Azimuth adaptive stabilization control system for optoelectronic pods

Abstract

The present application relates to the technical field of photoelectric stability control, and specifically discloses an azimuth adaptive stability control system for a photoelectric pod. The system comprises a carrier attitude sensing module, an image feature analysis module, a disturbance frequency spectrum identification module, a multi-mode controller and a servo drive module. By real-time identification of the frequency spectrum characteristics of carrier motion and image deviation, a suitable control strategy is dynamically selected and adapted, and the azimuth axis servo mechanism is driven to perform angle compensation, thereby achieving precise and differentiated compensation for wideband disturbances ranging from low-frequency drift to high-frequency noise, while avoiding self-shaking caused by excessive response of the servo mechanism to high-frequency noise. Furthermore, high-precision azimuth control ensures the continuity and positioning accuracy of target tracking, thereby overall enhancing the mission execution efficiency and reliability of the photoelectric pod.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of photoelectric stabilization control, and particularly relates to an azimuth adaptive stabilization control system for a photoelectric pod. BACKGROUND

[0002] In a moving carrier platform such as an aircraft or a ship, a photoelectric pod as a core visual perception and target tracking system directly determines the task execution efficiency in terms of imaging stability and target pointing accuracy. The photoelectric pod isolates disturbances caused by carrier attitude changes through internal optical sensors and stabilization mechanisms, thereby obtaining clear and stable image data in a dynamic environment.

[0003] The stabilization control system is a key technical link for the photoelectric pod to achieve high-performance imaging. The system calculates the carrier motion attitude in real time and drives the servo mechanism to perform reverse compensation to maintain the inertial space pointing stability of the optical viewing axis. The control performance directly affects the image jitter amplitude, tracking accuracy and overall response speed of the system.

[0004] The existing technology generally adopts a stabilization control strategy based on a fixed-bandwidth filter. Such a system has significant limitations in dealing with complex multi-frequency disturbances: the fixed-bandwidth design cannot take into account the differentiated compensation needs of low-frequency vibration and high-frequency noise, resulting in insufficient compensation for sustained low-frequency disturbances and causing image slow drift; at the same time, excessive response to high-frequency random noise easily introduces high-frequency jitter of the servo mechanism, which destroys image stability.

[0005] This contradiction not only reduces the signal-to-noise ratio of imaging, but also causes positioning error accumulation in the target tracking scene, seriously affecting the reliability and task completion of the photoelectric pod in a complex motion environment. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the contradiction that the existing photoelectric pod stabilization control system cannot take into account the low-frequency vibration compensation and high-frequency noise suppression due to the use of a fixed-bandwidth filter, and to provide an azimuth adaptive stabilization control system for a photoelectric pod. The system can autonomously adjust the frequency response and parameter configuration of the control loop according to the real-time dynamic characteristics of the carrier motion attitude and the optical sensor feedback, thereby realizing high-precision adaptive suppression of wide-band disturbances.

[0007] The technical solution of the present application is to construct an azimuth adaptive stabilization control system for a photoelectric pod, which includes a carrier attitude perception module, an image feature analysis module, a disturbance frequency spectrum identification module, a multi-modal controller and a servo drive module. The carrier attitude perception module is used to collect the angular velocity and angular acceleration data of the carrier on the pitch axis, roll axis and azimuth axis in real time;

[0008] The image feature analysis module is configured to receive the image sequence output by the optical sensor and extract a sequence of pixel coordinate offset of a preset reference point in the image. The disturbance spectrum recognition module is connected to the carrier attitude perception module and the image feature analysis module, and is configured to perform joint time-frequency analysis on the carrier motion data and the image offset data to recognize the energy distribution characteristics of the disturbance signal acting on the optoelectronic pod in the frequency domain.

[0009] The multi-modal controller is configured to dynamically select and generate corresponding control instructions from a plurality of preset control strategies based on the spectrum characteristics output by the disturbance spectrum recognition module. The servo drive module is configured to drive the azimuth axis servo mechanism of the optoelectronic pod to perform precise angle compensation actions in response to the control instructions of the multi-modal controller.

[0010] Further, the joint time-frequency analysis process of the disturbance spectrum recognition module includes the following steps. Firstly, the carrier angular velocity data and the image pixel offset data are synchronously sampled and aligned. Then, the short-time Fourier transform is applied to the synchronized data sequence to calculate the instantaneous energy values in a plurality of preset frequency bands. Then, according to the exceeding condition of the energy values of each frequency band relative to the preset energy threshold, the dominant frequency band of the current disturbance is determined. Finally, a spectrum characteristic vector containing the dominant frequency band identifier and the energy intensity level information is output to the multi-modal controller.

[0011] Further, three core control strategies are pre-stored in the multi-modal controller, which are low-frequency dominant control strategy, medium-frequency dominant control strategy and high-frequency dominant control strategy. The low-frequency dominant control strategy adopts proportional-integral-derivative control law, and the integral term coefficient is set to be 2 times the standard value, which is specially used to suppress slow drift disturbances with a frequency lower than 2 Hz.

[0012] The medium-frequency dominant control strategy adopts a lead-lag compensation network, the center frequency of which is automatically matched to the recognized dominant frequency of the disturbance, which is used to compensate for periodic vibrations with a frequency ranging from 2 Hz to 15 Hz. The high-frequency dominant control strategy enables a Kalman filter to optimally estimate the feedback signal of the servo mechanism and significantly reduces the control loop gain to suppress random noise and sensor measurement noise with a frequency higher than 15 Hz.

[0013] Further, the strategy selection logic of the multi-modal controller makes decisions based on the spectrum characteristic vector. When the spectrum characteristic vector indicates that the low-frequency band energy level exceeds threshold 1 and the high-frequency band energy level is lower than threshold 2, the low-frequency dominant control strategy is activated. When the spectrum characteristic vector indicates that there is a significant energy peak in the medium-frequency band and the peak frequency is within the range of 2 to 15 Hz, the medium-frequency dominant control strategy is activated. When the spectrum characteristic vector indicates that the high-frequency band energy level continuously exceeds threshold 3 for 500 milliseconds, the high-frequency dominant control strategy is activated.

[0014] Further, the specific method of the image feature analysis module extracting the pixel coordinate offset sequence is as follows. A reference point containing 128 pixels is set in the center area of the field of view of the optical sensor. The motion vector of the reference point between adjacent image frames is continuously calculated by the optical flow method. The motion vector sequence is filtered by 3-point sliding average filtering to suppress transient noise. Finally, the filtered pixel displacement sequence is output to the disturbance spectrum recognition module.

[0015] Further, the servo drive module includes a digital signal processor and a three-phase brushless DC motor. The digital signal processor receives digital control instructions from the multi-modal controller and converts them into pulse width modulation signals corresponding to the duty cycle. The three-phase brushless DC motor drives the azimuth axis transmission mechanism of the optoelectronic pod according to the pulse width modulation signal, and the position feedback is realized through a 22-bit absolute encoder, ensuring that the angle compensation accuracy is better than 0.005 degrees.

[0016] Further, the system also has a control performance evaluation module. This module continuously monitors the root mean square value of the pixel offset output by the image feature analysis module. When the root mean square value continuously falls below the preset stability threshold within a 1-second time window, a locking instruction is sent to the multi-modal controller to maintain the current control strategy. If the root mean square value exceeds the stability threshold, a re-identification instruction is sent to the multi-modal controller, triggering a new round of disturbance spectrum recognition and control strategy switching.

[0017] Further, the multi-modal controller uses a smooth transition algorithm during strategy switching. This algorithm weights and fuses the output instructions of the two control strategies before and after the transition within a 100-millisecond transition time. The weight coefficient linearly changes from 100% of the old strategy to 100% of the new strategy over time, thereby avoiding shocks or oscillations in the servo mechanism due to sudden changes in instructions.

[0018] Compared with the prior art, the beneficial effects of the present application are:

[0019] 1. The present application introduces a disturbance spectrum recognition module and a multi-modal controller to build an adaptive stable control architecture that can sense and respond to the disturbance spectrum characteristics of the carrier in real time. This system completely eliminates the limitations of fixed bandwidth filters, achieving precise and differentiated compensation for wideband disturbances from low-frequency drift to high-frequency noise. Specifically, by dynamically switching control strategies, it effectively eliminates the low-frequency slow drift of the image, while avoiding the self-shaking caused by the over-response of the servo mechanism to high-frequency noise.

[0020] 2. The image stability and clarity of the optoelectronic pod in complex motion environments are improved, and the continuity and positioning accuracy of target tracking are ensured through high-precision azimuth control, thereby enhancing the mission execution efficiency and reliability of the optoelectronic pod as a whole. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 is the overall technical scheme architecture diagram of the azimuth adaptive stabilization control system for the optoelectronic pod proposed in the application;

[0022] Figure 2 is the core principle framework diagram of the disturbance spectrum identification and multi-modal control collaborative work in the application;

[0023] Figure 3 is the logic flow framework diagram of the carrier attitude perception and image feature analysis joint data acquisition in the application;

[0024] Figure 4 is the joint time-frequency analysis processing flow framework diagram of the disturbance spectrum identification module in the application;

[0025] Figure 5 is the logic flow framework diagram of the multi-modal controller strategy selection and smooth transition in the application;

[0026] Figure 6 is the interaction diagram of the control performance evaluation module and system closed-loop feedback in the application. DETAILED DESCRIPTION

[0027] Please refer to the attached Figures 1 to 6 The azimuth adaptive stabilization control system for the optoelectronic pod is composed of a carrier attitude perception module, an image feature analysis module, a disturbance spectrum identification module, a multi-modal controller, a servo drive module, and a control performance evaluation module. These modules are connected with each other through a high-speed data bus and a real-time control bus, forming a closed-loop azimuth adaptive stabilization control system.

[0028] The carrier attitude perception module is responsible for collecting the motion data of the carrier in the pitch axis, the roll axis, and the azimuth axis. The image feature analysis module receives a continuous image sequence from the optical sensor and executes a preset image processing algorithm. The disturbance spectrum identification module simultaneously receives the carrier motion data from the carrier attitude perception module and the image offset data from the image feature analysis module, and performs joint time-frequency analysis operation.

[0029] The multi-modal controller dynamically selects and configures its internal control strategy according to the analysis results output by the disturbance spectrum identification module. The servo drive module converts the control instructions generated by the multi-modal controller into physical drive signals, which act on the azimuth axis servo mechanism of the optoelectronic pod. The control performance evaluation module continuously monitors the overall stability performance of the system and provides feedback signals to the multi-modal controller to maintain or adjust the control state.

[0030] The core components of the carrier attitude perception module are a three-axis micro-electro-mechanical system gyroscope and a three-axis micro-electro-mechanical system accelerometer. The gyroscope has a range of ±300 degrees per second, and its analog output signal is digitized by a 24-bit analog-to-digital converter with a sampling frequency of 1000 Hz. The accelerometer has a range of ±16 times the gravitational acceleration, and its output is also sampled by a 24-bit analog-to-digital converter with the same sampling frequency as the gyroscope. The digitized angular velocity and angular acceleration data are transmitted to a 32-bit microcontroller via a serial peripheral interface bus.

[0031] The microcontroller runs data preprocessing firmware to compensate for zero bias and temperature drift in the raw sensor data. The zero bias compensation values are stored in non-volatile memory and automatically loaded when the system is powered on. Temperature drift compensation is corrected in real time by a temperature sensor integrated inside the sensor module, with a sampling accuracy of 0.1 degrees Celsius.

[0032] The preprocessed carrier attitude data is packaged into a specific data frame format, with each frame containing a timestamp, pitch axis angular velocity, roll axis angular velocity, yaw axis angular velocity, pitch axis angular acceleration, roll axis angular acceleration, yaw axis angular acceleration, and a data checksum field. The data frame is broadcast to the system bus at a period of 1 millisecond via a controller area network bus.

[0033] Please refer to the attached Figure 3 The image feature analysis module is deployed on a dedicated image signal processor. The processor receives raw image data from the optical sensor through a mobile industry processor interface. The optical sensor has a resolution of 1920x1080 pixels and an output frame rate of 60 frames per second. During the initialization phase of the image feature analysis module, the system operator sets a reference point in the center of the sensor's field of view through a human-machine interface.

[0034] The reference point is a square region with a side length of 8 pixels, covering a total of 64 pixels, not the 128 pixels described in the invention content. This is a specific optimization of the present embodiment. The initial pixel coordinates of the reference point are recorded in the internal memory of the image signal processor.

[0035] For each new input image, the image feature analysis module performs an optical flow calculation algorithm to determine the motion of the reference point. Specifically, the algorithm uses the Lucas-Kanade method to select 16 feature points in the reference point area and calculate the motion vectors of these points between two adjacent frames of images. The calculation of the motion vector is based on the assumption of constant image brightness and gradient descent optimization.

[0036] The 16 calculated motion vectors are subjected to outlier rejection, any vector that differs from the median vector by more than 2 pixels is discarded. The remaining valid motion vectors are averaged to obtain a composite motion vector for the reference point, which contains the horizontal pixel displacement Δu and the vertical pixel displacement Δv. This motion vector sequence is then passed through a 3-point sliding average filter.

[0037] The filter window covers the results from the current frame and its previous two frames, and outputs a smoothed pixel displacement sequence. This sequence is also encapsulated into a data frame, containing the timestamp, horizontal displacement, vertical displacement, and data quality flag, and sent to the system bus via the Ethernet bus, with a transmission period of about 16.67 milliseconds, synchronized with the image frame rate.

[0038] Please refer to the attached Figure 4 The disturbance spectrum identification module runs on a high-performance digital signal processor. The core task of this module is to perform joint time-frequency analysis. First, the module receives carrier motion data from the carrier attitude perception module and image pixel offset data from the image feature analysis module from the system bus. Since the two data sources have different sampling rates and timestamps, the disturbance spectrum identification module first performs data synchronization and alignment operations to create a common time axis with a time resolution of 1 millisecond.

[0039] For image pixel offset data, linear interpolation is used to resample it to 1000 Hz to match the sampling rate of the carrier motion data. After data alignment, the module extracts two key data streams for subsequent analysis: carrier azimuth axis angular velocity data stream and image horizontal pixel offset data stream. The analysis window length is set to 512 sampling points, corresponding to a time length of 512 milliseconds. The window slides forward with an overlap rate of 50%, i.e. a new analysis is performed every 256 milliseconds.

[0040] Short-time Fourier transform is applied to the data in the window. The transform uses a Hanning window function with a window length of 256 sampling points and a fast Fourier transform point number of 1024. Through calculation, the complex spectrum of the azimuth axis angular velocity signal and the image horizontal pixel offset signal in multiple preset frequency bands is obtained. The preset frequency bands are divided into three main intervals:

[0041] The low frequency band covers 0.1 Hz to 2 Hz, the medium frequency band covers 2 Hz to 15 Hz, and the high frequency band covers 15 Hz to 50 Hz. For each frequency band, the instantaneous energy value is calculated, which is the sum of the squares of the amplitudes corresponding to all frequency points in the frequency band. Energy calculation is performed for both angular velocity signals and pixel offset signals, and the energy values of the two are weighted and summed according to the weight coefficients 0.6 and 0.4 to obtain a joint energy index.

[0042] The joint energy indicators of each band are compared with preset energy thresholds. Threshold 1 is set for the low frequency band, at 0.05 square meter per square second. Threshold 2 is set for the high frequency band, at 0.01 square meter per square second. Threshold 3 is also for the high frequency band, but for sustained judgment, at 0.008 square meter per square second. The decision logic for the dominant frequency band is as follows:

[0043] If the joint energy indicator of the low frequency band exceeds threshold 1, and the joint energy indicator of the high frequency band is below threshold 2, the current disturbance dominant frequency band is determined to be low frequency. If there is a local energy peak in the mid frequency band, and the peak frequency is confirmed to be within the range of 2 to 15 Hz, and its peak energy exceeds 2 times the average energy of the band, the current disturbance dominant frequency band is determined to be mid frequency.

[0044] If the joint energy indicator of the high frequency band exceeds threshold 3 for 500 milliseconds, the current disturbance dominant frequency band is determined to be high frequency. Finally, the disturbance spectrum recognition module outputs a spectrum feature vector. The vector is a three-dimensional array containing the low frequency energy level flag, the mid frequency dominant frequency value, and the high frequency energy sustained flag. This vector is transmitted to the multi-modal controller through serial peripheral interface communication.

[0045] The multi-modal controller is the decision-making core of the system, and its hardware platform is a field programmable gate array combined with a dual-core microprocessor. The field programmable gate array is responsible for high-speed logic judgment and signal routing, and the microprocessor is responsible for running complex control algorithms. The controller has pre-stored three core control strategies: low frequency dominant control strategy, mid frequency dominant control strategy, and high frequency dominant control strategy. Please refer to the attached Figure 2 and attached Figure 5 The strategy selection logic of the controller is completely based on the received spectrum feature vector.

[0046] When the spectrum feature vector indicates that the low frequency energy level flag is true, i.e. the low frequency band energy exceeds threshold 1 and the high frequency band energy is below threshold 2, the multi-modal controller activates the low frequency dominant control strategy. This strategy uses a proportional-integral-derivative control law. Its control output is composed of a proportional term, an integral term, and a derivative term. The integral term coefficient is set to 2 times the standard value, which is pre-tuned according to the servo mechanism model. The proportional term coefficient and the derivative term coefficient remain the standard value. The sampling period of the controller is 1 millisecond. The error signal is the difference between the desired azimuth angle position zero and the actual azimuth angle position currently fed back by the servo drive module.

[0047] When the spectral feature vector indicates that the mid-frequency dominant frequency value is valid, i.e. there is a peak frequency located in the range of 2 Hz to 15 Hz, the multi-modal controller activates the mid-frequency dominant control strategy. This strategy employs a lead-lag compensation network. The center frequency of the compensation network's transfer function is automatically matched to the identified dominant disturbance frequency. The network parameters are computed in real-time according to the zero-pole placement method, ensuring that a gain boost of approximately 10 decibels and a corresponding phase lead are provided at the dominant frequency to effectively compensate for the periodic vibration in that frequency band.

[0048] When the spectral feature vector indicates that the high-frequency energy persistence flag is true, i.e. the high-frequency band energy persists beyond the threshold of 3 for 500 milliseconds, the multi-modal controller activates the high-frequency dominant control strategy. The core of this strategy is to enable a discrete-time Kalman filter.

[0049] The Kalman filter performs optimal estimation on the feedback signal of the servo mechanism. The filter model contains system state equations and observation equations. The system states include azimuthal position and angular velocity. The process noise covariance matrix and the observation noise covariance matrix are configured according to the high-frequency noise characteristics, significantly reducing the gain of the control loop, typically reducing the proportional gain to 50% of the standard value, to suppress random noise and sensor measurement noise with frequencies higher than 15 Hz.

[0050] When switching control strategies, the multi-modal controller executes a smooth transition algorithm to avoid command shocks to the servo mechanism. The transition process lasts 100 milliseconds. During the transition period, the final output command of the controller is a weighted sum of the output of the old strategy and the output of the new strategy. The weight coefficients change linearly over time. At 0 milliseconds from the start of the transition, the old strategy weight is 100% and the new strategy weight is 0%. At 100 milliseconds from the end of the transition, the old strategy weight becomes 0% and the new strategy weight becomes 100%. The weight update period is 1 millisecond.

[0051] The servo drive module is responsible for converting control commands into physical actions. The core of the module includes a digital signal processor and a three-phase brushless DC motor. The digital signal processor receives digital control commands from the multi-modal controller, which is a 16-bit signed integer representing the desired current or torque. The digital signal processor internally runs a space vector pulse width modulation algorithm. The digital command is converted into six pulse width modulation signals, which control the six insulated gate bipolar transistors of the three-phase full-bridge inverter.

[0052] The switching frequency of the pulse width modulation signal is 20 kHz, and the duty cycle is linearly adjusted according to the control command, ranging from 0% to 100%. The three-phase brushless DC motor drives the azimuth axis transmission mechanism of the optical pod, which usually includes a harmonic reducer with a reduction ratio of 100:1. The position feedback of the motor is realized through a 22-bit absolute encoder. The encoder is directly installed on the motor rotor, and its output is a multi-turn absolute value signal, which is transmitted to the digital signal processor through an EnDat 2.2 interface.

[0053] The digital signal processor decodes the encoder signal to obtain high-precision azimuth position information with a resolution of 0.000086 degrees, ensuring that the angle compensation accuracy of the entire servo system is better than 0.005 degrees. The decoded position information is fed back to the multi-modal controller through the controller area network bus, forming a position closed loop.

[0054] Please refer to the attached Figure 6 The control performance evaluation module runs independently on the safety microcontroller and is responsible for monitoring the overall stability performance of the system. This module continuously reads the horizontal pixel offset sequence output by the image feature analysis module from the system bus. The module maintains a data buffer with a length of 1000 sampling points, corresponding to a time window of about 1 second. For the data in the buffer, the control performance evaluation module calculates the root mean square value of the pixel offset every 100 milliseconds. The formula for calculating the root mean square value is as follows:

[0055] ;

[0056] where, represents the number of data points in the current buffer, represents the pixel offset value, represents the average value of the pixel offset in the time window. The root mean square value obtained by calculation is compared with a preset stability threshold. The stability threshold is set according to the field of view angle and resolution of the optical sensor, and the typical value is 1.5 pixels.

[0057] The logic judgment rule is: if the root mean square value is lower than the stability threshold for 10 consecutive calculations, i.e. within 1 second, the control performance evaluation module sends a lock command to the multi-modal controller. The lock command instructs the multi-modal controller to maintain the control strategy currently being executed and suspend the periodic triggering of the disturbance spectrum identification module.

[0058] Conversely, if the RMS value exceeds the stability threshold in any 100 ms decision period, the control performance evaluation module sends a re-identification command to the multi-modal controller immediately. The re-identification command forces the disturbance spectrum identification module to perform a new joint time-frequency analysis, and can cause the multi-modal controller to switch control strategies. The instruction communication between the control performance evaluation module and the multi-modal controller is through a universal asynchronous receiver-transmitter, and the instruction frame contains instruction type and timestamp information.

[0059] The system power-up initialization process is as follows:

[0060] First, all modules perform self-checking, including sensor communication testing, memory read-write testing, and processor core testing. After passing the self-checking, the carrier attitude perception module starts to output data, and the image feature analysis module starts to capture and process images. The disturbance spectrum identification module waits to receive data of sufficient length before performing the first joint time-frequency analysis. The multi-modal controller activates the low-frequency dominant control strategy by default according to the first analysis result. The servo drive module enables the motor and enters the position control mode.

[0061] The control performance evaluation module starts to monitor the pixel shift. The entire system thus enters a stable adaptive control loop. The power management unit provides 5V and 12V DC power for the system, and has overcurrent and overvoltage protection functions. All key parameters, such as various threshold values, control parameters, filter coefficients, etc., are stored in ferroelectric random access memory, ensuring that power loss does not result in loss and supporting online updates. The system case adopts a shielding design to reduce the influence of electromagnetic interference on sensitive signals.

[0062] It should be noted that the relational terms herein, such as first and second, are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprising", "including" or any other variant thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed or inherent to such a process, method, article or device.

[0063] Although embodiments of the present application have been shown and described, it will be understood by those having ordinary skill in the art that various changes, modifications, alternatives, and variations can be made thereto without departing from the principles and spirit of the application, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An azimuth adaptive stabilization control system for an electro-optical pod, characterized by, include: The carrier attitude sensing module is used to collect the angular velocity and angular acceleration data of the carrier on the pitch axis, roll axis and azimuth axis in real time; The image feature analysis module is used to receive the image sequence output by the optical sensor and extract the pixel coordinate offset sequence of the preset reference point in the image; The disturbance spectrum identification module connects the carrier attitude sensing module and the image feature analysis module. It is used to perform joint time-frequency analysis on carrier motion data and image offset data to identify the energy distribution characteristics of the disturbance signal currently acting on the optoelectronic pod in the frequency domain. The multimodal controller dynamically selects and generates corresponding control commands from multiple preset control strategies based on the spectral characteristics output by the disturbance spectrum identification module. The servo drive module responds to the control commands of the multi-modal controller and drives the azimuth axis servo mechanism of the optoelectronic pod to perform precise angle compensation actions. The control performance evaluation module continuously monitors the root mean square value of the pixel offset output by the image feature analysis module. When the root mean square value is continuously lower than the preset stability threshold within a 1-second time window, a lock command is sent to the multimodal controller to maintain the current control strategy. If the root mean square value exceeds the stability threshold, a re-identification command is sent to the multimodal controller to trigger a new round of disturbance spectrum identification and control strategy switching. The multimodal controller has three pre-stored core control strategies: low-frequency dominant control strategy, medium-frequency dominant control strategy, and high-frequency dominant control strategy. The low-frequency dominant control strategy employs a proportional-integral-derivative control law, with the integral term coefficient set to twice the standard value to suppress slow drift disturbances with frequencies below 2 Hz. The mid-frequency dominant control strategy employs a lead-lag compensation network, whose center frequency is automatically matched with the identified dominant disturbance frequency to compensate for periodic vibrations in the frequency range of 2 Hz to 15 Hz. The high-frequency dominant control strategy enables the Kalman filter to make optimal estimation of the feedback signal of the servo mechanism and significantly reduces the control loop gain to suppress random noise and sensor measurement noise with frequencies higher than 15 Hz. The strategy selection logic of the multimodal controller makes decisions based on the spectral feature vector; when the spectral feature vector indicates that the low-frequency band energy level exceeds threshold 1 and the high-frequency band energy level is lower than threshold 2, the low-frequency dominant control strategy is activated. When the spectral feature vector indicates a significant energy peak in the mid-frequency band and the peak frequency is within the range of 2 to 15 Hz, the mid-frequency dominant control strategy is activated; when the spectral feature vector indicates that the energy level in the high-frequency band continuously exceeds the threshold 3 for 500 milliseconds, the high-frequency dominant control strategy is activated. The multimodal controller employs a smooth transition algorithm during strategy switching. The algorithm performs weighted fusion of the output commands of the two control strategies within a 100-millisecond transition time; the weight coefficient changes linearly from 100% of the old strategy to 100% of the new strategy over time, thereby avoiding shocks or oscillations in the servo mechanism due to sudden changes in commands.

2. The azimuthally adaptive stability control system for a photovoltaic pod of claim 1, wherein, The joint time-frequency analysis process of the disturbance spectrum identification module specifically includes: synchronously sampling and aligning the carrier angular velocity data and image pixel offset data; Apply short-time Fourier transform to the synchronized data sequence to calculate its instantaneous energy value in multiple preset frequency bands; determine the dominant frequency band of the current disturbance based on the extent to which the energy value of each frequency band exceeds the preset energy threshold; and output a spectral feature vector containing the dominant frequency band identifier and energy intensity level information to the multi-mode controller.

3. The azimuthally adaptive stability control system for a photovoltaic pod of claim 1, wherein, The specific method for the image feature analysis module to extract the pixel coordinate offset sequence includes: setting a reference point containing 128 pixels in the center region of the optical sensor's field of view; The motion vector of the reference point between adjacent image frames is continuously calculated using the optical flow method; the motion vector sequence is filtered by a 3-point moving average to suppress instantaneous noise; and the filtered pixel displacement sequence is output to the disturbance spectrum identification module.

4. The azimuth adaptive stabilization control system for an optoelectronic pod according to claim 1, characterized in that, The servo drive module includes a digital signal processor and a three-phase brushless DC motor; the digital signal processor receives digital control commands from the multi-mode controller and converts them into pulse width modulation signals with corresponding duty cycles; The three-phase brushless DC motor drives the orientation shaft transmission mechanism of the photoelectric pod according to the pulse width modulation signal. Its position feedback is achieved through a 22-bit absolute encoder to ensure that the angle compensation accuracy is better than 0.005 degrees.

5. The azimuth adaptive stabilization control system for an optoelectronic pod according to claim 1, characterized in that, The process by which the control performance evaluation module calculates the root mean square value of the pixel offset is as follows: Obtain the pixel offset sequence output by the image feature analysis module; calculate the root mean square value of these offsets within a 1-second time window; compare the calculated root mean square value with a preset stability threshold.

6. The azimuth adaptive stabilization control system for an optoelectronic pod according to claim 2, characterized in that, The joint time-frequency analysis process is pre-divided into three main intervals: The low-frequency band covers 0.1 Hz to 2 Hz, the mid-frequency band covers 2 Hz to 15 Hz, and the high-frequency band covers 15 Hz to 50 Hz. For each frequency band, its instantaneous energy value is calculated, which is the sum of squares of the amplitudes corresponding to all frequency points within that frequency band.

7. The azimuth adaptive stabilization control system for an optoelectronic pod according to claim 1, characterized in that, The carrier attitude sensing module includes a three-axis microelectromechanical system gyroscope and a three-axis microelectromechanical system accelerometer; the gyroscope has a range of ±300 degrees per second, and its analog output signal is digitally sampled by a 24-bit analog-to-digital converter with a sampling frequency of 1000 Hz. The accelerometer has a range of ±16 times the acceleration due to gravity, and its output is also sampled by a 24-bit analog-to-digital converter, with the sampling frequency being consistent with that of the gyroscope.

Images