A Multi-Source Joint Estimation Method for Micro-Vibration Compensation of Spectral Tracking Turntable
By using a multi-source joint estimation method, the problems of single observation source in spectral tracking turntable, such as difficulty in distinguishing motor harmonics from external disturbances, difficulty in matching resonance state with fixed compensation parameters, and lack of closed-loop correlation between exposure process and micro-vibration state, are solved. This method enables coordinated vibration judgment and dynamic updating of compensation parameters on the drive side, mechanical motion side, and imaging side, thereby improving image clarity and encoder residual feedback efficiency.
Patent Information
- Application Number
- CN202611522223.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-29
- Publication Date
- 2026-11-17
AI Technical Summary
Existing spectral tracking turntables have several problems in micro-vibration processing, including difficulty in distinguishing motor harmonics from external disturbances with a single observation source, difficulty in matching fixed compensation parameters with changing resonance states, and lack of closed-loop correlation between the exposure process and the micro-vibration state.
A multi-source joint estimation method is adopted. By acquiring servo current data, encoder data and short exposure image data, time alignment is performed based on a unified time base, servo current harmonic features, encoder residual features and image sharpness features are extracted, cross-validation is performed, state estimation and resonant frequency determination are performed, digital notch filter parameters are updated, and servo compensation and exposure control are performed.
It enables the responses from the driving side, mechanical motion side, and imaging side to jointly participate in vibration candidate judgment, and the compensation parameters are updated according to the resonance state. This solves the closed-loop correlation problem between the exposure process and the micro-vibration state, and improves the image clarity and encoder residual feedback efficiency.
Smart Images

Figure CN123111466A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of servo control, imaging control and micro-vibration compensation technology of spectral tracking turntable, specifically involving micro-vibration compensation technology of spectral tracking turntable based on multi-source joint estimation. Background Technology
[0002] Spectral tracking turntables are used for the stable acquisition of multispectral, hyperspectral, and long-distance images. Under vehicle-mounted or field operating conditions, the turntable's motion is easily affected by factors such as wind load, vehicle vibration, reducer periodic errors, and structural resonance. Even when the amplitude of these micro-vibrations is small, the encoder angle feedback may still be within acceptable limits, but long-integral spectra or long-distance images will already show a decrease in sharpness. Traditional image stabilization typically relies on inertial measurement information or image information for vibration judgment and imaging adjustment, and its processing effectiveness depends on whether the observed signal can accurately reflect the mechanical vibration state.
[0003] With the development of photoelectric tracking, digital servo, and edge computing technologies, existing solutions can collect vibration information and adjust the imaging process accordingly. Common methods include using inertial measurement units, acceleration information, or image feedback to characterize jitter. These solutions improve the automation of the process compared to methods that rely entirely on manual adjustments. However, software algorithms typically operate according to fixed procedures, and hardware layout or execution status is often passively input. When changes occur in the field, such as mechanical hysteresis, spectral band switching, or clock drift, it may still be necessary to reduce the frame rate or recalibrate. Furthermore, the physical sources of different observed signals are not entirely the same; changes in motor drive, mechanical transmission disturbances, and external random disturbances can produce similar variation characteristics.
[0004] However, existing micro-vibration processing still has two shortcomings. On the one hand, when using single inertial or image information for vibration judgment, it is difficult to simultaneously utilize the response relationships of the driving side, the mechanical motion side, and the imaging side. A single observation source cannot distinguish between motor harmonics and external disturbances, and changes unrelated to mechanical micro-vibration are easily included in the compensation process. On the other hand, software algorithms or servo parameters usually operate according to pre-set processes and ranges. When the reducer period error, structural resonance, or field disturbance state changes, fixed compensation parameters are difficult to match with the changing resonance state. Continuous compensation also requires consideration of the stable operation of the main servo closed loop.
[0005] Furthermore, existing image stabilization processes still lack sufficient coordination between turntable control and image acquisition, and the exposure process lacks a closed-loop correlation with the micro-vibration state. For long integral spectra or long-distance images, if significant micro-vibrations still exist within the exposure window, even if the encoder angle surface is stable, blurred or invalid acquisition results may occur. After compensation, if the image sharpness and encoder residuals cannot continue to participate in subsequent state estimation, it is difficult to continuously correct the processing based on the actual compensation results. Therefore, it is necessary to provide a multi-source joint estimation method for spectral tracking turntable micro-vibration compensation. Summary of the Invention
[0006] To address the problems in existing technologies, such as the difficulty in distinguishing between motor harmonics and external disturbances from a single observation source, the difficulty in matching fixed compensation parameters with changing resonance states, and the lack of closed-loop correlation between the exposure process and micro-vibration states, this invention proposes the following solutions: A multi-source joint estimation method for spectral tracking turntable micro-vibration compensation includes: S1. Acquire servo current data, encoder data, and short exposure image data of the spectral tracking turntable, and perform time alignment of the servo current data, encoder data, and short exposure image data based on a unified time reference; S2. Within the same state estimation period, extract servo current harmonic features from time-aligned servo current data, extract encoder residual features from time-aligned encoder data, and extract image sharpness features from time-aligned short-exposure image data. Based on the frequency consistency between the servo current harmonic features and encoder residual features, and combined with the changes in the image sharpness features, perform cross-validation to obtain vibration candidate frequencies and the three-source observation features corresponding to the vibration candidate frequencies. S3. Input the vibration candidate frequency and the corresponding three-source observation characteristics into the state estimator to perform state estimation for periodic vibration, structural resonance and random external disturbance, and obtain the angular micro-vibration estimation result and resonance frequency. S4. Based on the resonance frequency obtained from the continuous state estimation period, perform stability judgment. When the resonance frequency meets the preset stability condition, update the digital notch filter parameters or command feedforward shaping parameters according to the main servo closed-loop bandwidth constraint. S5. Based on the angle micro-vibration estimation result, generate the micro-vibration compensation amount for the next control cycle, and perform servo compensation on the spectral tracking turntable in combination with the updated digital notch filter parameters or command feedforward shaping parameters. S6. Based on the angle micro-vibration estimation result, predict the angle micro-vibration within the exposure window and obtain the root mean square jitter within the exposure window. Determine the exposure allowable angle jitter threshold based on the focal length, pixel size, and allowable blur amount of the imaging acquisition device. When the root mean square jitter exceeds the exposure allowable angle jitter threshold, delay long integral exposure or perform short exposure resampling. S7. Obtain the encoder residual and image sharpness after performing the servo compensation and exposure control, and input the encoder residual and image sharpness as the observations for the next state estimation cycle into the state estimator.
[0007] Furthermore, the time alignment based on a unified time reference described in S1 includes: resampling servo current data, encoder data, and short-exposure image data at different sampling rates according to the state estimation period, and determining the time age and data quality for each resampled data stream.
[0008] Furthermore, the cross-validation described in S2 includes: when the servo current harmonic characteristics and encoder residual characteristics show corresponding peak values in the same frequency range and the image sharpness decreases synchronously, the corresponding frequency is determined as a candidate vibration frequency; when only the servo current harmonic characteristics show peak values and the encoder residual characteristics and image sharpness characteristics do not show corresponding responses, the corresponding peak value is determined as an electrical disturbance.
[0009] Furthermore, the state estimation described in S3 includes: determining the observation weights of the corresponding three-source observation features based on the time age and data quality of each data source, and reducing the corresponding observation weights when the data quality of any observation source decreases, so as to update the angular micro-vibration estimation results and resonance frequencies based on the adjusted observation weights.
[0010] Furthermore, the updating of digital notch filter parameters or command feedforward shaping parameters as described in S4 includes: performing parameter updates after the resonant frequency obtained from the continuous state estimation period satisfies the frequency stability condition, and limiting the operating frequency band of the digital notch filter to not enter the main servo closed-loop bandwidth; and restoring the digital notch filter parameters or command feedforward shaping parameters before the update when the encoder residual increases after the parameter update.
[0011] Furthermore, the delayed long integral exposure or the short exposure resampling described in S6 includes: delaying the long integral exposure when the root mean square jitter within the predicted exposure window exceeds the exposure allowable angle jitter threshold, and re-acquiring the root mean square jitter within the delayed exposure window; and performing short exposure resampling when the re-acquired root mean square jitter still exceeds the exposure allowable angle jitter threshold.
[0012] Furthermore, the method also includes performing validity checks on each observation source; reducing the corresponding observation weight and tightening the micro-vibration compensation amplitude when any observation source fails; stopping the updating of digital notch filter parameters and command feedforward shaping parameters and performing fixed amplitude limiting compensation when only one observation source remains reliable; and resuming parameter updates when the failed observation source recovers and meets the preset recovery conditions.
[0013] Based on the same inventive concept, the present invention also proposes an electronic device, including a processor and a memory, wherein the memory stores a computer program, and the processor executes the above-described method when executing the computer program.
[0014] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the above-described method.
[0015] Based on the same inventive concept, the present invention also proposes a computer program product, including a computer program or instructions, which implement the above-described method when executed by a processor.
[0016] Compared with the prior art, the present invention has the following beneficial effects: By extracting servo current harmonic features, encoder residual features, and image sharpness features under a unified time reference, and performing cross-validation based on frequency consistency and image sharpness variation, the responses of the drive side, mechanical motion side, and imaging side can jointly participate in vibration candidate judgment, thereby reducing the impact of a single observation source on the estimation of micro-vibration caused by the inability to distinguish between motor harmonics and external disturbances.
[0017] By assessing the stability of the resonance frequency obtained from the continuous state estimation period, and updating the digital notch filter parameters or command feedforward shaping parameters according to the main servo closed-loop bandwidth constraint when the stability condition is met, the compensation parameters can be updated with the actual resonance state and the abnormal backoff condition is retained, thereby alleviating the problem that fixed compensation parameters are difficult to match with the changing resonance state.
[0018] By generating servo compensation and root mean square jitter within the exposure window based on the angular micro-vibration estimation results, and controlling long integral exposure or short exposure resampling according to the exposure allowable angular jitter threshold, the compensated encoder residual and image sharpness are fed back to the next state estimation cycle, so that control and imaging form feedback, thereby solving the problem of lack of closed-loop correlation between the exposure process and the micro-vibration state.
[0019] This invention has the capability to perform multi-source joint estimation, servo compensation, and exposure coordination control for the micro-vibration state of a spectral tracking turntable. It can output the micro-vibration compensation amount and control long integral exposure or short exposure re-sampling according to the jitter state of the predicted exposure window. It is applicable to fields such as multispectral monitoring, hyperspectral monitoring, vehicle-mounted spectral tracking, and field photoelectric tracking in oilfields. Attached Figure Description
[0020] Figure 1 This is a flowchart of the micro-vibration compensation method for the spectral tracking turntable described in the implementation method. Detailed Implementation
[0021] The technical solutions in the embodiments of the present invention will now be clearly and completely described in conjunction with the accompanying drawings.
[0022] Implementation Method 1 A multi-source joint estimation method for spectral tracking turntable micro-vibration compensation includes: S1. Acquire servo current data, encoder data, and short exposure image data of the spectral tracking turntable, and perform time alignment of the servo current data, encoder data, and short exposure image data based on a unified time reference; S2. Within the same state estimation period, extract servo current harmonic features from time-aligned servo current data, extract encoder residual features from time-aligned encoder data, and extract image sharpness features from time-aligned short-exposure image data. Based on the frequency consistency between the servo current harmonic features and encoder residual features, and combined with the changes in the image sharpness features, perform cross-validation to obtain vibration candidate frequencies and the three-source observation features corresponding to the vibration candidate frequencies. S3. Input the vibration candidate frequency and the corresponding three-source observation characteristics into the state estimator to perform state estimation for periodic vibration, structural resonance and random external disturbance, and obtain the angular micro-vibration estimation result and resonance frequency. S4. Based on the resonance frequency obtained from the continuous state estimation period, perform stability judgment. When the resonance frequency meets the preset stability condition, update the digital notch filter parameters or command feedforward shaping parameters according to the main servo closed-loop bandwidth constraint. S5. Based on the angle micro-vibration estimation result, generate the micro-vibration compensation amount for the next control cycle, and perform servo compensation on the spectral tracking turntable in combination with the updated digital notch filter parameters or command feedforward shaping parameters. S6. Based on the angle micro-vibration estimation result, predict the angle micro-vibration within the exposure window and obtain the root mean square jitter within the exposure window. Determine the exposure allowable angle jitter threshold based on the focal length, pixel size, and allowable blur amount of the imaging acquisition device. When the root mean square jitter exceeds the exposure allowable angle jitter threshold, delay long integral exposure or perform short exposure resampling. S7. Obtain the encoder residual and image sharpness after performing the servo compensation and exposure control, and input the encoder residual and image sharpness as the observations for the next state estimation cycle into the state estimator.
[0023] By using the above method, the three-source observation, vibration candidate judgment, state estimation, resonance parameter update, servo compensation, exposure control and post-compensation feedback are connected in series, so that the input and output of each step are interconnected, thus forming a micro-vibration estimation and compensation closed loop for the spectral tracking turntable.
[0024] Furthermore, the time alignment based on a unified time reference described in S1 includes: resampling servo current data, encoder data, and short-exposure image data at different sampling rates according to the state estimation period, and determining the time age and data quality for each resampled data stream.
[0025] By resampling three streams of data at different sampling rates and determining the time age and data quality, subsequent state estimation can obtain observational inputs with unified time relationships and quality criteria, thereby reducing the impact of asynchronous sampling on multi-source joint estimation.
[0026] Furthermore, the cross-validation described in S2 includes: when the servo current harmonic characteristics and encoder residual characteristics show corresponding peak values in the same frequency range and the image sharpness decreases synchronously, the corresponding frequency is determined as a candidate vibration frequency; when only the servo current harmonic characteristics show peak values and the encoder residual characteristics and image sharpness characteristics do not show corresponding responses, the corresponding peak value is determined as an electrical disturbance.
[0027] By examining servo current harmonics, encoder residuals, and changes in image sharpness, frequencies with corresponding responses are identified as vibration candidates, and electrical disturbances that only manifest as servo current peaks are excluded, thereby improving the specificity of vibration candidate judgment.
[0028] Furthermore, the state estimation described in S3 includes: determining the observation weights of the corresponding three-source observation features based on the time age and data quality of each data source, and reducing the corresponding observation weights when the data quality of any observation source decreases, so as to update the angular micro-vibration estimation results and resonance frequencies based on the adjusted observation weights.
[0029] By determining the observation weights of the three-source observation features based on the age of the time and the data quality, and reducing the corresponding observation weights when the quality of the observation source decreases, the state estimator can update the angular micro-vibration estimation results according to the actual observation reliability.
[0030] Furthermore, the updating of digital notch filter parameters or command feedforward shaping parameters as described in S4 includes: performing parameter updates after the resonant frequency obtained from the continuous state estimation period satisfies the frequency stability condition, and limiting the operating frequency band of the digital notch filter to not enter the main servo closed-loop bandwidth; and restoring the digital notch filter parameters or command feedforward shaping parameters before the update when the encoder residual increases after the parameter update.
[0031] By updating the digital notch filter parameters or command feedforward shaping parameters after the resonant frequency meets the stability condition, and combining the changes in the main servo closed-loop bandwidth and the updated encoder residual constraint parameters, the online update can take into account both the resonant state changes and the servo closed-loop stability.
[0032] Furthermore, the delayed long integral exposure or the short exposure resampling described in S6 includes: delaying the long integral exposure when the root mean square jitter within the predicted exposure window exceeds the exposure allowable angle jitter threshold, and re-acquiring the root mean square jitter within the delayed exposure window; and performing short exposure resampling when the re-acquired root mean square jitter still exceeds the exposure allowable angle jitter threshold.
[0033] By delaying long integral exposure when the root mean square jitter exceeds the exposure allowable angle jitter threshold, and performing short exposure resampling when the jitter still exceeds the threshold after the delay, the exposure action can be graded and adjusted according to the predicted micro-vibration state.
[0034] Furthermore, the method also includes performing validity checks on each observation source; reducing the corresponding observation weight and tightening the micro-vibration compensation amplitude when any observation source fails; stopping the updating of digital notch filter parameters and command feedforward shaping parameters and performing fixed amplitude limiting compensation when only one observation source remains reliable; and resuming parameter updates when the failed observation source recovers and meets the preset recovery conditions.
[0035] By detecting the effectiveness of the observation source, the corresponding weight is reduced and the compensation amplitude is tightened when the observation fails, and fixed amplitude compensation is performed when only one observation is reliable, so that the state estimation and parameter update under abnormal conditions have a clear fallback path.
[0036] The method described in this embodiment can be executed by a processor calling a computer program, which can be stored in a computer storage medium. When the computer program is executed by the processor, the above-described multi-source joint estimation method for spectral tracking turntable micro-vibration compensation can be implemented.
[0037] Implementation Method 2 This embodiment uses the micro-vibration compensation process of a spectral tracking turntable as a specific example, combined with... Figure 1 The first implementation method will be further explained below.
[0038] like Figure 1 As shown, the method described in this embodiment is executed in the following order: multi-source data time alignment, three-source feature extraction and cross-validation, micro-vibration state estimation, resonance parameter update, servo compensation, exposure control, and feedback correction.
[0039] In S1, servo current data, encoder data, and short-exposure image data from the spectral tracking turntable are acquired, and the three data streams are time-aligned based on a unified time reference. Optionally, the current / encoder sampling rate can be set within the range of 200 Hz-10 kHz; in one specific configuration, the servo current sampling rate can be 1-5 kHz, and the encoder raw count or speed differential sampling rate can be 500 Hz-2 kHz. The insertion interval for camera short-exposure observations can be 50-200 ms, and the short-exposure time can be 1-5 ms. The unified time reference can be provided by any of the PTP clock, FPGA counter, or other clock sources capable of providing unified timing for the three data streams. When resampling the three data streams at a state estimation frequency of 200 Hz, the servo current data can be processed into the RMS / FFT results required for subsequent harmonic feature extraction after anti-aliasing low-pass processing; the encoder data can be processed into subsequent encoder residual features through band-limited differential processing; and the image sharpness features can retain the most recent valid value and be aged by an additional time period. When the time alignment error exceeds 0.5 ms, the observation weight of the corresponding observation can be reduced, so that data with different sampling rates have comparable time relationships and data quality basis within the same state estimation period.
[0040] In S2, the time-aligned three-channel data are processed separately within the same state estimation period. The spectral window length can be set within 64-2048 points; in one vibration candidate identification configuration, the servo current fundamental wave and 2nd-8th harmonic energy E_i(f) can be extracted within a sliding window of 256-2048 points, the velocity residual spectrum E_e(f) is extracted from the encoder data, and the image sharpness feature E_b is formed by the gradient energy decrease and edge diffusion of the short-exposure image. Frequency consistency judgment is performed on the servo current harmonic feature and the encoder residual feature. When both appear with peak values within ±2% of the relative candidate frequency, and the sharpness of the short-exposure image decreases synchronously, the corresponding frequency is determined as the candidate frequency of structural or transmission vibration, and the corresponding three-source observation feature is retained; when only the servo current harmonic has a peak value and the encoder residual feature and image sharpness feature do not have a corresponding response, the corresponding peak value is determined as an electrical disturbance, and the digital notch filter parameters or command feedforward shaping parameters are not updated based on this peak value. Optionally, the cross-validation relationship can be expressed as follows:
[0041] Where PeakValid(f) represents the validity judgment result of candidate frequency f, I represents the indicator function, f_i represents the frequency corresponding to the peak value of servo current harmonics, f_e represents the frequency corresponding to the peak value of encoder residual, f represents the candidate frequency to be judged, BlurDrop represents the amount of image sharpness reduction during short exposure, and τb represents the image sharpness reduction judgment threshold.
[0042] In S3, the candidate vibration frequencies obtained in S2 and their corresponding three-source observation features are input into the state estimator to perform state estimation for periodic vibration, structural resonance, and random external disturbances. Optionally, the state vector can be represented as x=[jitter_periodic,jitter_resonance,jitter_random,f_res,damping], representing the periodic micro-vibration, resonant micro-vibration, random micro-vibration, resonant frequency, and damping-related state, respectively; the observation vector can be represented as z=[current_harmonic,encoder_residual,blur_metric], corresponding to the servo current harmonic characteristics, encoder residual characteristics, and image sharpness characteristics, respectively. The state estimator can be any of EKF, UKF, or a low-order observer, or other state estimators capable of low-order state estimation based on three-source observation features. The observation weights are adjusted according to data quality and time age. In one quality degradation setting, when the data quality of any observation is below 0.5 for five consecutive state estimation cycles, the corresponding observation weight can be reduced to 0.1, and the upper limit of the micro-vibration compensation amplitude can be tightened by 30%. When any observation source fails, the corresponding observation weight is reduced. When only one observation source remains reliable, the updating of digital notch filter parameters and command feedforward shaping parameters is stopped, and fixed amplitude limiting compensation is performed. Parameter updates resume after the failed observation source recovers and meets the recovery conditions. Optionally, the state estimation relationship can be expressed as follows:
[0043] Where, x k represents the state estimation result in the k-th state estimation period, x k+1 represents the state estimation result of the next state estimation cycle, A represents the state transition relationship, B represents the control input relationship, uk represents the control input of the kth state estimation cycle, K(Q) represents the observation gain determined according to the observation quality Q, zi represents the servo current harmonic observation, ze represents the encoder residual observation, and zb represents the image sharpness observation; the notch filter Q value mentioned later represents the notch filter quality factor, which has a different meaning from the observation quality Q.
[0044] In S4, the stability of the resonant frequency obtained from the continuous state estimation period is assessed, and the digital notch filter parameters or command feedforward shaping parameters are updated after the preset stability conditions are met. Optionally, the configurable update period for the notch filter center can be 0.1-5 s; in one update configuration, the resonant center frequency is evaluated every 1 s, and parameter updates are performed only when the vibration candidate peak lasts for 3 windows and the corresponding frequency change is less than 5%. The single update step size of the digital notch filter center frequency does not exceed 3% of the current center frequency, the Q value of the notch filter can be limited to 2-20, and the notch filter frequency band is kept out of the 0-0.7 times range of the main servo closed-loop bandwidth to reduce the impact of online updates on the main control bandwidth. After the parameter update, a 200 ms stability observation process can be set, and the updated encoder residual is obtained; when the encoder residual increases, the digital notch filter parameters or command feedforward shaping parameters before the update are restored. Thus, the resonant frequency obtained in S3 is directly used for control parameter updates, and the update result continues to enter the servo compensation process in S5.
[0045] In S5, the angular micro-vibration estimation result obtained in S3 forms the micro-vibration compensation amount for the next control cycle, and is combined with the updated digital notch filter parameters or command feedforward shaping parameters from S4 to perform servo compensation on the spectral tracking turntable. Optionally, the upper limit of the feedforward compensation amplitude can be set to 80% of the calibrated maximum micro-vibration, and the upper limit of the micro-vibration compensation amplitude can be set according to 1%-20% of the servo-allowed command, so that the compensation command remains within the servo execution range. The digital notch filter parameters are used to constrain the compensation processing of the corresponding resonant frequency band, and the command feedforward shaping parameters are used to map the angular micro-vibration estimation result to the compensation command for the next control cycle. When both digital notch processing and command feedforward shaping are configured simultaneously, the micro-vibration compensation relationship can be expressed as follows:
[0046] Where ucomp represents the control command after performing micro-vibration compensation, ucmd represents the control command before compensation, Fnotch represents the digital notch filtering relationship, and x This represents the state estimation result, Kff represents the command feedforward coefficient, and x represents the state estimation result. Jitter represents the angular micro-vibration estimate obtained from the state estimation result. The encoder residual generated after the servo controller performs micro-vibration compensation is reused as an observation in S7 for the next state estimation cycle.
[0047] In S6, based on the angular micro-vibration estimation result obtained in S3, the angular micro-vibration within the exposure window is predicted, and the root mean square jitter within the exposure window is obtained. The focal length of the imaging acquisition device is denoted as... The pixel size is denoted as Allow fuzzy quantities to be denoted as The integration time is denoted as Texp; the integration time Texp is used to record the time parameters of the current long integration exposure process. The allowable angular jitter threshold τj is determined based on the focal length, pixel size, and allowable blur amount; optionally, the allowable angular jitter threshold can be calculated according to the following formula:
[0048] in, This indicates the allowable angle jitter threshold for exposure. Indicates the allowable ambiguity. Indicates the pixel size. This represents the focal length of the imaging acquisition device. The root mean square jitter within the predicted exposure window is denoted as RMSjitter; optionally, the exposure gating relationship can be expressed as follows:
[0049] Here, ExposureGate represents the exposure gating judgment result, RMSjitter represents the root mean square jitter within the predicted exposure window, and τj represents the allowable angular jitter threshold. When RMSjitter is greater than τj, a long integral exposure is delayed, with a maximum delay time of one scan cycle, and the root mean square jitter within the delayed exposure window is reacquired. If the reacquired root mean square jitter is still greater than τj, a short exposure resampling is performed. Optionally, the multi-frame images obtained from the short exposure resampling can be fused in the backend, or other post-processing methods capable of handling the short exposure resampling results can be used. Thus, the angular micro-vibration estimation results output by S3 enter the servo compensation and exposure control execution branches respectively.
[0050] In S7, the encoder residual and image sharpness after servo compensation and exposure control are acquired, and these are used as the observation inputs for the state estimator in the next state estimation cycle. The compensated encoder residual characterizes the angular motion state after servo execution, and the compensated image sharpness characterizes the correspondence between the imaging result and the micro-vibration state. Together with the servo current observation reacquired in the next state estimation cycle, they form a new three-source observation input. The state estimator updates the observation weights according to the new data quality and time age, and re-observes the angular micro-vibration estimation result and resonant frequency. Optionally, the observation source recovery process can set up and down hysteresis for the dynamic threshold, and exit the conservative state after continuously meeting the recovery conditions. For processing that requires confirmation of the completion of control actions, any actual completion state from the MCU, FPGA, driver layer, or device feedback can be used as the admission information for subsequent processing, or other feedback methods that can confirm the completion state can be used. Thus, the estimation, parameter update, servo compensation, and exposure control results formed in S3 to S6 are fed back to the subsequent state estimation cycle.
[0051] In this embodiment, the electronic device includes a processor and a memory. The memory stores a computer program, and the processor executes the above-described method when executing the computer program. The computer program may also be stored in a computer-readable storage medium, or be executed by the processor as a computer program or instruction in a computer program product to implement the above-described method.
[0052] The embodiments of the present invention have been described above. These embodiments are used to help understand the technical solutions of the present invention and should not be construed as limiting the scope of the technical solutions of the present invention. Equivalent substitutions or conventional modifications made by those skilled in the art to the embodiments without departing from the technical concept of the present invention should all fall within the scope of the technical solutions of the present invention.
Claims
1. A method for compensating micro-vibration of a spectral tracking turntable using multi-source joint estimation, characterized in that, include: S1. Acquire servo current data, encoder data, and short exposure image data of the spectral tracking turntable, and perform time alignment of the servo current data, encoder data, and short exposure image data based on a unified time reference; S2. Within the same state estimation period, extract servo current harmonic features from time-aligned servo current data, extract encoder residual features from time-aligned encoder data, and extract image sharpness features from time-aligned short-exposure image data. Based on the frequency consistency between the servo current harmonic features and encoder residual features, and combined with the changes in the image sharpness features, perform cross-validation to obtain vibration candidate frequencies and the three-source observation features corresponding to the vibration candidate frequencies. S3. Input the vibration candidate frequency and the corresponding three-source observation characteristics into the state estimator to perform state estimation for periodic vibration, structural resonance and random external disturbance, and obtain the angular micro-vibration estimation result and resonance frequency. S4. Based on the resonance frequency obtained from the continuous state estimation period, perform stability judgment. When the resonance frequency meets the preset stability condition, update the digital notch filter parameters or command feedforward shaping parameters according to the main servo closed-loop bandwidth constraint. S5. Based on the angle micro-vibration estimation result, generate the micro-vibration compensation amount for the next control cycle, and perform servo compensation on the spectral tracking turntable in combination with the updated digital notch filter parameters or command feedforward shaping parameters. S6. Based on the angle micro-vibration estimation result, predict the angle micro-vibration within the exposure window and obtain the root mean square jitter within the exposure window. Determine the exposure allowable angle jitter threshold based on the focal length, pixel size, and allowable blur amount of the imaging acquisition device. When the root mean square jitter exceeds the exposure allowable angle jitter threshold, delay long integral exposure or perform short exposure resampling. S7. Obtain the encoder residual and image sharpness after performing the servo compensation and exposure control, and input the encoder residual and image sharpness as the observations for the next state estimation cycle into the state estimator.
2. The method according to claim 1, characterized in that, S1 describes time alignment based on a unified time reference, which includes: resampling servo current data, encoder data, and short-exposure image data at different sampling rates according to the state estimation period, and determining the time age and data quality for each resampled data stream.
3. The method according to claim 2, characterized in that, The cross-validation described in S2 includes: when the servo current harmonic characteristics and encoder residual characteristics show corresponding peak values in the same frequency range and the image sharpness decreases synchronously, the corresponding frequency is determined as a candidate vibration frequency; when only the servo current harmonic characteristics show peak values and the encoder residual characteristics and image sharpness characteristics do not show corresponding responses, the corresponding peak value is determined as an electrical disturbance.
4. The method according to claim 3, characterized in that, The state estimation described in S3 includes: determining the observation weights of the corresponding three-source observation features based on the time age and data quality of each data source, and reducing the corresponding observation weights when the data quality of any observation source decreases, so as to update the angular micro-vibration estimation results and resonance frequencies based on the adjusted observation weights.
5. The method according to claim 4, characterized in that, S4 describes updating the digital notch filter parameters or command feedforward shaping parameters, which includes: performing parameter updates after the resonant frequency obtained from the continuous state estimation period meets the frequency stability condition, and limiting the operating frequency band of the digital notch filter to not enter the main servo closed-loop bandwidth; and restoring the digital notch filter parameters or command feedforward shaping parameters before the update when the encoder residual increases after the parameter update.
6. The method according to claim 1, characterized in that, S6, which describes delaying long integral exposure or performing short exposure resampling, includes: delaying long integral exposure when the root mean square jitter within the predicted exposure window exceeds the exposure allowable angle jitter threshold, and re-acquiring the root mean square jitter within the delayed exposure window; and performing short exposure resampling when the re-acquired root mean square jitter still exceeds the exposure allowable angle jitter threshold.
7. The method according to claim 4, characterized in that, The method also includes detecting the effectiveness of each observation source; reducing the corresponding observation weight and tightening the micro-vibration compensation amplitude when any observation source fails; stopping the updating of digital notch filter parameters and command feedforward shaping parameters and performing fixed amplitude limiting compensation when only one observation source is reliable; The recovery parameters are updated when the failed observation source is recovered and the preset recovery conditions are met.
8. An electronic device comprising a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to perform the method according to any one of claims 1 to 7.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it performs the method according to any one of claims 1 to 7.
10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the method described in any one of claims 1 to 7.
