixel proportion (usually 20% based on historical data), and the proportion of continuous edge pixel segments is not less than the set continuous segment proportion (usually 80% based on historical data), the image quality is judged to be acceptable.

[0058] Skeleton extraction and fitting are performed on N consecutive frames (e.g., N=5). If high-frequency sampling is required, the number of frames N can be increased (e.g., to 7-10 frames). If faster response is required (e.g., in high-speed scenes), the number of frames N can be reduced. The number of frames N is dynamically and adaptively adjusted. If the environment is stable and the image is continuous and of good quality, N can be relaxed; otherwise, it can be tightened.

[0059] The difference between the extracted skeleton line in each frame and the previous frame is calculated to obtain the principal parameter change. The Laplacian variance of each frame image is used as the quality score. The image is considered to be stable in consecutive frames only when the principal parameter changes of all frames are lower than the preset morphology threshold and the quality score of each frame image is higher than the empirical threshold.

[0060] The main process of edge detection is to extract edges after image preprocessing and then perform core detection using the Canny edge detection algorithm.

[0061] Morphological processing first aligns the center of the structuring element with each pixel of the edge image. If a definite edge pixel exists within the coverage area of ​​the structuring element, the center pixel is set as an edge pixel. This operation fills in minor breaks in the edge contour (such as local edge loss caused by reflection from the cable surface) and enhances edge continuity. Then, the center of the structuring element is aligned with the pixels of the dilated image. Only when all pixels within the coverage area of ​​the structuring element are edge pixels is the center pixel set as an edge pixel. This operation removes edge burrs generated during dilation (such as fine edges formed by residual background noise) and shrinks the widened edges back to near their original width, restoring the true contour of the cable edge.

[0062] The catenary model fitting takes the skeleton point sequence (xi, yi) as input, where (xi, yi) represents the coordinates of the i-th discrete sampling point in the original skeleton point sequence. Each (xi, yi) lies in the domain of (x, y) and has a unique mapping point. Optimization algorithms such as the least squares method are used to fit all points to the catenary formula.

[0063]

[0064] In the fitting calculation, each xi is substituted into the model and compared with the actual yi to minimize the residual, determine the optimal parameters a and b, so that the error between the skeleton point and the model curve is minimized (the sum of squared residuals is minimized), and the swing amplitude is obtained by directly calculating the maximum lateral amplitude change of the skeleton line.

[0065] Using the fitted curve formula, calculate the y-coordinate of the lowest point of the cable and the y-coordinate of the line connecting the two ends at the lowest point. The sag (Sag) is defined as: Sag = y min -y line y min The y-coordinate represents the minimum point of the fitted curve; line This represents the y-coordinate value obtained by drawing a straight line through the two endpoints and finding the lowest point at the x-coordinate.

[0066] Preferably, the predicted tension is compared with the safe range, and the winding and unwinding speed of the winch is adjusted in real time. The specific steps are as follows: before the actual tension exceeds the boundary of the safe range, the predicted tension is compared with the safe range; if the predicted tension is greater than the maximum value of the safe range, the cable winding speed is reduced in the first step to relieve the tension. The first step means adjusting the winding and unwinding speed according to the set winding and unwinding speed.

[0067] If the tension is less than the minimum value of the safe range, then increase the cable laying speed in the first step to maintain the tension.

[0068] If the predicted tension is within the safe range, then the second step is used to extend and retract the tension, with the midpoint of the safe range as the target, so that the tension reaches the set midpoint of the safe range. The second step is slower than the extension and retraction speed of the first step. If the current speed can keep the tension stable within the set midpoint of the safe range, then the current speed is maintained.

[0069] At the same time, a hysteresis range is set at the maximum and minimum values ​​of the safety range, and a holding time for the adjustment speed is set. Only when the tension continuously crosses the hysteresis range or exceeds the holding time within the set time is it allowed to change the release speed, so as to avoid frequent adjustment switching in the critical state.

[0070] In this embodiment, compared to existing technologies that rely on real-time measured tension triggering and have significant lag, the winch winding speed is adjusted in real time by comparing the predicted tension with the safety range. When the predicted tension exceeds the maximum value or falls below the minimum value of the safety range, the winding speed can be adjusted in advance with a first step to alleviate or maintain tension, effectively preventing cable overload breakage or excessive loosening and tangling caused by abnormal tension. When the predicted tension is within the safety range, a smaller second step is used to fine-tune the tension, targeting the midpoint of the safety range, to stabilize the tension within the midpoint range. This solves the problems of drastic tension fluctuations, cable fatigue wear, and ineffective energy consumption caused by single-step adjustment. Simultaneously, setting a hysteresis range and adjustment speed maintenance time avoids frequent adjustment switching under critical conditions, reducing frequent winch wear, secondary cable oscillation, and drone attitude instability, significantly improving the timeliness, stability, and equipment reliability of winding control.

[0071] like Figure 3 As shown in the flowchart of a visual correction embodiment provided in this application, the specific process is as follows: When the system determines that the predicted tension is within a safe range, image recognition is initiated to monitor the physical shape of the cable. Based on the recognition result, two parallel correction logics are triggered: Slack reduction logic (for sag): If the detected sag exceeds a set value, the system retracts the cable at a third-step rate until the sag returns to the normal range; Vibration suppression logic (for sway): If the detected sway amplitude exceeds a set value, the system first retracts the cable at a third-step rate and observes the change in sway amplitude: If the sway returns to normal, the system continues to retract the cable at that rate; if the sway remains abnormal, a speed reduction strategy is adopted to prevent other problems that may be caused by over-tightening the cable.

[0072] Preferably, the winding and unwinding speed of the winch is visually corrected using loosening logic and vibration suppression logic. Specifically, when the predicted tension is within the safe range, the cable shape is identified by image recognition.

[0073] The specific process of modifying the loosening logic is as follows:

[0074] If the sag exceeds the set sag, it means the cable is too loose. Without touching the upper limit, take the cable in the third step until the sag returns to within the set sag. The third step is slower than the first step. The second and third steps are both slower than the first step. There is no need to compare their sizes; just make sure they are both slower than the first step.

[0075] The specific process of modifying the vibration suppression logic is as follows:

[0076] If the swing amplitude is greater than the set swing amplitude, it indicates that the cable has entered a vibration state. First, take the cable in the third step and observe the swing amplitude within the set time, and compare it with the set amplitude change range.

[0077] If the swing amplitude is within the set amplitude variation range and the tension does not exceed the safe range, continue to take in the third step in that direction; if the swing amplitude does not exceed the set amplitude variation range and the tension increases towards the maximum value of the safe range, then reduce the speed to prevent the vibration from intensifying.

[0078] In this embodiment, the limitations of traditional tension data perception are overcome, enabling precise identification of hidden problems such as tension safety but slack in form. Based on this, the slack-reduction logic actively eliminates slack, avoiding the risk of slack accumulation. It fills the gap in the dynamic monitoring of vibration state in traditional regulation, and the vibration suppression logic controls in stages accordingly to avoid aggravated vibration. It also provides precise basis for regulation, making small corrections only when sag or swing exceeds the limit, avoiding ineffective regulation, reducing winch wear and tension fluctuations, and balancing stability and economy. It realizes the transformation from passive tension regulation to active form and tension coordinated control.

[0079] Preferably, the corrected winch winding and unwinding speed is sent down and synchronized to the UAV speed adjustment system to adaptively match the winch winding and unwinding actions. The specific process is as follows: the corrected winch winding and unwinding speed (including whether the speed direction is winding or unwinding, the speed magnitude and execution cycle) is sent down to the winch controller, the winding and unwinding actions are executed according to the speed parameters, and the command sending timestamp is recorded and synchronized with the UAV timing.

[0080] After receiving synchronization information from the winch, the drone speed regulation system verifies the integrity of the speed parameters (confirming that the direction, speed, and period are complete) and processes the timing consistency (comparing its own timestamp with the sent timestamp; if the deviation exceeds the set time, it requests resynchronization).

[0081] By combining the drone's current flight speed, its spatial position relative to the winch (horizontal distance, vertical height), and cable shape recognition data, speed adjustment requirements adapted to the winch's movements are generated.

[0082] In this embodiment, deep coordination between the winch and the drone's cable deployment and retrieval actions can be achieved, solving the problems of asynchronous actions and parameter mismatches under traditional independent control. On the one hand, it ensures the reliability of command parameters and timing. By verifying the integrity of the speed parameters issued by the winch, erroneous actions caused by missing parameters are avoided. The processing of timing consistency can eliminate the hidden danger of asynchronous actions caused by timestamp deviations, ensuring precise coordination between winch deployment and drone flight in the time dimension. On the other hand, it enhances adaptability to complex scenarios. The drone, combining its own flight speed, relative spatial position to the winch, and cable morphology recognition data, can dynamically generate speed adjustment requirements adapted to the winch's actions, effectively coping with complex working conditions under dynamic interference such as wind fields. At the same time, it can also improve the overall safety and stability of the system, significantly reduce abnormal tension fluctuations during cable deployment and retrieval, reduce risks such as cable slack and vibration, and ensure the reliable operation of the tethered drone system in various scenarios.

[0083] Preferably, the process of issuing the corrected winch winding speed also includes evaluating the current cable status, specifically as follows:

[0084] After the cable winding and unwinding speed adjustment is issued and executed, the increment of safety margin, the decrease of sag, and the decrease of swing energy within the set time are obtained. After normalization, they are weighted and fused to obtain the comprehensive safety index.

[0085] If the overall safety index is positive, it indicates that the risk of the current state has decreased. Therefore, the adjustment is deemed effective, and the current correction direction can be maintained with a further increase in the step size within the allowable range.

[0086] If the overall safety index is negative, it indicates that the current risk level has increased. In this case, the adjustment is deemed invalid and the correction adjustment in that direction should be cancelled in a timely manner.

[0087] In this embodiment, the changes in each indicator before and after adjustment within a set time period are calculated:

[0088] =U after -U before ,

[0089] =S before -S after ,

[0090] =E before -E after ,

[0091] U=min(|T-Tmin|,|T-Tmax|),

[0092] Where T is the current cable tension, Tmin is the minimum tension (lower limit of the safe zone), and Tmax is the maximum tension (upper limit of the safe zone). This formula takes the smaller of the distance between the current tension and the lower limit of the safe zone and the distance between the current tension and the upper limit of the safe zone, and is used to measure the closest distance of the cable tension to the boundary of the safe zone, that is, the safety margin. For example, if the current tension is closer to Tmin, then U = |T - Tmin|, which reflects the safety margin of the tension from the lower limit boundary; conversely, U = |T - Tmax|, which reflects the safety margin from the upper limit boundary; ΔU is the change in safety margin, used to represent the change in the degree to which the system tension moves away from the safe boundary before and after adjustment. after The adjusted safety margin is defined as the minimum distance between the tension and the upper and lower limits of the safety range, U. before The initial safety margin is ΔU > 0, indicating an increase in safety margin; ΔS represents the change in sag, indicating the degree of improvement in cable tension.before S is the sag before adjustment (usually referring to the difference between the y-coordinate of the lowest point obtained from the catenary fitting and the y-coordinate of the line connecting the two ends at that point). after The adjusted sag is represented by ΔS > 0, indicating a reduction in sag (tighter); ΔE represents the change in oscillation energy, used to measure the improvement in cable vibration (or lateral energy) before and after adjustment. before E represents the oscillation energy before adjustment (such as the root mean square (RMS) of the lateral displacement, the sum of squares of the amplitude, etc.). after The adjusted oscillation energy, ΔE>0 represents a decrease in vibration energy (more stable).

[0093] The comprehensive safety index is calculated as J = ω1 × ΔU + ω2 × ΔS + ω3 × ΔE, where ω1, ω2, and ω3 are the weights of each item, satisfying ω1 + ω2 + ω3 = 1. J represents the comprehensive safety index, and ω1, ω2, and ω3 are the weighting coefficients for safety margin, sag, and oscillation energy, respectively, reflecting the system's emphasis on each physical index. Based on historical data, the actual performance of the system under different weight combinations (such as over-limit frequency, residual vibration, relaxation events, etc.) is tested through grid search. Among all weight combinations, the weight combination that minimizes the weighted total loss of over-limit frequency, residual vibration, and relaxation events is selected.

[0094] This state assessment process, through closed-loop feedback and quantitative indicators, overcomes the shortcomings of open-loop regulation in existing technologies. It achieves both accurate judgment of regulation effectiveness and dynamic optimization of correction strategies, significantly improving the safety, adaptability, and regulation efficiency of tethered UAV line release and take-up control.

[0095] Preferably, the speed adjustment requirements for generating the winch action are as follows: if the winch performs a cable winding action, the drone is controlled to adjust its flight position along the direction close to the winch at a speed that matches the winch winding rate. The speed values ​​that match the winch winding rate are approximately equal, the directions are perfectly aligned, and the timing is synchronized.

[0096] If the winch performs the cable release action, control the drone to adjust its flight position in a direction away from the winch at a speed that matches the winch's cable take-up rate, and maintain the drone's flight speed fluctuations to prevent the cable from slackening due to sudden speed changes disrupting the balance with the cable release rate.

[0097] If the cable is detected to be swinging laterally, the lateral speed of the drone is adjusted synchronously, and the winch vibration suppression logic is used to suppress the swing.

[0098] This embodiment overcomes the limitations of traditional winch and drone adjustments that lack precise coordination. By achieving near-equal speeds, perfectly aligned directions, and synchronized timing responses during cable retrieval, and by controlling drone speed fluctuations during cable release, it avoids cable breakage due to excessive tightness or tangling due to asynchronous actions, ensuring that tension remains stable within a safe range throughout the retrieval and release processes. Speed ​​parameter integrity verification (confirming direction, speed, and cycle) avoids erroneous adjustments caused by missing parameters. Combined with timing consistency processing (resynchronization when deviation exceeds limits), it solves the adjustment lag problem caused by traditional timestamp misalignment, ensuring precise temporal alignment between winch actions and drone responses. Compared to traditional adjustments relying solely on fixed logic, by combining real-time drone flight speed, relative winch spatial position, and cable morphology data, it can dynamically generate adaptation requirements. Especially when the cable swings laterally, synchronous adjustment of the drone's lateral speed, in conjunction with winch vibration suppression logic, effectively suppresses vibration, overcoming the limitations of traditional single-adjustment methods in handling dynamic interference and improving the system's operational reliability in complex scenarios such as wind farms.

[0099] like Figure 4 As shown in the figure, an embodiment of the present invention provides a structural schematic diagram of an anti-interference control system for tethered drone take-up and release lines based on image recognition, including: a data acquisition and alignment module, a tension prediction module, a take-up and release speed adjustment module, a visual correction module, and a module for issuing and executing commands and synchronizing with the drone.

[0100] The data acquisition and alignment module is used to acquire the temporal characteristics of timestamp alignment within a set time period when the UAV is performing aerial operations, and to project the wind speed along the cable direction to obtain the wind disturbance component along the cable; the tension prediction module is used to input the real-time measured tension, winch speed, and wind disturbance component into the tension prediction model, and output the predicted tension for the next moment; the winding and unwinding speed adjustment module is used to compare the predicted tension with the safe range and adjust the winding and unwinding speed of the winch in real time; the visual correction module is used to visually correct the winding and unwinding speed of the winch by combining image recognition of cable shape with loosening logic and vibration suppression logic; the command and execution synchronization module is used to send the corrected winding and unwinding speed of the winch and synchronize the target speed to the UAV. After receiving the target speed command, the UAV actively adjusts its own flight attitude to adaptively cooperate with the winding and unwinding actions of the winch.

[0101] In this embodiment, the data acquisition and alignment module aligns the time-series feature timestamps and calculates the wind disturbance component along the cable. Combined with the multi-parameter input of the tension prediction module, it improves the accuracy of tension prediction and avoids the prediction deviation caused by relying on only a single data point in the traditional method, providing a reliable basis for adjustment. The reeling and unloading speed adjustment module adjusts in advance based on the predicted tension, and the visual correction module combines image recognition of cable shape to correct the speed through loosening and vibration suppression logic, solving the problem that traditional methods cannot detect cable slack and vibration by relying solely on tension data, thus achieving dual control. The sending and execution synchronization module enables the speed of both to coordinate, and the drone actively adapts to the winch's movements, avoiding tension fluctuations caused by traditional independent control, enhancing the system's anti-interference capability, and ensuring the safe and stable operation of the tethered drone.

[0102] The following points need to be explained:

[0103] (1) The accompanying drawings of the embodiments of the present invention only involve the structures involved in the embodiments of the present invention. Other structures can refer to the general design.

[0104] (2) For clarity, the thickness of layers or regions is enlarged or reduced in the drawings used to describe embodiments of the invention, i.e., these drawings are not drawn to scale. It is understood that when an element such as a layer, film, region or substrate is referred to as being “above” or “below” another element, the element may be “directly” located “above” or “below” the other element or there may be intermediate elements.

[0105] (3) Where there is no conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0106] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. The scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for anti-interference control of tethered unmanned aerial vehicle (UAV) cable deployment and retrieval based on image recognition, characterized in that, include: S1. When the tethered drone is performing aerial operations, the time sequence features aligned with the timestamps within a set time period are obtained. The time sequence features include wind speed, wind direction, cable direction, measured tension, winch speed, and cable image. The wind speed is projected along the cable direction to obtain the wind disturbance component along the cable. The wind disturbance component along the cable is used to represent the disturbance intensity of the wind in the cable direction. S2 inputs the real-time measured tension, winch speed, and wind disturbance component into the tension prediction model and outputs the predicted tension for the next moment. S3 compares the predicted tension with the safe range and adjusts the winch's winding and unwinding speed in real time. The steps for comparing the predicted tension with the safe range and adjusting the winch's winding and unwinding speed in real time are as follows: Before the actual tension crosses the boundary of the safe zone, the predicted tension will be compared with the safe zone. If the predicted tension is greater than the maximum value of the safe range, the cable winding speed is reduced in the first step to relieve the tension. The first step means adjusting the winding and unwinding speed according to the set winding and unwinding speed. If the tension is less than the minimum value of the safe range, then increase the cable laying speed in the first step to maintain the tension; If the predicted tension is within the safe range, then the second step is used to extend and retract the tension with the midpoint of the safe range as the target, so that the tension reaches the set midpoint of the safe range. The second step is smaller than the extension and retraction speed of the first step. If the current speed can keep the tension stable within the set midpoint of the safe range, then the current speed is kept unchanged. At the same time, a hysteresis range is set at the maximum and minimum values ​​of the safety range, and the holding time of the adjustment speed is set. The speed of release and release is only allowed to be changed when the tension continuously crosses the hysteresis range or exceeds the holding time within the set time. S4, combining image recognition of cable shape, visually correct the winding and unwinding speed of the winch with anti-loosening logic and vibration suppression logic. The anti-loosening logic is used to suppress cable slack, and the vibration suppression logic is used to suppress the lateral sway of the cable caused by external interference or dynamic operation. S5 sends the corrected winch winding and unwinding speed down and executes it, and synchronizes it with the UAV's speed adjustment system to adaptively match the winch winding and unwinding actions.

2. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 1, characterized in that, The process involves inputting real-time measured tension, winch speed, and wind disturbance components into the tension prediction model, and outputting the predicted tension for the next moment. The collected data is denoised, outliers are removed, and normalized. A time-series sliding window is constructed to transform historical tension, velocity, and wind disturbance components into multi-time-series input features; Based on the system characteristics and data conditions, a state-space model was selected, and a recursive formula was established. The model inputs included the measured tension, winch speed, and wind disturbance component. The model is trained using a large amount of historical operating data to optimize weights or system parameters; Validate the model's prediction performance on independent test sets and in extreme scenarios; In actual operation, the latest set of time series features is input into the model, and the tension prediction value for the next moment is output in real time.

3. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 1, characterized in that, The output of the predicted tension at the next moment also includes wind disturbance component correction of the output predicted tension, the specific steps of which are as follows: After the prediction process is completed, the current cable sag is obtained, and the current cable sag is compared with the set sag based on the predicted tension. Based on the comparison results, wind disturbance component correction is performed. The wind disturbance component correction is used to reduce the impact of the deviation of ground and air wind speed on the predicted tension. If the actual sag measured is greater than the maximum value of the set sag threshold range, it is determined that the wind disturbance component is underestimated and the model does not cover the disturbance. Positive compensation is then performed on the model, i.e., the wind disturbance component is increased. The set sag threshold range represents the closed interval range formed by the minimum and maximum allowable values ​​of cable sag. If the actual sag measured is less than the minimum value of the set sag threshold range, it is determined that the wind disturbance component is overestimated and there are other tight disturbances. The model is then compensated in reverse, i.e., the wind disturbance component is reduced. When the sag recovers to the set sag threshold range, the correction of the wind disturbance component stops, and the sag of the next cycle is monitored until the sag is always within the sag threshold range within the set time period. This indicates that the cable has returned to a stable state. If the sag exceeds the sag threshold range continuously within the set time period, a tension prediction warning is issued. After each correction of the wind disturbance component, the corrected predicted tension is then checked for consistency with the safe range and system constraints. Once the consistency is verified, the predicted tension is used as the input for the main control adjustment.

4. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 3, characterized in that, The specific process for obtaining the current cable sag is as follows: A cable image with acceptable image quality and stable for several consecutive frames is acquired. Edge detection and morphological processing are then used. The edge detection is used to detect the edge contour of the cable in the image, and the morphological processing is used to connect the edge contour of the cable into continuous line segments, and then refine them into cable skeleton lines. The cable skeleton line is extracted as the geometric principal axis. The cable skeleton line is used to reflect the cable's direction, sag, and swing. The cable skeleton line is fitted into a catenary model. The catenary model is used to transform the discrete skeleton point cloud into a continuous curve that conforms to physical laws, thereby obtaining the cable's sag and swing amplitude. The sag is used to reflect the tension.

5. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 1, characterized in that, The process of visually correcting the winch's winding and unwinding speed using loosening and vibration suppression logic is as follows: When the predicted tension is within the safe range, image recognition of cable morphology is performed. The specific process of modifying the loosening logic is as follows: If the sag exceeds the set sag, it means the cable is too loose. Without touching the upper limit, take the cable in the third step until the sag is restored to within the set sag. The third step is slower than the first step. The second and third steps are both slower than the first step. There is no need to compare their sizes. Just make sure they are both slower than the first step. The specific process of modifying the vibration suppression logic is as follows: If the swing amplitude is greater than the set swing amplitude, it indicates that the cable has entered a vibration state. First, use the third step to reel in the cable and observe the swing amplitude within the set time, comparing it with the set amplitude variation range: If the swing amplitude is within the set amplitude variation range and the tension does not exceed the safe range, continue to take in the line in the third step in that direction; If the swing amplitude does not exceed the set amplitude variation range, and the tension increases towards the maximum value of the safe range, then the speed should be reduced.

6. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 1, characterized in that, The process of sending the corrected winch winding and unwinding speed down and synchronizing it with the UAV's speed adjustment system to adaptively coordinate with the winch winding and unwinding actions is as follows: The corrected winch winding and unwinding speed is sent to the winch controller, and the winding and unwinding actions are performed according to the corrected winch winding and unwinding speed. At the same time, the timestamp of the command is recorded and synchronized with the timing of the UAV. After receiving synchronization information from the winch, the UAV speed regulation system verifies the integrity of the speed parameters and processes the timing consistency. By combining the drone's current flight speed, its spatial position relative to the winch, and cable shape recognition data, speed adjustment requirements adapted to the winch's movements are generated.

7. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 6, characterized in that, The process of issuing the corrected winch winding and unwinding speed also includes evaluating the current cable status, specifically as follows: After the cable winding and unwinding speed adjustment is issued and executed, the increment of safety margin, the decrease of sag, and the decrease of swing energy within the set time are obtained. After normalization, they are weighted and fused to obtain the comprehensive safety index. If the overall safety index is positive, it indicates that the risk of the current state has decreased. Therefore, the adjustment is deemed effective, and the current correction direction can be maintained with a further increase in the step size within the allowable range. If the overall safety index is negative, it indicates that the current risk level has increased. In this case, the adjustment is deemed invalid and the correction adjustment in that direction should be cancelled in a timely manner.

8. The anti-interference control method for tethered UAV cable deployment and retrieval based on image recognition according to claim 6, characterized in that, The specific process for generating the speed adjustment requirements adapted to the winch motion is as follows: If the winch performs the cable winding action, control the drone to adjust its flight position in the direction close to the winch at a speed that matches the winch's cable winding rate; If the winch performs the cable release action, control the drone to adjust its flight position in a direction away from the winch at a speed that matches the winch's cable take-up rate, and maintain the drone's flight speed fluctuation; If the cable is detected to be swinging laterally, the lateral speed of the drone is adjusted synchronously, and the winch vibration suppression logic is used to suppress the swing.

9. An anti-interference control system for tethered UAV cable reeling and deployment based on image recognition, employing the anti-interference control method for tethered UAV cable reeling and deployment based on image recognition as described in any one of claims 1-8, characterized in that, include: The module includes a data acquisition and alignment module, a tension prediction module, a take-up and release speed adjustment module, a visual correction module, and a module for issuing and executing commands and synchronizing with the drone. The data acquisition and alignment module is used to acquire the temporal characteristics of timestamp alignment within a set time period when the UAV is performing aerial operations, and to project the wind speed along the cable direction to obtain the wind disturbance component along the cable. The tension prediction module is used to input the real-time measured tension, winch speed and wind disturbance component into the tension prediction model and output the predicted tension for the next moment. The winding and unwinding speed adjustment module is used to compare the predicted tension with the safe range and adjust the winding and unwinding speed of the winch in real time. The visual correction module is used to combine image recognition of cable shape and visually correct the winding and unwinding speed of the winch with loosening logic and vibration suppression logic. The module for issuing and executing the command and synchronizing with the UAV is used to issue and execute the corrected winch speed and synchronize the target speed to the UAV. After receiving the target speed command, the UAV actively adjusts its flight attitude to adaptively cooperate with the winch's winding and unwinding actions.

Images