Cross-scale composite wavefront sensing method based on deep learning model

Abstract

The invention discloses a cross-scale composite wavefront sensing method based on a deep learning model, and the method comprises the steps: achieving the combination of a Shack-Hartmann wavefront sensor (SHWFS) measurement system and an off-axis digital holography (OADH) measurement system through a composite wavefront sensing system, and carrying out the data fusion of wavefront measured by the composite wavefront sensing system through an MSWM-Net model based on a deep neural network, the advantages of the performance of the two measurement systems are complemented, and the high-confidence reconstructed wavefront is further obtained. When wavefront aberrations containing different spatial frequencies are measured, the method not only solves the problem of low phase precision in SHWFS measurement, but also solves the limitation that wavefront is wrapped in a # imgabs0 # period in OADH measurement. The feasibility and effectiveness of the method in wavefront detection are verified through tests. The method is easier to realize quantitative measurement of the atmospheric turbulence aberration, can provide design input parameters for an atmospheric turbulence aberration correction system, and can also be used for distortion wavefront detection of the atmospheric turbulence aberration correction system.

Description

Technical Field

[0001] The present invention belongs to the field of adaptive optics, and mainly relates to wavefront detection technology in adaptive optical systems, especially to cross-scale composite wavefront detection technology that takes into account both Shack-Hartmann wavefront detection systems and off-axis digital holographic wavefront detection systems, and particularly to a cross-scale composite wavefront sensing method based on a deep learning model. Background Art

[0002] Adaptive Optics (AO) technology uses a control system to drive active devices to correct wavefront distortion, effectively improving the performance of optical systems. It is widely used in astronomy, biomedicine, laser communications, etc. Wavefront sensors are an important part of adaptive optical systems. In the field of wavefront measurement, Shack-Hartmann Wavefront Sensor (SHWFS) is widely used due to its simple structure, high light energy utilization, and strong anti-noise interference ability. The Shack-Hartmann wavefront sensor is mainly composed of a microlens array and a charge coupled device (CCD). The basic principle is that the microlens array divides and samples the distorted wavefront and focuses the light beam on the CCD target surface. By calculating the centroid offset of the light spot between the distorted wavefront and the plane wavefront, the average slope of the distorted wavefront within the sub-aperture range is further calculated, and finally the wavefront reconstruction algorithm is used to restore the wavefront. However, when designing SHWFS, a trade-off between detection accuracy and dynamic range must be made. Off-axis digital holography (OADH) is an optical imaging technology that combines traditional optical holography with computer image processing technology. Compared with conventional imaging technology, OADH has the advantages of high precision, high temporal and spatial resolution, and high linearity. The basic principle is that the target detection wavefront interferes with the reference beam, the interference fringes are focused on the CCD camera plane, and the target detection wavefront is reconstructed according to the holographic reconstruction algorithm.

[0003] When measuring wavefront aberrations containing different scales (high and low frequencies), SHWFS can provide a large dynamic range, but its spatial resolution is limited due to its working principle. OADH can provide high spatial resolution, but its phase is limited. The traditional single wavefront measurement device cannot simultaneously meet the wavefront measurement requirements of dynamic range and high spatial resolution. Therefore, it is of great significance to carry out research on a cross-scale composite wavefront sensing method.

[0004] In recent years, deep learning technology has been used to transform input data into more abstract and high-level representations through nonlinear transformation, automatically learn corresponding features from massive data, and significantly reduce the need for modeling complex physical processes. Therefore, it has a strong advantage in processing high-dimensional data. At present, deep learning has been extended to the fields of digital holographic imaging, SHWFS wavefront detection, etc., such as digital holography (from hologram to phase and intensity image of the object), phase unwrapping (from wrapped phase to absolute phase), Shack-Hartmann (reconstruction of wavefront aberration phase or its Zernike coefficient), etc. In addition, deep learning can also automate and accelerate the analysis and processing of imaging data, thereby improving the efficiency and accuracy of data analysis. How to perform cross-scale composite wavefront sensing methods based on deep learning models is a problem that needs to be studied and solved. Summary of the invention

[0005] The purpose of this invention is to solve the problem that the existing traditional single measurement system cannot simultaneously meet the wavefront measurement requirements of large dynamic range and high spatial resolution. By analyzing the SHWFS and OADH wavefront detection characteristics for actual physical systems, a composite wavefront sensing system is constructed. A deep learning-based MSWM-Net model is proposed to fuse the wavefront data detected by the composite wavefront sensing system, and further obtain a high-confidence reconstructed wavefront.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is: a cross-scale composite wavefront sensing method based on a deep learning model, wherein the wavefront measurement is performed by a composite wavefront sensing system experimental device, and the composite wavefront sensing system experimental device comprises: a laser and an optical fiber beam splitter connected in sequence, and two branches are connected in parallel after the optical fiber beam splitter, which are a first branch composed of a first collimating lens barrel, an air outlet device, a fast reflector, a first beam splitter, a first beam reduction lens barrel, a microlens array, a first magnification matching system, and a first photoelectric detector, and a second branch composed of a second collimating lens barrel, a second beam splitter, a second beam reduction lens barrel, a second magnification matching system, and a second photoelectric detector, and a synchronous trigger device is connected after the first photoelectric detector and the second photoelectric detector, and the synchronous trigger device is used to realize the synchronous triggering of the laser and the first photoelectric detector and the second photoelectric detector; The cross-scale composite wavefront sensing method based on deep learning model includes the following steps: Step 1: First, calibrate the system. The process includes: turn on the laser, and the outgoing light beam passes through the fiber beam splitter to obtain two identical outgoing light beams, namely the first light beam and the second light beam, wherein the first light beam passes through the collimating lens barrel, the fast reflector and the first beam splitter respectively, and then enters the first beam reduction lens barrel, and then emits to the microlens array. After the first light beam is split by the microlens array, it is imaged on the first photodetector in the form of a calibration spot array through the first magnification matching system to form a calibration spot image; the second light beam passes through the second collimating lens barrel and then passes through the second beam splitter, and interferes with the first light beam passing through the first beam splitter, and is imaged on the second photodetector after passing through the second beam reduction lens barrel to form an interference fringe image; Step 2: extracting the calibration spot image on the first photodetector and the interference fringe image on the second photodetector; Step 3: Turn on the air outlet device, turn on the laser, and the outgoing light beam passes through the optical fiber beam splitter to obtain two identical outgoing light beams, namely the first light beam and the second light beam, wherein the first light beam passes through the first collimating lens barrel to the air outlet device, at this time, the first light beam is aberrated, and then passes through the fast reflector and the first beam splitter, the first beam splitter divides the first light beam into a third light beam and a fourth light beam, the fourth light beam passes through the second beam splitter, the third light beam enters the first beam reduction lens barrel, and then is emitted to the microlens array, after the third light beam is split by the microlens array, it is imaged on the first photodetector in the form of a dot matrix with a spatial offset relative to the calibration position through the first magnification matching system; the second light beam passes through the second collimating lens barrel and then passes through the second beam splitter, interferes with the fourth light beam, and then passes through the second beam reduction lens barrel and is imaged on the second photodetector to form an interference fringe image with aberration; Step 4: extract the distorted spot image on the first photodetector, calculate the offset of the distorted spot relative to the calibration position by the spot positioning technology, calculate the wavefront slope by using the offset, and then reconstruct the distorted wavefront by using the wavefront restoration algorithm; Step 5: Extract the interference fringe image on the second photodetector, perform Fourier transform on the interference fringe image to obtain a spectral image, use frequency domain filtering to obtain the +1 order term containing the reconstructed wavefront amplitude information, and then obtain the light field distribution in the focusing plane according to the scalar diffraction theory; Step 6: Input the reconstructed wavefront of the Shack-Hartmann wavefront measurement system and the reconstructed wavefront of the off-axis digital holographic wavefront measurement system into the MSWM-Net model to further obtain a high-confidence reconstructed wavefront.

[0007] The principle of the present invention is: a cross-scale composite wavefront sensing method based on a deep learning model, which collects the distorted wavefront at the same frequency through the Shack-Hartmann wavefront measurement system and the off-axis digital holographic wavefront measurement system, and then uses the MSWM-Net model to fuse the data of the two collected wavefronts to further obtain a high-confidence reconstructed wavefront. The device includes: a laser, a fiber beam splitter, a collimating lens barrel, a fast reflector, a beam splitter, a beam reduction lens barrel, a microlens array, a magnification matching system, a photoelectric detector, a synchronous trigger device, and an air outlet device. These optical devices constitute a composite wavefront sensing system, including a Shack-Hartmann wavefront measurement system and an off-axis digital holographic wavefront measurement system.

[0008] Compared with the prior art, the present invention has the following beneficial effects: (1) This paper proposes a new cross-scale composite wavefront sensing method that combines the system performance advantages of Shack-Hartmann wavefront sensing and off-axis digital holographic wavefront sensing. When measuring aberrations of different scales, this method breaks the low phase accuracy of SHWFS wavefront sensing and the problem of wrapping the wavefront in OADH measurement. Limitations of the cycle.

[0009] (2) The present invention proposes a deep neural network-based MSWM-Net model, which is a multi-input single-output efficient network structure composed of a neural network and a convolutional attention mechanism. By fusing the wavefront measured by the composite sensor, a high-confidence wavefront is further obtained. Compared with the existing neural network model, it is easier to implement data fusion.

[0010] (3) When measuring wavefront aberrations containing different spatial frequencies, the present invention has higher wavefront restoration accuracy than existing methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 This is a framework diagram of a cross-scale composite wavefront sensing method based on a deep learning model of the present invention; Figure 2 The experimental device diagram of the composite wavefront sensing system, wherein 1 is a laser, 2 is a fiber beam splitter, 3-1 is a first collimating lens barrel, 3-2 is a second collimating lens barrel, 4 is a fast reflector, 5-1 is a first beam splitter, 5-2 is a second beam splitter, 6-1 is a first beam reduction lens barrel, 6-2 is a second beam reduction lens barrel, 7 is a microlens array, 8-1 is a first magnification matching system, 8-2 is a second magnification matching system, 9-1 is a first photoelectric detector, 9-2 is a second photoelectric detector, 10 is a synchronous trigger device, and 11 is an air outlet device; Figure 3 Synchronous triggering system for laser and photodetector; Figure 4 Schematic diagram of the MSWM-Net model, where Figure 4 The Arabic numerals listed in are the number of channels, indicating the depth or dimension of the feature map; Figure 5 The reconstructed wavefront image is obtained using the cross-scale composite wavefront sensing method based on the deep learning model. DETAILED DESCRIPTION

[0012] In order to make the working principle and implementation process of the device of the present invention clearer, the present invention is further described below in conjunction with the accompanying drawings and specific implementation methods.

[0013] Figure 1 This is a framework diagram of a cross-scale composite wavefront sensing method based on a deep learning model. The specific implementation of this method is to first use Figure 2 The composite wavefront sensing system experimental device is used to measure the wavefront. The composite wavefront sensing system experimental device includes: a laser 1 and an optical fiber beam splitter 2 connected in sequence, two branches connected in parallel after the optical fiber beam splitter 2, which are a first branch consisting of a first collimating lens barrel 3-1, an air outlet device 11 for simulating atmospheric turbulence, a fast reflector 4, a first beam splitter 5-1, a first focusing lens barrel 6-1, a microlens array 7, a first magnification matching system 8-1, and a first photoelectric detector 9-1, and a second branch consisting of a second collimating lens barrel 3-2, a second beam splitter 5-2, a second focusing lens barrel 6-2, a second magnification matching system 8-2, and a second photoelectric detector 9-2. The first photoelectric detector 9-1 and the second photoelectric detector 9-2 are connected to a synchronous trigger device 10, and the synchronous trigger device 10 is used to realize the synchronous triggering of the laser 1 and the first photoelectric detector 9-1 and the second photoelectric detector 9-2. The experimental principle of the experimental device includes a Shack-Hartmann wavefront measurement system and an off-axis digital holographic wavefront measurement system. The photoelectric sensors corresponding to the Shack-Hartmann wavefront measurement system and the off-axis digital holographic wavefront measurement system are collected at the same frequency through a synchronous trigger device 10. Then, the collected spot array image and interference fringes are further processed to obtain the light field distribution map of the Shack-Hartmann reconstructed wavefront and the off-axis digital hologram respectively. Finally, the light field distribution map of the Shack-Hartmann reconstructed wavefront and the off-axis digital hologram is input into the MSWM-Net model.

[0014] In the embodiment of the present invention, the laser wavelength is 1024nm and the number of microlens arrays is 16. 16 (Ф4mm), sub-aperture size is 0.24mm×0.24mm, microlens focal length is 8.5mm, magnification matching system matching ratio , full aperture size is 3.8mm, pixel size is 15 , the sampling frequency is 50 25000Hz. The photoelectric detector takes MEMRECAM ACS-1M60 of Weiche Company as an example.

[0015] Among them, first, the air outlet device in the optical path is not turned on, and the laser 1 is turned on to calibrate the system. After the calibration, the air outlet device 11 is turned on again, and the composite sensor system is used to reconstruct the distorted wavefront with aberration.

[0016] When measuring the distorted wavefront, the synchronous trigger device 10 sends a synchronous trigger signal to the laser 1 and the first photodetector 9-1 and the second photodetector 9-2 at a preset frequency. After receiving the trigger signal, the laser 1 emits laser, and the first photodetector 9-1 and the second photodetector 9-2 start to collect images.

[0017] In one embodiment, the first magnification matching system 8 - 1 and the second magnification matching system 8 - 2 may be fixed imaging magnification systems or variable imaging magnification systems.

[0018] In one embodiment, the air outlet device 11 is a device for simulating and generating atmospheric turbulence, which can control the flow rate, and can also be a turbulence generator, etc.

[0019] The cross-scale composite wavefront sensing method based on deep learning model includes the following steps: Step 1: First, calibrate the system. The process includes: turn on the laser 1, and the outgoing light beam passes through the optical fiber beam splitter 2 to obtain two identical outgoing light beams, namely the first light beam O and the second light beam R. The first light beam O passes through the first collimating lens barrel 3-1, the fast reflector 4 and the first beam splitter 5-1 respectively, and then enters the first beam reduction lens barrel 6-1, and then emits to the microlens array 7. After the first light beam O is split by the microlens array 7, it is imaged on the first photodetector 9-1 in the form of a calibration spot array through the first magnification matching system 8-1 to form a calibration spot image; the second light beam R passes through the second collimating lens barrel 3-2 and the second beam splitter 5-2, and interferes with the first light beam O passing through the first beam splitter 5-1, and is imaged on the second photodetector 9-2 after passing through the second beam reduction lens barrel 6-2, forming an interference fringe image; When measuring the aberration, the synchronous trigger device 10 is activated at a preset frequency, thereby synchronizing the trigger signal to the laser 1 and the first photodetector 9-1 and the second photodetector 9-2. Figure 3 As shown, the laser 1 emits a laser pulse after receiving a trigger signal, and the first photodetector 9-1 and the second photodetector 9-2 start to open the electronic shutter pulse after recognizing the falling edge, and the exposure time begins. The first photodetector 9-1 and the second photodetector 9-2 start to transfer charge after recognizing the rising edge, and the exposure ends, and the image signal is output to the acquisition card at the same time.

[0020] Step 2: extracting the calibration spot image on the first photodetector 9-1 and the interference fringe image on the second photodetector 9-2; Step 3: Open the air outlet device 11. The laser 1 is turned on, and the outgoing light beam passes through the optical fiber beam splitter 2 to obtain two identical outgoing light beams, namely, the first light beam O and the second light beam R. The first light beam O passes through the first collimating lens barrel 3-1 and is emitted to the air outlet device 11. At this time, the first light beam O is aberrated. Then, it passes through the fast reflecting mirror 4 and the first beam splitter 5-1. The first beam splitter 5-1 splits the first light beam O into a third light beam O_1 and a fourth light beam O_2. The fourth light beam O_2 passes through the second beam splitter 5-2. The third light beam O_1 enters the first beam reduction lens barrel 6-1 and then is emitted to the microlens array 7. After the third light beam O_1 is split by the microlens array 7, it is imaged on the first photodetector 9-1 in the form of a dot matrix with a spatial offset relative to the calibration position through the first magnification matching system 8-1; the second light beam R passes through the second collimating lens barrel 3-2 and then passes through the second beam splitter 5-2, interferes with the fourth light beam O_2, and then passes through the second beam reduction lens barrel 6-2 and is imaged on the second photodetector 9-2 to form an interference fringe image with aberrations. Step 4: Extract the distorted spot array image on the first photodetector 9-1, calculate the offset of the distorted spot relative to the calibration position through the spot positioning technology, use the offset to calculate the wavefront slope, and then use the wavefront restoration algorithm to reconstruct the distorted wavefront.

[0021] In one embodiment, the spot positioning technology includes a weighted centroid method, a threshold centroid method, a matched filter method, a registration algorithm, or any other method that can locate the spot position.

[0022] The wavefront slope calculation process is described below: Microlens arrays The complex amplitude of the light field on the subaperture focal plane at the position for: , in, , is the position information of the corresponding subaperture, is the complex amplitude corresponding to the first light beam O in step 1, , is the sub-aperture calibration position information, is an imaginary unit, is the laser wavelength, for Figure 1 Focal length of microlens array 7, wave number .

[0023] The centroid position of the focused light spot in the x direction is: , in, is the centroid position of the focused light spot in the x direction, for The conjugate of To take the real part of a complex number.

[0024] The centroid position of the focused light spot in the y direction is: , in, is the centroid position of the focused light spot in the y direction, for The conjugate of To take the real part of a complex number.

[0025] The incident wavefront slope in the x and y directions: , in, , is the wavefront slope of the incident wavefront in the x and y directions, , Indicates The centroid position of the focused spot of the incident light wave of each sub-aperture, and Indicates the centroid position of the calibration light wave focusing spot.

[0026] The distorted wavefront is then reconstructed using a wavefront restoration algorithm. In one embodiment, the wavefront restoration algorithm includes a pattern method, a regional method, and other methods that can achieve wavefront restoration using the sub-spot array data of the Shack-Hartmann wavefront sensor. The Shack-Hartmann wavefront sensor uses the pattern method to reconstruct the wavefront, and further obtains the reconstructed wavefront of the Shack-Hartmann wavefront measurement system. The reconstructed Zernike polynomial coefficients are: , Where D is the number of effective sub-apertures; R is the wavefront restoration matrix, which is used to construct the restoration matrix for calculating the aberration mode coefficients using the centroid offset or slope data of the light spot within the sub-aperture. The matrix can be generated in advance as a system configuration; M is the order of the reconstructed Zernike polynomial; for( , ) corresponds to the wavefront slope matrix; a is the reconstructed Zernike polynomial coefficient.

[0027] Step 5: Extract the interference fringe image on the second photodetector 9-2, perform Fourier transform on the interference fringe image to obtain a spectral image, use frequency domain filtering to obtain the +1 order term containing the reconstructed wavefront amplitude information, and then obtain the focal plane light field distribution based on the scalar diffraction theory, and further obtain the reconstructed wavefront of the Shack-Hartmann wavefront measurement system.

[0028] Step 6: Input the reconstructed wavefront of the Shack-Hartmann wavefront measurement system and the reconstructed wavefront of the off-axis digital holographic wavefront measurement system into Figure 4In the MSWM-Net model shown in FIG, a high-confidence reconstructed wavefront is further obtained, such as Figure 5 shown.

[0029] The MSWM-Net model of an embodiment of the present invention is described below. The MSWM-Net model is a multi-input single-output efficient network structure composed of a neural network and a convolutional attention mechanism. The framework of the model is based on the typical U-Net structure, including a convolutional layer, a maximum pooling layer, an upsampling layer (Convtranspose2d), and a convolutional attention mechanism (CBAM) module. Each convolutional layer is followed by a batch normalization layer and a LeakyReLu activation function to accelerate training. Among them, Convtranspose2d reduces the number of feature channels by half and cascades with the corresponding cropped features from the downsampling block. CBAM generates image features with attention mechanism weights from channel and spatial dimensions to improve the feature extraction capability of the network. Figure 4 The general framework of the model is shown in Figure 1, where the Arabic numerals listed are the number of channels, indicating the depth of the feature map. The downsampling process is accompanied by an increase in the number of channels, which better represents the attributes and features of the input image, thereby improving the accuracy of segmentation. The upsampling process reduces the number of channels, allowing the network to reduce computational complexity while maintaining segmentation performance, thereby improving the inference speed.

[0030] The above description is only a specific implementation of the present invention, but the protection scope of the present invention is not limited thereto. Any person familiar with the technology can understand and think of any changes or substitutions within the technical scope disclosed by the present invention, which should be included in the scope of the present invention.

Claims

1. A cross-scale composite wavefront sensing method based on a deep learning model, characterized in that: The wavefront measurement is performed by a composite wavefront sensor system experimental device, which comprises: a laser (1) and an optical fiber beam splitter (2) connected in sequence, and two branches connected in parallel after the optical fiber beam splitter (2), which are respectively composed of a first collimating lens barrel (3-1), an air outlet device (11) for simulating atmospheric turbulence, a fast reflector (4), a first beam splitter (5-1), a first beam reduction lens barrel (6-1), a microlens array (7), a first magnification matching system (8-1), and a first photoelectric detector ( A first branch is composed of a second collimating lens barrel (3-2), a second beam splitter (5-2), a second beam reduction lens barrel (6-2), a second magnification matching system (8-2), and a second photoelectric detector (9-2), wherein the first photoelectric detector (9-1) and the second photoelectric detector (9-2) are connected to a synchronous trigger device (10), and the synchronous trigger device (10) is used to achieve synchronous triggering of the laser (1) and the first photoelectric detector (9-1) and the second photoelectric detector (9-2); The cross-scale composite wavefront sensing method based on deep learning model includes the following steps: Step 1: First, calibrate the system. The process includes: turning on the laser (1), and the outgoing light beam passes through the optical fiber beam splitter (2) to obtain two identical outgoing light beams, namely the first light beam and the second light beam, wherein the first light beam passes through the first collimating lens barrel (3-1), the fast reflector (4) and the first beam splitter (5-1) respectively, and then enters the first beam reduction lens barrel (6-1), and then is emitted to the microlens array (7). After the first light beam is split by the microlens array (7), it is imaged on the first photodetector (9-1) in the form of a calibration light spot array through the first magnification matching system (8-1), forming a calibration light spot image; the second light beam passes through the second collimating lens barrel (3-2) and then passes through the second beam splitter (5-2), and interferes with the first light beam passing through the first beam splitter (5-1), and then passes through the second beam reduction lens barrel (6-2) and then is imaged on the second photodetector (9-2), forming an interference fringe image; Step 2: extracting a calibration spot image on the first photodetector (9-1) and an interference fringe image on the second photodetector (9-2); Step 3: Turn on the air outlet device (11), turn on the laser (1), and the outgoing light beam passes through the optical fiber beam splitter (2) to obtain two identical outgoing light beams, namely the first light beam and the second light beam. The first light beam passes through the first collimating lens barrel (3-1) and is emitted to the air outlet device (11). At this time, the first light beam has an aberration, and then passes through the fast reflection mirror (4) and the first beam splitter (5-1). The first beam splitter (5-1) splits the first light beam into a third light beam and a fourth light beam. The fourth light beam passes through the second beam splitter (5-2), and the third light beam enters The third light beam is split by the microlens array (7) and then imaged on the first photodetector (9-1) in the form of a dot matrix with a spatial offset relative to the calibration position through the first magnification matching system (8-1); the second light beam passes through the second collimating lens barrel (3-2) and then passes through the second beam splitter (5-2), interferes with the fourth light beam, and then passes through the second beam reducing lens barrel (6-2) and then imaged on the second photodetector (9-2) to form an interference fringe image with aberration; Step 4: extracting the distorted spot image on the first photodetector (9-1), calculating the offset of the distorted spot relative to the calibration position by using the spot positioning technology, calculating the wavefront slope by using the offset, and then reconstructing the distorted wavefront by using the wavefront restoration algorithm; Step 5: extract the interference fringe image on the second photodetector (9-2), perform Fourier transform on the interference fringe image to obtain a spectral image, use frequency domain filtering to obtain the +1 order term containing the reconstructed wavefront amplitude information, and then obtain the light field distribution in the focusing plane according to the scalar diffraction theory; Step 6: Input the reconstructed wavefront of the Shack-Hartmann wavefront measurement system and the reconstructed wavefront of the off-axis digital holographic wavefront measurement system into the MSWM-Net model to further obtain a high-confidence reconstructed wavefront.

2. The cross-scale composite wavefront sensing method based on a deep learning model according to claim 1, characterized in that: The first magnification matching system (8-1) and the second magnification matching system (8-2) are fixed imaging magnification systems or variable imaging magnification systems.

3. The cross-scale composite wavefront sensing method based on deep learning model according to claim 1, characterized in that: The wind outlet device (11) is a device for simulating and generating atmospheric turbulence, or a turbulence generator.

4. The cross-scale composite wavefront sensing method based on a deep learning model according to claim 1, characterized in that: The spot positioning technology includes one of the weighted centroid method, the threshold centroid method, the matched filter method, and the registration algorithm.

5. The cross-scale composite wavefront sensing method based on deep learning model according to claim 1, characterized in that: The wavefront restoration algorithm includes one of the pattern method and the regional method.

6. The cross-scale composite wavefront sensing method based on deep learning model according to claim 1, characterized in that: The MSWM-Net model is a multi-input single-output efficient network structure consisting of a neural network and a convolutional attention mechanism.

Images