High-precision error compensation method for optical axis of continuous zoom lens

Abstract

This application discloses a high-precision error compensation method for the optical axis of a continuous zoom lens, relating to the field of lens inspection. The continuous zoom system is disassembled into independent components and mounted on a rotating stage centering fixture. Static center offset data for each component is calculated by rotating and acquiring the image point trajectory circle. Sampling parameters are set according to the zoom ratio of the continuous zoom system and the movement speed of each component along the optical axis during zooming, while simultaneously acquiring dynamic optical axis offset data of the components. A nonlinear mapping relationship is constructed between the static center offset data and the dynamic optical axis offset data, and a center offset-optical axis error coupling model is established. Based on the deviation compensation amount, the actuator is driven to adjust the position and / or angle of the components, and the optical axis accuracy index is verified after compensation. This solution achieves deep coupling between static and dynamic data, improving the accuracy of error tracing and the specificity of compensation; ensuring coverage of key error locations during zooming; supporting real-time compensation, and enhancing the stability and accuracy of the system during zooming.

Description

Technical Field

[0001] This application relates to the field of lens inspection, and in particular to a method for high-precision error compensation of the optical axis of a continuous zoom lens. Background Technology

[0002] The optical axis accuracy of a continuous zoom optical system has a decisive impact on image quality. Traditional optical axis detection and compensation methods typically consist of two separate stages: first, static center offset detection is performed on the lens components to obtain basic data such as eccentricity and tilt angle; then, during the zoom process, a sampling detection method is used, for example, measuring only the focal length endpoint, and calibration is achieved through mechanical structure adjustments.

[0003] While some related technologies achieve optical axis detection throughout the zoom process, they fail to deeply couple static component center offset data with dynamic optical axis offset data, leading to difficulties in error tracing and a lack of targeted compensation schemes. Other schemes introduce temperature compensation strategies, but their sampling density is insufficient to cover key error locations during zooming. Specifically, static and dynamic data are disconnected, making it impossible to accurately identify error sources; low sampling density makes it difficult to capture the continuous variation of optical axis errors in high zoom ratio systems; compensation processes largely rely on offline calibration, failing to correct dynamic errors in real time during zooming and becoming susceptible to interference from ambient temperature fluctuations and mechanical transmission clearances; system stability significantly decreases under extreme high and low temperature conditions; the contribution analysis of error sources is unclear, model construction lacks quantitative basis, and generalization ability is limited; furthermore, for aspherical and freeform lens systems, there is a lack of effective technical adaptation solutions, severely restricting their application scope. Summary of the Invention

[0004] This application provides a high-precision error compensation method for the optical axis of a continuous zoom lens, which achieves deep coupling of static and dynamic data, improves the accuracy of error tracing and the pertinence of compensation; by optimizing sampling parameters, it ensures the coverage of key error positions during zooming; and it supports real-time compensation, enhancing the stability and accuracy of the system during zooming.

[0005] The method disassembles the continuous zoom system into independent components and installs them on a rotary table centering fixture. By rotating and collecting the image point trajectory circle, the static center offset data of each component is calculated, including the eccentricity, tilt angle and coaxiality error between components. The established component center offset database is used to quantify the contribution of fixed error, dynamic error and coupling error.

[0006] The sampling parameters are set according to the zoom ratio of each component of the continuous zoom system and the moving speed along the optical axis during the zoom process. The dynamic optical axis offset data of the components is collected simultaneously, as well as the ambient temperature data including the component position, optical axis offset, and ambient temperature. A nonlinear mapping relationship is constructed between static center offset data and dynamic optical axis offset data, and a center offset-optical axis error coupling model is established. The error coupling model is used to predict the output deviation compensation amount based on the optical axis deviation generated by monitoring during zooming. The actuator is driven to adjust the position and / or angle of the components based on the deviation compensation amount, and the optical axis accuracy index is verified after compensation.

[0007] Specifically, after installing the independent components into the rotary table centering fixture, level them using a dial indicator and align the rotary table axis with the component reference axis. Using an autocollimator and laser interferometer, the image point trajectory circle was acquired by rotating the component 360°, and the eccentricity, tilt angle and coaxiality error between components were calculated for each component.

[0008] Specifically, when the zoom ratio of the continuous zoom system is <40:1, the sampling density is ≥50 points / full zoom range, and the sampling interval is ≤2% of the zoom travel; when the zoom ratio is >40:1, the sampling density is ≥80 points / full range. When the component moving speed is ≤10mm / s, the sampling frequency is ≥100Hz; when the component moving speed is >10mm / s, the sampling frequency is dynamically adjusted according to v×10Hz / mm, where v is the component moving speed.

[0009] Specifically, the mapping function of the error coupling model is expressed as follows:

[0010] Among them For optical axis deviation, For data centered on the constituents, This is the zoom position. Ambient temperature; This is due to mechanical transmission error; These are the weighting coefficients. This is the random error term.

[0011] Specifically, the weighting coefficients are obtained by fitting static center-biased data using the least squares method, with a goodness of fit R² ≥ 0.98. The error coupling model is based on the BP neural network algorithm to learn the dynamic error law.

[0012] Specifically, during the compensation phase, actuators are used to perform compensation actions, including a four-dimensional adjustment frame, an electric turntable, and a voice coil motor. First, the four-dimensional adjustment frame is used to handle displacement deviations, then the electric turntable is used to handle angular deviations, and finally the voice coil motor is used to perform overtravel compensation. When the compensation amount exceeds the travel of the actuator, a graded compensation mechanism and action timing control are activated, first through coarse adjustment of the mechanical structure, and then through fine adjustment of the piezoelectric ceramic.

[0013] Specifically, the method is compatible with continuous zoom systems with a zoom ratio ≥30:1, an operating temperature range of -40 to 85℃, and is compatible with spherical, aspherical, and freeform lens systems. When an aspherical / freeform lens is used, the curvature radius correction coefficient of the component center offset detection is directly adjusted.

[0014] Specifically, a high-speed camera is used to acquire target images, and a laser interferometer is used to measure wavefront aberration to verify optical axis runout, optical axis parallelism deviation, and wavefront aberration. If at least one compensation error exceeds the set threshold, update the weight coefficients of the error model and the hidden layer parameters of the BP neural network.

[0015] Specifically, the zoom process is operated by a zoom drive module, which includes a servo motor, a moving guide rail, and a position encoder, controlling the movement of components on the moving guide rail for zooming.

[0016] Specifically, the zoom process monitors the optical axis deviation in real time through the optical axis monitoring module, which includes a high-speed camera, a crosshair target, and a laser tracker, and dynamically sets the sampling parameters according to the zoom ratio and the component movement speed.

[0017] The beneficial effects of the technical solution provided in this application include at least the following: This solution achieves deep coupling of static and dynamic data and real-time error correction by collecting static center offset data and dynamic optical axis offset data, establishing nonlinear mapping relationships and error coupling models, and driving the actuator to compensate in real time. This achieves deep coupling of static and dynamic data, improves the accuracy of error tracing and the pertinence of compensation; by optimizing sampling parameters, it ensures the coverage of key error positions during zooming; and by supporting real-time compensation, it enhances the stability and accuracy of the system during zooming. Attached Figure Description

[0018] Figure 1 This is a flowchart of the high-precision error compensation method for the optical axis of a continuous zoom lens provided in the embodiments of this application; Figure 2 A schematic diagram of a possible structure for static center deviation detection is shown; Figure 3 This is a schematic diagram of a configuration that performs high-density sampling throughout the zoom process in one possible form; Figure 4 This is a schematic diagram of the composition of the zoom drive module; Figure 5 The algorithm flowchart of the high-precision error compensation method for the optical axis of a continuous zoom lens provided in the embodiments of this application is shown; Figure 6 A schematic diagram of the compensation execution mechanism is shown in one possible scenario; Figure 7A schematic diagram of the overall structure of this system is shown; Figure 8 A comparison chart of optical axis deviation curves throughout the zoom range is shown; Figure 9 A comparison diagram of optical axis runout under extreme temperatures is shown.

[0019] Reference numerals: 1-Autocollimator, 2-Micrometer, 3-Centering fixture, 4-Rotating stage, 5-Laser interferometer, 6-Optical element, 7-Temperature / vibration monitoring module, 8-Laser tracker, 9-High-speed camera, 10-Four-dimensional adjustment frame, 11-Electromagnetic shielding module, 12-Precision optical bench, 13-Cross target, 14-Position encoder, 15-Guide rail, 16-Servo motor, 17-Electric rotary table, 18-Voice coil motor, 19-Industrial computer, 20-Rear fixing group, 21-Compensation group, 22-Magnification group, 23-Front fixing group. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0021] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0022] Traditional continuous zoom optical systems typically treat static center offset detection of lens components and optical axis calibration during zooming as independent steps in optical axis detection and compensation. This leads to a disconnect between static and dynamic data, making error tracing difficult and compensation schemes less targeted. Furthermore, existing technologies generally suffer from low zoom sampling density, inability to capture the continuous variation of optical axis error throughout the zoom range, and offline calibration methods, making real-time error correction during zooming difficult. They are also highly susceptible to environmental temperature and mechanical transmission clearances, making them unsuitable for extreme high and low temperature conditions. Optical axis accuracy stability is poor, and the contribution of error sources is unclear. Model construction lacks quantitative basis, resulting in insufficient generalization ability. There are also no clear technical solutions for adapting to aspherical and freeform lenses, limiting their application scope.

[0023] To address this, this application provides a method for high-precision error compensation of the optical axis of a continuous zoom lens, such as... Figure 1 As shown, this method can be summarized in the following steps: S1. The continuous zoom system is disassembled into independent components and installed on the rotary table centering fixture. The static center offset data of each component is calculated by rotating and collecting the image point trajectory circle, including the eccentricity, tilt angle and coaxiality error between components. The established component center offset database is used to quantify the contribution of fixed error, dynamic error and coupling error. A continuous zoom system is an optical system whose focal length can be continuously varied within a certain range to achieve imaging at different magnifications. This system typically consists of multiple optical components that move along the optical axis during the zoom process.

[0024] The image point trajectory circle refers to the trajectory formed on the image plane by the imaging point of an eccentric optical element as it rotates. By analyzing the characteristics of this trajectory circle, the eccentricity and tilt angle of the element can be calculated. Specifically, a combination of an autocollimator and a laser interferometer can be used to accurately detect the static center offset data of the optical element.

[0025] Static center offset data refers to the deviation of the optical center of each optical component from the mechanical center when it is at rest. This mainly includes eccentricity, tilt angle, and coaxiality error in the assembly between components. This data can be used for subsequent network model training and error compensation.

[0026] First, the continuous zoom system is disassembled into independent components such as the zoom group, compensation group, front fixed group, and rear fixed group. Ultrasonic cleaning (e.g., 500W power, cleaning time 3~5min) is used to remove surface impurities. After cleaning, it is installed on a precision rotary table centering fixture for subsequent testing.

[0027] The control components rotate on a fixture, and the image point trajectory circle is acquired using an optical inspection device to calculate the static center offset data of each component. This static center offset data includes the eccentricity, tilt angle, and coaxiality error between components. Subsequently, a component center offset database is established to quantify the contribution of fixed error, dynamic error, and coupling error to optical axis deviation. For example, in practical operation, optical elements such as lenses and prisms, or their assemblies, in a continuous zoom system can be disassembled as independent components. These independent components can be mounted on a high-precision rotary table and slowly rotated manually or electrically. During rotation, a high-resolution camera with a collimating light source can be used to record the trajectory of the image point of the component at different rotation angles. By performing image processing and mathematical analysis on these trajectory data, the degree of deviation of the optical center of each component from its mechanical center, i.e., the eccentricity and tilt angle, can be calculated. At the same time, by comparing the optical axes between different components, their coaxiality error can be evaluated. These calculated error data are then stored in a structured database for subsequent querying and analysis.

[0028] S2. Set sampling parameters according to the zoom ratio of each component of the continuous zoom system and the moving speed along the optical axis during the zoom process, and simultaneously collect dynamic optical axis offset data of the components, as well as ambient temperature data including component position, optical axis offset, etc. Zoom ratio is the ratio between the longest and shortest focal lengths of a continuous zoom system, reflecting the system's zoom capability. Optical axis movement speed is the rate at which each component in a continuous zoom system moves along the optical axis during zooming.

[0029] The process sets corresponding sampling parameters based on the zoom ratio and the movement speed of each component along the optical axis during zooming. Based on this, dynamic optical axis offset data of the components, as well as multi-dimensional data including component position, optical axis offset, and ambient temperature, are collected.

[0030] Specifically, during zooming, a series of zoom positions can be pre-defined, and data can be collected at these positions. The sampling parameters can be adjusted according to the characteristics of the zoom system. For example, a lower sampling density can be used for systems with a relatively small zoom ratio, while a lower sampling frequency can be used for systems with slow component movement. At each sampling point, optical sensors can be used to monitor the optical axis offset and tilt angle in real time, while simultaneously recording the precise position of the current component on the optical axis, and ambient temperature data can be obtained through a temperature sensor. This dynamically acquired data, combined with the previously established static center-biased database, provides comprehensive input for the subsequent construction of the error model.

[0031] S3. Construct a nonlinear mapping relationship between static center offset data and dynamic optical axis offset data, and establish a center offset-optical axis error coupling model. The error coupling model is used to predict the output deviation compensation amount based on the optical axis deviation generated by monitoring during zooming. Next, a nonlinear mapping relationship is constructed between static center-bias data and dynamic optical axis offset data, and a center-bias-optical axis error coupling model is established. This error coupling model is used to predict the output deviation compensation amount based on the monitored optical axis deviation during zooming. For example, machine learning algorithms such as multinomial regression, support vector machines, or simple multilayer perceptrons can be used to establish this nonlinear mapping relationship. The model input can include static center-bias data (such as eccentricity, tilt angle), the current zoom position, real-time ambient temperature, and other factors that may affect optical axis deviation (such as mechanical transmission status). By training on a large amount of historical data, the model can learn and capture the complex nonlinear relationship between these input variables and optical axis deviation. Once the model is trained, during actual zooming, when an optical axis deviation is detected and exceeds a preset threshold, the model can predict the precise compensation amount that needs to be adjusted based on the current input conditions.

[0032] S4. Based on the deviation compensation amount, drive the actuator to adjust the position and / or angle of the components, and verify the optical axis accuracy index after compensation.

[0033] The actuator can be various types of precision drive devices, such as high-precision motors and adjustment frames. These can receive commands from the control system and precisely move or rotate the optical components. When the error coupling model outputs a deviation compensation amount, the control system converts this compensation amount into a control signal for the actuator, thereby enabling fine-tuning of the corresponding component. For example, if a lateral offset of the optical axis is predicted, the actuator can drive the component to move slightly in a direction perpendicular to the optical axis; if a tilt of the optical axis is predicted, the actuator can drive the component to rotate at a slight angle.

[0034] After the actuator has been adjusted, the optical axis accuracy needs to be verified again using optical testing equipment, such as measuring the optical axis runout or wavefront aberration. If the verification result still does not meet the preset optical axis accuracy target, the model parameters can be iteratively optimized or further compensated and adjusted according to the actual situation, thereby forming a closed-loop control system.

[0035] In summary, this embodiment integrates static center offset detection with high-density dynamic sampling throughout the zoom range and constructs a center offset-optical axis error coupling model, achieving accurate source tracing and real-time compensation for optical axis errors in continuous zoom lenses. This method effectively solves the limitations of traditional technologies, such as the disconnect between static and dynamic data, low sampling density, and offline compensation. It significantly improves the optical axis accuracy and stability of continuous zoom systems under complex conditions and expands the applicability of the technology to different types of lens systems.

[0036] Figure 2 A schematic diagram of a possible structure for static center offset detection is shown, which can be operated in combination using an autocollimator 1, a micrometer 2, a centering fixture 3, a precision rotary stage 4, and a laser interferometer 5. For example, a combination detection method using an autocollimator (model: ZygoGPIXP, measurement range ±1000μrad, measurement accuracy ±0.01μrad) + a laser interferometer (model: Agilent5529A, wavelength stability ±0.01nm, measurement accuracy ≤0.01μm) is used. The image point trajectory circle is acquired by rotating the component 360° (rotation step size 1°), and the eccentricity (accuracy ≤0.05μm), tilt angle (accuracy ≤0.1μrad), and coaxiality error between components are calculated for each component.

[0037] After installing the independent component (i.e., optical element 6) onto the rotary table centering fixture, level it using a dial indicator (leveling accuracy ≤ 0.02 μm), aligning the rotary table axis with the component's reference axis, with a coaxiality error ≤ 0.1 μm. When using an autocollimator and laser interferometer for inspection, rotate the component 360° (rotation step 1°) to acquire the image point trajectory circle, calculating the eccentricity (accuracy ≤ 0.05 μm), tilt angle (accuracy ≤ 0.1 μrad), and coaxiality error between components for each component. When reassembling the component after disassembly, use locating pins and laser alignment for assembly assistance, resetting the coaxiality error to ≤ 0.08 μm. Establish a component center offset database, recording the error parameters, position coordinates, and error source contribution of each component (fixed error accounts for 40%–60%, dynamic error accounts for 20%–30%, and coupling error accounts for 10%–30%).

[0038] Figure 3 This is a schematic diagram of a configuration for high-density sampling throughout the zoom process in one possible form. It adopts a combination of a precision optical bench 12 (model: NewportMM-4000, positioning accuracy ±0.05μm) + a four-dimensional adjustment frame 10 (model: ThorlabsMAX313D, repeatability ±0.02μm) + a high-speed camera 9 (model: PhantomV2512, frame rate ≥100fps, pixel size 3.75μm) + a laser tracker 8 (model: LeicaAT960-MR, measurement accuracy ±0.5μm / m) + a temperature / vibration monitoring module 7 (temperature sensor accuracy ±0.1℃, vibration sensor measurement range 0.1~1000Hz) + an electromagnetic shielding module 11 (shielding effectiveness ≥80dB, frequency range 30MHz~1GHz), and the sampling parameters are dynamically set according to the zoom ratio and the component movement speed.

[0039] To ensure high-precision and high-density sampling across various zoom ratios and to guarantee that sampling points cover error inflection points, this application sets a sampling density of ≥50 points per full zoom range and a sampling interval of ≤2% of the zoom travel when the zoom ratio of the continuous zoom system is <40:1. This aims to ensure that a sufficient number of sampling data points can be acquired in continuous zoom systems with relatively low zoom ratios, and to ensure that the distance between sampling points is small enough to capture subtle error changes that may occur during zooming.

[0040] When the zoom ratio is greater than 40:1, the sampling density is ≥80 points / full range. This is to address the more complex and frequent error variations that may occur in high zoom ratio continuous zoom systems during zooming. High zoom ratio systems typically require higher optical axis accuracy, and the error variation patterns may be more drastic.

[0041] When the zoom element's moving speed is ≤10mm / s, the sampling frequency is ≥100Hz. This is designed to ensure that data can still be acquired at a sufficiently high frequency when the zoom element moves at a low speed, thereby capturing transient or slowly changing errors that may occur during low-speed movement. When the zoom element's moving speed is >10mm / s, the sampling frequency is dynamically adjusted by v×10Hz / mm to ensure dynamic matching between the sampling frequency and the zoom element's moving speed. This addresses the challenge of rapid error changes during high-speed zooming and effectively captures error inflection points. Here, v represents the zoom element's moving speed.

[0042] This process can employ a dual-trigger mechanism of displacement sensor + image recognition, and be precisely controlled in conjunction with a zoom drive module to achieve synchronous acquisition of multi-dimensional data such as component position, optical axis offset (X / Y direction), ambient temperature, and transmission mechanism status (gap, wear). Figure 4 This is a schematic diagram of the composition of the zoom drive module. The servo motor 16 controls the component to generate displacement on the precision guide rail 15, and uses the position encoder 14 to control the specific position.

[0043] In one possible implementation, the mapping function of the error coupling model can be expressed as follows:

[0044] Among them This is the optical axis deviation (offset + tilt angle). For data centered on the constituents, This is the zoom position. Ambient temperature; This is due to mechanical transmission error; These are the weighting coefficients. This represents the random error term (e.g., ≤0.02μm). The weighting coefficients can be obtained by fitting static centrally biased data using the least squares method, achieving a goodness of fit R² ≥ 0.98.

[0045] Figure 5 The flowchart illustrates the algorithm of the high-precision optical axis error compensation method for continuous zoom lenses provided in this application embodiment. Static data, dynamic data, and system parameters are input into the model. After removing outliers and normalizing the data, the dataset is divided, for example, into a 7:2:1 ratio of training set, validation set, and test set. Then, the least squares method is used to fit the static error (i.e., the error model), calculating the corresponding weight coefficients and the predicted static optical axis deviation. The error coupling model learns the dynamic error rules based on the BP neural network algorithm.

[0046] In one possible implementation, the BP neural network can be designed as a three-layer structure: 6 neurons in the input layer, 12 neurons in the hidden layer, and 1 neuron in the output layer. The activation function is a combination of ReLU and Sigmoid. Training is performed with 12,000 input samples (covering a temperature range of -40 to 85°C and a zoom range of 0 to 100%).

[0047] The number of neurons in the input layer matches the number of input features in the model, while the number of neurons in the hidden layer provides sufficient computational power to learn complex nonlinear relationships. The number of neurons in the output layer corresponds to the optical axis deviation value that needs to be predicted. The activation functions used are a combination of ReLU and Sigmoid. ReLU (Rectified Linear Unit) is typically used in the hidden layers to accelerate training and alleviate the vanishing gradient problem; the Sigmoid function is often used in the output layer, especially when the output needs to be mapped to a specific range, providing a smooth and bounded output. This combination of activation functions helps improve the learning efficiency and predictive performance of the neural network.

[0048] Figure 6 A schematic diagram of a possible compensation actuator is shown. During the compensation phase, the actuator performs compensation actions and includes a four-dimensional adjustment frame 10, an electric turntable 17, and a voice coil motor 18. For example, this embodiment uses piezoelectric ceramics to drive the four-dimensional adjustment frame (displacement accuracy ≤ 0.01 μm), a precision electric turntable (angle accuracy ≤ 0.05 μrad), and a voice coil motor to drive the compensation group (response frequency ≥ 1 kHz).

[0049] The three-way collaborative control strategy is as follows: the piezoelectric ceramic adjustment frame prioritizes handling micro-displacement deviations (≤1μm), then the precision electric turntable handles angular deviations, and finally the voice coil motor handles large stroke compensation (the part exceeding >1μm). The action timing interval is ≤2ms to avoid mechanism conflicts.

[0050] Compensation process: Real-time acquisition of optical axis deviation at the current position during zooming → Input error model to calculate compensation amount → Drive adjustment component position / angle → Feedback on the optical axis state after compensation, iteratively optimizing parameters. When the compensation amount exceeds the actuator stroke (maximum stroke of piezoelectric ceramic 50μm, maximum stroke of voice coil motor 100μm), a graded compensation mechanism is activated: first, coarse adjustment is performed through mechanical structure (accuracy ≤0.5μm), and then fine adjustment is performed through piezoelectric ceramic. This optimized actuator combination, operation sequence, and graded compensation mechanism effectively solve the problems of low compensation efficiency, insufficient accuracy, and difficulty in handling large stroke deviations.

[0051] Specifically, the piezoelectric ceramic four-dimensional adjustment frame, with its extremely high displacement accuracy (≤0.01μm), is prioritized for handling minute displacement deviations (≤1μm), ensuring the basic accuracy of optical axis alignment and avoiding the accumulation of minor errors. Subsequently, the electric rotary table, with its excellent angular accuracy (≤0.05μrad), is specifically used to handle tilt deviations of the optical axis, ensuring precise parallelism and preventing mutual interference between displacement and angular adjustments. Finally, the voice coil motor, with its high response frequency (≥1kHz) and relatively large stroke capacity, is responsible for performing stroke compensation beyond the stroke range (>1μm), effectively addressing the problem of large stroke deviations that are difficult to handle in traditional methods. This step-by-step, type-based compensation strategy allows each actuator to function in its area of ​​expertise, significantly improving the efficiency and accuracy of compensation. Furthermore, when the monitored compensation amount exceeds the effective stroke range of a single actuator, this application innovatively employs a graded compensation mechanism and action timing control. First, coarse adjustment is performed using mechanical structures (accuracy ≤ 0.5μm) to quickly reduce large deviations to a controllable range, avoiding the problem of high-precision actuators being unable to compensate due to insufficient stroke. Then, fine adjustment is performed using piezoelectric ceramics to further improve compensation accuracy. This tiered compensation strategy ensures that the system can achieve high-precision optical axis compensation even under extreme deviation conditions, greatly expanding the applicability and robustness of the compensation method. Ultimately, this significantly improves the overall accuracy stability, real-time performance, and adaptability to complex working conditions of the continuous zoom lens's optical axis, thereby guaranteeing continuous optimization of image quality during zooming.

[0052] Figure 7 The diagram shows the overall structure of the system, including the outermost electromagnetic shielding module housing. All other modules and components are housed within this housing, and a precision optical bench is used for external detection of the crosshair target. During compensation, the response time is kept ≤10ms, and the number of single compensation iterations is ≤3 to avoid over-adjustment that could lead to system instability.

[0053] Traditional methods are difficult to adapt to high zoom ratio systems, extreme temperature conditions, and aspherical and freeform lens systems, resulting in limited application range and insufficient accuracy. In particular, in the inspection of non-standard lenses, the quantization of center offset data is inaccurate, affecting the reliability of the error model.

[0054] To address this, this application proposes a high-precision error compensation method for the optical axis of a continuous zoom lens. This method is compatible with continuous zoom systems with a zoom ratio ≥ 30:1, an operating temperature range of -40 to 85℃, and is suitable for spherical, aspherical, and freeform lens systems. When an aspherical / freeform lens is used, the curvature radius correction coefficient of the component center offset detection is directly adjusted (aspherical correction coefficient k = 1.2-1.5, freeform correction coefficient k = 1.5-1.8).

[0055] Specifically, this method is compatible with continuous zoom systems with a zoom ratio of ≥30:1. High zoom ratio systems typically mean that optical components have a larger travel distance and more complex motion trajectories during zooming, which significantly increases the dynamics and cumulative nature of optical axis errors.

[0056] Meanwhile, this method can adapt to extreme environments with an operating temperature range of -40 to 85°C. Temperature change is one of the important factors causing optical axis drift in optical systems, which can cause thermal expansion of optical materials, changes in refractive index, and thermal deformation of mechanical structures.

[0057] Furthermore, this method is adaptable to spherical, aspherical, and freeform lens systems. Traditional center offset detection methods are mainly designed for spherical lenses. For aspherical and freeform lenses, due to their complex surface shapes, accurately quantifying center offset error is extremely challenging. This method expands its compatibility with various optical surfaces by introducing a correction coefficient.

[0058] When adapted with aspherical or freeform lenses, this method ensures detection accuracy by directly adjusting the curvature radius correction coefficient for center offset detection. For aspherical lenses, the surface curvature varies radially, no longer a single radius. Therefore, an aspherical correction coefficient k needs to be introduced for center offset detection, with a value ranging from 1.2 to 1.5. This correction coefficient can be obtained through theoretical calculation or experimental calibration based on the specific design parameters of the aspherical lens (such as aspherical coefficient, taper coefficient, etc.), and is used to equate the response of the aspherical surface under specific detection conditions to a spherical response, thereby accurately assessing its center offset. For freeform lenses, the surface curvature may change in different directions and lacks rotational symmetry, making its geometric complexity far greater than that of aspherical lenses. Therefore, a freeform correction coefficient k needs to be introduced for center offset detection, with a value ranging from 1.5 to 1.8. Determining this correction factor may require more complex algorithms, such as those based on finite element analysis or high-precision 3D scanning data, to accurately reflect the optical properties of the freeform surface during the detection process and ensure the accuracy of the center offset data.

[0059] The above solution can achieve the following beneficial effects: 1. Improved accuracy: The optical axis runout has been reduced from >5μm to ≤2μm using traditional methods, resulting in an accuracy improvement of over 60%. The accuracy improvement reaches 78% for the 40x zoom system, meeting the requirements of high-end optical systems. 2. Full coverage: High-density sampling (≥50 points / full range, ≥80 points / full range for high zoom ratio systems) avoids missing key error locations and is compatible with complex systems with zoom ratios ≥30:1; 3. Closed-loop control: Achieve a closed-loop process of "detection-modeling-compensation-verification", with accurate error source tracing, highly targeted compensation, and quantitative analysis of error sources to improve the model's generalization ability; 4. Environmental adaptability: Combining temperature / vibration monitoring and electromagnetic shielding modules, it can work stably under extreme conditions of -40~85℃, improving optical axis stability by 80% and significantly enhancing anti-interference capabilities; 5. Increased efficiency: Integrated detection and compensation eliminates the need for repeated disassembly and assembly, improving efficiency by 50%; 6. Application Expansion: Adaptable to spherical, aspherical, and freeform lens systems, expanding the scope of application of the technology.

[0060] The following example, using a 40x zoom infrared reconnaissance lens, illustrates the implementation process of this solution in detail: S1 component testing: The lens was disassembled into zoom group 22 (3 lenses), compensation group 21 (2 lenses), front fixed group 23 (3 lenses), and rear fixed group 20 (5 lenses). After ultrasonic cleaning (500W, 4min), the lens was installed and positioned. The maximum eccentricity of the zoom group was measured to be 0.08μm, the tilt angle was 0.15μrad, the fixed error accounted for 52%, the dynamic error accounted for 28%, and the coupling error accounted for 20%. A component center eccentricity database was established.

[0061] S2 sampling design: zoom range 8~320mm, sampling density 80 points / full range, sampling interval 4mm; component movement speed 8mm / s, sampling frequency 80Hz; synchronously collects data such as temperature (25℃, -40℃, 85℃) and transmission gap (0.02mm); adopts a dual trigger mechanism with a synchronization error of 0.8ms.

[0062] S3 model construction: BP neural network was used for training, with 12,000 input samples (covering the temperature range of -40 to 85℃ and the zoom range of 0 to 100%). The model prediction accuracy was 97%. The weight coefficients were w1=0.32, w2=0.25, w3=0.18, w4=0.15, w5=0.07, and w6=0.03.

[0063] S4 Real-time Compensation: During zooming, the optical axis deviation is fed back to the industrial computer 19 in real time. After calculating the compensation amount, the piezoelectric ceramic adjustment frame 10 is driven to adjust the position of the zoom group. The compensation response time is 8ms. Under extreme temperatures, the collaborative compensation mechanism is activated, and the piezoelectric ceramic and the voice coil motor 18 work together to complete the compensation.

[0064] S5 Performance Verification: At room temperature (25℃): after compensation, the optical axis runout is 1.8μm, the optical axis parallelism deviation is 0.4μrad, and the wavefront aberration RMS is 0.06λ. Low temperature (-40℃): optical axis runout 1.9μm, optical axis parallelism deviation 0.45μrad, wavefront aberration RMS 0.065λ; High temperature (85℃): optical axis runout 1.95μm, optical axis parallelism deviation 0.48μrad, wavefront aberration RMS 0.068λ; After 1000 hours of continuous operation: optical axis runout 1.88μm, model prediction accuracy 94%.

[0065] Comparison test (same lens):

[0066] Figure 8 A comparison chart of optical axis deviation curves throughout the zoom range is shown. This chart compares the trend of optical axis deviation changes between this method and the traditional optical axis error compensation method in a continuous zoom system throughout the entire zoom range.

[0067] 1. Horizontal axis: zoom travel (0% to 100%), covering the entire zoom process from the shortest focal length to the longest focal length; 2. Vertical axis: Optical axis deviation, unit μm; 3. Curve Explanation: Traditional method curves: large deviation fluctuations, many inflection points, high peak values, and deviations generally greater than 5μm throughout the process, with more significant deviations in high zoom ratio scenarios; This solution completely eliminates optical axis jump during zooming by using high-density sampling throughout the process and center offset-optical axis error coupling compensation. The accuracy is improved by more than 60% compared with traditional methods, and it is perfectly adapted to high zoom ratio optical systems.

[0068] Figure 9 The figure shows a comparison of optical axis runout under extreme temperatures. This figure is a comparison curve of the measured optical axis runout between the method and the traditional optical axis compensation method within the extreme temperature range of 40℃~80℃.

[0069] 1. Horizontal axis: Ambient temperature, covering 13 temperature points: 40℃, 30℃, 20℃, 10℃, 0℃, 10℃, 20℃, 30℃, 40℃, 50℃, 60℃, 70℃, and 80℃; 2. Vertical axis: Optical axis runout, unit μm; 3. Curve Description: The optical axis runout of the curve obtained by the traditional method fluctuates drastically with temperature, reaching a peak of 14μm in the low temperature region (40℃) and still exceeding 10μm in the high temperature region (80℃), with extremely poor stability; the optical axis runout of the curve obtained by this method is stable between 1.8μm and 1.95μm throughout the entire process, which is far better than the accuracy index of ≤2μm. 4. Data Conclusion: The proposed solution improves the optical axis stability by 80% compared to traditional methods under extreme operating conditions of 40~85℃, and can meet the high and low temperature adaptation requirements of high-precision optical systems such as automotive, aerospace, and military reconnaissance.

[0070] In summary, this solution completely eliminates optical axis runout during zooming by using high-density sampling throughout the process and center offset-optical axis error coupling compensation. The accuracy is improved by more than 78% compared to traditional methods, the detection efficiency is improved by 60%, it is perfectly adapted to high zoom ratio optical systems, and its stability under extreme conditions is significantly better than traditional technologies.

[0071] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A method for high-precision error compensation of the optical axis of a continuous zoom lens, characterized in that, The method includes: The continuous zoom system is disassembled into independent components and installed on a rotary table centering fixture. The static center offset data of each component is calculated by rotating and collecting the image point trajectory circle, including the eccentricity, tilt angle and coaxiality error between components. The established component center offset database is used to quantify the contribution of fixed error, dynamic error and coupling error. The sampling parameters are set according to the zoom ratio of each component of the continuous zoom system and the moving speed along the optical axis during the zoom process. The dynamic optical axis offset data of the components and the ambient temperature data including the component position, optical axis offset amount and other data are collected simultaneously. A nonlinear mapping relationship is constructed between static center offset data and dynamic optical axis offset data, and a center offset-optical axis error coupling model is established. The error coupling model is used to predict the output deviation compensation amount based on the optical axis deviation generated by monitoring during zooming. Based on the deviation compensation amount, the actuator is driven to adjust the position and / or angle of the components, and the optical axis accuracy index is verified after compensation.

2. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, After installing the independent components into the rotary table centering fixture, level them using a dial indicator and align the rotary table axis with the component reference axis. Using an autocollimator and laser interferometer, the image point trajectory circle was acquired by rotating the component 360°, and the eccentricity, tilt angle and coaxiality error between components were calculated for each component.

3. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, When the zoom ratio of the continuous zoom system is <40:1, the sampling density is ≥50 points / full zoom range, and the sampling interval is ≤2% of the zoom travel; when the zoom ratio is >40:1, the sampling density is ≥80 points / full range. When the component moving speed is ≤10mm / s, the sampling frequency is ≥100Hz; when the component moving speed is >10mm / s, the sampling frequency is dynamically adjusted according to v×10Hz / mm, where v is the component moving speed.

4. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 3, characterized in that, The mapping function of the error coupling model is expressed as follows: Among them For optical axis deviation, For data centered on the constituents, This is the zoom position. Ambient temperature; This is due to mechanical transmission error; These are the weighting coefficients. This is the random error term.

5. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 4, characterized in that, The weighting coefficients are obtained by fitting static center-biased data using the least squares method, with a goodness of fit R² ≥ 0.

98. The error coupling model is based on the BP neural network algorithm to learn the dynamic error law.

6. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, The compensation phase uses actuators to perform compensation actions, including a four-dimensional adjustment frame, an electric turntable, and a voice coil motor. First, the four-dimensional adjustment frame is used to handle displacement deviations, then the electric turntable is used to handle angular deviations, and finally the voice coil motor is used to perform overtravel compensation. When the compensation amount exceeds the travel of the actuator, a graded compensation mechanism and action timing control are activated, first through coarse adjustment of the mechanical structure, and then through fine adjustment of the piezoelectric ceramic.

7. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, The method is compatible with continuous zoom systems with a zoom ratio ≥30:1, an operating temperature range of -40 to 85℃, and is suitable for spherical, aspherical, and freeform lens systems. When an aspherical / freeform lens is used, the curvature radius correction coefficient of the component center offset detection is directly adjusted.

8. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, High-speed cameras were used to acquire target images, and laser interferometers were used to measure wavefront aberrations to verify optical axis runout, optical axis parallelism deviation, and wavefront aberrations. If at least one compensation error exceeds the set threshold, update the weight coefficients of the error model and the hidden layer parameters of the BP neural network.

9. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 1, characterized in that, The zoom process is operated by a zoom drive module, which includes a servo motor, a moving guide rail, and a position encoder, controlling the movement of components on the moving guide rail for zooming.

10. The method for high-precision error compensation of the optical axis of a continuous zoom lens according to claim 9, characterized in that, The zoom process monitors the optical axis deviation in real time through the optical axis monitoring module, which includes a high-speed camera, a crosshair target, and a laser tracker, and dynamically sets the sampling parameters according to the zoom ratio and the component movement speed.

Images