A flight system feedback control method, device, equipment and storage medium

Abstract

The present application relates to the technical field of image detection, and discloses a flight system feedback control method, device, equipment and storage medium; the method comprises the following steps: performing feature extraction on a visual image based on a preset maximum absolute gradient evaluation function to obtain a definition sequence; performing analysis on the definition sequence according to a preset offline calibration inverse function and a preset adaptive threshold rule to obtain a defocus amount sequence and an adaptive analysis result; performing compliance analysis on the defocus amount sequence according to the adaptive analysis result and a preset upper limit of focusing time to obtain a compliance analysis result; and generating a flight feedback control signal according to system matrix parameters, system states, a feedforward instruction and the compliance analysis result; the definition sequence of the image is extracted efficiently, the offline calibration and the adaptive threshold are combined to quickly adapt to a complex scene, the upper limit of focusing time is relied on to avoid noise misjudgment, and finally the flight feedback control signal is generated, so that the focusing response speed under a high-speed flight working condition is effectively improved.

Description

Technical Field

[0001] This invention relates to the field of image detection technology, and in particular to a flight system feedback control method, apparatus, device, and storage medium. Background Technology

[0002] In the field of high-speed online visual inspection of high-generation flat panel displays, the in-flight autofocus module is crucial for ensuring the identification of microscopic defects in the panel, and its focusing performance directly affects the inspection accuracy and efficiency of the production line. Existing autofocus control systems have significant technical shortcomings: image sharpness evaluation methods are simplistic, edge feature characterization is incomplete, and defocus calculations often rely on fixed mapping relationships, making them difficult to adapt to dynamically changing working conditions; focusing thresholds are mostly fixed settings, making them highly susceptible to interference from ambient noise and motion disturbances, leading to misjudgments; and there is a lack of focusing time constraints, resulting in significant control response lag. Furthermore, the control system is not optimized in conjunction with the dynamic characteristics of the actuators and the system state, lacks a feedforward compensation mechanism, has weak suppression capabilities against multi-source disturbances such as panel speed fluctuations and mechanical vibrations, and its macro-micro drive strategies are simplistic and crude. These problems collectively result in insufficient focusing accuracy, slow response speed, and poor operational stability, failing to meet the high-precision and robust focusing requirements of G8.5 and larger large-size OLED and LCD panels in short-time-sequence in-flight inspection, severely limiting the large-scale engineering application of high-end panel intelligent inspection equipment. Summary of the Invention

[0003] In order to overcome the shortcomings of the prior art, the present invention aims to provide a flight system feedback control method, device, equipment and storage medium.

[0004] A flight system feedback control method includes: acquiring visual images and extracting features from the visual images based on a preset maximum absolute gradient evaluation function to obtain a sharpness sequence; analyzing the sharpness sequence according to a preset offline calibration inverse function and a preset adaptive threshold rule to obtain a defocus sequence and adaptive analysis results; performing compliance analysis on the defocus sequence according to the adaptive analysis results and a preset focusing time upper limit to obtain compliance analysis results; acquiring system matrix parameters, system state, and feedforward commands; and generating flight feedback control signals based on the system matrix parameters, system state, feedforward commands, and compliance analysis results.

[0005] Furthermore, the step of analyzing the sharpness sequence according to a preset offline calibration inverse function and a preset adaptive threshold rule to obtain a defocus sequence and adaptive analysis results includes: converting the sharpness sequence according to the offline calibration inverse function to obtain a defocus sequence; collecting panel motion speed parameters and calculating the panel motion speed parameters according to a preset dynamic threshold calculation formula to obtain a dynamic threshold; and analyzing the defocus sequence according to the adaptive threshold rule and the dynamic threshold to obtain adaptive analysis results.

[0006] Furthermore, the step of performing compliance analysis on the defocus quantity sequence based on the adaptive analysis results and the preset focusing time upper limit to obtain compliance analysis results includes: optimizing the defocus quantity sequence based on the pre-trained macro-micro dynamic mapping model and the adaptive analysis results to obtain an optimized defocus quantity sequence; and performing compliance analysis on the optimized defocus quantity sequence based on the adaptive threshold rules and the focusing time upper limit to obtain compliance analysis results.

[0007] Furthermore, the step of generating flight feedback control signals based on system matrix parameters, system state, feedforward commands, and compliance analysis results includes: if the compliance analysis result indicates that the optimized defocus sequence is compliant, then acquiring the original grating signal; calculating the error of the original grating signal based on a preset initial calibration offset and a pre-trained error compensation model to obtain the equivalent displacement error; acquiring the original defocus amount and calculating the difference between the equivalent displacement error and the original defocus amount to obtain the corrected defocus amount; generating piezoelectric commands based on system matrix parameters, system state, and feedforward commands; and generating flight feedback control signals based on the corrected defocus amount and the piezoelectric commands.

[0008] Further, the step of calculating the error of the original grating signal according to the preset initial calibration offset and the pre-trained error compensation model to obtain the equivalent displacement error includes: filtering the original grating signal to obtain a filtered grating signal; performing zero-drift calibration on the filtered grating signal according to the initial calibration offset to obtain a zero-drift calibration grating signal; acquiring real-time temperature change values ​​and performing temperature compensation on the zero-drift calibration grating signal according to the real-time temperature change values ​​to obtain a compensated grating signal; performing differential operation on the compensated grating signal to obtain the motion speed; and calculating the error of the motion speed according to the error compensation model to obtain the equivalent displacement error.

[0009] Furthermore, the step of generating piezoelectric commands based on system matrix parameters, system state, and feedforward commands includes: solving a preset discrete algebraic equation based on a preset sampling period and system matrix parameters to obtain a positive definite solution matrix; calculating the system matrix parameters and the positive definite solution matrix based on a preset discrete gain calculation formula to obtain a discrete gain; calculating a correction command based on the discrete gain and system state; and generating piezoelectric commands based on the correction command and feedforward commands.

[0010] Furthermore, the step of generating a flight feedback control signal based on the corrected defocus amount and piezoelectric command includes: calculating the optimized defocus amount sequence based on a preset focus scoring mapping function to obtain a focus scoring sequence; predicting the optimized defocus amount sequence and focus scoring sequence according to a preset predictive control method to obtain a future score prediction sequence and a piezoelectric ceramic drive increment sequence; and generating a flight feedback control signal based on the corrected defocus amount, piezoelectric command, future score prediction sequence, and piezoelectric ceramic drive increment sequence.

[0011] Furthermore, a flight system feedback control device includes: a feature extraction module for acquiring visual images and extracting features from the visual images based on a preset maximum absolute gradient evaluation function to obtain a sharpness sequence; a first analysis module for analyzing the sharpness sequence according to a preset offline calibration inverse function and a preset adaptive threshold rule to obtain a defocus sequence and an adaptive analysis result; a second analysis module for performing compliance analysis on the defocus sequence according to the adaptive analysis result and a preset focusing time upper limit to obtain a compliance analysis result; a parameter acquisition module for acquiring system matrix parameters, system status, and feedforward commands; and a signal generation module for generating flight feedback control signals according to the system matrix parameters, system status, feedforward commands, and compliance analysis results.

[0012] Furthermore, a flight system feedback control device includes: a memory and at least one processor, the memory storing instructions; at least one processor invokes the instructions in the memory to cause the flight system feedback control device to perform the various steps of the flight system feedback control method as described above.

[0013] Furthermore, a computer-readable storage medium stores instructions that, when executed by a processor, implement the various steps of the flight system feedback control method described above.

[0014] In the technical solution of this invention, the visual image sharpness sequence is extracted through the maximum absolute gradient evaluation function, which can comprehensively reflect the image focus status and is computationally efficient. Relying on the offline calibration inverse function and adaptive threshold rules, the adaptive analysis results of the defocus amount sequence and working conditions can be quickly obtained, adapting to complex motion scenarios. In combination with the upper limit of focus time, compliance analysis can be carried out to determine the focus status in multiple dimensions, avoid noise misjudgment, and ensure the timeliness of judgment. Finally, the flight feedback control signal is generated by comprehensively considering system parameters, system status, and compliance analysis results, which can compensate for lag in advance, optimize control timing, and improve the focus response speed, positioning accuracy, and system robustness under high-speed flight conditions. Attached Figure Description

[0015] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a first flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 2 This is a second flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 3 A third flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 4 This is a fourth flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 5 A fifth flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 6 A sixth flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 7 A seventh flowchart of a flight system feedback control method provided in an embodiment of the present invention; Figure 8 This is a schematic diagram of the structure of a flight system feedback control device provided in an embodiment of the present invention; Figure 9 This is a schematic diagram of the structure of a flight system feedback control device provided in an embodiment of the present invention. Detailed Implementation

[0016] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0017] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the flight system feedback control method of the present invention includes: 101. Acquire visual images and extract features from the visual images based on the preset maximum absolute gradient evaluation function to obtain a sharpness sequence; In this embodiment, visual images (1280×1024 pixels) Evaluation function using Maximum Absolute Gradient (MAG): In the formula, The image sharpness value at the current time t; The number of pixels per row in the image. The number of pixels in each column of the image; For horizontal gradient, For vertical gradient, The method involves: first, acquiring visual images with a resolution of 1280×1024 pixels in real time as the raw data for sharpness evaluation; then, extracting features from the image based on the preset Maximum Absolute Gradient (MAG) evaluation function, calculating the horizontal, vertical, and diagonal gradients of each pixel in the image, taking the maximum value of the gradients in the three directions for each pixel, and performing mean normalization on the maximum gradient values ​​of all pixels in the entire image to obtain the sharpness value at the current time t; continuously calculating the sharpness values ​​of multiple frames of visual images according to the time series to finally form a complete sharpness sequence, providing basic quantitative data for subsequent offline calibration and conversion of defocus amount; 102. Analyze the sharpness sequence according to the preset offline calibration inverse function and the preset adaptive threshold rule to obtain the defocus sequence and adaptive analysis results; In this embodiment, by analyzing the sharpness sequence through offline calibration of the inverse function, the defocus sequence can be quickly obtained from the image sharpness. At the same time, the working condition adaptation analysis is completed by combining the adaptive threshold rule, and the adaptive analysis results are output, providing a basis for macro and micro drive allocation. The overall solution takes into account the real-time performance of defocus calculation, and the dynamic threshold can be adapted to different motion conditions, effectively improving the system's adaptability to complex scenes and laying a reliable data foundation for subsequent focus control. 103. Perform compliance analysis on the defocusing amount sequence based on the adaptive analysis results and the preset focus time limit to obtain the compliance analysis results; In this embodiment, by combining the adaptive analysis results with the upper limit of focusing time, a multi-dimensional compliance analysis of the defocus sequence is performed, which can comprehensively determine the focusing status from the aspects of focusing accuracy, process stability and response time. Relying on adaptive rules can effectively avoid the misjudgment problem caused by noise interference of fixed thresholds, while following the upper limit constraint of focusing time to ensure that the judgment process is fast and efficient. This enables the system to stably complete the focus judgment and control switching even under high-speed motion conditions. 104. Obtain system matrix parameters, system status, and feedforward instructions; 105. Generate flight feedback control signals based on system matrix parameters, system status, feedforward commands, and compliance analysis results; In this embodiment, flight feedback control signals are generated by integrating system matrix parameters, real-time system status, feedforward commands, and compliance analysis results. These signals can match the dynamic characteristics of the system, compensate for response lag in advance based on feedforward commands, and ensure reasonable control timing by combining compliance results. This effectively improves the response speed and positioning accuracy of focus control and enhances the stability and robustness of the system under high-speed flight conditions. In this embodiment, the visual image sharpness sequence is extracted using the maximum absolute gradient evaluation function, which can comprehensively reflect the image focus status and is computationally efficient. Relying on the offline calibration inverse function and adaptive threshold rules, the adaptive analysis results of the defocus amount sequence and working conditions can be quickly obtained, adapting to complex motion scenarios. In conjunction with the upper limit of focus time, compliance analysis can be carried out to determine the focus status in multiple dimensions, avoid noise misjudgment, and ensure the timeliness of the judgment. Finally, the flight feedback control signal is generated by comprehensively considering system parameters, system status, and compliance analysis results, which can compensate for lag in advance, optimize control timing, and improve the focus response speed, positioning accuracy, and system robustness under high-speed flight conditions.

[0018] Please see Figure 2 In a second embodiment of the feedback control method for a flight system according to the present invention, step 102 includes: 201. Transform the sharpness sequence according to the offline calibration inverse function to obtain the defocus sequence; In this embodiment, a monotonic mapping relationship between sharpness and defocus is established through offline calibration. The sharpness sequence {F(t)} continuously acquired within the control period is converted point by point into the corresponding defocus sequence {d(t)} using the offline calibration inverse function. The expression of the offline calibration inverse function is as follows: In the formula, This represents the real-time defocus amount corresponding to the current time t. This is the positive mapping function established through calibration experiments during the offline phase. It is the inverse function of the forward mapping function, also known as the "offline calibration inverse function," and its function is to determine the resolution value from the known resolution value. Inversely solve for the corresponding defocus amount By applying the inverse function to the sharpness value F(t) at consecutive sampling times, the corresponding defocus sequence can be reconstructed point by point from the sharpness sequence, providing continuous and reliable deviation input for subsequent macro-micro dual-drive focusing control. 202. Collect panel motion speed parameters and calculate the panel motion speed parameters according to the preset dynamic threshold calculation formula to obtain the dynamic threshold; In this embodiment, the expression for the dynamic threshold calculation formula is as follows: , In the formula, Dynamic threshold (unit: mm); The instantaneous speed of the panel (derived from the panel's motion speed parameters, unit: m / s); Stepper motor response time (unit: s, typically 10~15ms); To ensure a safety margin (0.1~0.2mm), a dynamic threshold is generated based on the panel movement speed, stepper motor response time, and safety margin. This effectively overcomes the shortcomings of fixed thresholds, which cannot adapt to different motion conditions. Under high-speed conditions, the dynamic threshold automatically increases, reserving sufficient response time for the stepper motor macro-motion and avoiding focusing failure due to insufficient piezoelectric ceramic travel. Under low-speed conditions, the threshold automatically decreases, switching to micro-adjustment in advance, balancing focusing response speed and control stability, and enhancing the system's adaptability to complex motion scenarios. 203. Analyze the defocus sequence based on the adaptive threshold rule and dynamic threshold to obtain adaptive analysis results; In this embodiment, the adaptive threshold rule is a rule that dynamically adjusts the focus threshold based on the system's real-time motion state, disturbance intensity, and defocusing trend (e.g., adjusting the dynamic threshold). The defocus sequence is adjusted with a dynamic threshold based on adaptive threshold rules. Point-by-point comparison yields adaptive analysis results, clarifying the drive allocation strategy at each moment; macro-motion stage control ( : If the defocus deviation exceeds the effective range of micro-motion adjustment, priority is given to allocating macro-motion to the stepper motor. Utilizing the stepper motor's large stroke and high thrust characteristics, the defocus deviation is quickly reduced, addressing the need for rapid focusing under large deviations, while avoiding the problem of piezoelectric ceramics failing to converge quickly due to stroke limitations; micro-motion stage control ( ): Once the defocus deviation is determined to have entered the micro-motion adjustment range, piezoelectric ceramic micro-motion is prioritized for allocation, achieving high-precision focusing convergence; simultaneously, flight speed feedforward correction is introduced, calculating the feedforward compensation amount based on real-time flight speed to compensate for defocus changes caused by platform movement in advance, improving focusing response speed under high-speed conditions; adaptive threshold dynamic adaptation: the adaptive threshold rule adjusts the dynamic threshold in real time based on flight speed, defocus change rate, and system noise level. For example, when the flight speed increases, the dynamic threshold is automatically increased. Extend the duration of the macro-motion phase; automatically reduce the dynamic threshold when the defocus deviation decreases. Switch to micro-adjustment in advance to balance response speed and adjustment accuracy; In this embodiment, the solution converts the sharpness sequence point by point into a defocus sequence through offline calibration of the inverse function, providing precise input for focus control. A dynamic threshold is constructed by combining panel movement speed, motor response time, and safety margin, which can adaptively match different motion conditions, solving the problem of poor applicability of fixed thresholds. Combined with adaptive threshold rules, the dynamic threshold size is optimized in real time, enabling intelligent analysis of the defocus sequence and allocation of drive strategies. Simultaneously, a flight speed feedforward correction is introduced to compensate for defocus offset caused by platform movement. The overall solution balances focusing speed under large defocus and control accuracy under small defocus, effectively improving focusing response speed and stability in high-speed motion scenarios, enhancing the system's adaptability to complex conditions, and ensuring efficient, reliable, and accurate focusing.

[0019] Please see Figure 3 In a third embodiment of the feedback control method for a flight system according to the present invention, step 103 includes: 301. The defocus sequence is optimized based on the pre-trained macro-micro dynamic mapping model and adaptive analysis results to obtain an optimized defocus sequence. In this embodiment, the macro-micro dynamic mapping model is a model based on LSTM pre-training. By pre-training the LSTM macro-micro dynamic mapping model and combining the adaptive analysis results, the original defocus quantity sequence is time-series optimized to obtain an optimized defocus quantity sequence that is more suitable for macro-micro dual-drive control. The LSTM model possesses excellent temporal feature learning capabilities. In the offline phase, it undergoes pre-training using extensive operating data to learn the dynamic mapping patterns between the temporal changes in defocus amount, panel motion speed, macro / micro motion switching rules, and system disturbances. During online operation, the original defocus amount sequence and adaptive analysis results (drive allocation strategy, real-time motion status, etc.) are input into the model, which performs temporal correction, noise suppression, and deviation compensation, ultimately outputting a smooth, optimized defocus amount sequence that meets control requirements. This approach leverages the long short-term memory characteristics of the LSTM network to capture the temporal change trend and dynamic correlation features of the defocus amount sequence. Simultaneously, it integrates prior information from the adaptive analysis results, such as macro / micro motion allocation, motion speed, and disturbance intensity, to specifically correct measurement noise, transmission lag, and transient errors in the original defocus amount sequence. Furthermore, based on the different drive characteristics of large macro motion strokes and high micro motion precision, it dynamically adjusts the defocus amount values ​​and gradients, eliminating abnormal jump points to make the defocus amount sequence more closely match the actual motion and actuator response patterns, thus achieving temporal optimization of the defocus amount. 302. Perform compliance analysis on the optimized defocus sequence based on the adaptive threshold rule and the upper limit of focusing time to obtain the compliance analysis results; In this embodiment, the adaptive threshold rule dynamically adjusts the focus threshold based on the system's real-time motion state, disturbance intensity, and defocus change trend, rather than using a fixed threshold. This adapts to changes in high-speed focusing conditions and avoids misjudgments caused by noise or instantaneous jitter that can easily affect fixed thresholds. The principle of compliance analysis is to combine the optimized defocus sequence with the adaptive threshold. The upper limit of the focus time is checked to determine whether the defocusing amount is continuous. The condition |d(t)| ≤ And within the maximum focusing time If the focus is achieved within ≤30ms, it is determined whether the focus is complete and the compliance analysis result is output. The compliance analysis result is a comprehensive compliance judgment conclusion on the optimized defocus sequence. In essence, it is to determine whether the current focus state meets the triple constraints of "accuracy met, stability maintained, and time controlled" and outputs a clear control instruction: either complete the focus within 30ms and terminate the adjustment, or continue the control iteration of the next sampling cycle. In this embodiment, a pre-trained LSTM macro-micro dynamic mapping model is used in conjunction with adaptive analysis results to optimize the defocus sequence. Relying on the temporal feature learning capability of LSTM, the defocus variation pattern can be captured, effectively suppressing measurement noise, transmission lag, and transient errors in drive switching, and eliminating abnormal jump points. This makes the optimized defocus sequence more consistent with the macro-micro dual-drive characteristics, providing a smooth and accurate input for focus control. At the same time, compliance analysis is carried out by combining adaptive threshold rules with the upper limit of focus time. The dynamic threshold can be adapted to high-speed focusing conditions. The overall solution not only improves the reliability and adaptability of the defocus data, but also achieves accurate compliance judgment of the focusing process, improving the stability, robustness, and control efficiency of the focusing system in high-speed motion scenarios, and ensuring that the focusing process is completed efficiently and reliably.

[0020] Please see Figure 4 In a fourth embodiment of the flight system feedback control method of the present invention, step 105 includes: 401. If the compliance analysis result shows that the optimized defocus sequence is compliant, then collect the original grating signal; In this embodiment, the LSTM model is used to optimize the defocus sequence, which can suppress errors, eliminate outliers, and make the data more suitable for macro and micro drives; compliance analysis is carried out by combining adaptive threshold rules and time constraints to avoid misjudgment and improve system stability and efficiency. 402. Calculate the error of the original grating signal based on the preset initial calibration offset and the pre-trained error compensation model to obtain the equivalent displacement error; In this embodiment, by combining the preset initial calibration offset and the error compensation model to perform error calculation on the original grating signal, the inherent zero drift deviation of the system can be eliminated simultaneously, and the dynamic displacement error caused by the flexible deformation of the mechanism, load change and motion damping can be calculated to obtain the accurate equivalent displacement error. This method can comprehensively correct the static and dynamic measurement deviations of the grating signal, improve the displacement detection accuracy, provide a reliable error correction basis for subsequent defocusing optimization and focus control, and effectively improve the control accuracy and operational stability of the focusing system under high-speed motion conditions. 403. Collect the original defocus amount, and calculate the difference between the equivalent displacement error and the original defocus amount to obtain the corrected defocus amount; In this embodiment, the corrected defocus amount (i.e., the corrected defocus amount) is finally obtained using a compensation formula: ; In the formula, To correct the out-of-focus amount, This refers to the original defocus amount obtained through measurement (e.g., direct reading from a grating ruler or sensor); The equivalent displacement error is calculated by using an error compensation model to calculate the equivalent displacement error caused by the flexible load and motion speed. Then, the error is removed from the original measured displacement to obtain the corrected defocus amount. This method can effectively compensate for the measurement deviation caused by the flexible deformation and dynamic damping of the mechanism, improve the accuracy of defocus amount detection, provide reliable input for subsequent state estimation and robust control, and thus enhance the focusing stability and control robustness of the system under high-speed flight conditions. 404. Generate piezoelectric commands based on system matrix parameters, system state, and feedforward commands; In this embodiment, piezoelectric commands are generated by combining system matrix parameters, system state, and feedforward commands. This can match the dynamic characteristics and motion requirements of the system, compensate for dynamic deviations and response lags in advance, make the piezoelectric ceramic drive commands more in line with actual working conditions, improve the response speed and positioning accuracy of micro-motion control, and ensure the stability of the high-speed focusing process. 405. Generate flight feedback control signals based on the defocus correction amount and piezoelectric commands; In this embodiment, the flight feedback control signal is generated by combining the correction of defocus amount and piezoelectric command to realize focus control. The piezoelectric ceramic action can be adjusted to compensate for defocus deviation in real time, improve the response speed and positioning accuracy of micro-motion control, and effectively ensure the stability and control accuracy of the focusing system under high-speed flight conditions. In this embodiment, after compliance analysis determines that the defocus sequence is compliant, the original grating signal is acquired. Optimizing the defocus sequence using an LSTM model can effectively suppress errors and eliminate outliers, making it suitable for macro-micro dual-drive control. Compliance analysis combined with adaptive thresholds and time constraints can avoid misjudgments and improve system stability and efficiency. The equivalent displacement error is calculated by initial calibration offset and error compensation model, which can eliminate system zero drift deviation and correct dynamic measurement deviations caused by flexible deformation of the mechanism, load changes, and motion damping. The difference between the original defocus and the equivalent displacement error is used to obtain the corrected defocus, improving the defocus detection accuracy. Combined with system matrix parameters, system state, and feedforward commands, piezoelectric commands are generated to match the system's dynamic characteristics and compensate for response lag. Then, flight feedback control signals are generated based on the corrected defocus and piezoelectric commands to achieve focus control, real-time compensation for defocus deviation, and improved response speed and positioning accuracy of micro-motion control under high-speed flight conditions, enhancing the stability, accuracy, and robustness of the focusing system.

[0021] Please see Figure 5 In the fifth embodiment of the flight system feedback control method of the present invention, step 402 includes: 501. Filter the original grating signal to obtain the filtered grating signal; In this embodiment, the original grating signal is subjected to Kalman filtering to obtain a filtered grating signal, suppressing noise interference in the original grating signal and restoring the true characteristics of the signal. First, a state prediction model is established based on the dynamic characteristics of the grating signal. Based on the filtered signal at the previous time step, the theoretical value of the grating signal at the current time step is predicted. Then, the original grating signal at the current time step is acquired and compared with the predicted value to calculate the observation residual. Finally, the Kalman gain is calculated by combining the preset process noise variance and observation noise variance. The predicted value is then corrected using this Kalman gain to obtain the filtered grating signal at the current time step. Through continuous iteration, random noise in the original grating signal is effectively filtered out, improving signal stability and accuracy, and providing reliable input for subsequent signal processing. 502. Perform zero-drift calibration on the filtered grating signal based on the initial calibration offset to obtain the zero-drift calibration grating signal; In this embodiment, the zero-drift calibration formula is as follows: , In the formula, To calibrate the grating signal for zero drift, This is the initial calibration offset. The filter grating signal is calibrated with an initial calibration offset to eliminate constant zero drift error caused by inherent zero-point offset of the grating sensor and system assembly deviation. This effectively avoids the continuous accumulation of zero drift, which can lead to systematic deviations in subsequent displacement and velocity calculations. This calibration method is efficient and easy to implement, improving the reference accuracy and consistency of the grating signal and providing a reliable signal basis for subsequent temperature compensation, differential velocity measurement, and equivalent displacement error calculation. 503. Acquire real-time temperature change values ​​and perform temperature compensation on the zero-drift calibration grating signal based on the real-time temperature change values ​​to obtain the compensation grating signal; In this embodiment, the formula for calculating temperature compensation is as follows: In the formula, The coefficient of thermal expansion is This represents the real-time temperature change value. To compensate for the grating signal, real-time temperature changes are collected, and the grating signal after zero-drift calibration is compensated by combining the thermal expansion coefficient of the grating material. This effectively eliminates the measurement error caused by temperature drift and thermal expansion and contraction, improves the accuracy and stability of displacement detection, and provides a reliable input for subsequent defocus calculation and focus control. This enhances the system's adaptability to different ambient temperatures and ensures the robustness of focus control under high-speed flight conditions. 504. Perform differential operations on the compensation grating signal to obtain the motion speed; In this embodiment, specifically, the compensation grating signal at two adjacent sampling times is acquired, the difference between the two (i.e., the displacement change) is calculated, and then the displacement change is divided by the time interval between the two sampling times (i.e., the sampling period Ts) to obtain the motion speed. This operation quickly calculates the instantaneous speed of the moving part by capturing the real-time changes of the grating signal. The operation is simple and can provide speed input for subsequent error compensation and focus control. 505. Calculate the error of the motion velocity based on the error compensation model to obtain the equivalent displacement error; In this embodiment, the expression for the error compensation model is as follows: In the formula, This is the flexibility stiffness coefficient; The load (which can be measured by force / strain sensors or estimated online by an extended state observer); The damping coefficient; The motion speed is used as the reference. By using an error compensation model, combined with the flexible stiffness coefficient, damping coefficient, load and motion speed to calculate the equivalent displacement error, the measurement deviation caused by the flexible deformation and dynamic damping of the mechanism can be effectively compensated, the defocus detection accuracy can be improved, and the focusing stability and control robustness of the system under high-speed flight conditions can be enhanced. In this embodiment, the original grating signal is denoised using Kalman filtering, combined with zero-drift calibration and temperature compensation, effectively eliminating measurement errors caused by noise, temperature drift, and thermal expansion and contraction, thus improving signal reliability. Motion speed is obtained through differential calculation, providing support for subsequent error compensation. Simultaneously, an error compensation model is used to correct measurement deviations caused by flexible deformation and load changes, eliminating interference factors in the original displacement measurement. The overall process balances signal purity, control accuracy, and environmental adaptability, adapting to the requirements of high-speed flight conditions. It effectively compensates for deviations caused by mechanism flexibility and damping, comprehensively improving the stability and environmental adaptability of the focusing system, providing reliable input for subsequent control stages, and ensuring the efficient and stable operation of the overall system.

[0022] Please see Figure 6 In the sixth embodiment of the flight system feedback control method of the present invention, step 404 includes: 601. Solve the preset discrete algebraic equations according to the preset sampling period and system matrix parameters to obtain the positive definite solution matrix; In this embodiment, the expression for the discrete algebraic equation is as follows: , , , In the formula, Let A be the sampling period, and A be the state transition matrix of the continuous system. Given a discrete state matrix, calculate the matrix exponent of the continuous system state matrix to obtain the discretized state transition matrix (i.e., the discrete state matrix), which describes the change of the system state within one sampling period. Discrete state matrix The transpose of the matrix, It is a positive definite solution matrix; Let be the perturbation input matrix in a continuous system. As a discrete input matrix, the continuous disturbance input is discretized through integration to obtain the cumulative effect of the disturbance signal on the system state within one sampling period (i.e., the discrete input matrix). Discrete input matrix The transpose of the matrix; Let be the integral variable, representing the time from 0 to the sampling period. The time is used to calculate the cumulative effect of the input signal on the system state of a continuous system within one sampling period; For the control input matrix of a continuous system, To obtain the discretized control input matrix, the continuous control input is discretized to obtain the cumulative effect of the control signal on the system state within one sampling period (i.e., the control input matrix). The preset H∞ performance index (disturbance suppression level) determines the system's ability to suppress external disturbances. It is the identity matrix; For the output matrix of a continuous system, an evaluation index for control performance is defined; For the output matrix The transpose matrix; by using the preset sampling period and system matrix parameters, the positive definite solution matrix is ​​obtained by solving the discrete algebraic equation, which can theoretically guarantee the stability of the system and have the preset H∞ disturbance suppression capability. Discretization and equation solving are completed offline, without occupying online computing resources, meeting the short response time requirements of flight focusing, providing a reliable foundation for subsequent feedback control, and effectively improving focusing accuracy and system robustness. 602. Calculate the system matrix parameters and positive definite solution matrix according to the preset discrete gain calculation formula to obtain the discrete gain; In this embodiment, the discrete gain calculation formula is the discrete Riccati gain calculation formula, and the expression of the discrete gain calculation formula is as follows: In the formula, The discrete gain is obtained by using the discrete Riccati gain calculation formula based on the system matrix parameters and the positive definite solution matrix obtained offline. Theoretically, this can guarantee the stability and disturbance suppression capability of H∞ robust feedback control. This calculation can be completed offline without complex calculations online, providing a directly usable control gain for the generation of subsequent correction commands, realizing error correction of feedforward commands, taking into account the system response speed and robustness, and effectively improving the focusing control accuracy and system reliability. 603. Calculate based on discrete gain and system state to obtain correction instructions; In this embodiment, this step is the implementation stage of the H∞ robust feedback controller. It employs a discretized robust control law and generates correction commands online using only simple matrix multiplication. The controller, based on the solved discrete gain, performs multiplication calculations with the real-time system state estimate, outputting correction commands in the opposite direction to the error. This is directly used to dynamically correct the feedforward commands, fully adapting to the stringent time constraints of OLED panel flight focusing. This achieves real-time suppression of model uncertainties, external disturbances, hysteresis errors, and unmodeled dynamics. The expression for the simple matrix multiplication operation is as follows: , In the formula, This is a correction instruction used to correct errors in feedforward instructions. For system state estimation, the negative sign indicates that the control action is opposite to the error direction, realizing negative feedback correction; firstly, the system state at the current moment is predicted based on the system state estimate and control input at the previous moment, and then the residual between the predicted output and the actual measured output is corrected by the pre-designed observer gain feedback to obtain the system state estimate containing the internal state information of the system that cannot be directly measured, providing a reliable state signal for subsequent negative feedback control; 604. Generate piezoelectric commands based on correction and feedforward commands; In this embodiment, the calculation expression for the piezoelectric command is: , In the formula, For feedforward instructions, The piezoelectric command is generated by superimposing feedforward and correction commands. It organically combines the fast predictive feedforward of LSTM-MPC with the robust feedback correction of H∞. It can quickly track defocus changes and shorten response time through feedforward to meet the requirements of short time window focusing. At the same time, it can suppress disturbances, nonlinearity and model errors through feedback correction to improve focusing accuracy and stability. The superposition of the two makes the piezoelectric command both fast and robust, which can realize micro-motion control under high-speed flight conditions and effectively improve the overall focusing performance and reliability. In this embodiment, a positive definite solution matrix is ​​obtained by solving discrete algebraic equations offline, providing theoretical stability and disturbance suppression capabilities for the system. All calculations are completed offline, without consuming online resources, thus meeting the short response time requirements for flight focusing. Discrete gain is calculated based on the positive definite solution matrix, and correction commands are generated online only through matrix multiplication to achieve dynamic error correction of feedforward commands, effectively suppressing model uncertainty, external disturbances, and hysteresis errors. Finally, the feedforward command and correction command are superimposed to generate piezoelectric commands, organically combining the fast predictive feedforward of LSTM-MPC with H∞ robust feedback correction. This not only allows for rapid tracking of defocus changes and shortens the response time to meet the short time window focusing requirements through feedforward, but also suppresses disturbances, nonlinearities, and model errors through feedback correction, improving focusing accuracy and stability. This makes the piezoelectric commands both fast and robust, achieving precise and stable micro-motion control under high-speed flight conditions, improving overall focusing performance and system reliability.

[0023] Please see Figure 7 In the seventh embodiment of the flight system feedback control method of the present invention, step 405 includes: 701. Calculate the optimized defocus sequence based on the preset focus score mapping function to obtain the focus score sequence; In this embodiment, the expression for the focus rating mapping function is: In the formula, The optimized real-time defocus amount corresponding to the current control moment is a real physical quantity that reflects the actual defocus state of the system; The focus score corresponding to the current control moment is a continuous evaluation index used for control optimization. The monotonic nonlinear focus scoring mapping function, which is obtained by pre-calibration offline, can establish a stable, distortion-free, and local extremum-free one-to-one correspondence between the amount of defocus and the focus sharpness, thus avoiding the problem that the focus scoring evaluation function is prone to getting trapped in local optima. 702. Based on the preset predictive control method, the optimized defocusing sequence and focus scoring sequence are predicted to obtain the future scoring prediction sequence and the piezoelectric ceramic drive increment sequence. In this embodiment, the prediction control method is a prediction control method based on an LSTM-MPC fusion prediction network. The prediction steps specifically include: continuously acquiring 5 to 7 frames of visual images and generating an optimized defocus sequence accordingly. and the focus rating sequence: ; Indicates the current time and continuous An optimized defocus sequence consisting of optimized defocus amounts at each historical moment. This is the current control time, expressed in units of sampling time. for Optimize the defocus amount corresponding to the control time. for Optimize the defocus amount corresponding to the control time. For sequence length, Indicates the current time and continuous A sequence of focus ratings composed of focus ratings from historical moments. for The focus score corresponding to the control moment. for The focus score corresponding to the control moment is used as the input feature of the optimized defocus sequence and focus score sequence as LSTM-MPC fusion prediction network. The LSTM network is used to model the hysteresis characteristics of the actuator and the dynamic response of the system to complete the forward prediction and output the predicted focus score value for the next j steps. and piezoelectric ceramic driven incremental The LSTM-MPC fusion prediction network takes the defocusing sequence and focus score sequence as input. First, the LSTM network extracts the temporal features of defocusing and focus score changes at multiple historical moments, learns and fits the hysteresis nonlinearity of the piezoelectric ceramic actuator, the system dynamic response delay, and the disturbance law caused by the high-speed movement of the panel during flight focusing, transforming the nonlinear dynamics that cannot be directly modeled into predictable temporal relationships, thereby outputting the focus score prediction values ​​at multiple future moments, forming the future score prediction sequence. Then, the prediction result is fed into the MPC model predictive controller, with the focus score rapidly approaching the target value as the optimization objective and the piezoelectric driving quantity not changing abruptly and not exceeding the range as the constraint to construct the optimization objective function. The optimal control increment is solved by rolling time domain, and finally outputs the piezoelectric ceramic driving increment sequence that can ensure focusing accuracy, suppress vibration, and match the dynamic characteristics of the system. In a further embodiment, the optimization objective function of the model predictive control (MPC) is constructed based on the future score prediction sequence, and its expression is: In the formula, The objective function value for model predictive control. For the future j-th step, reference focus score is used to determine the target's focus. Through the focus scoring function Mapping results in (usually) (or the optimal value within the tolerance). For the prediction time domain (prediction steps). To control the time domain (actual number of optimization steps). The control smoothing weight coefficient for step j is used to penalize sudden changes in the driving quantity and suppress system vibration. The piezoelectric ceramic driving increment includes voltage increment or current increment, which satisfies the driver amplitude and rate of change constraints. 703. Generate flight feedback control signals based on the corrected defocusing amount, piezoelectric command, future score prediction sequence, and piezoelectric ceramic drive increment sequence; In this embodiment, the defocus correction amount is used as feedback of the system's true defocus state, the piezoelectric command is used as the basic feedback control output, the future score prediction sequence is used as feedforward guidance of the focusing trend, and the piezoelectric ceramic drive incremental sequence is used as a constraint condition for smoothing and vibration suppression. The four types of signals are weighted and fused according to the control logic and limited to form a flight feedback control signal, which is used to directly drive the piezoelectric ceramic actuator to complete the high-precision focusing action. In this embodiment, the optimized defocus amount is converted into a stable and continuous focus score sequence through an offline calibrated monotonic nonlinear focus score mapping function, providing a reliable evaluation basis for subsequent control. An LSTM-MPC fusion prediction network is employed to fully learn the system's dynamics, hysteresis characteristics, and disturbance patterns, predict the focusing trend, and solve for the smoothed optimal piezoelectric drive increment, effectively suppressing vibration and overshoot. Finally, the corrected defocus amount, piezoelectric command, predicted score, and drive increment are fused to form a flight feedback control signal, achieving a balance between feedforward prediction and feedback correction, error compensation and smoothing constraints. This enhances the system's robustness to nonlinearity, load changes, temperature drift, and vibration, adapting to high-speed flight detection conditions of high-generation OLED panels, and demonstrating strong real-time control and a high degree of engineering sophistication.

[0024] The above describes a flight system feedback control method according to an embodiment of the present invention. The following describes a flight system feedback control device according to an embodiment of the present invention. Please refer to [link / reference]. Figure 8 One embodiment of a flight system feedback control device according to the present invention includes: Feature extraction module 1 is used to acquire visual images and extract features from the visual images based on a preset maximum absolute gradient evaluation function to obtain a sharpness sequence; The first analysis module 2 is used to analyze the sharpness sequence according to the preset offline calibration inverse function and the preset adaptive threshold rule to obtain the defocus amount sequence and adaptive analysis results. The second analysis module 3 is used to perform compliance analysis on the defocus amount sequence based on the adaptive analysis results and the preset focus time upper limit, so as to obtain the compliance analysis results. Parameter acquisition module 4 is used to acquire system matrix parameters, system status, and feedforward instructions; Signal generation module 5 is used to generate flight feedback control signals based on system matrix parameters, system status, feedforward commands, and compliance analysis results; In this embodiment, the visual image sharpness sequence is extracted using the maximum absolute gradient evaluation function, which can comprehensively reflect the image focus status and is computationally efficient. Relying on the offline calibration inverse function and adaptive threshold rules, the adaptive analysis results of the defocus amount sequence and working conditions can be quickly obtained, adapting to complex motion scenarios. In conjunction with the upper limit of focus time, compliance analysis can be carried out to determine the focus status in multiple dimensions, avoid noise misjudgment, and ensure the timeliness of the judgment. Finally, the flight feedback control signal is generated by comprehensively considering system parameters, system status, and compliance analysis results, which can compensate for lag in advance, optimize control timing, and improve the focus response speed, positioning accuracy, and system robustness under high-speed flight conditions.

[0025] Figure 9This is a schematic diagram of the structure of a flight system feedback control device 900 provided in an embodiment of the present invention. This flight system feedback control device 900 can vary significantly due to different configurations or performance. It may include one or more central processing units (CPUs) 910 (e.g., one or more processors) and a memory 920, and one or more storage media 930 (e.g., one or more mass storage devices) storing application programs 933 or data 932. The memory 920 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the flight system feedback control device 900. Furthermore, the processor 910 may be configured to communicate with the storage media 930 and execute a series of instruction operations in the storage media 930 on the flight system feedback control device 900 to implement the steps of the flight system feedback control method provided in the above-described method embodiments.

[0026] A flight system feedback control device 900 may further include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Server, MacOSX, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 9 The illustrated structure of a flight system feedback control device does not constitute a limitation on a flight system feedback control device. It may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0027] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a flight system feedback control method.

[0028] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0029] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0030] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A feedback control method for a flight system, characterized in that, include: Visual images are acquired, and features are extracted from the visual images based on a preset maximum absolute gradient evaluation function to obtain a sharpness sequence. The sharpness sequence is analyzed based on the preset offline calibration inverse function and the preset adaptive threshold rule to obtain the defocus sequence and adaptive analysis results. Based on the adaptive analysis results and the preset focus time limit, a compliance analysis is performed on the defocus amount sequence to obtain the compliance analysis results; The compliance analysis of the defocus quantity sequence based on the adaptive analysis results and the preset focus time upper limit, to obtain the compliance analysis results, includes: The defocus sequence is optimized based on the pre-trained macro-micro dynamic mapping model and adaptive analysis results to obtain an optimized defocus sequence. Compliance analysis was performed on the optimized defocus sequence based on adaptive threshold rules and focus time limits to obtain compliance analysis results; Obtain system matrix parameters, system status, and feedforward instructions; Flight feedback control signals are generated based on system matrix parameters, system status, feedforward commands, and compliance analysis results.

2. The flight system feedback control method as described in claim 1, characterized in that, The step of analyzing the sharpness sequence based on a preset offline calibration inverse function and a preset adaptive threshold rule to obtain a defocus sequence and adaptive analysis results includes: The sharpness sequence is transformed using the offline calibration inverse function to obtain the defocus sequence; Collect panel motion speed parameters and calculate the panel motion speed parameters according to the preset dynamic threshold calculation formula to obtain the dynamic threshold; The defocus sequence is analyzed based on adaptive threshold rules and dynamic thresholds to obtain adaptive analysis results.

3. The flight system feedback control method as described in claim 1, characterized in that, The generation of flight feedback control signals based on system matrix parameters, system state, feedforward commands, and compliance analysis results includes: If the compliance analysis result indicates that the optimized defocus sequence is compliant, then the original grating signal is acquired; The original grating signal is subjected to error calculation based on the preset initial calibration offset and the pre-trained error compensation model to obtain the equivalent displacement error. The original defocus amount is collected, and the difference between the equivalent displacement error and the original defocus amount is calculated to obtain the corrected defocus amount. Piezoelectric commands are generated based on system matrix parameters, system status, and feedforward commands; flight feedback control signals are generated based on the defocus correction amount and piezoelectric commands.

4. The flight system feedback control method as described in claim 3, characterized in that, The step of calculating the error of the original grating signal based on the preset initial calibration offset and the pre-trained error compensation model to obtain the equivalent displacement error includes: The original grating signal is filtered to obtain the filtered grating signal; The filtered grating signal is zero-drift calibrated based on the initial calibration offset to obtain a zero-drift calibrated grating signal; Real-time temperature change values ​​are acquired, and the zero-drift calibration grating signal is temperature compensated based on the real-time temperature change values ​​to obtain the compensated grating signal. Differential operations are performed on the compensation grating signal to obtain the motion speed; the motion speed error is calculated according to the error compensation model to obtain the equivalent displacement error.

5. The flight system feedback control method as described in claim 3, characterized in that, The generation of piezoelectric commands based on system matrix parameters, system state, and feedforward commands includes: The preset discrete algebraic equations are solved according to the preset sampling period and system matrix parameters to obtain the positive definite solution matrix; The system matrix parameters and positive definite solution matrix are calculated according to the preset discrete gain calculation formula to obtain the discrete gain; The correction command is calculated based on the discrete gain and system state; the piezoelectric command is generated based on the correction command and the feedforward command.

6. The flight system feedback control method as described in claim 3, characterized in that, The process of generating flight feedback control signals based on the defocus correction amount and piezoelectric commands includes: The focus score sequence is calculated based on a preset focus score mapping function to obtain the focus score sequence. The optimized defocusing sequence and focus scoring sequence are predicted according to the preset predictive control method to obtain the future scoring prediction sequence and the piezoelectric ceramic driving increment sequence. The flight feedback control signal is generated based on the corrected defocusing amount, piezoelectric command, future score prediction sequence, and piezoelectric ceramic drive increment sequence.

7. A flight system feedback control device, characterized in that, include: The feature extraction module is used to acquire visual images and extract features from the visual images based on a preset maximum absolute gradient evaluation function to obtain a sharpness sequence. The first analysis module is used to analyze the sharpness sequence according to the preset offline calibration inverse function and the preset adaptive threshold rule to obtain the defocus sequence and adaptive analysis results. The second analysis module is used to perform compliance analysis on the defocus sequence based on the adaptive analysis results and the preset focus time limit, in order to obtain the compliance analysis results. The specific steps include: The defocus sequence is optimized based on the pre-trained macro-micro dynamic mapping model and adaptive analysis results to obtain an optimized defocus sequence. Compliance analysis was performed on the optimized defocus sequence based on adaptive threshold rules and focus time limits to obtain compliance analysis results; The parameter acquisition module is used to acquire system matrix parameters, system status, and feedforward instructions. The signal generation module is used to generate flight feedback control signals based on system matrix parameters, system status, feedforward commands, and compliance analysis results.

8. A flight system feedback control device, characterized in that, The flight system feedback control device includes: a memory and at least one processor, wherein the memory stores instructions; at least one processor invokes the instructions in the memory to cause the flight system feedback control device to perform the steps of the flight system feedback control method as described in any one of claims 1-6.

9. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the flight system feedback control method as described in any one of claims 1-6.

Images