An unmanned aerial vehicle anti-disturbance flight control method for multi-modal insulator detection

Abstract

The application discloses a UAV anti-disturbance flight control method for multi-modal insulator detection, and relates to the technical field of intelligent inspection of unmanned aerial vehicle power transmission lines; electric field information is acquired through an all-fiber space electric field sensor; when the electric field intensity gradient exceeds a threshold value, the weight of a magnetic compass in fusion calculation is adaptively reduced to obtain real attitude data resistant to electromagnetic interference; based on an extended state observer and a self-disturbance control algorithm, wind deviation and other system total disturbances are estimated and compensated to control the UAV to stably hover; in the stable hovering state, the data acquisition of a vision camera and the all-fiber space electric field sensor is triggered synchronously by a flight control main clock, the hovering position and the real attitude data are associated with the vision image and a one-dimensional electric field waveform signal, and spatiotemporal alignment multi-modal inspection data corresponding to the stable state of the flight control is obtained, so that image motion blur and spatiotemporal misalignment of multi-modal data are reduced.

Description

Technical Field

[0001] This application relates to the field of intelligent inspection technology of UAV power transmission lines, and in particular to a UAV anti-disturbance flight control method for multi-mode insulator detection. Background Technology

[0002] Insulators are the core components of transmission lines, providing electrical insulation and mechanical support. Their operating condition directly affects the safety and stability of the power grid. Due to long-term exposure to complex outdoor environments, insulators are not only prone to external physical defects such as damage and cracks, but also to internal, hidden electrical degradation such as zero values ​​and partial discharges. If these are not investigated and addressed in a timely manner, they can easily lead to line flashovers and large-scale power outages.

[0003] In recent years, with the rapid development of drone and artificial intelligence technologies, using drones equipped with visual sensors for inspection has become mainstream. However, visual images alone can only detect external physical damage to insulators and are difficult to detect internal electrical degradation defects. Therefore, multimodal detection technology that integrates visual images with spatial electric field signals has become an important direction for comprehensive health assessment of insulators, and high-quality, spatiotemporally aligned multimodal data acquisition is the fundamental prerequisite for realizing this technology.

[0004] However, UAVs designed for multimodal insulator inspection face extremely harsh physical environments in actual high-voltage inspection operations. Existing flight control systems and data acquisition mechanisms are insufficient to meet the demands for high-quality multimodal data acquisition. While some attempts have been made to obtain accurate data, they still face the following serious challenges in practical applications: First, strong alternating electromagnetic fields exist around high-voltage transmission lines. When a UAV approaches an insulator at close range, traditional airborne magnetic compasses are easily affected by electromagnetic distortion, causing divergence in the UAV's heading angle and attitude calculations. Second, high-altitude environments in the field are often accompanied by complex and variable gusts and shear airflow. Due to the lack of dynamic compensation mechanisms for unknown external wind deviation disturbances, existing UAVs experience high-frequency shaking and displacement when hovering for filming. This lack of anti-disturbance capability not only leads to severe motion blur in the acquired visual images, significantly degrading the performance of advanced backend visual target detection models, but also introduces a fatal data misalignment problem. Multimodal detection requires extremely high spatiotemporal alignment accuracy of data. Existing systems mostly rely on software system timestamps for asynchronous post-processing alignment. Under conditions of severe shaking of the machine and system processing delays, spatiotemporal misalignment at the millisecond level or more can easily occur, causing the one-dimensional high-frequency electric field signal and the two-dimensional visual target to be unable to be accurately mapped in spatial physical location, directly undermining the reliability of the back-end multimodal joint decision.

[0005] In summary, existing UAV flight control technologies cannot guarantee extremely high-precision stable hovering of UAVs in strong electromagnetic and complex wind deflection environments. Furthermore, they lack a multi-source data absolute synchronization mechanism at the underlying hardware level, resulting in spatial misalignment and motion ambiguity in the acquired inspection data. There is an urgent need for an anti-disturbance flight control method that can systematically solve the aforementioned problems of electromagnetic disturbances, wind deflection disturbances, and data spatiotemporal misalignment. Through a technical closed loop of anti-disturbance sensing, anti-disturbance hovering, and hardware synchronous acquisition, a high-quality data source guarantee can be provided for the multi-modal comprehensive detection of insulators. Summary of the Invention

[0006] To address the technical problems of existing UAV flight control technology, which cannot guarantee extremely high-precision stable hovering under strong electromagnetic and complex wind deflection environments, and lacks a hardware-level multi-source data absolute synchronization mechanism, resulting in spatial misalignment and motion ambiguity in the acquired inspection data, this invention proposes a UAV anti-disturbance flight control method for multimodal insulator detection. This method utilizes a cross-layered collaborative closed loop consisting of anti-electromagnetic interference attitude calculation, anti-wind deflection precise hovering, and hardware-level synchronous triggering. It provides extremely clean attitude through electric field gradient adaptive filtering, achieves physical-level anti-shake through an extended state observer and active disturbance rejection control, and uses hardware pulse forced alignment of exposure and sensor signals, providing progressive support. Ultimately, this provides unambiguous and spatiotemporally aligned multimodal inspection data for insulator defect detection that corresponds to the stable state of the UAV flight control.

[0007] To solve the technical problem, the technical solution of the present invention is as follows: A disturbance-resistant flight control method for UAVs for multimodal insulator detection includes the following steps: Step 1: Obtain true attitude data resistant to electromagnetic interference based on multi-source sensor fusion, wherein the multi-source sensors include a magnetic compass and a full-fiber spatial electric field sensor based on the Pockels effect. When the electric field intensity gradient measured by the full-fiber spatial electric field sensor exceeds a threshold, the weight of the magnetic compass in the fusion solution is adaptively reduced to obtain the true attitude data. Step 2: Using the real attitude data obtained in Step 1 as state feedback information, the extended state observer is used to estimate the total system disturbance, including wind deflection. Based on the active disturbance rejection control algorithm, the compensation amount is output according to the total system disturbance, and motor control commands are generated to control the UAV to enter the wind-resistant stable hovering state and determine the hovering position of the UAV. Step 3: After entering the wind-resistant stable hovering state in Step 2, the master clock of the flight control system generates a global synchronization trigger pulse, which is simultaneously sent to the vision camera and the all-fiber spatial electric field sensor. The vision camera is controlled to perform global exposure to obtain a visual image and determine the exposure time window of the vision camera. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse, and one-dimensional electric field waveform data is continuously cached through hardware FIFO buffer. According to the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by hardware FIFO. The hovering position of the UAV and the actual attitude data corresponding to the acquisition time are associated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current flight control stable hovering state.

[0008] Furthermore, step 1 includes: The system acquires the one-dimensional electric field sequence measured by the all-fiber space electric field sensor in real time, and calculates the electric field intensity gradient based on the one-dimensional electric field sequence. ; When the electric field intensity gradient Greater than the set safety threshold At that time, calculate the adaptive penalty coefficient. The adaptive penalty coefficient With maximum amplitude value The adaptive penalty coefficient When the maximum amplitude value is reached Previously, with the electric field intensity gradient The increase is not a nonlinear monotonically increasing phenomenon; Using the adaptive penalty coefficient The measurement noise covariance matrix in the Kalman filter is updated, the noise variance of the corresponding magnetic compass measurement item in the measurement noise covariance matrix is ​​amplified, the weight of the magnetic compass in the fusion solution is reduced, and the fusion solution is performed based on the updated measurement noise covariance matrix to obtain the true attitude data. Wherein, when the adaptive penalty coefficient Exceeding the preset maximum amplitude value At that time, the adaptive penalty coefficient will be... Restricted to ; The adaptive penalty coefficient Calculate according to the following formula:

[0009] Where, min It is a function that takes the minimum value. This is the sensitivity adjustment coefficient, with a value ranging from 0.5 to 3.0, preferably 1.5; The maximum penalty limit value ranges from 5.0 to 20.0, with 10.0 being the preferred value. The electric field intensity gradient is calculated as the ratio of the difference in electric field intensity at adjacent sampling times to the sampling time interval. The safety threshold is determined based on electric field gradient samples collected within the first 10 seconds after the UAV takes off under conditions of no significant electromagnetic interference. Specifically:

[0010] in, The mean of the electric field gradient samples under undisturbed conditions. The standard deviation of the electric field gradient sample under interference-free conditions is denoted by k, which is a threshold amplification factor ranging from 3 to 5, preferably 3. The above values ​​are based on the fact that when k is between 3 and 5, it can effectively cover normal background electric field fluctuations and reduce the probability of false triggering caused by random noise. When the value is between 0.5 and 3.0, the adaptive penalty coefficient can maintain high sensitivity to sudden increases in electric field gradient, while avoiding abrupt changes in the measurement noise covariance matrix. When the value is between 5.0 and 20.0, the equivalent weight of the magnetic compass can be reduced to 1 / 20 to 1 / 5 of the original weight, which is preferable. When the value is 10.0, the minimum equivalent weight of the magnetic compass is about 10% of the original weight. This can suppress the influence of strong electromagnetic distortion on attitude calculation and avoid the instability of attitude calculation caused by completely discarding magnetic compass information.

[0011] Furthermore, step 2 includes: The aerodynamic wind deflection affecting drone hovering, the center of gravity shift caused by sensor mounting, and the internal nonlinear coupling are all considered as the total system disturbance. The total system disturbance is estimated in real time by using an extended state observer to obtain an estimated value of the total system disturbance. Based on the actual attitude data and the estimated total disturbance of the system, the disturbance compensation amount is calculated using an active disturbance rejection control algorithm. State feedback is performed based on the actual attitude data, and a basic control quantity is generated according to the state error between the expected hovering state and the actual hovering state of the UAV. The basic control quantity is then perturbed and compensated according to the perturbation compensation quantity to obtain the compensated control quantity. Based on the compensated control quantity, motor control commands are generated to maintain stable hovering of the UAV. The drone is controlled to enter a wind-resistant stable hovering state according to the motor control command, and the hovering position of the drone is determined.

[0012] Furthermore, step 3 includes: After the UAV enters the wind-resistant stable hovering state, the hardware timer of the flight control system generates a global synchronization trigger pulse as the master clock; The global synchronization trigger pulse is simultaneously sent to the external trigger interface of the vision camera and the acquisition module of the all-fiber spatial electric field sensor via a dedicated parallel hardline for transmitting the trigger pulse. The global synchronization trigger pulse controls the vision camera to start global exposure, obtain a visual image, and determine the exposure time window of the vision camera according to the exposure start time and exposure duration. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse, and the one-dimensional electric field waveform data is continuously cached through hardware FIFO buffer. Based on the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by the hardware FIFO. The hovering position and actual attitude data of the UAV corresponding to the acquisition time are correlated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current stable hovering state of flight control.

[0013] A disturbance-resistant flight control system for UAVs designed for multimodal insulator detection includes: An anti-electromagnetic interference attitude calculation unit is used to obtain real attitude data against electromagnetic interference based on multi-source sensor fusion. The multi-source sensors include a magnetic compass and a pockmarked all-fiber spatial electric field sensor. When the electric field intensity gradient measured by the all-fiber spatial electric field sensor exceeds a threshold, the anti-electromagnetic interference attitude calculation unit adaptively reduces the weight of the magnetic compass in the fusion calculation to obtain the real attitude data. The active disturbance rejection hovering control unit is used to receive the real attitude data output by the anti-electromagnetic interference attitude calculation unit, use an extended state observer to estimate the total system disturbance including wind deflection, and generate motor control commands based on the active disturbance rejection control algorithm according to the total system disturbance to control the UAV to enter the wind deflection-resistant stable hovering state and determine the hovering position of the UAV. The hardware-level synchronous acquisition unit is used to generate a global synchronization trigger pulse by the main control clock of the flight control system after the UAV enters the wind-resistant stable hovering state. The global synchronization trigger pulse is sent to the vision camera and the all-fiber spatial electric field sensor at the same time. The vision camera is controlled to perform global exposure to obtain a visual image and determine the exposure time window of the vision camera. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse. One-dimensional electric field waveform data is continuously cached through hardware FIFO. According to the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by hardware FIFO. The hovering position of the UAV and the actual attitude data corresponding to the acquisition time are associated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporal aligned multimodal inspection data corresponding to the current flight control stable hovering state. The output of the anti-electromagnetic interference attitude calculation unit is connected to the input of the self-disturbance rejection hovering control unit. The self-disturbance rejection hovering control unit determines whether the UAV has entered the wind-resistant stable hovering state and uses the determination result as the synchronous acquisition trigger condition of the hardware-level synchronous acquisition unit.

[0014] Furthermore, the anti-electromagnetic interference attitude calculation unit includes: The electric field gradient calculation module is used to calculate the electric field intensity gradient based on the one-dimensional electric field sequence measured by the all-fiber space electric field sensor. ; The penalty coefficient calculation module is used when the electric field intensity gradient... Greater than the set safety threshold At that time, calculate the adaptive penalty coefficient. The adaptive penalty coefficient With maximum amplitude value The adaptive penalty coefficient When the maximum amplitude value is reached Previously, with the electric field intensity gradient The increase is not a nonlinear monotonically increasing phenomenon; The covariance update module is used to utilize the adaptive penalty coefficient. Update the measurement noise covariance matrix in the Kalman filter, amplify the noise variance of the corresponding magnetic compass measurement item in the measurement noise covariance matrix, and reduce the weight of the magnetic compass in the fusion solution; The attitude fusion calculation module is used to perform fusion calculation based on the updated measurement noise covariance matrix to obtain the true attitude data.

[0015] Furthermore, the self-disturbance-resistant hovering control unit includes: The extended state observer module is used to unify the aerodynamic wind deflection, center of gravity shift caused by sensor mounting, and internal nonlinear coupling that affect the hovering of the UAV as the total system disturbance, and to estimate the total system disturbance in real time to obtain the estimated value of the total system disturbance. The disturbance compensation control module is used to calculate the disturbance compensation amount using an active disturbance rejection control algorithm based on the actual attitude data and the estimated total disturbance of the system, perform state feedback based on the actual attitude data, generate a basic control quantity based on the state error between the expected hovering state and the actual hovering state of the UAV, perform disturbance compensation on the basic control quantity based on the disturbance compensation amount to obtain the compensated control quantity, generate motor control commands for maintaining stable hovering of the UAV based on the compensated control quantity, and control the UAV to enter the wind-resistant stable hovering state based on the motor control commands. The hovering position determination module is used to determine the hovering position of the UAV after the UAV enters the wind-resistant stable hovering state.

[0016] Furthermore, the hardware-level synchronous acquisition unit includes: The master clock trigger module is used to generate a global synchronization trigger pulse by the hardware timer of the flight control system after the UAV enters the wind-resistant stable hovering state; The synchronous acquisition control module is used to simultaneously send the global synchronous trigger pulse to the external trigger interface of the vision camera and the acquisition module of the all-fiber spatial electric field sensor via a dedicated parallel hardline for transmitting the trigger pulse. It controls the vision camera to perform global exposure, obtain a visual image, and determine the exposure time window of the vision camera. At the same time, it triggers the all-fiber spatial electric field sensor to record the trigger time corresponding to the global synchronous trigger pulse, and continuously caches one-dimensional electric field waveform data through a hardware FIFO buffer. Based on the trigger time and the exposure time window, it locks and extracts the corresponding one-dimensional electric field waveform signal from the one-dimensional electric field waveform data cached in the hardware FIFO. The data association module is used to associate the hovering position and the actual attitude data of the UAV at the time of acquisition with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current stable hovering state of flight control. The all-fiber spatial electric field sensor includes a hardware FIFO buffer, which is used to continuously buffer the one-dimensional electric field waveform data sampled by the all-fiber spatial electric field sensor, and lock the one-dimensional electric field waveform data corresponding to the exposure time window according to the trigger time corresponding to the global synchronization trigger pulse and the exposure time window.

[0017] An electronic device, comprising: One or more processors; Memory; and computer programs stored in said memory; When the computer program is executed by the one or more processors, it causes the electronic device to perform any one of the above-described methods for anti-disturbance flight control of unmanned aerial vehicles (UAVs) for multi-mode insulator detection.

[0018] A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements, as described above, a UAV anti-disturbance flight control method for multimodal insulator detection.

[0019] This application has the following advantages: This application constructs a technical solution combining anti-electromagnetic interference attitude calculation, wind-resistant precise hovering, and hardware-level synchronous triggering. Through the coordinated operation of these three components, reliable attitude data is first ensured for the UAV in a strong electromagnetic environment. Then, stable hovering is maintained in a complex wind-resistant environment. Finally, hardware-level synchronous acquisition of visual images and electric field waveform signals is completed within the stable hovering window. The UAV hovering position and actual attitude data corresponding to the acquisition time are correlated with the visual image and one-dimensional electric field waveform signal, thereby effectively reducing image motion blur and spatiotemporal misalignment of multimodal data.

[0020] This application introduces an adaptive Kalman filter mechanism triggered by electric field intensity gradient. When the UAV approaches a high-voltage transmission line and detects an anomaly in the electric field gradient, the system adaptively reduces the weight of the magnetic compass in the attitude fusion solution and avoids filter divergence under extreme conditions through a maximum penalty amplitude limiting mechanism, thereby improving the stability and reliability of attitude solution in strong electromagnetic environments.

[0021] This application employs an extended state observer and an active disturbance rejection control algorithm to estimate and compensate for wind deflection, load center of gravity shift, and internal nonlinear coupling as a unified total system disturbance. This enables the UAV to maintain stable hovering under complex airflow conditions in the field, providing stable data acquisition conditions for the visual camera and the all-fiber space electric field sensor.

[0022] This application employs a flight controller master clock to generate a global synchronization trigger pulse, synchronously controlling the global exposure of the vision camera and locking the electric field waveform of the all-fiber spatial electric field sensor, unlike post-processing alignment methods that rely on software timestamps. This hardware-level synchronization mechanism enables the exposure time window to correspond with the one-dimensional electric field waveform signal at the source of acquisition, improving the spatiotemporal consistency of multimodal data.

[0023] Therefore, this application can provide a stable, synchronous, and reliable data foundation for multimodal comprehensive detection of external physical defects and internal electrical degradation defects of insulators in high-voltage transmission line inspection scenarios where strong electromagnetic interference and complex wind deflection disturbances coexist, and has high engineering application value. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating the UAV anti-disturbance flight control method for multi-mode insulator detection provided in this application embodiment; Figure 2 A schematic diagram of the module structure of an unmanned aerial vehicle (UAV) anti-disturbance flight control system for multimodal insulator detection provided in an embodiment of this application; Figure 3 This is a diagram of the electromagnetic interference-resistant attitude calculation control logic based on adaptive Kalman filtering proposed in this application embodiment; Figure 4 This is a block diagram illustrating the principle of active disturbance rejection precise hovering control based on an extended state observer proposed in an embodiment of this application. Figure 5 This is a timing diagram illustrating the multimodal data hardware-level spatiotemporal synchronization triggering proposed in an embodiment of this application; Figure 6 A flowchart illustrating an optional joint decision logic for back-end insulator defects, as proposed in an embodiment of this application; Figure 7 The following are comparison diagrams of the transient response of the UAV under the combined disturbance proposed in this embodiment of the application; wherein, (a) is the transient response diagram of yaw angle, (b) is the transient response diagram of lateral displacement, and (c) is the comparison diagram of control output smoothness. Figure 8 This is a comparison diagram of the three-dimensional spatial trajectory of the UAV under extreme combined disturbances proposed in this embodiment of the application. Detailed Implementation

[0026] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] Example 1: This application provides a method for anti-disturbance flight control of unmanned aerial vehicles (UAVs) for multi-mode insulator detection, the overall process of which is as follows: Figure 1As shown. During insulator inspection, the UAV first acquires multi-source sensing data and performs anti-electromagnetic interference attitude calculation based on the spatial electric field intensity gradient; then, it uses an extended state observer to estimate the total system disturbance and achieves stable hovering through active disturbance rejection control; after the UAV enters a stable hovering state, the flight control master clock generates a global synchronization trigger pulse to control the visual camera and the all-fiber spatial electric field sensor to perform hardware-level spatiotemporal synchronization acquisition. Specifically, the steps include: Step S1: Obtain real attitude data with electromagnetic interference resistance by fusing multiple source sensors based on adaptive Kalman filtering.

[0028] Specifically, this embodiment constructs a quaternion-based... The nonlinear attitude dynamics model, in which, The scalar part (real part) of the quaternion represents the cosine component of the rotation angle; The vector part (imaginary part) of the quaternion represents the three components in the direction of the rotation axis.

[0029] Let the state vector of the system be ,in, For the random walk zero bias of the gyroscope, This is the transpose. The system's measurement vector is... Where a and m are the triaxial measurements from the airborne accelerometer and magnetic compass, respectively. This is a transpose.

[0030] The conventional Extended Kalman Filter (EKF) calculates the Kalman gain matrix. At that time, its measurement noise covariance matrix R is a constant matrix. When the UAV approaches the high-voltage cable, the strong magnetic field causes a large observation error in the magnetic compass measurement equation. To address this, this embodiment introduces an all-fiber space electric field sensor based on the Pockels effect. In terms of hardware topology, the front-end probe of this sensor is fixed to the extension of the UAV arm using a metal-free all-dielectric material, and its rear-end photoelectric detection module is connected to the serial port receiving pin of the flight control board via an RS422 high-speed serial port, transmitting the one-dimensional electric field sequence E(t) back in real time at a frequency of 100Hz.

[0031] In the flight control main loop, the system calculates the electric field intensity gradient in real time. ,in, For the current moment The instantaneous value of the measured electric field intensity; For the previous sampling time The instantaneous value of the electric field intensity; The sampling time interval; To avoid false triggering caused by a fixed threshold, a safety distortion threshold is used. Initialization is performed using a sliding window dynamic baseline method, which involves sampling the electric field background noise in an interference-free environment for the first 10 seconds after the UAV takes off, and then tuning the baseline. .like Figure 3 As shown, when At that time, the system constructed an exponential adaptive regulation law with a limit:

[0032] in, This is the sensitivity adjustment coefficient. The maximum penalty limit is set to prevent the matrix condition number from deteriorating. In this embodiment, the maximum penalty limit is preferably set to 10.0.

[0033] In the measurement update phase of the filtering process, this embodiment decouples and dynamically reconstructs the measurement noise covariance matrix R. Let the initial measurement noise matrix be:

[0034] in, This is the initial measurement noise covariance matrix, used to characterize the uncertainty of the measurement data from each sensor; These represent the accelerometer readings at... Measurement noise variance in three directions These represent the magnetic compass at... Measurement noise variance in three directions. The larger the variance value, the higher the uncertainty of the corresponding sensor measurement results, and the lower its weight in the Kalman filter fusion solution; the smaller the variance value, the higher the reliability of the corresponding sensor measurement results, and the higher its weight in the fusion solution.

[0035] The updated dynamic covariance matrix is ​​then rewritten as:

[0036] in, For the first The updated dynamic measurement noise covariance matrix at each sampling time point. (Compared to the initial matrix) compared to, Only the three-axis noise variance corresponding to the magnetic compass is amplified, while the three-axis noise variance corresponding to the accelerometer remains unchanged. This is because the alternating electromagnetic field near high-voltage transmission lines primarily affects the magnetic compass measurement results, while its direct impact on the accelerometer's measurement items is relatively small. Therefore, this embodiment uses an adaptive penalty coefficient. The noise variance of the magnetic compass is amplified, that is... Updated to This allows for dynamic adjustment of the reliability of magnetic compass measurements. This indicates that a diagonal matrix is ​​constructed using the elements within the parentheses as the main diagonal elements, with all off-diagonal elements being 0. The use of a diagonal matrix form signifies that the measurement channels are independent of each other; that is, the accelerometer triaxial measurement and the magnetic compass triaxial measurement participate in the fusion calculation according to their respective noise variances during the filtering update process.

[0037] Subsequently, the new information covariance matrix is ​​calculated:

[0038] in, The information covariance matrix is ​​used to characterize the degree of uncertainty between the predicted and actual measurements. For the first The measurement matrix or measurement Jacobian matrix at each sampling time is used to describe the mapping relationship between system state quantities and sensor measurements; For the first The prior state estimation error covariance matrix at each sampling time is used to characterize the uncertainty of the state prediction result; for The transpose of the matrix; This is the dynamic measurement noise covariance matrix.

[0039] The Kalman gain matrix is ​​calculated as follows:

[0040] in, This is the Kalman gain matrix, used to determine the degree of correction that the current measurement data makes to the state estimation results; This is the inverse of the new information covariance matrix. When the electric field intensity gradient... Increase and exceed the safety threshold When, adaptive penalty coefficient As it increases, The noise variance of the corresponding three-axis measurement terms of the magnetic compass increases synchronously, thereby affecting the new information covariance matrix. The uncertainty of related measurement channels of the magnetic compass increases. Due to... and Related, The greater the uncertainty corresponding to the magnetic compass channel, the greater the Kalman gain matrix. The smaller the correction weight of the corresponding magnetic compass measurement, the better. Therefore, when a strong electromagnetic distortion risk is detected, the system can automatically reduce the influence of magnetic compass data in attitude fusion calculation, avoiding direct divergence in heading angle and attitude calculations caused by abnormal magnetic field measurements.

[0041] like Figure 3As shown, IMU data and magnetic compass data first enter the state prediction, i.e., time update stage, to obtain the prior state estimate and prior error covariance at the current sampling time; simultaneously, the all-fiber space electric field sensor continuously acquires space electric field data and calculates the electric field intensity gradient. .when At that time, the system calculates the adaptive penalty coefficient. and according to Increasing the noise variance corresponding to the magnetic compass measurement terms yields the updated dynamic measurement noise covariance matrix. ;when At this time, no magnetic compass penalty update is performed, maintaining the normal measurement noise configuration. Subsequently, the state prediction results and the dynamic measurement noise covariance matrix are jointly input into the measurement update stage to calculate the innovation covariance matrix. and Kalman gain The system completes the fusion and calculation of IMU data and magnetic compass data, and finally outputs realistic attitude data that is resistant to electromagnetic interference.

[0042] This dynamic covariance reconstruction mechanism does not directly delete magnetic compass data, but rather reduces its reliability in the fusion solution by increasing the variance of magnetic compass measurement noise. When the electric field gradient recovers to a safe range... It can be restored to a lower level, and the magnetic compass measurement weight is restored accordingly, thus taking into account both robustness under strong electromagnetic interference and attitude calculation accuracy under normal inspection conditions.

[0043] Step S2: Perform precise hovering control of UAV against wind deflection based on active disturbance rejection control of extended state observer.

[0044] Specifically, after acquiring the actual attitude data, this embodiment constructs a position loop control law based on an Extended State Observer (ESO) to address the hovering and drifting problem of UAVs in complex wind fields. Taking the position state in the altitude direction as an example in the single-axis hovering dynamics model of the UAV, let... The second-order hovering dynamics equation represents the actual position of the drone in the altitude direction:

[0045] in, This refers to the actual position of the drone in the altitude direction. This refers to the actual speed of the drone in the altitude direction. The linear acceleration of the drone in the altitude direction; For motor control input, To control the gain, This is due to external gusts of wind. This includes the total disturbance caused by unmodeled system dynamics and external airflow shear.

[0046] like Figure 4 As shown, this embodiment expands the total disturbance into a new state variable. A third-order discrete nonlinear extended state observer (ESO) is constructed, and its state equations are derived as follows:

[0047] in, and These represent the state variables of the extended state observer. , and The first derivative with respect to time characterizes the rate of change of the position estimation state, the velocity estimation state, and the total disturbance estimation state, respectively. Estimate the error for the observer; This is an estimate of the system control gain, used to characterize the motor control input. The intensity of the effect on the acceleration of the drone's motion, and When using normalized control input, The value range is 0.8 to 1.2, preferably 1.0; when using actual dynamic parameters, It can be calibrated based on the drone's mass, the motor's maximum thrust, and the input-acceleration response under hovering conditions.

[0048] and These are the three state variables of the extended state observer, where, To determine the actual location The estimated value, To determine the actual speed The estimated value, To the total disturbance The real-time estimated value of the total system disturbance. Factors include external gust disturbances, airflow shear, center of gravity shifts caused by sensor mounting, unmodeled system dynamics, and internal nonlinear couplings. This is addressed by expanding the total disturbance into a state variable. It can transform wind deflection disturbances, which are originally difficult to measure directly, into state variables that can be estimated and compensated in real time. The observer estimation error is used to characterize the deviation between the position estimate and the actual measurement. , , These are the gain parameters of the extended state observer, used to adjust the convergence speed of position estimation, velocity estimation, and disturbance estimation, respectively. To reduce the complexity of independent tuning of multiple parameters, this embodiment uses a bandwidth parameterization method to determine the observer gain, i.e.:

[0049] in, The observer bandwidth ranges from 15 to 60 rad / s, with 30 rad / s being preferred. A larger observer bandwidth results in a faster ESO response to external disturbances, but excessive bandwidth amplifies measurement noise and causes control output jitter; conversely, insufficient bandwidth leads to delayed disturbance estimation, making it difficult to compensate for gusts in a timely manner. Therefore, considering the UAV hovering control cycle, sensor sampling frequency, and airframe disturbance rejection stability, a bandwidth of 30 rad / s is preferred. It is set to 30 rad / s to balance disturbance estimation speed and control smoothness. This is a nonlinear error feedback function used to employ approximate linear adjustment when the error is small and nonlinear adjustment when the error is large. Its expression is:

[0050] in, This is a non-linear exponential parameter, with a value range of 0.25 to 0.8; This is the linear interval width parameter, with a value ranging from 0.005 to 0.05, preferably 0.01. (Setting...) The purpose is to avoid high-frequency chattering or numerical singularities when the error is close to zero; setting The purpose is to enhance the system's ability to correct for large nonlinear errors. In this embodiment, The value ranges from 0.5 to 0.8, with 0.5 being the preferred value. The value ranges from 0.25 to 0.5, with 0.25 being preferred. This value ensures the convergence speed of the perturbation estimation while reducing the observer's sensitivity to measurement noise.

[0051] like Figure 4 As shown, the target desired position is given by the visual detection or trajectory planning module. First, the tracking differentiator TD is input, which smooths the desired position command and generates the corresponding desired position and velocity signals. The extended state observer ESO then adjusts the signal based on the actual hovering position of the UAV. and control input Real-time estimation of location status Speed ​​state and total system disturbance NLSEF calculates the initial control input based on the error between the desired state and the ESO estimated state. The total disturbance estimate obtained from ESO is then introduced into the control variable as a disturbance compensation term to obtain the final control input. The final control input is applied to the disturbance caused by external gusts of wind. The hovering dynamics model of the UAV, which does not have internal dynamic influences modeled, is used to obtain the actual hovering position and feed the actual position back to the ESO, thereby forming a closed-loop disturbance rejection control.

[0052] In the Nonlinear State Error Feedback (NLSEF) stage, the position error is calculated based on the set desired position v. With speed error Generate the initial control law:

[0053] in, The desired hovering position of the drone. The desired speed. When the drone performs a hovering mission, =0; When the drone is performing a low-speed tracking and inspection task. The desired velocity given for the target trajectory. Position tracking error represents the difference between the desired position and the observed position; The velocity tracking error represents the difference between the expected velocity and the observed velocity. This is the initial control law before disturbance compensation, used to generate basic control quantities based on position and velocity errors; , These are the position error feedback gain and the velocity error feedback gain, respectively. For ease of engineering tuning, this embodiment uses a closed-loop control bandwidth. Perform parameterized configuration, that is: ,

[0054] in, The closed-loop control bandwidth ranges from 3 to 10 rad / s, with 5 rad / s being preferred. Correspondingly, The value range is 9 to 100, preferably 25; The value range is 6 to 20, preferably 10. This value is based on the following: when... When the position error is too small, the convergence speed is slow, making it difficult to suppress hovering drift caused by gusts in time; when If the value is too large, the motor control commands change too rapidly, which can easily cause high-frequency vibration of the machine body. Therefore, a better approach is to... = 5rad / s, to achieve a balance between hovering accuracy and control smoothness.

[0055] Finally, by introducing disturbance feedforward compensation, the final control quantity output to the underlying power distribution matrix is:

[0056] in, This is the control quantity that is ultimately output to the underlying power distribution module; z1 is the initial control quantity calculated based on position error and velocity error; z2 is the total system disturbance estimated by ESO. This is the estimated control gain. This is a perturbation feedforward compensation term used to offset the effects of external gusts, center of gravity shifts, and unmodeled dynamics on the hovering state of the UAV.

[0057] Specifically, when external wind disturbance When the UAV undergoes a small displacement, the extended state observer can estimate the total disturbance z3 in real time through the change in position error, and introduce this disturbance estimate in the form of feedforward compensation. If the disturbance causes the drone to deviate in a certain direction, then the compensation term... This generates a control effect opposite to the direction of the disturbance, thereby reducing or offsetting the position drift caused by wind deflection. Therefore, this embodiment does not rely solely on position error feedback for hysteresis correction, but performs feedforward compensation immediately after the disturbance is estimated by the ESO, thereby improving the hovering stability of the UAV in complex airflow environments.

[0058] In this embodiment, the flight control cycle is preferably 10 ms, and the ESO updates the disturbance estimate once per control cycle. Combined with an algorithm processing lag of less than 12 ms, the system can complete the disturbance estimation and compensation output within approximately 10 to 30 ms. This range of values ​​can meet the real-time requirements for wind deflection, shake reduction, and stable imaging during close-range insulator inspection by UAVs.

[0059] Step S3: Based on hardware-level synchronous triggering of the flight controller's main clock, acquire sub-millisecond spatiotemporally aligned multimodal insulator inspection data. Specifically, this embodiment abandons the high-latency software system timestamp and instead utilizes a high-precision hardware timer on the flight controller motherboard as the global clock source. In the flight controller firmware, the interrupt priority of this timer is configured to the highest (preemption priority is 0). When the flight controller determines that the aircraft's position and attitude errors are both less than the set threshold, i.e., the UAV enters a stable hovering state, it triggers a hardware interrupt and synchronously outputs a TTL trigger pulse through a GPIO pin configured in push-pull output mode. This hardware pulse directly drives the external trigger interface of the vision camera through parallel hardwired connections, and simultaneously triggers the acquisition module of the all-fiber spatial electric field sensor. Figure 5 As shown, the flight control master clock synchronous trigger pulse, the visual camera global exposure time window, and the one-dimensional electric field waveform signal are all under a unified time reference. The exposure start time is determined by the synchronous trigger pulse, and the corresponding signal interval is extracted from the continuously acquired one-dimensional electric field waveform according to the exposure time window.

[0060] To achieve sub-millisecond precision alignment between the visual camera's exposure time window and the one-dimensional electric field waveform, this embodiment employs a dual mechanism combining hardware FIFO buffering and timestamp backtracking: On the one hand, the all-fiber space electric field sensor continuously samples at a fixed frequency (preferably 1kHz). Each sampling point is given a global timestamp by the hardware timer of the flight control master clock and stored in the hardware FIFO (first-in, first-out) buffer queue embedded in the sensor. On the other hand, when the GPIO pin outputs a TTL trigger pulse, the rising edge of this pulse simultaneously triggers two actions: the first directly drives the global shutter of the vision camera to open, and the camera records the exposure start timestamp T_exp_start of that frame; the second triggers the FIFO buffer freezing mechanism of the all-fiber spatial electric field sensor, freezing the data before and after the time T_exp_start. Electric field waveform data is locked within the time window.

[0061] Subsequently, the flight control main board reads the frozen FIFO buffer data through the SPI interface, and extracts the electric field waveform sequence within the time interval [T_exp_start, T_exp_start + T_shutter] based on the camera exposure time T_shutter (read from the camera register), as multimodal feature data that is precisely aligned with the image of that frame.

[0062] The above mechanism does not rely on operating system software timestamps, completely avoids thread scheduling delays and bus transmission congestion, ensures absolute physical alignment of heterogeneous sensor data on the time axis, and controls the time synchronization error within 500 microseconds (preferably less than 50 microseconds).

[0063] After step S3 is completed, multimodal data with absolute temporal alignment between the visual camera and the all-fiber spatial electric field sensor can be obtained. As an optional implementation, a back-end detection model can be used to process the above data. The following section combines... Figure 6 An exemplary implementation of a possible joint decision is provided. It should be noted that this method is not a necessary technical feature for achieving the anti-disturbance flight control objective of this invention.

[0064] like Figure 6 As shown, spatiotemporally aligned multimodal inspection data is first acquired, which includes a two-dimensional visual image and a one-dimensional electric field waveform signal corresponding to the exposure time window of the two-dimensional visual image. The two-dimensional visual image and the one-dimensional electric field waveform signal are then input into the visual image defect detection branch and the one-dimensional electric field waveform anomaly detection branch, respectively, for parallel processing.

[0065] In the visual image defect detection branch, a two-dimensional visual image is input into a deep learning-based target detection network. Through multi-level feature extraction and attention mechanisms, external physical defects such as damage and cracks of the insulator are detected, and the first defect confidence score and the corresponding bounding box position are output.

[0066] In the one-dimensional electric field waveform anomaly detection branch, the one-dimensional electric field waveform signal corresponding to the exposure time window of the two-dimensional visual image is processed. The partial discharge pulse or high-frequency distortion component in the one-dimensional electric field waveform signal is extracted using a time-series one-dimensional convolutional network or a threshold determination algorithm. The electrical degradation probability of the insulator is calculated, and the electrical degradation probability is output as the second defect confidence level.

[0067] The confidence levels of the first and second defects are input into the joint decision module. The two defect detection results are jointly judged by weighted fusion, logical combination or DS evidence theory to realize the comprehensive analysis of the external physical defect information and internal electrical degradation information of the insulator, and output the comprehensive health status report of the insulator.

[0068] By using the above-mentioned parallel detection and joint decision-making method, it is possible to identify the external physical defects of the insulator using two-dimensional visual images, and to characterize the internal electrical degradation state of the insulator using one-dimensional electric field waveform signals that are spatiotemporally aligned with the two-dimensional visual images, thereby achieving comprehensive detection of the physical and electrical states of the insulator.

[0069] For example, the actual inspection workflow of a UAV anti-disturbance flight control system for multimodal insulator detection provided in this application embodiment is as follows: The ground control station issues the insulator inspection route, the UAV takes off and flies according to the predetermined waypoint. During the approach to the high-voltage tower, the all-fiber spatial electric field sensor continuously monitors the spatial electric field. Once an abnormal jump in the electric field gradient is detected, the module M1 at the bottom layer of the flight control instantly triggers an adaptive Kalman filter to reduce the weight of the magnetic compass and prevent the aircraft from crashing or yawing due to magnetic declination distortion. When the target waypoint is reached directly in front of the insulator, the active disturbance rejection controller of module M2 quickly resists the body sway caused by high-altitude shear wind, so that the body always maintains high-precision attitude stability. Within microseconds after confirming extremely stable hovering, the hardware trigger pin of module M3 emits a square wave pulse to control the camera and the all-fiber spatial electric field sensor to synchronously perform instantaneous data acquisition. Subsequently, this absolutely spatiotemporally aligned multimodal data is sent to an edge computing device (such as NVIDIA Jetson NX) mounted on the UAV arm, where the back-end edge computing device performs feature extraction. Ultimately, the edge computing device only sends the coordinates and defect confidence levels of high-risk insulators that are identified as "damaged" or "discharging" to the ground station via the 4G / 5G link, completing an extremely high-quality and efficient closed-loop detection.

[0070] Example 2: This embodiment provides an application scenario in which the aforementioned UAV anti-disturbance flight control method for multimodal insulator detection is applied. Specifically, the implementation process includes the following: First, the experimental environment and benchmark settings were established, constructing a simulation platform that incorporates the six-degree-of-freedom closed-loop dynamics of the UAV. To closely resemble real industrial inspection scenarios, the integral step size of the control system was set to 10ms, and a 12ms algorithm inference lag term was forcibly injected to simulate the latency of real airborne edge computing hardware. Considering the characteristics of the ultra-high voltage transmission environment, a composite extreme disturbance was introduced: on the one hand, a continuous wind field with a base wind speed of 0.5m / s was generated using the Dryden turbulence model, and nonlinear shear gusts were superimposed as aerodynamic disturbances; on the other hand, within a specific time window of 20 to 40 seconds, alternating electromagnetic distortion that decays exponentially with relative distance was injected to serve as non-Gaussian information interference.

[0071] To demonstrate the superiority of this invention, the following classic and cutting-edge control methods were selected as comparative baselines: conventional cascaded control (Cascade-PID), passive adaptive filter control (IAEKF-PID), and sliding mode variable structure control (EKF-SMC). The overall performance comparison results are shown in Table 1.

[0072] The following industry-standard quantitative indicators are used for evaluation: Root Mean Square Error (RMSE): Unit: degrees (°), used to evaluate the purity of attitude calculation under extreme geomagnetic storms, the smaller the value, the better; Maximum lateral deviation: Unit: meters (m), used to quantify the transient obstacle avoidance safety boundary and wind resistance and anti-drift capability of UAV under the impact of strong shear gusts. The smaller the value, the better. Control energy consumption integral: used to quantitatively evaluate the smoothness of the work done by the underlying actuator; an excessively large value indicates severe high-frequency mechanical chattering.

[0073] Table 1 Overall Performance Comparison Results

[0074] The comparative experimental results show that the method of this invention achieves the best overall performance in key indicators under multimodal composite disturbances. Regarding antimagnetic accuracy, the yaw angle RMSE of the method of this invention is reduced to 0.7380°, while the error of traditional cascade control and sliding mode control both exceed 2.6°, indicating that this invention is significantly superior to both. In terms of wind resistance and drift prevention, the method of this invention strictly controls the maximum lateral deviation to 0.2901m, a reduction of approximately 56.9% compared to 0.6738m for sliding mode control, and a significant reduction of approximately 80.3% compared to 1.4763m for traditional cascade control. Regarding control smoothness, although the energy consumption integral value of the method of this invention is 0.1420, slightly higher than the comparative algorithm, this is achieved through reasonable dynamic compensation to obtain extremely high trajectory stiffness, effectively avoiding the high-frequency mechanical chatter inherent in sliding mode control algorithms—which often manifests as severe sawtooth fluctuations on a three-dimensional trajectory—thus providing an extremely stable working environment for the airborne vision gimbal. The above results fully demonstrate the effectiveness, feasibility, and engineering applicability of the overall architecture of this invention.

[0075] To verify the necessity of the various innovative modules in this invention, namely the electric field feedforward adaptive filtering module S1 in the information layer and the extended state observer feedforward anti-disturbance module S2 in the physical layer, an ablation experiment was conducted, and the experimental results are shown in Table 2.

[0076] Table 2 Comparison of ablation experimental data

[0077] The analysis will be conducted using yaw angle RMSE and maximum lateral deviation as examples: In this embodiment, A, B, and C in the ablation experiment are temporary codes: A refers to the complete method (architecture of this invention, including S1+S2), B refers to adaptive filtering only (S1-Only, S2 removed), and C refers to feedforward disturbance rejection only (S2-Only, S1 removed).

[0078] (1) Comparison between A and B (removal of physical layer feedforward disturbance rejection mechanism): After removing the extended state observer, the maximum lateral deviation deteriorated sharply from 0.2901m to 1.4712m. This result shows that relying solely on the clean attitude at the front end cannot resist the impact of the non-steady gusts at the bottom layer, and the feedforward disturbance compensation module of the physical layer is the key to ensuring the anti-drift hovering of the UAV at high altitude.

[0079] (2) Comparison between A and C (removal of the adaptive filtering mechanism of the information layer): After removing the adaptive filtering triggered by the electric field, although the bottom anti-interference module controls the lateral deviation to 0.3057m, the RMSE of the yaw angle calculation quickly diverges to 2.7160°, and the bottom controller generates severe mechanical chattering during this process. This shows that the unidirectional empowerment of the front-end ultra-clean attitude is the core prerequisite for cutting off the penetration of magnetic storm distortion into the physical dynamics equation.

[0080] (3) Comprehensive analysis: The complete model achieved excellent results in both electromagnetic divergence resistance and wind drift resistance, and the long-term yaw drift slope remained stable at 0.0157. This indicates that there is a significant positive synergistic effect between the information perception layer and the physical control layer, and each module is indispensable, which significantly improves the robustness of the UAV in complex ultra-high pressure environments.

[0081] This embodiment also proposes a typical complex disturbance scenario and a visualization analysis process for the control time domain, providing the time domain transient response and the three-dimensional space flight trajectory, respectively.

[0082] from Figure 7 It can be seen that the method of the present invention has good transient disturbance rejection performance under combined disturbance conditions. For example... Figure 7 As shown in (a), when the strong alternating magnetic distortion enters the effective range, the yaw angle error of the traditional cascaded control exhibits significant oscillations. The adaptive filtering method suffers from a certain response lag, while the method of this invention can promptly reduce the impact of abnormal magnetic compass measurements on attitude calculation, keeping the yaw angle error at a low level. Figure 7 As shown in (b) of the diagram, within the range of non-stationary shear gusts, traditional cascaded control produces significant lateral displacement accumulation, while sliding mode control can suppress lateral drift to some extent. The method of this invention exhibits smaller lateral displacement and faster recovery speed, demonstrating better wind deflection resistance. Figure 7 As shown in (c), the control output of sliding mode control exhibits significant high-frequency sawtooth fluctuations, while the control output of the method of this invention is continuous and relatively smooth, reducing high-frequency control chattering while ensuring anti-interference capability. Figure 7 The gray area in (c) is the gust disturbance area.

[0083] from Figure 8 As can be seen, in terms of three-dimensional spatial trajectory tracking, the spatial flight trajectory of the method of this invention is smooth and stable, accurately following the reference path. In contrast, traditional cascaded control exhibits obvious lateral bulge divergence (maximum lateral deviation exceeding 1.4m), while sliding mode control, although not exhibiting lateral divergence, results in significant high-frequency sawtooth jitter in both altitude and lateral directions due to discontinuous control inputs. This directly verifies that the method of this invention achieves a balance between high precision and "jatter-free" performance.

[0084] Finally, simulation analysis was conducted on the system's real-time performance and control lag tolerance. The cross-layer collaborative algorithm proposed in this invention, under harsh experimental conditions including a 12ms system hardware lag and a 10ms discrete sampling period, maintained a stable response in its single closed-loop operation and compensation mechanism. These simulation results demonstrate that the algorithm's processing efficiency and compensation capability are theoretically sufficient to handle the evolution cycle of high-frequency aerodynamic disturbances. This preliminarily verifies that the technical solution of this invention can meet the hard real-time control requirements of close-range inspection in ultra-high pressure field applications, and possesses good potential for industrial application.

[0085] Example 3: like Figure 2 As shown, this embodiment proposes a UAV anti-disturbance flight control system for multi-mode insulator detection. The system is applied to a UAV anti-disturbance flight control method for multi-mode insulator detection. The system includes: An anti-electromagnetic interference attitude calculation unit is used to obtain real attitude data against electromagnetic interference based on multi-source sensor fusion. The multi-source sensors include a magnetic compass and a pockmarked all-fiber spatial electric field sensor. When the electric field intensity gradient measured by the all-fiber spatial electric field sensor exceeds a threshold, the anti-electromagnetic interference attitude calculation unit adaptively reduces the weight of the magnetic compass in the fusion calculation. The active disturbance rejection hovering control unit is used to receive the real attitude data output by the anti-electromagnetic interference attitude calculation unit, use an extended state observer to estimate the total system disturbance including wind deflection, and generate motor control commands based on the active disturbance rejection control algorithm according to the total system disturbance output compensation amount to control the UAV to enter the wind deflection-resistant stable hovering state. The hardware-level synchronous acquisition unit is used to generate a global synchronization trigger pulse by the master control clock of the flight control system after the UAV enters the wind-resistant stable hovering state. The global synchronization trigger pulse is sent to the visual camera and the all-fiber spatial electric field sensor at the same time. The visual camera is controlled to perform global exposure, and the all-fiber spatial electric field sensor is triggered to lock and extract the one-dimensional electric field waveform signal corresponding to the exposure time window. The result is to generate spatiotemporally aligned multimodal data containing visual images and one-dimensional electric field waveform signals. The output of the anti-electromagnetic interference attitude calculation unit is connected to the input of the self-disturbance rejection hovering control unit, and the hovering state output of the self-disturbance rejection hovering control unit serves as the synchronous acquisition trigger condition for the hardware-level synchronous acquisition unit.

[0086] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0087] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A disturbance-resistant flight control method for UAVs for multi-mode insulator detection, characterized in that, Includes the following steps: Step 1: Obtain true attitude data resistant to electromagnetic interference based on multi-source sensor fusion, wherein the multi-source sensors include a magnetic compass and a full-fiber spatial electric field sensor based on the Pockels effect. When the electric field intensity gradient measured by the full-fiber spatial electric field sensor exceeds a threshold, the weight of the magnetic compass in the fusion solution is adaptively reduced to obtain the true attitude data. Step 2: Using the real attitude data obtained in Step 1 as state feedback information, the extended state observer is used to estimate the total system disturbance, including wind deflection. Based on the active disturbance rejection control algorithm, the compensation amount is output according to the total system disturbance, and motor control commands are generated to control the UAV to enter the wind-resistant stable hovering state and determine the hovering position of the UAV. Step 3: After entering the wind-resistant stable hovering state in Step 2, the master clock of the flight control system generates a global synchronization trigger pulse, which is simultaneously sent to the vision camera and the all-fiber spatial electric field sensor. The vision camera is controlled to perform global exposure to obtain a visual image and determine the exposure time window of the vision camera. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse, and one-dimensional electric field waveform data is continuously cached through hardware FIFO buffer. According to the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by hardware FIFO. The hovering position of the UAV and the actual attitude data corresponding to the acquisition time are associated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current flight control stable hovering state.

2. The UAV anti-disturbance flight control method for multi-mode insulator detection according to claim 1, characterized in that, Step 1 includes: The system acquires the one-dimensional electric field sequence measured by the all-fiber space electric field sensor in real time, and calculates the electric field intensity gradient based on the one-dimensional electric field sequence. ; When the electric field intensity gradient Greater than the set safety threshold At that time, calculate the adaptive penalty coefficient. The adaptive penalty coefficient With maximum amplitude value The adaptive penalty coefficient When the maximum amplitude value is reached Previously, with the electric field intensity gradient The increase is not a nonlinear monotonically increasing phenomenon; Using the adaptive penalty coefficient The measurement noise covariance matrix in the Kalman filter is updated, the noise variance of the corresponding magnetic compass measurement item in the measurement noise covariance matrix is ​​amplified, the weight of the magnetic compass in the fusion solution is reduced, and the fusion solution is performed based on the updated measurement noise covariance matrix to obtain the true attitude data. Wherein, when the adaptive penalty coefficient Exceeding the preset maximum amplitude value At that time, the adaptive penalty coefficient will be... Restricted to ; The adaptive penalty coefficient Calculate according to the following formula: in, This is the sensitivity adjustment coefficient, with a value ranging from 0.5 to 3.0; The maximum penalty limit value ranges from 5.0 to 20.

0. The electric field intensity gradient is calculated as the ratio of the difference in electric field intensity at adjacent sampling times to the sampling time interval. The safety threshold is determined based on electric field gradient samples collected within the first 10 seconds after the UAV takes off under conditions of no significant electromagnetic interference. Specifically: in, The mean of the electric field gradient samples under undisturbed conditions. denoted as the standard deviation of the electric field gradient sample under interference-free conditions, and k is the threshold amplification factor, ranging from 3 to 5.

3. The UAV anti-disturbance flight control method for multi-mode insulator detection according to claim 1, characterized in that, Step 2 includes: The aerodynamic wind deflection affecting drone hovering, the center of gravity shift caused by sensor mounting, and the internal nonlinear coupling are all considered as the total system disturbance. The total system disturbance is estimated in real time by using an extended state observer to obtain an estimated value of the total system disturbance. Based on the actual attitude data and the estimated total disturbance of the system, the disturbance compensation amount is calculated using an active disturbance rejection control algorithm. State feedback is performed based on the actual attitude data, and a basic control quantity is generated according to the state error between the expected hovering state and the actual hovering state of the UAV. The basic control quantity is then perturbed and compensated according to the perturbation compensation quantity to obtain the compensated control quantity. Based on the compensated control quantity, motor control commands are generated to maintain stable hovering of the UAV. The drone is controlled to enter a wind-resistant stable hovering state according to the motor control command, and the hovering position of the drone is determined.

4. The UAV anti-disturbance flight control method for multi-mode insulator detection according to claim 1, characterized in that, Step 3 includes: After the UAV enters the wind-resistant stable hovering state, the hardware timer of the flight control system generates a global synchronization trigger pulse as the master control clock; The global synchronization trigger pulse is simultaneously sent to the external trigger interface of the vision camera and the acquisition module of the all-fiber spatial electric field sensor via a dedicated parallel hardline for transmitting the trigger pulse. The global synchronization trigger pulse controls the vision camera to start global exposure, obtain a visual image, and determine the exposure time window of the vision camera according to the exposure start time and exposure duration. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse, and the one-dimensional electric field waveform data is continuously cached through hardware FIFO buffer. Based on the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by the hardware FIFO. The hovering position and actual attitude data of the UAV corresponding to the acquisition time are correlated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current stable hovering state of flight control.

5. A UAV anti-disturbance flight control system for multi-mode insulator detection, characterized in that, include: An anti-electromagnetic interference attitude calculation unit is used to obtain real attitude data against electromagnetic interference based on multi-source sensor fusion. The multi-source sensors include a magnetic compass and a pockmarked all-fiber spatial electric field sensor. When the electric field intensity gradient measured by the all-fiber spatial electric field sensor exceeds a threshold, the anti-electromagnetic interference attitude calculation unit adaptively reduces the weight of the magnetic compass in the fusion calculation to obtain the real attitude data. The active disturbance rejection hovering control unit is used to receive the real attitude data output by the anti-electromagnetic interference attitude calculation unit, use an extended state observer to estimate the total system disturbance including wind deflection, and generate motor control commands based on the active disturbance rejection control algorithm according to the total system disturbance to control the UAV to enter the wind deflection-resistant stable hovering state and determine the hovering position of the UAV. The hardware-level synchronous acquisition unit is used to generate a global synchronization trigger pulse by the main control clock of the flight control system after the UAV enters the wind-resistant stable hovering state. The global synchronization trigger pulse is sent to the vision camera and the all-fiber spatial electric field sensor at the same time. The vision camera is controlled to perform global exposure to obtain a visual image and determine the exposure time window of the vision camera. At the same time, the all-fiber spatial electric field sensor is triggered to record the trigger time corresponding to the global synchronization trigger pulse. One-dimensional electric field waveform data is continuously cached through hardware FIFO. According to the trigger time and the exposure time window, the corresponding one-dimensional electric field waveform signal is locked and extracted from the one-dimensional electric field waveform data cached by hardware FIFO. The hovering position of the UAV and the actual attitude data corresponding to the acquisition time are associated with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporal aligned multimodal inspection data corresponding to the current flight control stable hovering state. The output of the anti-electromagnetic interference attitude calculation unit is connected to the input of the self-disturbance rejection hovering control unit. The self-disturbance rejection hovering control unit determines whether the UAV has entered the wind-resistant stable hovering state and uses the determination result as the synchronous acquisition trigger condition of the hardware-level synchronous acquisition unit.

6. The UAV anti-disturbance flight control system for multi-mode insulator detection according to claim 5, characterized in that, The anti-electromagnetic interference attitude calculation unit includes: The electric field gradient calculation module is used to calculate the electric field intensity gradient based on the one-dimensional electric field sequence measured by the all-fiber space electric field sensor. ; The penalty coefficient calculation module is used when the electric field intensity gradient... Greater than the set safety threshold At that time, calculate the adaptive penalty coefficient. The adaptive penalty coefficient With maximum amplitude value The adaptive penalty coefficient When the maximum amplitude value is reached Previously, with the electric field intensity gradient The increase is not a nonlinear monotonically increasing phenomenon; The covariance update module is used to utilize the adaptive penalty coefficient. The measurement noise covariance matrix in the Kalman filter is updated, and the noise variance of the corresponding magnetic compass measurement item in the measurement noise covariance matrix is ​​amplified to reduce the weight of the magnetic compass in the fusion solution. The attitude fusion calculation module is used to perform fusion calculation based on the updated measurement noise covariance matrix to obtain the true attitude data.

7. A UAV anti-disturbance flight control system for multi-mode insulator detection according to claim 5, characterized in that, The active disturbance rejection hovering control unit includes: The extended state observer module is used to unify the aerodynamic wind deflection, center of gravity shift caused by sensor mounting, and internal nonlinear coupling that affect the hovering of the UAV as the total system disturbance, and to estimate the total system disturbance in real time to obtain the estimated value of the total system disturbance. The disturbance compensation control module is used to calculate the disturbance compensation amount using an active disturbance rejection control algorithm based on the actual attitude data and the estimated total disturbance of the system, perform state feedback based on the actual attitude data, generate a basic control quantity based on the state error between the expected hovering state and the actual hovering state of the UAV, perform disturbance compensation on the basic control quantity based on the disturbance compensation amount to obtain the compensated control quantity, generate motor control commands for maintaining stable hovering of the UAV based on the compensated control quantity, and control the UAV to enter the wind-resistant stable hovering state based on the motor control commands. The hovering position determination module is used to determine the hovering position of the UAV after the UAV enters the wind-resistant stable hovering state.

8. A UAV anti-disturbance flight control system for multi-mode insulator detection according to claim 5, characterized in that, The hardware-level synchronous acquisition unit includes: The master clock trigger module is used to generate a global synchronization trigger pulse by the hardware timer of the flight control system after the UAV enters the wind-resistant stable hovering state; The synchronous acquisition control module is used to simultaneously send the global synchronous trigger pulse to the external trigger interface of the vision camera and the acquisition module of the all-fiber spatial electric field sensor via a dedicated parallel hardline for transmitting the trigger pulse. It controls the vision camera to perform global exposure, obtain a visual image, and determine the exposure time window of the vision camera. At the same time, it triggers the all-fiber spatial electric field sensor to record the trigger time corresponding to the global synchronous trigger pulse, and continuously caches one-dimensional electric field waveform data through a hardware FIFO buffer. Based on the trigger time and the exposure time window, it locks and extracts the corresponding one-dimensional electric field waveform signal from the one-dimensional electric field waveform data cached in the hardware FIFO. The data association module is used to associate the hovering position and the actual attitude data of the UAV at the time of acquisition with the visual image and the one-dimensional electric field waveform signal to obtain spatiotemporally aligned multimodal inspection data corresponding to the current stable hovering state of flight control. The all-fiber spatial electric field sensor includes a hardware FIFO buffer, which is used to continuously buffer the one-dimensional electric field waveform data sampled by the all-fiber spatial electric field sensor, and lock the one-dimensional electric field waveform data corresponding to the exposure time window according to the trigger time corresponding to the global synchronization trigger pulse and the exposure time window.

9. An electronic device, characterized in that, include: One or more processors; Memory; and the computer program stored in the memory; When the computer program is executed by the one or more processors, the electronic device performs the unmanned aerial vehicle anti-disturbance flight control method for multi-mode insulator detection as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the UAV anti-disturbance flight control method for multi-mode insulator detection as described in any one of claims 1 to 4.

Images