Self-adaptive imaging dual-wavelength reflective Fourier laminated microscopy system

Abstract

The invention provides a self-adaptive imaging dual-wavelength reflection-type Fourier lamination microscopic system, and the system comprises a bright field and dark field cooperative flexible light source module which is used for improving the imaging quality, reducing the image collection time, and improving the resolution of a reconstructed image; the high-speed self-adaptive optical module is used for detecting image aberration in real time and compensating the aberration in real time by controlling a deformable mirror; the microscopic imaging module is used for realizing super-resolution image reconstruction and realizing depth three-dimensional reconstruction through the image of the dual-wavelength channel; the computing control platform is used for configuring a plurality of high-performance general GPUs and controlling software operation of other modules; and the image reconstruction algorithm module is used for realizing efficient and universal multi-mode image 2D and 3D reconstruction. According to the method, a deep learning-based FPM network fused with an imaging physical model is combined, the data acquisition amount and acquisition time are remarkably reduced, meanwhile, the reconstruction structure and the reconstruction stability are optimized, the universality of the FPM to different imaging targets is improved, and efficient and universal multi-mode image reconstruction is achieved.

Description

Technical Field

[0001] The present invention belongs to the field of computational imaging and deep learning, and specifically relates to an adaptive imaging dual-wavelength reflective Fourier stack microscopy system. Background Art

[0002] The development of modern optics has posed three major challenges to optical microscopy and imaging systems: (1) breaking through the diffraction limit of optical imaging; (2) the bottleneck of spatial bandwidth product; and (3) the high cost of high-end optical systems. Computational optical imaging is an emerging imaging method that achieves specific imaging functions and characteristics by jointly optimizing optical systems and signal processing. Through the organic combination of optics and algorithms, it provides new means and new ideas for breaking through many restrictive factors in traditional imaging systems. Among them, Fourier Ptychographic Microscopy (FPM) is a large-field-of-view, high-resolution, quantitative phase computational microscopy technology developed in recent years. This technology integrates the concepts of phase recovery and synthetic aperture. The sample is illuminated by plane waves at different angles and imaged through a low numerical aperture objective lens, resulting in the frequency spectrum of the object on the back focal plane of the objective lens being shifted to corresponding different positions, and some frequency components that originally exceeded the numerical aperture of the objective lens can be transmitted to the imaging surface for imaging.

[0003] However, existing FPM usually uses a fixed LED light source, which has low optical efficiency and poor flexibility. Long exposure time is required to obtain images with sufficient contrast during dark field illumination. Since FPM usually needs to collect dozens or even hundreds of images, the long exposure time leads to low image acquisition efficiency of existing FPM. Like other optical systems, the image quality of FPM is also affected by the aberration of the optical system. Although FPM can obtain and compensate for the optical aberration of the system through algorithms during image reconstruction, this method has a very large amount of calculation, which further increases the computational complexity of FPM image reconstruction and reduces the image reconstruction efficiency. At the same time, the typical FPM uses transmissive illumination, which is only suitable for thin samples with good light transmittance, and cannot be applied to non-transparent samples with a certain thickness (micrometer level). Moreover, the illumination light source of the existing system mostly uses visible light. If it is used for rough surfaces at the micro-nano scale, the visible light will be reflected (the surface of the object acts as a mirror), making it difficult for the tiny surface feature information to be captured by the camera. Summary of the invention

[0004] The object of the present invention is to provide an adaptive imaging dual-wavelength reflection Fourier stacking microscopy system.

[0005] The technical scheme of the present invention is as follows:

[0006] An adaptive imaging dual-wavelength reflective Fourier stack microscopy system, the system comprising:

[0007] A flexible light source module that supports bright field and dark field collaboration, which is used to improve imaging quality, reduce image acquisition time, and increase the resolution of reconstructed images. It also supports flexible illumination of deep ultraviolet (DUV) light sources.

[0008] High-speed adaptive optics module, used to detect image aberration in real time and compensate for aberration in real time by controlling the deformable mirror, thus reducing the complexity of image reconstruction and the amount of computation required for image reconstruction;

[0009] Microscopic imaging module, used to achieve super-resolution image reconstruction and deep three-dimensional reconstruction through dual-wavelength channel images;

[0010] The computing control platform is used to configure multiple high-performance general-purpose GPUs, use parallel computing to provide sufficient computing power for image reconstruction, and centrally control the software operation of other modules;

[0011] Image reconstruction algorithm module, used to achieve efficient and universal multi-modal image 2D and 3D reconstruction.

[0012] Furthermore, the flexible light source module for coordinated bright field and dark field is composed of a lifting bracket, a rotating mirror, a bright field light source and a dark field light source, wherein the bright field light source and the dark field light source are coordinated through the motion control of the rotating mirror and the lifting bracket to achieve controllable multi-angle lighting conditions using a small number of light sources.

[0013] Furthermore, the system uses a high optical efficiency DUV LED package as a light source, and the high optical efficiency DUV LED package is sequentially composed of a substrate, a light-emitting chip, an aspherical reflective chip, a housing, a micro-nano surface homogenizer, and a protective glass.

[0014] Furthermore, the high-speed adaptive optical module is composed of a wavefront sensor, a beam splitter and a deformable mirror; wherein, the light path is introduced into the wavefront sensor through the beam splitter to detect the system aberration in real time, and the aberration is further compensated in real time through feedback control of the deformable mirror.

[0015] Furthermore, the microscopic imaging module is composed of a stage, an objective lens, a beam splitter and a camera.

[0016] Furthermore, the image reconstruction algorithm consists of three main parts: data synthesis, residual attention module, and Transformer module;

[0017] Among them, data synthesis synthesizes the original data into an amplitude tensor and a phase tensor to reduce the amount of network calculation;

[0018] The residual attention module is used to effectively extract features from images and suppress the negative problems caused by the increase in the number of convolutional layers;

[0019] Transformer module, used to enhance the network's ability to extract multi-scale information.

[0020] Furthermore, the residual attention module extracts image features through residual learning and channel attention mechanism, wherein residual learning accelerates the convergence of the network and solves the network degradation problem caused by too many network layers, and the channel attention mechanism adaptively allocates weights according to the importance of different channel features, thereby emphasizing important channels and suppressing unnecessary channels.

[0021] Furthermore, the image reconstruction algorithm designs the loss function by combining space and frequency domains.

[0022] Compared with the prior art, the present invention has the following advantages:

[0023] The present invention supports the use of high optical efficiency LED packages as light sources, and uses flexible lighting strategies to significantly improve the resolution limit of FPM reconstruction images, while achieving programmable adjustment of imaging resolution without changing the FPM objective lens. In addition, the dual-wavelength difference frequency technology is combined to solve the problem of deep three-dimensional reconstruction of thick samples that is difficult to achieve with traditional FPM.

[0024] For opaque or reflective samples such as wafers and semiconductor packages, a reflective microscopy system combined with a flexible illumination source provides high degrees of freedom and high contrast illumination and imaging capabilities.

[0025] In order to solve the problem that FPM needs to collect a large number of low-resolution images as the basis for reconstruction, resulting in large data collection volume and long calculation time, an adaptive optical module is added to the imaging system to compensate for imaging aberrations at the front end, thereby improving the resolution of the reconstructed image and reducing the processing time of the algorithm. In addition, the FPM network based on deep learning that integrates the imaging physical model can significantly reduce the amount of data collection and acquisition time, while optimizing the reconstruction structure and reconstruction stability, improving the versatility of FPM for different imaging targets and realizing efficient and universal multi-mode image reconstruction. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the embodiments of the invention. Where appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0027] Figure 1 A schematic diagram of the adaptive imaging dual-wavelength reflective Fourier stack microscopy system of the present invention is shown;

[0028] Figure 2A schematic diagram of a high optical efficiency DUV LED package of the present invention is shown;

[0029] Figure 3 The schematic diagram of the working principle of the high-speed adaptive optics of the present invention and the comparative schematic diagram of the control method are shown;

[0030] Figure 4 A schematic diagram of the image acquisition method driven by a reconstruction target of the present invention is shown;

[0031] Figure 5 The schematic diagram of the dual-wavelength difference frequency based on the optical filter of the present invention is shown;

[0032] Figure 6 A strategy diagram of deep learning-based reconstruction of the present invention is shown;

[0033] Figure 7 The structure diagram of the FPM reconstruction network based on deep learning of the present invention is shown;

[0034] Figure 8 A schematic diagram of the residual attention module structure of the present invention is shown;

[0035] Fig. 9 The schematic diagram of the Transformer module structure of the present invention is shown;

[0036] Fig.10 A schematic diagram showing the reconstruction of micro surface defects of a silicon wafer according to the present invention is shown;

[0037] Fig.11 A framework diagram of an adaptive imaging dual-wavelength reflective Fourier stacking microscopy system is shown. DETAILED DESCRIPTION

[0038] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0039] The present invention provides an adaptive imaging dual-wavelength reflective Fourier stacking microscopy system, in which a split bright field and dark field light source is designed. With the assistance of a mechanical structure with freely adjustable illumination angle and height, the light source lighting mode is optimized, and combined with dual-wavelength imaging technology, a large field of view and high-resolution three-dimensional information reconstruction of non-transparent / surface reflective samples is achieved. Specifically, it includes: Figure 1 The five modules shown are: a flexible light source module for coordinated bright field and dark field (M1), a high-speed adaptive optics module (M2), a microscopic imaging module (M3), a computing control platform (M4), and image reconstruction software (M5).

[0040] Since the imaging resolution and image quality of FPM are highly dependent on the light source module, the flexible light source that cooperates with bright field and dark field can improve the imaging quality, reduce the image acquisition time and improve the resolution of the reconstructed image, and can support flexible illumination of deep ultraviolet light source to achieve super-resolution detail illumination; the high-speed adaptive optics module can detect image aberrations in real time, and compensate for the aberrations in real time by controlling the deformable mirror, reducing the complexity of the image reconstruction problem and the amount of calculation for image reconstruction; the microscopic imaging module supports low-cost microscopic imaging systems using the DUV band, and realizes super-resolution three-dimensional image reconstruction through images of dual-wavelength channels; the computing control platform is equipped with multiple high-performance general-purpose GPUs, which use parallel computing to provide sufficient computing power for image reconstruction, and centrally control the software operation of other modules; in order to improve the versatility of FPM for different imaging targets and the stability of image reconstruction, the image reconstruction algorithm and software module expand the physical model of reflective FPM imaging, and combine it with the deep neural network (DNN) in the image reconstruction process to achieve efficient and general multi-mode image (2D+3D) reconstruction.

[0041] In addition, in order to meet the requirements of microscopic imaging of samples with different surface features, the light source band used in the present invention can be freely selected and replaced. If there are micrometer or nanometer scale structures on the sample surface, a short-wavelength light source such as ultraviolet light or even deep ultraviolet light can be used. Under the illumination of this wavelength, more directional scattering can occur on the tiny surface features, thereby solving the information collection problem of the corresponding features, and at the same time, there is a corresponding improvement in resolution. For simplicity, the technical details of the present invention are described below using the DUV light source as an example, and light sources of other bands follow the same technical route.

[0042] The structural design of the present invention is as follows Figure 1 As shown, the high optical efficiency DUV LED package, on the one hand, fully utilizes the light emitted from the light-emitting chip at a large divergence angle through the aspherical reflective surface to reduce light energy loss, and on the other hand, replaces the traditional beam shaping lens with a micro-nano surface homogenizer to make the output light distribution more uniform. The high optical efficiency DUV LED package makes the light energy it emits concentrated and evenly distributed, and can increase the optical efficiency of the reflective FPM system by more than ten times compared to ordinary large divergence angle LEDs. For the DUV flexible lighting solution, the present invention uses a light source with a wavelength of 210-200nm.

[0043] For dark-field light sources, different illumination angles for imaging targets can be achieved through a combination of a programmable control lifting mechanism and a two-dimensional rotating reflector. Increasing the illumination angle of the light source can increase the numerical aperture of the FPM illumination, thereby increasing the imaging equivalent numerical aperture of the FPM system (the imaging equivalent numerical aperture is the sum of the numerical aperture of the objective lens and the numerical aperture of the illumination). Therefore, this flexible light source module design can achieve programmable adjustment of the imaging resolution without changing the FPM objective lens. In terms of the spectrum, the increase in the dark-field illumination numerical aperture is equivalent to expanding the high-frequency boundary of the dark-field spectrum, thereby increasing the resolution limit of the FPM reconstructed image ( Figure 1 on the right).

[0044] The present invention introduces a high-speed adaptive optics (AO) module based on a wavefront sensor and a deformable mirror (DM) in the FPM to achieve real-time compensation of aberrations ( Figure 3 ). This module is placed in the relay mirror group of the FPM imaging optical path, and the optical path is guided into the wavefront sensor through a beam splitter to detect the system aberration in real time, and the aberration is further compensated in real time through feedback control of the DM. The existing AO modules are mainly designed for the visible light band and the near-infrared band. The main work of the present invention in this regard is to design and process the microlens array, the core component of the wavefront sensor, for the DUV band and to adopt an image sensor that supports the DUV band. Based on the Zernike function representation of the aberration, wavefront reconstruction can be achieved by solving linear equations, and then mapping the Zernike coefficients to the parameter space of the DM to control and drive it to achieve aberration compensation. This sensing adaptive optical control method is as shown in Figure 3 As shown above. The present invention uses an end-to-end neural network to implement feedback control of AO, that is, the image collected by the wavefront sensor is used as the input of DNN and the DM control signal is directly output.

[0045] The present invention adopts Figure 4 The imaging target driven image acquisition method demonstrated, that is, the position and illumination angle of the light source are determined according to the spectral characteristics of the imaging target to be reconstructed, and the ideal position and illumination angle of the light source are achieved through programming control of the flexible light source module.

[0046] The optical imaging model of reflective FPM can be described as follows:

[0047] Assume that the distribution function of the imaging target is o(r) = Ae iφ(r) , the illumination of each LED on the imaging target plane can be regarded as a monochromatic plane wave with a wavelength of λ, and its incident angle depends on the physical position and illumination direction of the LED. Therefore, when the lth LED is lit, the wavefront emitted from the imaging target surface can be expressed as:

[0048]

[0049] where ξ l is the spatial frequency vector of the LED illumination direction. In this way, it is equivalent to introducing a frequency shift in the frequency domain plane of the imaging target

[0050]

[0051] Due to the limited aperture of the objective lens, the object wavefront is equivalent to passing through a low-pass filter once, and the filtering degree depends on the objective pupil function P(u). After passing through the imaging lens (Fourier transform), finally, on the imaging plane, the low-resolution image collected can be expressed as

[0052] y l (r) = |F -1 {P(u)F{y(r)}(u - ξ l )}| 2 #(3)

[0053] For the case where multiple LEDs are lit simultaneously, since the LED light sources are incoherent with each other, the collected image is the intensity weighted superposition of each LED illumination

[0054]

[0055] where the non-negative scalar c l is the relative brightness of the l-th LED ( Figure 5 ).

[0056] According to the above optical propagation model, with the idea of solving the non-linear inverse problem, the complex field of the imaging target is reconstructed. The distribution function of the imaging target and the image obtained by multiplexed illumination of multiple LEDs are discretized respectively, and are denoted as where p < q (the reconstructed high-resolution image has a better spatial bandwidth product). In this way, the reconstruction of the super-resolution image can be expressed as an optimization problem:

[0057]

[0058] where, Ω k is the index set when the k-th multiple LEDs are lit. And the discretized FPM system propagation model in the above formula is A l = F H P l F, where is the matrix expression of the pupil function with a frequency shift of ξ l . F and F H are the matrix expressions of the Fourier transform and the inverse Fourier transform respectively

[0059] Although the above model covers the propagation process of the sample complex field, in order to reconstruct it in three dimensions, the imaging target is required to meet the "thin sample" assumption (the phase changes gradually within the range of 2π). For the three-dimensional reconstruction of the reflection pattern imaging target, the present invention intends to add additional physical conditions at the illumination light source end, such as adding an adjustable UV filter to the DUV camera or a dual DUV camera with bandpass filtering, and introducing difference frequency technology ( Figure 5 ), achieving deep three-dimensional reconstruction.

[0060] In the dual-wavelength difference frequency technology, the calculation formula for the synthetic wavelength is:

[0061]

[0062] With λ 1 =210nm,λ 2 =215nm as an example, its synthetic wavelength is about 9μm. For samples with a depth smaller than this scale, the three-dimensional image of the sample can be calculated by reconstructing two two-dimensional images and using their corresponding composite phases:

[0063]

[0064] It should be pointed out that, like other difference frequency methods, due to the periodicity of the phase, the 3D reconstruction is non-unique when the measurement depth exceeds the synthetic wavelength.

[0065] The present invention proposes a solution based on deep learning, such as Figure 6 As shown in the figure, a deep neural network is used to reconstruct a high-resolution image from a series of low-resolution images generated by the FPM system in an end-to-end manner. This solution can reduce the acquisition of original images and avoid iterative high-complexity operations by leveraging the strong generalization of deep neural networks, thus significantly reducing the reconstruction speed.

[0066] The FPM reconstruction network structure proposed by the present invention is as follows Figure 7 As shown in the figure, the input is a series of low-resolution images captured by sequential illumination of LED array light sources, and the output is the reconstructed high-resolution amplitude and phase. The network consists of three main parts: data synthesis, residual attention module and Transformer module. Data synthesis synthesizes the FPM raw data into an amplitude tensor and a phase tensor to reduce the computational complexity of the network; the residual attention module combines residual connection and attention mechanism to effectively extract features from the image and suppress the negative problems caused by the increase in the number of convolutional layers; the Transformer module enhances the network's ability to extract multi-scale information by providing a larger receptive field.

[0067] Data synthesis preprocesses a series of intensity images collected by FPM, synthesizes them in Fourier space, and converts them into dual-channel complex amplitude images as input through inverse Fourier transform. This avoids directly extracting features from hundreds of intensity images and reduces the amount of network calculations.

[0068] The residual attention module is as follows Figure 8 As shown in the figure, it introduces residual learning and channel attention mechanism to extract image features. Residual learning can accelerate the convergence of the network and solve the network degradation problem caused by too many network layers. The channel attention mechanism adaptively allocates weights according to the importance of different channel features, so as to emphasize important channels and suppress unnecessary channels.

[0069] Transformer module such as Fig. 9 As shown in the figure, it is used to expand the network receptive field and enhance the expressive power of deep features. A large receptive field is very important for the network to accurately understand the scene and reconstruct the image. In order to expand the receptive field of the convolutional neural network, the general solution is to introduce a dilated convolution. However, its effect is limited and may cause a grid effect. For this reason, the network introduces a powerful Transformer module, in which each Transformer layer has a global receptive field.

[0070] The present invention collects actual wafer sample data on an existing experimental platform, then uses a traditional iterative algorithm to reconstruct the data to obtain the corresponding high-resolution complex amplitude as the true value, and finally divides the training set and the test set. Specifically, the number of LED units in the programmable LED array in the experimental platform is first adjusted to 100 and 200 at equal intervals, respectively, and then the same 1200 samples are taken, and the samples are adjusted to 512×512 pixels after preprocessing. The samples obtained from 200 LED units further use an interactive projection framework to iteratively update the Fourier spectrum, and after convergence, the amplitude and phase of the high-resolution image are obtained, which are used as the true value. Among the 1200 samples obtained from 100 LED units, 900 samples are used as training sets and 300 samples are used as test sets. In addition, in order to avoid overfitting during the training process, the present invention uses rotation, flipping, translation, cropping and adding noise to enhance the data of the training set.

[0071] The present invention adopts a combination of space and frequency domains to design the loss function. In the space domain, the L 1 Norm and SSIM (Structure Similarity Index Measure) are built. Using L 1The norm can make the network learn more detailed information and improve the sparsity of the network, avoid the network from overfitting the training data, and also increase the robustness of the network to outliers, such as noise. The addition of SSIM can make the network focus on the macroscopic structural information in the image. At the same time, in the frequency domain, there will be obvious position differences between high-frequency information and low-frequency information. For example, the high-frequency part corresponding to high-resolution information will be farther away from the center of the spectrum, while the low-frequency information will be in the center of the spectrum. Compared with the spatial domain, it is easier to distinguish the high-resolution information of the sample in the frequency domain.

[0072] Example 1

[0073] For traditional reflective microscopy systems, due to the magnification requirements, the field of view and resolution are often not guaranteed at the same time. However, by using the reflective Fourier stack microscopy system mentioned in the present invention, a high spatial resolution can be achieved while ensuring a large field of view. For reflective micro-surface samples, a DUV light source can be used to obtain more imaging information, further improving the resolution. Fig.10 The figure shows a simulation schematic diagram of microscopic imaging of silicon wafer surface defect detection. (a) shows the microscopic imaging effect of the silicon wafer surface under a low magnification objective lens, and (b) is an enlarged view of the field of view shown in the box in (a); (c) is a high-resolution image obtained by preliminary simulation using the algorithm mentioned in the present invention, and (d) is an enlarged view of the field of view shown in the box.

[0074] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. An adaptive imaging dual-wavelength reflective Fourier stacking microscopy system, characterized in that: The system comprises: A flexible light source module that supports bright field and dark field collaboration, which is used to improve imaging quality, reduce image acquisition time, and increase the resolution of reconstructed images. It also supports flexible illumination of deep ultraviolet light sources. High-speed adaptive optics module, used to detect image aberration in real time and compensate for aberration in real time by controlling the deformable mirror, thus reducing the complexity of image reconstruction and the amount of computation required for image reconstruction; Microscopic imaging module, used to achieve super-resolution image reconstruction and deep three-dimensional reconstruction through dual-wavelength channel images; The computing control platform is used to configure multiple high-performance general-purpose GPUs, use parallel computing to provide sufficient computing power for image reconstruction, and centrally control the software operation of other modules; Image reconstruction algorithm module, used to achieve efficient and universal multi-modal image 2D and 3D reconstruction.

2. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The bright field and dark field coordinated flexible light source module is composed of a lifting bracket, a rotating mirror, a bright field light source and a dark field light source, wherein the bright field light source and the dark field light source are coordinated through the motion control of the rotating mirror and the lifting bracket to achieve controllable multi-angle lighting conditions using a small number of light sources.

3. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The system uses a high optical efficiency DUV LED package as a light source, and the high optical efficiency DUV LED package is sequentially composed of a substrate, a light-emitting chip, an aspherical reflective chip, a housing, a micro-nano surface homogenizing sheet, and a protective glass.

4. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The high-speed adaptive optical module is composed of a wavefront sensor, a beam splitter and a deformable mirror. The beam splitter is used to guide the optical path into the wavefront sensor to detect the system aberration in real time, and the deformable mirror is further feedback-controlled to compensate the aberration in real time.

5. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The microscopic imaging module is composed of an object stage, an objective lens, a beam splitter and a camera.

6. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The image reconstruction algorithm consists of three main parts: data synthesis, residual attention module and Transformer module; Among them, data synthesis synthesizes the original data into an amplitude tensor and a phase tensor to reduce the amount of network calculation; The residual attention module is used to effectively extract features from images and suppress the negative problems caused by the increase in the number of convolutional layers; Transformer module, used to enhance the network's ability to extract multi-scale information.

7. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 6, characterized in that: The residual attention module extracts image features through residual learning and channel attention mechanism, wherein residual learning accelerates the convergence of the network and solves the network degradation problem caused by too many network layers, and the channel attention mechanism adaptively allocates weights according to the importance of different channel features, thereby emphasizing important channels and suppressing unnecessary channels.

8. The adaptive imaging dual-wavelength reflective Fourier stacking microscopy system according to claim 1, characterized in that: The image reconstruction algorithm designs the loss function by combining space and frequency domains.

Images