Mosaic mirror fine co-phase error detection system and method based on new pupil plane information
Patent Information
- Application Number
- CN202611150395.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-31
- Publication Date
- 2026-08-28
AI Technical Summary
传统如相位恢复(PR)或相位差法(PD)等探测手段,由于依赖焦面/离焦面图像的反复迭代求解,存在计算耗时长、易陷入局部最优的瓶颈,难以实现实时共相
[0011]This invention introduces a mask and pyramidal device into the splicing mirror optical system. After the reflected beam is extracted by a local pupil and spatially modulated, a coherently superimposed modulated pupil surface image containing phase difference information is formed on the detector. This image is then used as input to a convolutional neural network, which, relying on the network's powerful nonlinear mapping capability, directly calculates and outputs a high-precision translation error prediction value. Therefore, this invention proposes a novel pupil surface detection principle. The modulated light intensity distribution exhibits a clear deflection characteristic as the translation error changes, thus providing a physical basis for the neural network to distinguish the positive and negative directions of the error at the optical hardware level. In addition, the single-device design significantly reduces the difficulty of system assembly and adjustment and the risk of error propagation, improving the stability and reliability of the entire system and facilitating the engineering realization of ultra-large aperture telescopes. Furthermore, relying on the powerful feature extraction and fitting capabilities of the convolutional neural network, an end-to-end direct mapping is established from the "two-dimensional modulated pupil surface image" to the "one-dimensional translation error value," requiring only milliseconds of forward inference time to accurately output the error prediction value.
Smart Images

Figure CN122651300A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of space optics and optoelectronic measurement technology, specifically involving the cross-application of wavefront sensing, image processing and deep learning technologies, and particularly a splicing mirror precision co-phase error detection system and method based on novel pupil plane information. Background Technology
[0002] With the relentless pursuit of high resolution and light-gathering capabilities in space exploration and astronomical observation, increasing the aperture of the primary mirror of optical telescopes has become an inevitable trend. However, once the aperture of a single primary mirror reaches the 8-meter range, its high-precision grinding and launch transportation have reached their engineering and physical limits due to limitations in material properties, manufacturing cycle, and the envelope size of the launch fairing. The mosaic mirror technology, by discretizing the equivalent large pupil into modular sub-mirrors, effectively avoids the manufacturing risks and high costs of ultra-large single mirrors and overcomes the launch volume limitations through a folding and unfolding mechanism, becoming a core approach to realizing ultra-large aperture optical systems. Furthermore, the mosaic structure has significant advantages in reducing the overall system weight and improving the flexibility of long-term operation and maintenance. Nevertheless, the challenge of mosaic mirror technology shifts from "optical processing" to "pose control." If there are phase errors (especially translation errors) between sub-mirrors, the system's imaging performance will be severely degraded. Therefore, how to achieve high-precision phase error detection has become a key bottleneck restricting the mosaic mirror optical system from achieving diffraction-limited imaging performance.
[0003] In the fine co-phase stage, translation errors are typically within a single wavelength range. Traditional detection methods, such as phase retrieval (PR) or phase difference (PD), rely on iterative solutions using focal / defocal images, resulting in computationally intensive calculations and a tendency to get trapped in local optima, making real-time co-phase detection difficult. While existing deep learning-based co-phase detection methods achieve rapid error prediction by learning the focal point spread function, they are limited by the inherent symmetry of the focal image with respect to positive and negative translation errors, making the network prone to "directional degeneracy" and unable to distinguish error directions. Furthermore, because pure focal information loses intuitive phase modulation details during Fourier transform, its physical representation capability is significantly weaker than pupil information, constituting a fundamental limitation of existing intelligent co-phase detection technologies. Summary of the Invention
[0004] To overcome the technical problems existing in the background art, this invention proposes a system and method for detecting the precise co-phase error of a stitched mirror based on novel pupil plane information. This method uses the modulated pupil plane image as the input feature of a convolutional neural network, relying on the powerful nonlinear feature extraction capability of deep learning to directly calculate and output high-precision translation error prediction values, providing an optimized technical approach for the wavefront detection system of a stitched mirror.
[0005] The first aspect discloses a system for detecting the co-phase error of a splicing mirror based on novel pupil plane information, the system comprising:
[0006] The system comprises at least two stitching sub-mirrors, a mask, a focusing lens, a pyramid, an imaging lens, and a camera. The mask, under illumination from a parallel wavefront, performs local pupil extraction and spatial confinement on a reflected beam carrying translational errors between any two adjacent stitching sub-mirrors to form a transmitted beam. The focusing lens, through Fourier transform, converges the transmitted beam onto the physical vertex of the pyramid. The pyramid is positioned at the focal plane of the stitching mirror's precise co-phase error detection system for left-right beam splitting and phase modulation, forming a separated beam. The imaging lens passes the separated beam through an inverse Fourier transform lens to form an outgoing beam, which is then re-imaged on the camera's detection target surface to obtain a modulated pupil image.
[0007] The second aspect discloses a method for detecting the precise co-phase error of a stitched mirror based on novel pupil plane information, applied to the precise co-phase error detection system for a stitched mirror based on novel pupil plane information disclosed in the first aspect or any possible implementation thereof, the method comprising:
[0008] Acquire the modulated pupil plane image between the two stitched sub-mirrors;
[0009] The obtained modulated pupil image is input into a pre-trained convolutional neural network to obtain the translation error prediction value between the two stitched sub-mirrors. The pre-trained convolutional neural network includes alternating convolutional and pooling layers, fully connected layers, and an output layer. The convolutional layers are used to extract the spatial light intensity features of the modulated pupil image. The pooling layers are used to perform spatial feature downsampling on the spatial light intensity features to obtain image features. The fully connected layers are used to map the extracted image features to a one-dimensional regression space, and the output layer outputs the translation error prediction value.
[0010] As can be seen from the above technical solutions, the present invention has the following beneficial effects:
[0011] This invention introduces a mask and pyramidal device into the splicing mirror optical system. After the reflected beam is extracted by a local pupil and spatially modulated, a coherently superimposed modulated pupil surface image containing phase difference information is formed on the detector. This image is then used as input to a convolutional neural network, which, relying on the network's powerful nonlinear mapping capability, directly calculates and outputs a high-precision translation error prediction value. Therefore, this invention proposes a novel pupil surface detection principle. The modulated light intensity distribution exhibits a clear deflection characteristic as the translation error changes, thus providing a physical basis for the neural network to distinguish the positive and negative directions of the error at the optical hardware level. In addition, the single-device design significantly reduces the difficulty of system assembly and adjustment and the risk of error propagation, improving the stability and reliability of the entire system and facilitating the engineering realization of ultra-large aperture telescopes. Furthermore, relying on the powerful feature extraction and fitting capabilities of the convolutional neural network, an end-to-end direct mapping is established from the "two-dimensional modulated pupil surface image" to the "one-dimensional translation error value," requiring only milliseconds of forward inference time to accurately output the error prediction value. Attached Figure Description
[0012] Figure 1 This is a schematic diagram of the architecture of a splicing mirror precision co-phase error detection system based on novel pupil plane information provided by the present invention.
[0013] Figure 2 This is a schematic diagram of a common type of splicing primary mirror in engineering, including (a) a regular hexagonal sub-mirror, (b) a circular sub-mirror, and (c) a fan-shaped sub-mirror.
[0014] Figure 3 This is a flowchart of the method for detecting the co-phase error of a splicing mirror based on novel pupil plane information proposed in this invention.
[0015] The components include: 1. splicing sub-mirror; 2. mask plate; 3. focusing lens; 4. pyramid; 5. imaging lens; 6. camera; 7. convolutional neural network module; and 8. control system. Detailed Implementation
[0016] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be thorough and complete.
[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a particular 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 a sequence other than that illustrated or described herein. Furthermore, the terms “comprising” and “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.
[0018] It should be noted that when an element is referred to as being "fixed to" another element, it can be directly on the other element or there may be an intervening element. When an element is considered to be "connected to" another element, it can be directly connected to the other element or there may be an intervening element. The terms "vertical," "horizontal," "left," "right," "up," "down," and similar expressions used herein are for illustrative purposes only and are not intended to indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.
[0019] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. The term "and / or" as used herein includes any and all combinations of one or more of the related listed items.
[0020] The technical problem this invention aims to solve is as follows: In the fine co-phase detection stage of large-aperture mosaic telescopes, traditional iterative algorithms are computationally time-consuming and struggle to meet real-time closed-loop control requirements. Furthermore, existing deep learning-based pure focal plane detection methods are limited by the physical mechanism of the Fourier transform of the spatial light field, easily losing a large amount of phase constraint information. This invention proposes a fine co-phase error detection system and method for mosaic mirrors based on novel pupil plane information. This method uses the modulated novel pupil plane image as input features to a convolutional neural network (CNN). Leveraging the powerful nonlinear mapping capabilities of deep learning, it directly calculates and outputs high-precision translation error predictions, achieving rapid, high-precision, full-range single-shot acquisition and detection of nanometer-level translation errors between sub-mirrors.
[0021] The technical solution adopted by this invention to solve the above-mentioned technical problems is as follows: a pyramidal device is placed at the focal plane position of the front-mounted stitching mirror optical system. By rationally designing the geometric parameters of the pyramid, the pupil images from different stitching sub-mirrors form an overlapping region after passing through the device. This overlapping region contains the phase difference information between the stitching sub-mirrors. To meet the nanometer-level accuracy requirements of the fine co-phase stage, this invention further introduces deep learning-based sensing technology. Specifically, a novel pupil plane image containing a small amount of co-phase error is acquired and input into a pre-trained convolutional neural network model. The nonlinear mapping relationship constructed by the network is used to directly regress the co-phase error between the stitching sub-mirrors.
[0022] In one embodiment, the present invention provides a splicing mirror precision co-phase error detection system based on novel pupil plane information, such as... Figure 1 As shown, this system mainly includes:
[0023] The system comprises at least two stitching sub-mirrors 1, a mask 2, a focusing lens 3, a pyramid 4, an imaging lens 5, and a camera 6. The mask 2, under illumination from a parallel wavefront, performs local pupil extraction and spatial confinement on the reflected beam carrying translational errors between any two adjacent stitching sub-mirrors 1 to form a transmitted beam. The focusing lens 3 converges the transmitted beam onto the physical vertex of the pyramid 4 under Fourier transform. The pyramid 4 is positioned at the focal plane of the stitching mirror optical system for left-right beam splitting and phase modulation to form a separated beam. The imaging lens 5 passes the separated beam through an inverse Fourier transform lens to form an outgoing beam, which is then re-imaged on the detection target surface of the camera 6 to obtain a modulated pupil image.
[0024] The specific workflow of the system is as follows: Under the illumination of the parallel wavefront, the reflected beam carrying the translational error between the splicing sub-mirrors 1 first passes through the mask plate 2 for local pupil extraction and spatial constraint across the edge; subsequently, the transmitted beam, under the Fourier transform of the focusing lens 3, converges on the physical vertex (edge) of the pyramid 4. After left and right beam splitting and phase modulation by the pyramid 4, the separated beams are emitted again through the imaging lens 5 (inverse Fourier transform lens), and finally re-imaged on the detection target surface of the camera 6. In this process, due to the precise matching of the parameters of the mask plate and the pyramid, the mask plate 2 includes two sub-apertures of the same size, and the pupil surface patterns of the two sub-apertures partially overlap in the central region of the detector, forming a coherent superimposed light intensity pattern with physical constraints, which is then acquired by the camera 6 to obtain the modulated pupil surface image.
[0025] In addition, such as Figure 2 As shown, common types of primary mirrors used in engineering splicing include regular hexagonal sub-mirrors (such as...). Figure 2 As shown in (a)), the circular sub-mirror (such as...) Figure 2 (b) and sector sub-mirrors (e.g.) Figure 2In (c) of this embodiment, the splicing sub-mirror can be any shape.
[0026] Furthermore, based on the principles of optical system design, a novel optical system based on pyramidal beam splitting is constructed. This process requires precise design and matching of key optical parameters such as the mask size, lens focal length, and pyramidal base angle to ensure the transmitted light field forms a specific modulation pattern. In one example, the mask has two sub-apertures, each 6mm × 3mm in size, with a center-to-center distance of 28.2mm. The focusing lens 3 has a focal length f1 = 3m, the imaging lens 5 has a focal length f2 = 0.1m, and the pyramidal base angle is 0.78°.
[0027] In addition, the system also includes an actuator corresponding to the splicing sub-mirror 1, which applies a known relative translation to the splicing sub-mirror 1 to simulate the co-phase error generated in actual operation.
[0028] In another embodiment, the system further includes a convolutional neural network module 7 and a control system 8, wherein the control system 8 is used to generate feedback control commands based on the translation error prediction values output by the convolutional neural network module 7, drive the actuators on the back of the corresponding stitching sub-mirrors to perform pose adjustment, and realize the co-phase stitching between the stitching sub-mirrors, wherein the convolutional neural network module 7 is used to output the corresponding translation error prediction values based on the modulation pupil plane image acquired by the camera 6.
[0029] Among them, the input light field E of the mask in The following formula (1) can be used to express this:
[0030] (1)
[0031] in, These are pupil plane coordinates. and These represent the amplitude distributions of the two sub-apertures, It is the imaginary unit. It is the wavelength of the incident light. This represents the optical path difference between the two sub-apertures.
[0032] Furthermore, the light field distribution of the modulated pupil image is expressed as shown in the following formula (2):
[0033] (2)
[0034] in, This represents the light field distribution of the modulated pupil image. and The amplitude distributions of the two sub-apertures of mask 2 are represented respectively. It should be noted that, to distinguish between the two sub-apertures, they are divided into the first sub-aperture and the second sub-aperture. The amplitude distribution corresponding to the first sub-aperture is shown below. The amplitude distribution corresponding to the second sub-aperture is T A and T B These represent the amplitude distributions for the first sub-aperture. Amplitude distribution of the second sub-aperture Perform a Hilbert transform.
[0035] This shows that the modulated light intensity distribution exhibits a clear deflection characteristic as the translation error changes, thus providing a physical basis for the neural network to distinguish the positive and negative directions of the error at the optical hardware level.
[0036] This invention employs a single-device design, placing a pyramidal device at the focal plane of the front-mounted stitching mirror optical system. By rationally designing the pyramid's geometric parameters, the pupil images from different stitching sub-mirrors form an overlapping region after passing through this device. This region contains the phase difference information between the sub-mirrors. This significantly reduces the difficulty of system assembly and adjustment, as well as the risk of error propagation, improving the stability and reliability of the entire system and facilitating the engineering realization of ultra-large aperture telescopes.
[0037] In one embodiment, the present invention provides a method for detecting the precise co-phase error of a stitched mirror based on novel pupil plane information, applied to the precise co-phase error detection system for stitched mirrors based on novel pupil plane information disclosed in the above embodiments, such as... Figure 3 As shown, the method includes:
[0038] S101. Obtain the modulation pupil plane image between the two stitched sub-mirrors;
[0039] S102. The obtained modulated pupil image is input into a pre-trained convolutional neural network to obtain the translation error prediction value between the two stitched sub-mirrors; wherein, the pre-trained convolutional neural network includes alternating convolutional layers, pooling layers, fully connected layers and an output layer, the convolutional layers are used to extract the spatial light intensity features of the modulated pupil image, the pooling layers are used to perform spatial feature downsampling on the spatial light intensity features to obtain image features, the fully connected layers are used to map the extracted image features to a one-dimensional regression space, and the output layer outputs the translation error prediction value.
[0040] Specifically, in the online inference and closed-loop control stage: during the actual operation of the stitching sub-mirror telescope, the novel modulation pupil plane image acquired by the camera 6 in a single real-time acquisition is input into the convolutional neural network module 7. The convolutional neural network can quickly and accurately predict the current co-phase translation error value. Subsequently, the control system 8 (host computer) generates feedback control commands based on the co-phase translation error value, driving the actuator on the back of the stitching sub-mirror to perform high-precision pose adjustment, ultimately realizing the co-phase stitching between sub-mirrors.
[0041] The pre-trained convolutional neural network includes an input layer, alternating convolutional and pooling layers, a fully connected layer, and an output layer. The input layer produces a modulated pupil image mirror. The convolutional layers extract the spatial intensity features of the modulated pupil image. The pooling layers downsample the spatial intensity features to obtain image features. The fully connected layer maps the extracted image features to a one-dimensional regression space, establishing an end-to-end direct mapping from the "two-dimensional modulated pupil image" to the "one-dimensional translation error value". The output layer outputs the predicted translation error value.
[0042] In one embodiment, the training process of the pre-trained convolutional neural network includes:
[0043] Obtain a training sample dataset, wherein the training sample dataset includes an image dataset and a label dataset, the image dataset being captured modulation pupil images, and the label dataset being translation error values corresponding to the modulation pupil images in the image dataset;
[0044] The modulated pupil image from the image dataset is input into the convolutional neural network to be trained to obtain the corresponding translation error prediction value.
[0045] Calculate the loss value between the predicted translation error value and the corresponding translation error value in the label dataset;
[0046] The convolutional neural network to be trained is optimized based on the loss value until the network converges, thus obtaining a pre-trained convolutional neural network.
[0047] Specifically, in the offline training phase: by actively driving the stitching sub-mirrors to traverse different known translation errors, the above optical acquisition process is repeatedly operated to construct a dataset containing a large number of "novel pupil images - translation error labels", and then the dataset is fed into the convolutional neural network module 7 for end-to-end model training until the network converges and learns the mapping relationship between light intensity deflection and the polarity and value of translation error.
[0048] Specifically, in the fine co-phase error detection stage, using one stitched sub-mirror as a reference, 10,000 sets of random translation errors within the range of [-λ / 2, λ / 2] are introduced into another stitched sub-mirror. 10,000 modulated pupil plane images captured by camera 6 are collected and the corresponding co-phase translation error values are recorded, and image datasets and label datasets are constructed respectively.
[0049] Figure 1 In the splicing mirror optical system shown, there is a common phase error between the splicing sub-mirrors 1. When different translation errors (labels) are applied to one of the splicing sub-mirrors 1, the modulation pupil image captured by the camera 6 also changes accordingly. That is, the image and label values are in one-to-one correspondence.
[0050] The methods for constructing image datasets and label datasets can include: creating one folder to store all named images (image dataset); and creating another folder to store all label data containing image indexes (label dataset), with a one-to-one correspondence between the two, that is, one image corresponds to one data label.
[0051] Configure the deep learning environment, build the convolutional neural network, and initialize the network structure parameters, such as network structure hyperparameters, trainable weights, and bias initialization. Network structure parameter initialization refers to the process of instantiating the network computation graph in computer memory and assigning initial values to all trainable and non-trainable parameters before model training begins. Specifically, this includes hyperparameter instantiation, trainable weight initialization, and bias initialization. Hyperparameter instantiation mainly refers to defining the image pixel size of the input layer (e.g., M×N), the kernel size of each convolutional layer (e.g., 3×3 or 5×5), the number of feature channels, the stride, and the padding, as well as the number of nodes in the fully connected layers.
[0052] 80% of the image dataset was selected as the training set for the convolutional neural network model to learn the nonlinear mapping relationship between the modulated pupil image and the co-phase error value. 20% of the image dataset was selected as the validation set to measure the accuracy and generalization of the neural network model.
[0053] In one embodiment, the loss value between the predicted translation error value and the corresponding translation error value in the label dataset is calculated using the mean squared error loss function.
[0054] The convolutional neural network (CNN) is optimized and its parameters are adjusted by calculating the loss value between the predicted translation error and the given label. Specifically, the corresponding relationship is extracted based on a pre-defined string parsing rule. For example, for image data named "00001.png_0.025", the algorithm automatically locates and extracts the index value "0.025" attached to the filename as the given label. Subsequently, the difference between the predicted common phase error value and the given label value is calculated using the mean squared error (MSE) loss function to obtain the network loss value. Based on this loss value, the backpropagation algorithm is used to calculate the change in the network gradient, and a pre-defined optimizer (such as Adam or SGD optimizer) is used to iteratively update and fine-tune the learnable parameters inside the CNN. The learnable parameters specifically include: the kernel weights and biases in each convolutional layer used to extract spatial light intensity features, and the weight matrix and biases in the fully connected layers used for feature dimension transformation. Pooling layers only perform spatial downsampling and do not participate in parameter updates. Through multiple rounds of iterative training, the loss value is continuously minimized until the network model converges.
[0055] In one embodiment, the invention further includes:
[0056] The phase error value between the stitching sub-mirrors 1 is compensated in reverse by the predicted translation error value, thereby completing the phase stitching between the stitching sub-mirrors 1.
[0057] After the training network converges, only a single frame of the modulated pupil image needs to be applied to the trained convolutional neural network, and the network will output the predicted co-phase error value. The control system 8 performs reverse compensation on the co-phase error values between the stitched sub-mirrors to complete the co-phase stitching between the sub-mirror segments.
[0058] In one embodiment, the invention further includes:
[0059] When there are two or more stitched sub-mirrors, the modulated pupil plane image between any two adjacent stitched sub-mirrors is obtained, and the corresponding relative translation error prediction value is output through a pre-trained convolutional neural network.
[0060] Any central stitching sub-mirror is designated as the global zero-phase reference. Through cascading operations, the predicted relative translation error values are uniformly converted into the absolute translation error of each stitching sub-mirror relative to the reference.
[0061] Based on the calculated absolute translation error, control commands are generated to drive the full array actuators to adjust synchronously, thereby completing the global fine co-phase stitching of the multi-segment sub-mirror system.
[0062] The core principle of this invention is also applicable to multi-sub-mirror stitching primary mirror systems containing three or more stitched sub-mirrors. For multi-sub-mirror stitching systems, a discretized detection logic of "pairwise incidence and cross-edge extraction" is adopted, and the specific process is as follows:
[0063] (1) Reference mapping: During the offline training phase, the local light field at the junction of two adjacent sub-mirrors is extracted using the dual aperture of the mask plate, and the neural network is trained to master the mapping relationship between the local modulated pupil image and the relative translation error between the sub-mirrors;
[0064] (2) Traversal detection: In the online closed-loop stage, each adjacent splicing sub-mirror pair passes through the optical modulation optical path in sequence or in parallel. The camera captures the pupil plane image at each junction and inputs it into the network to quickly calculate the relative translation error between all adjacent sub-mirror pairs.
[0065] (3) Error calculation: The control system designates any central sub-mirror as the global zero-phase reference. Through cascaded operation, the above relative errors are uniformly converted into the absolute translation error of each sub-mirror relative to the reference.
[0066] (4) Global closed loop: The control system generates instructions based on the calculated absolute translation error, drives the full array actuators to adjust synchronously, and completes the global fine co-phase splicing of the multi-sub-mirror system in one go.
[0067] Therefore, this invention leverages the powerful feature extraction and fitting capabilities of convolutional neural networks to establish an end-to-end direct mapping from "two-dimensional modulated pupil surface image" to "one-dimensional translation error value". After the network training converges, it can accurately output the error prediction value in only milliseconds of forward inference time.
[0068] This application also provides an electronic device, which includes a processor and a memory. The memory stores at least one instruction or at least one program. The at least one instruction or at least one program is loaded by the processor and executed by the processor. This method provides a splicing mirror precision co-phase error detection method based on novel pupil plane information.
[0069] Furthermore, the electronic device may participate in or include the apparatus or system provided in the embodiments of this application. The electronic device may include one or more processors (processors may include, but are not limited to, processing devices such as microprocessors (MCUs) or programmable logic devices (FPGAs), memory for storing data, and transmission devices for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the I / O interface), a network interface, a power supply, and / or a camera.
[0070] It should be noted that the aforementioned one or more processors and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits can be implemented wholly or partially as software, hardware, firmware, or any other combination. Furthermore, the data processing circuits can be a single, independent processing module, or wholly or partially integrated into any other element within a device (or mobile device). As involved in the embodiments of this application, the data processing circuit serves as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).
[0071] The memory can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the method described in the embodiments of this application. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the above-mentioned data processing method. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to electronic devices via a network. Examples of the above-mentioned networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.
[0072] The transmission device is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the device's communication provider. In one example, the transmission device includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device may be a Radio Frequency (RF) module, used for wireless communication with the Internet.
[0073] The display can be, for example, a touchscreen liquid crystal display (LCD), which allows users to interact with the user interface of an electronic device (or mobile device).
[0074] This application embodiment also provides a computer storage medium storing at least one instruction or at least one program, which is loaded and executed by a processor to implement the splicing mirror precision co-phase error detection method based on novel pupil plane information provided in the above method embodiment.
[0075] Optionally, in this embodiment, the aforementioned computer storage medium may be located at at least one of the multiple network servers in a computer network. Optionally, in this embodiment, the aforementioned storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.
[0076] This application also provides a computer program product or computer program, which includes computer instructions stored in a computer storage medium. The processor of an electronic device reads the computer instructions from the computer storage medium and executes the computer instructions, causing the electronic device to perform the stitching mirror precision co-phase error detection method based on novel pupil plane information provided in the above-described method embodiments.
[0077] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0078] It should be understood that the above description of the preferred embodiments is quite detailed, but it should not be considered as a limitation on the scope of protection of this invention. Those skilled in the art, under the guidance of this invention, can make substitutions or modifications without departing from the scope of protection of the claims of this invention, and all such substitutions or modifications fall within the scope of protection of this invention. The scope of protection of this invention should be determined by the appended claims.
Claims
1. A system for detecting phase co-occurrence error in splicing mirrors based on novel pupil plane information, characterized in that, The system includes: At least two splicing sub-mirrors (1), a mask plate (2), a focusing lens (3), a pyramid (4), an imaging lens (5), and a camera (6) are provided. The mask plate (2) is used to extract and spatially confine the reflected beam carrying translational errors between any two adjacent splicing sub-mirrors (1) under the illumination of a parallel wavefront to form a transmitted beam through the edge of the local pupil. The focusing lens (3) is used to converge the transmitted beam to the physical vertex of the pyramid (4) under the action of Fourier transform. The pyramid (4) is set at the focal plane position of the splicing mirror fine co-phase error detection system for left and right beam splitting and phase modulation to form a separated beam. The imaging lens (5) is used to form an outgoing beam by passing the separated beam through an inverse Fourier transform lens to re-image on the detection target surface of the camera (6) to obtain a modulated pupil surface image.
2. The splicing mirror precision co-phase error detection system based on novel pupil plane information according to claim 1, characterized in that, The mask plate (2) includes two sub-apertures of the same size, wherein each sub-aperture is used to extract and spatially confine the local pupil of the reflected beam carrying translational error between any two adjacent splicing sub-mirrors (1).
3. The splicing mirror precision co-phase error detection system based on novel pupil plane information according to claim 2, characterized in that, The two sub-apertures have a size of 6mm×3mm and a center distance of 28.2mm. The focal length of the focusing lens (3) is 3m, the focal length of the imaging lens (5) is 0.1m, and the base angle of the pyramid (4) is 0.78°.
4. The splicing mirror precision co-phase error detection system based on novel pupil plane information according to claim 1, characterized in that, The system also includes: a convolutional neural network module (7) and a control system (8); The control system (8) is used to generate feedback control commands based on the translation error prediction value output by the convolutional neural network module (7), drive the actuators on the back of the corresponding stitching sub-mirrors (1) to adjust their pose, and realize the co-phase stitching between the stitching sub-mirrors (1). The convolutional neural network module (7) is used to output the translation error prediction value based on the modulated pupil image acquired by the camera (6).
5. The splicing mirror precision co-phase error detection system based on novel pupil plane information according to claim 2, characterized in that, The light field distribution of the modulated pupil image is expressed by the following formula: , in, This represents the light field distribution of the modulated pupil image. and These represent the amplitude distributions of the first and second sub-apertures of the mask plate (2), respectively. It is the wavelength of the incident light. The optical path difference between the two sub-apertures. and These represent the amplitude distributions for the first sub-aperture. Amplitude distribution of the second sub-aperture Perform a Hilbert transform.
6. A method for detecting the precise co-phase error of a stitched mirror based on novel pupil plane information, applied to the precise co-phase error detection system of a stitched mirror based on novel pupil plane information according to any one of claims 1-5, characterized in that, The method includes: Obtain the modulation pupil plane image between the two splicing sub-mirrors (1); The obtained modulated pupil image is input into a pre-trained convolutional neural network to obtain the translation error prediction value between the two spliced sub-mirrors (1); wherein, the pre-trained convolutional neural network includes alternating convolutional layers and pooling layers, fully connected layers and an output layer, the convolutional layer is used to extract the spatial light intensity features of the modulated pupil image, the pooling layer is used to perform spatial feature downsampling on the spatial light intensity features to obtain image features, the fully connected layer is used to map the extracted image features to a one-dimensional regression space, and the output layer outputs the translation error prediction value.
7. The method for detecting the precise co-phase error of a splicing mirror based on novel pupil plane information according to claim 6, characterized in that, The training process of the pre-trained convolutional neural network includes: Obtain a training sample dataset, wherein the training sample dataset includes an image dataset and a label dataset, the image dataset being captured modulation pupil images, and the label dataset being translation error values corresponding to the modulation pupil images in the image dataset; The modulated pupil image from the image dataset is input into the convolutional neural network to be trained to obtain the corresponding translation error prediction value. Calculate the loss value between the predicted translation error value and the corresponding translation error value in the label dataset; The convolutional neural network to be trained is optimized based on the loss value until the network converges, thus obtaining a pre-trained convolutional neural network.
8. The method for detecting the precise co-phase error of a splicing mirror based on novel pupil plane information according to claim 7, characterized in that, The calculation of the loss value between the predicted translation error value and the corresponding translation error value in the label dataset includes: The loss value between the predicted translation error and the corresponding translation error value in the labeled dataset is calculated using the mean squared error loss function.
9. The method for detecting the precise co-phase error of a splicing mirror based on novel pupil plane information according to claim 6, characterized in that, The method further includes: The phase error between the splicing sub-mirrors (1) is compensated in reverse by the predicted translation error value, and the phase splicing between the splicing sub-mirrors (1) segments is completed.
10. The method for detecting the precise co-phase error of a splicing mirror based on novel pupil plane information according to claim 6, characterized in that, The method further includes: When there are two or more stitching sub-mirrors (1), the modulation pupil plane image between any two adjacent stitching sub-mirrors (1) is obtained respectively, and the corresponding relative translation error prediction value is output through a pre-trained convolutional neural network. Any central stitching sub-mirror (1) is designated as the global zero-phase reference. Through cascading operations, the predicted relative translation error is uniformly converted into the absolute translation error of each stitching sub-mirror (1) relative to the reference. Based on the calculated absolute translation error, a control command is generated to drive the full array actuator to adjust synchronously and complete the global fine cophase splicing of the multi-segment sub-mirror (1) system.
